J I, , ‘ I “3 "g t . The Impact of Comprehensive [131 TIONAL HEALTH INS URANQEJ on Demand for Health Manpower u.s. DEPOSITORY JUN 1 01977 flaw—1' {Gm/M fw/efio‘ t/ - . . a“? ‘ U.S. DEPARTMENT OF HEALTH. EDUCATION. AND WELFARE . Public Health Service - Health Resources Administration ' i ,4. Health Manpower References The Impact of Comprehensive LNA TIONAL HEALTH INS URANCZQ on Demand for Health Manpower July 1976 DHEW Publication No.(HRA) 77-102 US. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Resources Administration Bureau of Health Manpower J £11838qu 5 i CONTENTS Page ‘. it»... Introduction ........................................................................ 1 Tables ............................................................................ vii Figures ........................................................................... viii Summary ........................................................................... 2 I. Description of the Comprehensive Health Insurance Plan ........................... 3 II. Methodological Overview ..................................................... 5 III. Limitations of the Analysis .................................................... 9 IV. The Impact of Chip Upon Demand for Health Services ............................. 13 V. Supplementary Insurance .................................................... 19 VI. Estimated Demand Change Among Income Groups ............................... 21 A. Short-Term Hospital Services .............................................. 22 B. Medical Office Services ................................................... 23 C. Community Pharmacy Services ............................................. 24 D. Dental Services .......................................................... 25 VII. Estimates of the Manpower Impact of Chip ...................................... 27 A. Physicians .............................................................. 29 B. Dentists ................................................................ 29 C. Registered Nurses ....................................................... 30 D. Pharmacists ............................................................ 30 E. Allied Health Manpower .................................................. 30 VIII. The Possible Shift From Inpatient to Outpatient Services .......................... 37 IX. The Impact of Deductibles .................................................... 39 X. Conclusions ................................................................ 43 XI. Development of the Methodology .............................................. 47 A. Projection of the Population ............................................... 47 B. Price Elasticity Assumptions ............................................... 50 C. Elasticity of Demand Estimates ............................................. 51 D. Gross Effective Coinsurance ............................................... 55 XII. Summary of Biases .......................................................... 61 R A 410 .7 U555 1976 FUEL TABLES Table Number Page 1. The cost sharing requirements of CHIP .............................................. 4 2. Estimated percent change in health care demand resulting from CHIP, by supplementation assumption, elasticity assumption. and health service category - linear model ............... 14 3. Estimated percent change in health care demand resulting from CHIP, by supplementation assumption, elasticity assumption, and health service category - semilog model. ............ 14 4. Estimated percent change in health care demand resulting from CHIP, by supplementation assumption, elasticity assumption, and health service category - doublelog model ........... 14 5. Estimated percentage change in health care demand resulting from CHIP, by income and health service category - no supplementation assumed .............................. 21 6. Estimated percentage change in health care demand resulting from CHIP, by income and health service category - uniform 40% reduction in coinsurance ...................... 21 7. Estimated percentage change in health care demand resulting from CHIP, by income and health service category - average 25% reduction in coinsurance scaled to low income groups ........ 21 8. Estimated percentage change in health care demand resulting from CHIP, by income and health service category — average 25% reduction in coinsurance scaled to high income groups ....... 21 9. Estimated percentage change in health care demand resulting from CHIP, by income and health service category - uniform 25% reduction in coinsurance ............................... 21 10. Estimates of health manpower requirements under CHIP by type and health service category: 1970 ........................................................... 27 11. Estimates of the percentage change in overall requirements for health manpower due to 12. 13. 14. 15. 16. 17. . the introduction of CHIP, by elasticity assumption, supplementation assumption and manpower type - figures reflect any estimated decrease in demand for short-term hospital services ...... 28 Estimates of the percentage change in overall requirements for health manpower due to the introduction of CHIP, by elasticity assumption, supplementation assumption and manpower type — estimates assume no decrease in demand for short-term hospital services ............. 29 Estimated percent reductions in care demands with various coinsurance rates due to CHIP deductibles .......................................................... 40 Estimated percentage change in health care demand resulting from CHIP, by income and health service category, assuming decreases in demand due to deductibles provisions - no supplementation assumed ........................................................ 40 Estimates of the population by age, sex, and income: 1976 ............................. 48-49 Selected price elasticity assumptions by health service category ......................... 55 Estimated average coinsurance before CHIP by age, health service category, and income ..... 56 vi FIGURES Figure Number Page 1. General logic of the CHIP demand and manpower requirements model - methodology flowchart. 6 2. Demand curves of the health services demand models ................................... 7 3. Estimated percent changes in demand for selected health services upon the introduction of CHIP — results are based on the semilog model and lower elasticity assumption ............. 15 4. Estimated percent changes in demand for selected health services upon the introduction of CHIP - results are based on the semilog model and higher elasticity assumption ............. 16 5. Estimated percent changes in requirements for health manpower upon the introduction of CHIP - assuming no supplementatary insurance ...................................... 32 6. Estimated percent changes in requirements for health manpower upon the introduction of CHIP, assuming a uniform 25% reduction in CHIP coinsurance due to supplementation ...... 33 7. Estimated percent changes in requirements for health manpower upon the introduction of CHIP, assuming a uniform 40% reduction in CHIP coinsurance due to supplementation ...... 34 8. Estimated percent changes in requirements for health manpower upon the introduction of CHIP, assuming a “hybrid" coinsurance reduction (variable supplementation) .............. 35 9 An example of elasticity curves relating demand to price for health services ............... 52 10. Estimation of the care demand adjustment factor ..................................... 53 vii lilllhllllll Ill. INTRODUCTION The serious consideration given by the Con- gress to various National Health Insurance (NHI) proposals has created a pressing need to evaluate their possible impact on the demand for health care and the requirements for health manpower. Due to the complexity of the interactions which accompany any expansion or reordering of health care demands, it is important to have as much understanding as possible of the probable impact of NHI at the time when alternate proposals are still being formulated and discussed. This is especially important since the impact of NHI will undoubtedly be felt within a short period after its implementation. Although some general analyses of National Health Insurance have been done, comprehensive studies of the possible impact of specific NHI proposals have been notably absent. This is because economic processes in the area of health are not well defined and the subject areas to be addressed by an analysis of NHI are lacking both in pertinent data and in generally acceptable theoretical perspectives. The specific NHI proposal considered in this study is the former administration proposal of February, 1974, generally known as the “Com- prehensive Health Insurance Plan (CHIP)."* While this proposal presents several difficulties in quantification of its impact because of the dif— ferent treatment afforded different income groups, the proposal is probably representative of any future comprehensive NHI bills sponsored by the administration, and may well be the pro— totype of the model for any final legislation. The ‘ The Administration did not submit a NHI bill to the 94th Congress. A bill identical to CHIP, known as the “Comprehen- sive Health Insurance Act of 1974", is contained in H.R. 4747, and is currently under consideration by Congress. present report provides analyses and information on the immediate impact of an NHI plan such as CHIP on certain types of manpower under a set of specific assumptions. It should be clearly recognized that such analyses afford only provi- sional requirements estimates. and these must be interpreted within the limitations of the methodology and assumptions used in their development. While tentative estimates of CHIP impact are presented, the intent of this study was as much to test a methodology and to discuss some of the issues implicit to NHI and health manpower planning as on the development of specific manpower estimates. The principal work for this study was done in late 1974 at the time when CHIP was still under active consideration by the Congress and re ceived limited circulation as Bureau report No. 75-136 (3-1475). This report was prepared in the Bureau of Health Manpower (formally the Bureau of Health Resources Development), Health Resources Ad— ministration, by James M. Cultice and Roger B. Cole of the Bureau’s Manpower Analysis Branch (then the Resource Analysis Staff), Howard V. Stambler. Chief. Special assistance was provided by Ann C. Lawlor of the University of Virginia, then on assignment to the Manpower Analysis Branch. Extensive and special computer pro- gramming assistance was provided by Joan C. Lane and Harold J. King of the Data Systems Branch, HRA, Leonard Sokolower, Chief. The authors are grateful for the assistance of Dr. Allen Dobson of Health Services Administration, DHEW, and Dr. Gail Wilensky, of Health Resources Administration, DHEW, for comments on draft versions of the study. Special thanks to Margaret A. King of MAB, who typed the report in its various draft stages. are in order. SUMMARY This summary provides an overview of the Manpower Analysis Branch's study conducted to assess the probable immediate consequences of the Comprehensive Health Insurance Plan (CHIP), earlier submitted by the Administration, on the demand for selected health services and re- quirements for health manpower.* The specific health service areas studied were short—term hospitals, medical and dental office services, and pharmacy services. Estimated increases in the demand for these services were translated into gross requirements estimates for physicians (MD's and DOS), dentists, registered nurses, pharmacists, and allied health manpower. The study used baseline data consisting of 1970 health services utilization rates by service category and population characteristic; a popula- tion projected to 1976 according to these characteristics; measures of average coinsurance before CHIP for each of these services; and a distribution of health manpower serving each health service area in 1970. The method was to determine the status of present health insurance coverage under the existing private and public systems and to project estimates of 1976 average coinsurance rates expressing the proportions of total expenditures paid out—of-pocket for the respective health services. By using the coin- surance rate as the insurance plan variable and the price elasticity of demand as the basic parameter, the overall change in demand subse- quent to CHIP for each of the health service areas was estimated using three types of demand models. “ The Administration did not submit a NHI bill to the 94th Congress. Various hypotheses about supplementary in- surance acting in conjunction with CHIP were developed and tested in the models. These hypotheses serve to evaluate a wide range of possible demand outcomes in terms of reduced consumer cost—sharing requirements under CHIP, and therefore some of the contingencies, in addi- tion to CHIP, which might act to put stress on the health care system. The results indicate that the impact of CHIP 0n the demand for health care and the consequent re- quirements for health manpower would vary significantly for different types of services. De- mand for short-term hospital inpatient services could very well diminish under CHIP, but this projection must be qualified by the prospects of private supplemental insurance intervening, by way of the upgrading of CHIP plans to current levels of coverage where those differences occur, and other investments in private health in— surance supplemental to CHIP. CHIP is seen as having the potential to greatly enhance coverage for medical office services, and a relatively sudden increase in demand on the order of 20 percent could be generated in that sec- tor. The projected impact of CHIP on demand for dental services is seen as moderate because of the limitation of coverage to children under 13 years of age. However, if the demand for dental care is quite high with respect to price, additional dental manpower requirements under comprehensive National Health Insurance could be substantial. Large-scale increases in the demand for phar- macy services under a program such as CHIP are not foreseen. But again, if price responsiveness is much higher than the available data indicate then demand for pharmacy services could rise con- siderably. l. Description Of The Comprehensive Health Insurance Plan The national health insurance proposal earlier submitted by the Administration, and designated as the Comprehensive Health Insurance Plan (CHIP), would establish a three-part voluntary health insurance program relying extensively on private insurers. The proposal is structured in three subplans: ll Employee Health Insurance Plan (EHIP); 2) Assisted Health Insurance Plan (AHIP); and 3) an expanded Medicare. Covered services of all the plans would be the same. Broad guidelines, including eligibility standards and reimbursable services, would be established in Federal legislation, with the specific legislation regulating private insurance carriers, prepaid group insurance plans, and self-insured employers, enacted by the States. Review of the quality of medical services would be left to Pro- fessional Standards Review Organizations. Under the Employee Health Insurance Plan, employers would be required to offer full-time employees the choice of a basic health insurance plan and Health Maintenance Organization coverage for those persons under 65 years of age. Coverage, deductibles, and maximum liability provisions would have to meet Federal standards. For those covered, 65 percent of the premium costs would be paid by the employer for the first three years of the program, and 75 percent of the premiums thereafter. Under the Assisted Health Insurance Plan, States would contract with private insurance car- riers to offer health insurance coverage to all residents with incomes under $7,500, to the unemployed, to high-risk employer groups, and to high medical risks, including the disabled. This program Would generally replace the coverage now provided under Medicaid, with the exception of services not covered under CHIP beyond basic benefits, e.g., care in nursing homes, home care and care in mental insitutions for those under 21 or over 65 years of age. Medicare would be retained in CHIP, with benefits altered to conform with those offered under the other subplans. Notably, the new coverage would include benefits for outpatient drugs and mental health care, and depending on income, enrollees would have a total maximum liability of no more than $750. Coinsurance under Medicare would be no higher than 20 percent, with a maximum $100 deductible (with an addi— tional $50 for outpatient drugs). Beneficiaries would be charged a $90 annual premium. Benefits under all the above plans would in- clude unlimited inpatient and outpatient hospital and physician's services (excluding physical ex- aminations), outpatient prescription drugs (sub- ject to deductibles), extended care up to 100 days per year, home health services up to 100 visits per year, and mental health services limited to 30 days inpatient care, and 30 outpatient visits to a community mental health center or private prac— titioner (the latter not to exceed 15 Visits). Also included would be preventive pediatric care; postnatal and maternity services; eye, ear, and dental care to age 13; and family planning ser- vices. Prosthetics, ambulance services, x-rays, laboratory services, blood and blood products, and dialysis equipment and supplies would be in- cluded. All persons covered under CHIP would be issued a “healthcard” — an identification card — as evidence of coverage. Participating care pro- viders would bill the insurance carrier for reim- bursable services. The carrier, in turn, would bill enrollees for necessary cost-sharing. Enrollee costs, including premiums, deductibles, coin- surance and maximum liability provisions, would vary with income and would differ somewhat be- tween families and unrelated individuals. These costs for families and individuals are summarized in Table 1. Table l The cost sharing requirements of CHIP Income class All individuals ..... All families ....... Individuals Cost sharing Amount (per person) Maximum family liability2/ Annual income of premium 1/ Regular Drug Coinsurance deductible deductible rate Percent Amount Employee health care plan $240 $150 $50 25 — - $1,050 600 150 50 25 --- 1.500 Assisted health insurance plan O-under $1.750 0 0 0 10 6 O-under $105 $1.750-under 3.500 0 $50 $25 15 9 $158-under 315 3.500-under 5.250 $120 100 50 20 12 420-under 630 5.250-under 7.000 240 150 50 25 15 788-under 1,050 7,000 and over 360 150 50 25 - - - $1.050 O-under $2,500 0 0 0 10 6 O-under $150 $2,500-under 5,000 0 $50 $25 15 9 $225-under 450 5.000-under 7,500 $300 100 50 20 12 BOO-under 900 7.500-under 10,000 600 150 50 25 15 1.125-under 1,500 10,000 and over 900 150 50 25 - - - $1,500 Federal plan for the aged 2/ O-under $1.750 0 O 0 10 6 0-under $105 $1.750-under 3.500 0 $50 $25 15 9 $158-under 315 3,500-under 5,250 $90 100 50 20 1 2 420-under 630 5.250 and over 90 100 50 20 _ - _ $750 1/ Total estimated national average premium at 1975 levels. For employee plan and. in some cases. for assisted plan. the employee and employer would share payment of the premium. 2/ Under Federal plan for the aged. eligible persons would be treated as individuals. regardless of family status. Source: Waldman, Saul.National Health Insurance Proposals. Provisions of Bills Introduced in the 93rd Congress, as of February 1974: Office of Research and Statistics. Social Security Administration. DHEW Publication No. (SSA) 74-11920, February 1974. ll. Methodological Overview The methodology used to quantify the impact of CHIP was based upon available data on utiliza- tion of health care, by service area, and man- power requirements, by manpower types, over all health care categories. It consisted of three general stages of analysis. The first stage en- compassed a projection of the population ac- cording to income in the target year 1976, the year in which the CHIP plan would become effec- tive. In the second stage, population projections were used to update the base year (1970) utiliza- tion levels to the baseline (1976) of the insurance plan implementation. Utilization is measured in visits for all services except pharmacy where prescription volume is the measure of utilization. Specifically, for each of 50 population subgroups (classified by age, sex, and income) a utilization rate was obtained. On the assumption that these utilization rates would be the same in 1976 as they were in 1970, a projection of the total popula- tion’s utilization of care in 1976 was obtained by cross multiplying the projected size of each population subgroup by its associated utilization rate. The level of utilization associated with CHIP was then calculated using (a) estimated baseline (pre-CHIP) average coinsurance levels for the overall population; (b) CHIP-related coinsurance levels; and (c) assumptions regarding the price elasticities of demand by health care area. To assess how demand might be altered by the acquisition of supplemental health insurance, the CHIP coinsurance rates were modified to suggest various degrees and forms of supplementation acting to reduce the cost-sharing requirements of the plan. These illustrative assumptions were ap- plied to reduce the average CHIP coinsurance rates, by varying amounts, uniformly across all income groups and health service categories, and differentially by health service to indicate risk aversion according to the severity of the expen— diture risk. The supplementary insurance ad- justments were also weighted toward both ex- tremes of income level in the model in an effort to indicate how demand might be affected should supplementation appear biased to favor either low or high income groups. Demand change for health services at various levels of detail was then computed from the ratios of the amount of care utilized at the base period (i.e., pre-CHIP) to that projected to occur after the onset of CHIP. The third stage of the analysis was an estima~ tion of the resultant changes in manpower re- quirements due to changes in demand for services associated with CHIP. The relationships between utilization and manpower were assumed to be those of 1970 and remain constant over time. A schemata of this methodology is shown in Figure 1. Estimates of the increases in the demand for health services that might result from CHIP were prepared under two sets of assumptions as to the amount of sensitivity which the demand for health services has for changes in the net price of health care to the consumer (i.e., price elasticity) and three alternate technical assumptions as to the shape of the demand curve. Mathematical models of the shape of the demand curve are needed because typically data are only available for changes over a small segment of the demand curve and the models have the function of relating this partial knowledge to portions out- side of the area for which data were collected and analyzed. As approximations of the true demand curve for the various health care categories, three models—linear, semilogarithmic, and doublelogarithmic—were specified. (More technical descriptions of these demand models are given later in the report.) In all of the quan- tifications, data limitations required two basic assumptions. First, all three models were derived on the assumption that the insured population has the same “demand function" as the uninsured population. and second, the average coinsurance rate (i.e., the cost to the consumer) was the only independent variable in the three models. By virtue of its mathematical properties, the linear model provides the most conservative estimate of the demand for health care. In this model, the price elasticity of demand decreases as the utilization of a particular health service in- creases. The rationale for this is that the time cost and travel cost of seeking extra health care, and the decreasing marginal utility of utilizing ad- ditional units of health care services, cause de- mand increases to lessen as dollar price lessens and more care is obtained. The linear model assumes the shape of the demand curve to be a straight line which intersects the horizontal axis (quantity of service demanded). This is important to note since even at zero price there would be a Figure 1 General logic of the CHIP demand and manpower requirements model-methodology flowchart ESTIMATED 1970 AVER- AGE COINSUR- ANCE LEVELS AVERAGE COIN~ SURANCE LEVELS FOR OVERALL POPULATION IN 1976 ELASTICITY DEMAND MEASURES BY CURVE TYPE OF ASSUMP- HEALTH CARE TIONS CHIP-RELATED ADJUSTMENTS AVERAGE 00- ' FOR SUPPLE- INSURANCE MENTATION LEVELS ASSUMPTIONS 1970 POPU- Eggéggfig 1976 DIS- LATION BY TRIBUTION AGE/SEX/ GREASE IN 01“ POPULATION INCOME PERSONAL BY AGE/SEX INCOME , BASE BASE 1976 , UTILIZATION UTILIZATION RATES LEVELS 1976 POPULATION DISTRIBUTION BY AGE/SEX/ INCOME CHIP CARE UTILIZATION UTILIZATION LEVELS RATES UNDER CHIP 1970 DISTRIBUTION OF HEALTH MANPOWER PERCENT CHANGE IN DEMAND FOR SERVICES UNDER CHIP PERCENT CHANGE IN MANPOWER ‘ ’ REQUIREMENTS DUE TO CHIP limit to the amount of some health service people would demand. Figure 2 illustrates the linear de- mand curve and the other two demand curves described below.1 The doublelog model describes a hyperbola which, when graphed on a natural scale, never in- tersects the horizontal or vertical axes, so that as price approaches zero, demand for a health ser- vice will approach infinity. This is because the doublelog assumption leads to price elasticity be- ing constant at every point along the curve. This ‘ For the development of these models see: Lien-fu Huang and Elwood Shomo (Robert Nathan Associates). Assessment and Evaluation of the Impact of Archetypal National Health Insurance Plans on U.S. Health Manpower Requirements. DHEW Publication No. (HRA)75-1. DHEW/HRA/BHRD, July 1974. Figure 2 model is less realistic than the linear or semilog approximations of the true demand curve only where out-of—pocket costs are very low (about 15 percent or less). The semilog model more closely approaches the linear model in shape (describing a curve between the hyperbola of the doublelog model and a straight line) and characteristics. This curve also intersects the horizontal axis, and like the linear model, the effect of price change on demand is dif- ferent at different levels of out-of—pocket costs. That is, there will be a different value for the price elasticity at any point along the curve. Results from the semilog and linear models would normally be similar, although the linear model always generates more conservative estimates due to its mathematical properties. Demand curves of the health services demand models‘ Health Services Price (P) 10-L 10-— 5 — 5 — 0% 0 Quantity Demanded (Q) Linear Semilog ‘ Adapted from Huang. Lien-fu, and Shomo. Elwood W. (Robert R. Nathan Associates). Assessment and Evaluation of the Impact of Archetypal National Health Insurance Plans on U.S. Health Manpower Requirements. Reprinted by BHRD. HRA, DHEW. July. 1974. 10-— Dou blelog «nun-a...— Ill. Limitations Of The Analysis Although the estimates developed in this report are believed to provide reasonable forecasts of the impact of an NHI plan such as CHIP, there are a number of difficulties associated with the analysis. Most of the pro- blems may be summarized into two catergories— those of data constraints and those resulting from the lack of a generalized mathematical model. While specific data constraints are treated in ap- propriate sections elsewhere, an underlying prob— lem of considerable importance which must be ad- dressed is that of inadequacies in the utilization data base. The most serious shortcomings are the possible distortive effects of using 1970 data on utilization to analyze the implications of an NHI proposal designed for implementation in 1976, and the compromises that were made in the utilization base by aggregating data within each care catergory. Data and time limitations precluded making a determination of changes occuring between 1970 and 1976 in the rate at'which the population will seek health services—changes that are of course highly relevant in assessing the effective impact of an NHI proposal. The level of disaggregation within each health service catergory was largely decided by the availability of data for broad care categories from the National Health Interview " Survey.2 Explict treatment of care requirements and therefore of medical and other specialty re- quirements is only possible with some definition of services, and while diagnosis-level service categories would be preferred, the data source are not yet adequate. Utilization patterns can be expected to have shifted in significant ways in the last five years through changes in the availability of private health insurance, with upward trends toward more third party coverage through comprehen— sive group health plans. Also, there have been substantial increases in the proportion of total ex- penditures for health care met through govern— ment sources in recent years. Although definitive studies are not available, there is good reason to suspect that the general freeze on wages and prices that prevailed from August, 1971, through 2 Source - National Center for Health Statistics, data from National Health Interview Surveys, years 1969-70 (unpub~ lished special tabulations). April, 1974, when controls on prices were finally lifted from the health care market, may have stimulated increases in utilization whose effects still linger. This is especially likely since price controls on major health services persisted after wage controls were removed. Not only did overall changes probably result from the imposition of economic controls, but differences in demographic patterns of utilization may have developed mainly through easier access to the health care market for lower and middle income persons, and for young adults and the elderly with less discretionary income. From the preceding discussion, it is clear that a model which simply assumes the continuation of 1970 age—sex—income—specific care utilization rates is simplistic, essentially only a minor step beyond a simple physician-population ratio model; it can— not mirror the complex changes that are taking place in the health care system today. What is re- quired is a generalized model of the health care system. The model should be comprehensive; it should be sensitive to the interactions of the variety of economic, cultural, and demographic variables which affect the demand for health ser- vices, and it should reflect the ability and disposi- tion of suppliers to respond to the changes in de- mand, and the manner in which they do so. Relatedly, the model should be responsive to policy decisions about desired levels of care and cost-sharing, the levels of demand suppression and rationing which may be generated, and the social costs of the system.3 The lack of a generalized model means that the analysis presented here has been restricted to the total change in visit demand in response to dollar price changes. There are many conse- quences of this. No analyses of changes in the kinds of visits made or of visit content are pos- sible, nor can any real assessment of time cost ef- fects, market interactions, or the effects of these factors on special segments of the population such as the disadvantaged be made. Another limitation of the analysis is that relating to utilization of health services. The analysis will not directly or adequately revealthe 3 The Manpower Analysis Branch is currently sponsoring development of a comprehensive interactive model of the medical care system to replace the model used for this analysis. effects of economic factors on the demand for ser- vices, nor does it reflect the constraints placed on the consumer by events within, and related to, the health care system. Economic factors obvious— 1y include the direct and indirect money costs of health services to the consumer and the ability of the consumer to pay these costs, but also they in- clude the intangible nonmoney costs such as loss of leisure time, travel time to obtain care, waiting time to appointment, and office queues. While measures of these nondollar costs are very dif— ficult to isolate, they are becoming more amenable to quantification as the state of health service and resource data improves.‘ These price variables affect not only a person's decision to seek health services, but also the amount and type of service he uses. The outcome of such deci- sions is influenced by the person's ability to meet these cost—a function of his income and in- surance coverage, and the value he places on the time, income, and other valuables foregone to ob» tain care. The need to introduce a supply response com- ponent into the model is most important; its omis— sion prevents refining projections of the man- power requirements estimates and thus diminishes the ability to attribute a meaningful interpretation to the demand results. If the pres‘ ent health manpower supply were to respond suf- ficiently to produce all of the additional care that might be generated by the implementation of CHIP, then no numerical increases in manpower requirements would be expected. However, it is highly possible that the generation of care de- mand might be controlled by adjustments in the health care market, which make before—and-after comparisons of supply adequacy tenuous. Re- quirements should be modeled within a given price, quantity, and quality of care structure. If, after NHI, the amount of time spent providing a unit of service were cut or if the costs (both monetary and nonmonetary) of acquiring care were increased then it could be concluded that there had been a real increase in the need for ad- ‘ The effect of not explicitly treating nondollar costs does not create a specific bias, as such, because these costs in the post- CHIP period are dependent upon the care supply response to demand increases. That is, while the nondollar costs presently have a certain rationing effect, these costs after CHIP would depend upon the adequacy of the supply response and the ex- tent to which nondollar costs would be required to ration de- mand to the level of the available care supply. Nondollar costs are compensatory and indeterminant except through reference to supply response. 10 ditional manpower. The possibility of this situa— tion arising from conditions of National Health In- surance is very real and its potential for conflict appreciable. The exigencies of the health care market argue that very discriminative and regressive forms of rationing would not be tolerable for long, and the practical residual capacity for productivity expansion in the pre- sent delivery system, while largely unknown, is not thought to be great. It is likely that most of the comprehensive NHI proposals now under review have the potential for creating substantial disequilibrium in the health care market. Mechanisms acting to restore some form of stability might include increased money costs to the consumer, inconvenience in obtaining care, changes in case mix“, lowered follow-up visit rates, and others. The emergence and degree of activeness of such rationing devices will, in significant measure, determine the kinds of ser- vices that are delivered and the segment of the population receiving care. The modeling of the ef- fects of these factors on the utilization of health services is necessary to fully assess the impact of social economic proposals such as CHIP, aimed at reducing economic barriers confronting those of lower income, and others currently unable to ob— tain adequate financial protection against the costs of health care. The model should at the same time be capable of analyzing complementary pro- grams designed to ameliorate the total cost of health care and improve the supply and distribu— tion of health manpower and services. Whereas information on the effects of task delegation to allied health manpower and other forms of man- power substitution and productivity enhance- ment, Health Maintenance Organizations, and 5 An interesting study that succinctly illustrates the implica- tions of some of these points was prepared by John Rafferty of the National Center for Health Services Research, HRA, DHEW (“Enfranchisement and Rationing: The Effects of Medicare on Discretionary Hospital Use." Health Services Research, Vol. 10, No., 1, Spring 1975, pp. 51-62.) Rafferty found indications that Medicare resulted in a change in hospital case-mix proportion, characterized by a decrease in the propor- tion of discretionary admissions among patients under age 65. Conversely, there appears to have been a shift toward a larger discretionary component in the Medicare admissions. Overall, an increase was observed in the proportion of total hospital care going to the elderly. Thus, indications are that a change in insurance coverage for one subpopulation - the elderly - may have had significant allocative effects on the rest of the popula- tion, with the rationing of beds (and services) occurring discriminatively against those facing the higher costs, i.e., the under-65 population. progress in technological innovation has only recently begun to accumulate sufficiently to per- mit general analyses, these factors are vitally im- portant to the successful modeling of NH1 con- tingencies, and future MAB modeling efforts are expected to include them. In summary, it is important to remember that this report is partial victim to the current state of the art regarding health service and resource in- formation. There are many projects now in prog- ress, however, that should markedly improve the data base of future efforts. These include surveys of ambulatory care facilities and other care delivery settings, household interviews, and health examination surveys from the National Center for Health Statistics’ National Health Surveys, and data from professional associations. One most important addition of data from the perspective of this report will be those to come from the RAND Health Insurance Study.6 The ex» perimental portion of this study is expected to ° A detailed description and critique of this study is found in Inquiry, Vol. XI, No. 1, March, 1974. 11 yield valuable information on, among other mat- ters, consumer responses to changes in the price paid for health care, some data on equilibrating mechanisms which might arise consequent to in- creases in demand for outpatient physician ser- vices, investments in private health insurance supplemental to a public plan, and information on the effect of insurance on health status. A seren- dipitous benefit from this study is data collected from the RAND nationwide telephone survey of 9,000 office-based physicians engaged in primary care, undertaken for the purpose of selecting sites from which to conduct the health insurance experiment. This surveygives promise of pro- viding much useful information on physician pro- ductivity factors, accessibility in terms of physi- cians accepting new patients, waiting time, and other characteristics. In summary, the data col— lected from the experiment are expected to make a substantial contribution toward resolving some of the chronic problems in evaluating the interac- tions of health care needs, demands, prices, man- power supply and its responsiveness to price and demand, and other equilibrating mechanisms. IV. The Impact Of CHIP Upon Demand for Health Services The estimates of changes in the demand for health services which might occur within a short time after the implementation of CHIP must be examined in the context of their underlying assumptions. The demand projections developed for this report, in essence, describe changes in de- mand that could be generated given changes in the relative liability of the population for the costs of selected health services before and after CHIP, and the extent to which additional pay- ment sources, supplemental to CHIP, may in- fluence demand for these services. It is important to note at the outset that no single demand estimate presented in this report is advanced as being “the best guess" about what demand for health services, or requirements for health manpower, might arise under CHIP. In- stead the estimations should properly be viewed as describing a range within which actual demand change under CHIP would be expected to fall, should a given set of conditions prevail.While results from the semilog model are most fre- quently discussed because the semilog model is thought to be the least prone among the models toward either overstatement or understatement of demand change throughout a broad range of consumer price changes, there is no empirical basis for choosing any one demand curve shape as being the best predictor of the demand/price rela- tionship. In this regard, the discussion largely pertains to estimates based on the “higher" price elasticity assumptions; this series is favored as it points out the upper bounds of the present de- mand estimates, which are the most crucial and important to identify. Within this context, it must be emphasized that the present study is strictly limited to the impact of CHIP as a prototypical NHI plan and does not include any adjustments for other concurrent changes in the health care system. It does not ad- just for any increases in the role of HMO's or other changes in delivery, such as increased home care services or decreased use of inpatient psychiatric care. It does not attempt to evaluate changes in care delivery such as the increased use of physician's assistants and nurse practitioners, ' nor does it attempt to address quality-of—care issues. Most importantly, this study assumes that 13 the introduction of CHIP leads to no increase in the market price of health services or, in other words, that the supply of health services is perfectly elastic. Any degree of inelasticity of supply would lead to higher prices or other costs of care use and a smaller CHIP-induced increase in the demand for health services. This in turn would lead to a smaller increase in manpower re- quirements. The study does attempt to measure the change in a care system (one similar to that now existing but with a 1976 population) which would occur if CHIP were introduced and no other changes occured. The quantification idealize the situation at equilibrium immediately preceding CHIP and the corresponding situation immediately after implementation of CHIP. As such, they ignore the demand surge that normally occurs when a population obtains new access to health care. Finally, the estimates which speak of an increased demand for specific types of health manpower should be interpreted to mean an in— crease in that type of manpower or in some other type of manpower which is substitutive. In all probability any supply response to demand in- creases will take the form of increased reliance upon modern forms of task delegation and team effort. Thus, the manpower requirements forecasts which are derived from estimated changes in demand should be thought of in terms of services rather than specific types of man- power. The demand estimates presented in Tables 2 through 4, and graphically portrayed in Figures 3 and 4, are aggregated estimates (derived from population subgroups) which represent the per- cent increase in demand associated with the im- plementation of CHIP when compared with the base year (1976) utilization rate estimates. To repeat from earlier, these figures were derived by multiplying each population subgroup (ac— cording to age. sex, and income) by its estimated care utilization rate before and after CHIP, and then dividing the summed utilization after CHIP by the summed utilization before CHIP. The results which assume ”no supplementa— tion" reflect the full effect of the CHIP coin- surance rates, and are not measured with ad- justments for the effects of deductibles. These results for the initial CHIP impact are mainly con- ceived to be a set of “framework" demand estimates which serve as a base structure to which alternative forecasts assuming various degrees of reduced consumer cost-sharing obliga- tions under CHIP can be compared. They do not necessarily represent any final or conclusive estimates. Table 2 Estimated percent change in health care demand resulting from CHIP, by supplementation assumption. elasticity assumtion, and health service category Linear Model Supplementation Health service category assumption Medical Short-term Pharmacy Dental office hospital services office Lower elasticity assumptions No supplementation . . . 7.8 —9.1 5.0 2.5 Public assistance ...... 8.1 -8.3 5.0 2.5 Uniform 25% reduction ......... 9.4 -4.8 5.5 2.8 Uniform 40% reduction ......... 10.3 -2.3 5.8 2.9 Hybrid series ......... 9.4 -2.3 5.3 2.7 Higher elasticity assumptions No supplementation . . . 16.8 -22.7 10.7 15.7 Public assistance ...... 17.4 -20.8 10.7 15.8 Uniform 25% reduction ......... 20.1 -12.0 11.8 17.3 Uniform 40% reduction ......... 22.1 -5.6 12.4 18.2 Hybrid series ......... 20.1 -5.6 11.4 16.6 Table 3 Estimated percent change in health care demand resulting from CHIP. by supplementation assumption. elasticity assumtion, and health service category Semilog Model Supplementation Health servrce category assumption Medical Short-term Pharmacy Dental office hospital services office Lower elasticity assumptions No supplementation . . . 8.2 -8.5 5.1 2.6 Public assistance ...... 8.5 -7.8 5.1 2.7 Uniform 25% reduction ......... 9.8 -4.6 5.6 2.9 Uniform 40% reduction ......... 10.8 ~2.2 6.0 3.1 Hybrid series ......... 9.8 -2.2 5.4 2.8 Higher elasticity assumptio ns No supplementation . . . 18.3 -19.4 11.3 22.8 Public assistance ...... 19.1 -17.8 11.3 23.1 Uniform 25% reduction ......... 22.3 -10.8 12.5 26.2 Uniform 40% reduction ......... 24.7 -5.1 1 3.2 28.3 Hybrid series ......... 22.3 -5. 1 1 2.0 24.8 14 Table 4 Estimated percent change in health care demand resulting from CHIP. by supplementation assumption, elasticity assumtion. and health service category Doublelog Model Health service category Supplementation assumption Medical Short-term Pharmacy Dental office hospital services office Lower elasticity assumptions No supplementation . . . 12.6 -5.2 9.4 4.8 Public assistance ...... 13.5 —4.7 9.4 5.0 Uniform 25% reduction ......... 17.2 -3.0 11.6 6.1 Uniform 40% reduction ......... 21.0 -1.3 13.4 7.1 Hybrid series ......... 17.2 ~1.3 10.6 5.6 Higher elastic ity assumptions No supplementation . . . 29.2 -12.3 21.2 54.5 Public assitsance ...... 31.4 ~11.0 21.2 57.0 Uniform 25% reduction ......... 40.8 -7.1 26.6 80.0 Uniform 40% reduction ......... 50.6 -2.8 30.9 105.0 Hybrid series ......... 40.8 -2.8 24.2 68.0 As is seen in the tables and from the bar charts, CHIP is expected to have widely differing conse- quences across the health services. Because ap- proximately 90 percent of patient expenditures for short-term hospital inpatient care is now paid by private health insurance or government pro- grams, replacement of existing coverage by CHIP would mean that average coinsurance (i.e., the payment by the consumer) would rise 10 percent to about 21 percent and, under the semilog assumption (Table 3), demand for these services would be expected to drop by about 10 to 20 per- cent. However, there are several reasons for believing that this demand displacement will not happen in full force. One strong possibility is that most people covered at that time under a plan with coinsurance provisions more favorable to them than CHIP. either would not enroll in CHIP, or would have their CHIP coverage brought up to the standard of their existing policy. Those with poorer coverage, or no coverage at all, though. would be better off under CHIP, and their de- mand for inpatient care would be expected to rise. The implications of this initial result and some possible interventions are discussed in Chapter VIII. When the linear model is used (Table 2), the range of initial estimates for short-term hospital care is from a decline of 9 percent under the lower elasticity assumption to a drop of 23 percent results are not seen as significantly different under the higher elasticity assumption. These from those recorded using the semilog model. FIGURE 3 ESTIMATED PERCENT CHANGES IN DEMAND FOR SELECTED HEALTH SERVICES UPON THE INTRODUCTION OF CHIP. RESULTS ARE BASED ON THE SEMILOG MODEL AND LOWER ELASTICITY ASSUMPTION. (ESTIMATES DO NOT REFLECT ADJUSTMENTS FOR DEDUCTIBLES IMPACT) NO SUPPLEMENTARY INSURANCE ASSUMED UNIFORM 25% REDUCTION IN CHIP COINSURANCE io‘g do . PERCENT ,. O CHANGE o .. DUE TO SUPPLEMENTATION 30-r UNIFORM 40% REDUCTION IN CHIP COINSURANCE m DUE To SUPPLEMENTATION \g “HYBRID” COINSURANCE REDUCTION (VARIABLE & SUPPLEMENTATION ~ SEE TEXT) 25—— 20—— 15~~ 10—— ' N 5" § i: \ MEDICAL DENTAL PHARMACY OFFICE OFFICE SERVICES SERVICES SERVICES 5.-.. 10___ SHORT-TERM HOSPITAL SERVICES 15-—- ml- 15 PERCENT CHANGE 30T 10—— 15- —_ 20- J— MEDICAL OFFICE SERVICES FIGURE 4 ESTIMATED PERCENT CHANGES IN DEMAND FOR SELECTED HEALTH SERVICES UPON THE INTRODUCTION OF CHIP. RESULTS ARE BASED ON THE SEMILOG MODEL AND HIGHER ELASTICITY ASSUMPTION. (ESTIMATES DO NOT REFLECT ADJUSTMENTS FOR DEDUCTIBLES IMPACT) %i§% NO SUPPLEMENTARY INSURANCE ASSUMED UNIFORM 25% REDUCTION IN CHIP COINSURANCE DUE TO SUPPLEMENTATION UNIFORM 40% REDUCTION IN CHIP COINSURANCE DUE TO SUPPLEMENTATION “HYBRID” COINSURANCE REDUCTION (VARIABLE SUPPLEMENTATION - SEE TEXT) \ N KQVR \h x; \ {Y x \“S. \( \f \- >< 7» x I \‘v w] ,\ DENTAL OFFICE SERVICES SHORT-TERM HOSPITAL SERVICES 16 PHARMACY SERVICES The doublelog model (Table 4) yields a range of estimates from a 5 percent decline to a 12 percent drop under the lower and higher elasticity series respectively. If indeed, demand were to respond to price change in this fashion then these latter estimates would be the most reasonable of the three sets. A different picture is seen in the medical office category. Here, demand is predicted to rise by 8 to 18 percent in the initial estimates according to the semilog model. The linear model produces nearly identical estimates, ranging from 8 to 17 percent, while the doublelog model shows de- mand ranging from 13 percent to almost 30 per- cent from the lower to the higher elasticity series. These results using the doublelog model may be spurious though, due to the implicit assumption of constant price elasticity (see Con- clusions section and technical appendix). It is worth a cautionary note at this point that the de- mand changes evaluated for each sector are estimated independently; it is thus unrealistically assumed for simplification of the analysis that a change in the level of consumption in one health service area has no direct bearing on the level of consumption in another. Where the consumption of services is closely related, as for example inpa- tient hospital care and medical office care, it is useful to examine the projected demand in one in relation to associated changes in demand for the other. This has been done in a later section discussing the possible shift from inpatient hospital to ambulatory care. The estimations for physician office services presented here should be viewed as conservative. A comprehensive NHI program such as CHIP will probably exert a strong influence on demand for these services because coverage for ambulatory care is presently weak compared to coverage for hospital expenses. While the physician‘s office would probably not be swamped with patients demanding services because of CHIP, heavy and sudden demand pressures could arise soon after the program’s onset. This would expectably arise from the standpoint of reduced patient cost- shar- ing burdens, and the release of any backlog of unmet health needs postponed because of finan- cial inabilities, and even from less obvious fac- tors, such as the ease of access through the availability of a “healthcard” which could be used much like credit cards to charge reimbursable health services. It is relevant to look at data on the Canadian government-sponsored compulsory health in- 17 surance program in this regard. The Canadian ex- perience after the introduction of a universal health insurance plan in Quebec found an 83- per- cent increase in waiting time for a doctor’s ap- pointment, increases in office waiting time, and some dissatisfaction in the quality of care after “Medicare".7 Overall, a 17-percent increase in physician office visits, accompanied by a 59- per- cent decline in home visits and a 14-percent decline in telephone contacts were reported for the 1-year period following introduction of the program in Quebec. Because of similarities in the overall structure of the health system of the US. and Canada prior to implementation of the lat- ter's universal health plan, there is substantial reason not to discount the possibilities of somewhat the same experience occuring here under NHI.” Demand for pharmacy services is assumed to be relatively unresponsive to changes in con- sumer price. Since only one elasticity estimate (-0.07) for this service was found in the literature (see technical appendix), an arbitrary elasticity value of -0.15 was assumed for the “high" series. Demand for pharmacy services is estimated to in- crease between 5 and 11 percent in the initial CHIP impact results using both the linear and semilog models, and between 9 and 21 percent when the doublelog model is used. Even though the present insurance coverage for phar- maceuticals is relatively small, the availability of coverage under CHIP is not expected to lead to large increases in demand because of its observed relative insensitivity to price change. But this forecast could be greatly altered if complemen- tary effects from increased physician office utilization resulted in proportionately heavier prescription volume, with physicians prescribing for more patients, and more frequently as their patients became better able to afford medica- tions. Because of the wide variation in the price elasticity coefficients used for dental services in the model, the initial demand estimates differ greatly between the lower and higher elasticity series. Results range from about 3 percent to 17 percent using the linear model, and from about 3 7 Enterline, P.E., et aL The Distribution of Medical Services Before and After “Free" Medical Care ~ The Quebec ExA perience. New England Journal of Medicine. Vol. 289, No. 22, Nov. 29, 1973, pp. 117478. " See: National Health Insurance: Can We Learn from Canada? Spiros Andreopoulos, ed. NY. John Wiley & Sons, Inc.. 1975. 273 pages. percent to 23 percent using the semilog model. The doublelog model produces results ranging from 5 percent to over 50 percent when demand is assumed to be relatively elastic ( = -1.0). Although price elasticity for dental services has been found to be relatively high and existing in- surance coverage for the service is small, there is not expected to be much of an increase in the de- mand for dental care under a program like CHIP. The principal factor working to relieve demand from this sector would be the limitation of coverage to children under 13 years of age.9 However, if the doublelog curve assumption most accurately represents the demand curve for den- “ Although excluded from coverage under the terms of the CHIP legislation, dental benefits were extended to the entire population in the model on a “test" basis. Results from the linear and semilog models ranged from about 12 percent to nearly 100 percent between the two elasticity series. The 18 tal services over a wide range of prices, then the possibility of sizable demand increases within the dental setting exists. Investments in private sup- plemental insurance would aggravate the situa- tion. If, on the other hand, the linear or semilog model together with relatively inelastic demand better describes consumer reaction to price change for dental services, moderate (and more manageable) demand increases would be ex- pected and substantial supplementation could oc- cur before the increases would cause concern. More in-depth discussion of the possibilities is provided in the following section specifically deal- ing with supplemental protection under CHIP. possibility of extensive supplementation of the CHIP benefits coupled with strong demand sensitivity to price changes sug- gests the possibility of extraordinary demand increases if coverage were extended to everyone and therefore suggests that limited coverage would be a wise precaution in the initial phases of a comprehensive NHI program. V. Supplementary Insurance There are a variety of forms which supplemen- tation of CHIP benefits would probably assume. One of the most significant would derive from the unwillingness of persons to accept substantially lower coverage, especially for hospital inpatient services, than they presently possess. There is a large number of persons who enjoy comprehen- sive, generous plans as part of their employee benefits who would not willingly accept a signifi- cant “deliberalization” of their insurance coverage. Secondly, some persons who were in- eligible for higher health insurance benefits through employment plans would voluntarily choose to purchase supplemental coverage; this would take place either because the people would currently be paying less in direct cost for their in- surance coverage and would feel that they could continue paying the same amount, or because they would not be willing to accept less protec— tion than they currently possess. A third source of supplementation would result from the pro- bable continued existence of public assistance and public health programs which provide and pay for health services beyond those which are obtained through private sources and insurance coverage. Persons receiving the benefits of such programs would also be paying for a smaller pro— portion of the total costs of their health care con- sumption out-of—pocket than is suggested by the specified CHIP coinsurance level. In essence then, their protection against their total health service expenditures would be supplemented beyond the CHIP provisions. Uniform supplementation of the CHIP in- surance benefits he, a proportionate reduction in the effective coinsurance rates for all services) produces different effects among the different services. As is seen in Table 3 and from Figure 4, when using the semilog model under the higher elasticity assumption, uniform 25 percent sup- plementation causes a change from an 18—percent increase registered in the unsupplemented basis estimates to a 22-percent increase with sup- plementation in the projected demand change for medical office services; a change from a 19- percent to an ll-percent decline for short-term hospital services; and a change from an 11- percent increase to a 13-percent increase in de- mand for pharmacy services. Demand for dental 19 services rises from 23 percent to 26 percent under these conditions. An assumption of a uniform 40-percent reduction in effective coin- surance rates raises these estimates even higher of course, as is indicated in the tables and bar charts. The impact of uniform supplementation thus would appear to be greatest in the hospital sector where restoration of about the original (pre-CHIP) level of demand could very well be achieved. Additional increases in the demand for dental services with extra insurance in effect would probably be moderate, but the overall de- mand induced by NHI could be large. The estimate for increases in demand for pharmacy services would be elevated slightly, but it would remain in a moderate range. The estimate of de- mand increases for medical office services ap- proaches 25 percent after a 40-percent reduction in coinsurance. Substantial supplementation (on the order of 25 or 40 percent) could thus lead to very high increases in the requirements for physi- cians. Although greater utilization of auxiliary manpower (where smaller increases in re- quirements are projected) would help to alleviate the impact of large increases in demand for medical services, the potential for dislocation ex- ists. Probably the most logical explanation for the present distribution of insurance coverage is the larger financial risk faced by the consumer if he incurs expenditures in an inpatient setting where care is most costly. Since hospitalization implies the presence of acute medical needs and the possibility of major medical bills, the consumer may be most anxious to insure against that even- tuality. Under CHIP, the rationale for insuring against the most costly type of service remains. CHIP’s uniform coverage provisions mean that the health consumer faces potentially higher financial burdens when he acquires hospital inpa- tient services according to the terms of the bill. For these reasons, it is expected that the degree of supplementation will be the highest in the hospital inpatient sector. A “hybrid" sup- plementation series was developed that reflects this expectation, with an assumption of a 40- percent reduction in the effective coinsurance rates for hospital inpatient services. Additional insurance protection would logically be next greatest for services obtained in the medical of— fice sector (25 percent average coinsurance reduc- tion in the hybrid series). Supplementation of on- ly 15 percent was assumed in this series for coverage of dental services and pharmacy ser- vices, where consumer’s expenditures are apt to be more discretionary or financially less substan— tial. Because the hybrid supplementation series is weighted toward inpatient hospital coverage, the most noticeable impact occurs in the demand estimates for short-term hospital services, where it restores most of the demand that was initially shifted out of that sector when the population's protection against hospital costs was largely reduced according to the model’s initial estimates. As Table 3 and Figure 4 show, when a 25-percent supplementation of medical office coverage is assumed, the “hybrid" series raises the demand increase estimates from 18 to 22 per- cent in that sector. Less impact is felt in the phar- macy and dental office sectors where the demand rises from 11 and 12 percent. and from 23 to 25 percent, respectively. An interesting question concerns the potential changes in demand increases should supplemen- tation occur predominantly among the highest or lowest income populations. If the pattern of sup- plementation of CHIP benefits is assumed to resemble the existing pattern of insurance coverage, then supplementation would probably be most prevalent among the high—income groups. This would be especially likely if the continuation of high-benefit group employee plans were a ma— jor component of total post-CHIP supplementa- tion. Because of such factors as different age structures and utilization rates among income groups, the “scaling" of supplementation to in- come produces some differences in the aggregate demand change estimates between the high and low income-scaled versions (see Table 7 and 8). Scaling supplementation toward the high income groups, in contrast to results from biasing coin— 20 surance reductions toward lower income groups, has the effect of generating: (a) slightly lower ag- gregate demand estimates for short-term hospital services, and pharmaceuticals, (b) similar estimates for medical office services, and (c) slightly higher estimate for dental services. If supplementation were scaled to low-income groups, short-term hospital utilization would be greater, and CHIP would have little net impact in that sector. There would be a decrease in overall demand for dental services. One form of supplementary protection which would be held most extensively by low-income groups is public assistance and public health pro- grams. If these programs continued to provide sources through which the poor could acquire health services, then their post-CHIP effective coinsurance rates would actually be lower than those specified in the legislation. Crude estimates of the magnitude of these “free" sources of care were made for this study, and with the assump— tion that they would be continued in force, the im— pact of their continuation as a form of supplemen- tation was estimated. Although they migh play an important role in the acquisition of services by the poor, they would not contribute substantially to the overall demand for health service. In the semilog, higher elasticity estimates (Table 3). with no other supplementation assumed, estimates of the effect of the continued availabili- ty of public assistance programs show a minor im— pact on medical office (raising the demand estimates only from 18 to 19 percent) and on hospital services (changing the estimates from a 19—percent decline to an 18-percent drop). Negligi- ble impacts were projected for dental and phar- macy services because of the small role of public programs in these sectors of the health care market. The possible effect of public assistance programs is best considered as a form of sup- plementation scaled heavily toward the low- income groups. They might then serve as compen- sation for other forms of supplementary in~ surance more available to higher income persons. VI. Estimated Demand Change Among In- come Groups Tables 5 through 9 show the estimated relative demand changes among different income classes for health services. These results are based upon the linear and semilog models under the “higher" alternate price elasticity assumptions, with elasticity held constant across income classes. Estimates are presented without assumed sup- plementation of the CHIP coinsurance rates in Table 5, and estimates which contain different levels of reduced cost-sharing liabilities under CHIP, in Tables 6 through 9. The succeeding discussion refers specifically to results from the semilog model, but pertains as well to the linear model results. Table 5 Estimated percentage change in health care demand resulting from CHIP. by income and health service category 1/ N0 supplementation assumed Health service category Income Medical Short-term Pharmacy Dental office hospital services office All incomes ......... 18.3 -19.4 11.3 22.8 Under $2.500 ......... 25.4 -1.1 13.4 11.3 $2,500-$4,999 ........ 20.2 -11.5 12.0 13.7 $5.000-$7.499 ........ 17.6 —21.2 11.2 17.9 $7,500-$9.999 ........ 16.4 -22.3 10.9 20.1 $10,000 and over ..... 17.1 -25.0 10.7 24.6 1/ Estimates are from the semilog model. “higher" elasticity series. Table 6 Estimated percentage change in health care demand resulting from CHIP. by income and health service category 1/ Uniform 40% reduction in coinsurance Table 7 Estimated percentage change in health care demand resulting from CHIP, by income and health service category 1/ Average 25% reduction in coinsurance scaled to low income groups Health service category Income Medical Short-term Pharmacy Dental office hospital services office All incomes ......... 23.6 -5.7 13.0 24.6 Under $2,500 ......... 35.0 22.1 16.2 15.0 $2.500-$4.999 ........ 35.0 22.1 16.2 21.9 $5,000-$7.499 ........ 35.0 22.1 16.2 27.1 $7.500-S9,999 ........ 32.1 14.2 15.4 31.8 $10.000 and over ..... 17.1 -25.0 10.7 24.6 1/ Estimates are from the semilog model, “higher” elasticity series. Table 8 Estimated percentage change in health care demand resulting from CHIP. by income and health service category 1/ Average 25% reduction in coinsurance scaled to high income groups Health service category Income Medical Short-term Pharmacy Dental office hospital services office All incomes ......... 22.3 -10.9 12.5 26.6 Under $2,500 ........ 25.4 -1.1 13.4 11.3 $2.500-$4.999 ........ 20.2 -11.5 12.0 13.7 $5,000-$7.499 ........ 18.7 -19.2 11.7 17.9 $7.500—$9.999 ........ 19.3 -16.6 12.0 20.1 $10,000 and over ..... 22.9 -11.3 12.6 29.4 1/ Estimates are from the semilog model. “higher” elasticity series. Table 9 Estimated percentage change in health care demand resulting from CHIP. by income and health service category 1/ Uniform 25% reduction in coinsurance Health service category Health service category Income Income Medical Short-term Pharmacy Dental Medical Short-term Pharmacy Dental office hospital services office office hospital services office All incomes ......... 24.7 -5.1 13.2 28.3 All incomes ......... 22.3 -10.8 12.5 26.2 Under $2,500 ........ 29.1 7.4 14.5 12.7 Under $2.500 ........ 27.7 4.1 14.1 12.1 $2.500-S4.999 ........ 25.9 0.2 13.6 16.7 $2,500-S4.999 ........ 23.7 —4.4 13.0 15.5 $5.000-$7,499 ........ 24.2 -6.6 1 3.2 21.3 $5.000-S7.499 ........ 21.7 -1 2.4 12. 5 20.0 $7,500-$9.999. . ..... 23.5 -7.0 13.0 25.4 $7.500-$9,999 ........ 20.8 -13.1 12.2 23.4 $10,000 and over ..... 24.0 -8.8 12.9 30.7 $10.000 and over ..... 21.3 -15.2 12.0 28.3 1/ Estimates are from the semilog model, “higher” elasticity series. 21 1/ Estimates are from the semilog model, “higher” elasticity series. A. Short-Term Hospital Services In general, the direction of change across in- come categories in the initial impact estimates (Table 5) can be anticipated, with emergent de- mand decreasing as income (and the proportion of the population already having insurance) rises. As has been discussed, demand for short-term hospital services would be sharply depressed if CHIP coverage for this service actually replaced the insurance coverage estimated to be in effect prior to CHIP. The expected price differentials are predicted to operate most strongly in the highest income groups where out-of—pocket costs are now lowest and where the cost-sharing re- quirements are highest under CHIP. It should be noted that the decreases in demand for hospital services shown for the individual income groups become much less with declining income; this is a function of the relationship between mildly regressive tendencies in present prices to the consumer for these services and the more pro- gressive CHIP income-scaled coinsurance rates. These results suggest that while the overall pic- ture in the short-term hospital setting is one of rising coinsurance liabilities and sharply curtail- ed demand, the effect would be more pronounced toward persons in the higher income groups who could most afford to make up the price differen- tial and who would most likely already be enroll- ed in the private group health insurance plans that would not undergo “deliberalization” under CHIP or similar comprehensive NHI plans. At the same time, it can be seen that significant demand decreases may occur, given no intervening fac- tors, in the lower middle income groups-— decreases that on the average may not be much less than those of the two highest income groups. These groups include persons who might be con- sidered “medically needy" by 1976, who probably will not have alternative sources of more ade- quate insurance, and who may likely be actuarial- ly sicker than persons of higher income. While it may reasonably be supposed that some of these persons will obtain proper and sufficient health care through ambulatory care facilities which they can better afford, some persons may indeed be denied access to required inpatient hospital services by a price barrier in the form of higher cost-sharing requirements. Were supplementation to occur to the extent that average coinsurance under CHIP were reduced by 40 percent, however, the differences 22 between income groups would diminish con— siderably. As Table 6 shows, demand for short- term hospital services by the lowest income group would change from a decrease of about 1 percent in the initial estimates to an increase of 7 percent, while the demand estimated for the highest income group would rise even more sharply, from a drop of 25 percent in the initial estimates, to a decline of less than 9 percent. Greater demand change from the unsup- plemented CHIP impact estimates to those assuming supplementary insurance would thus be expected for the higher income groups than for the lower. This is because the net price difference between pre~CHIP conditions and the coin- surance rates specified by CHIP is predicted to be greater in the higher income groups where ex- isting hospitalization insurance coverage is more extensive and the price under CHIP would be much higher. Conversely, the initial CHIP impact (not assuming supplemental coverage) is ex- pected to be felt less by the lower income groups where the consumer price for hospital services under CHIP would be only slightly lower than their insurance protection at baseline is predicted to be. Uniform supplementary insurance would then produce a greater relative increase in de— mand for these services among the higher income persons most affected by the initial reduction in price due to CHIP. There is a theoretical concern that is worth raising here also. It will be recalled that the linear and semilog health service demand models used in this analysis have the common property of declining price elasticity with decreases in price, i.e., coinsurance (and increases in utilization). The upper income groups, being subject to the higher CHIP coinsurance rates, are thus more price- responsive than the lower income groups after the 40-percent uniform coinsurance reduction, as price approaches zero. This is a fundamental prin- cipal underlying all of the coinsurance reduction results and the effect is more pronounced as baseline average coinsurance rates become higher. The results of the income-scaled coinsurance reductions, which assume that obtainment of sup- plemental insurance will occur most heavily among either the lower or the higher income groups, are shown in Tables 7 and 8, and reflect an average 25-percent reduction in coinsurance under the higher elasticity assumptions. Were supplementation weighted entirely toward the upper income groups insuring against their characteristically greater expenditure risk, either through independent purchases of addi— tional insurance or through enrichment of CHIP policies to pre—CHIP levels (as would seem likely to happen), short—term hospital demand decreases in the higher income classes would be about half of what is estimated without supplementation; from about a 19-percent drop to an 11-percent decline (Table 8). The overall demand change for short-term hospital services under this assump- tion is of similar magnitude. On the other hand, if CHIP coinsurance reductions were weighted en- tirely to favor low income groups, as shown in Table 7, overall demand change for short-term hospital services would appear to be about that seen in the uniform 40 percent reduction series— close to a 14-percent decrease from the baseline service utilization (Table 5). This is largely a func- tion of the greater numbers of Medicare eligibles in the lower income groups, and the generally higher short-term hospital utilization rates among those of lesser means. Differences in demand changes between in- come groups tend to be moderate for the re- mainder of the service categories, reflecting the use of a variety of free or low—cost medical clinic and social services by those of lower income, which generally makes baseline effective coin- surance progressively less as income declines and thus tends to produce a close positive relation- ship between coinsurance before and after CHIP among the income groups. B. Medical Office Services Although the model forecasts slightly higher demand for medical office services as income falls, and there is support for concluding this strictly in terms of income-related differences in altered financial protection, there are good reasons for viewing this forecast with caution. In- creases in demand for health services by lower in- come persons may be reduced by less willingness or ability to perceive need for medical services, accompanied by less propensity to consult a physician in some cases. Also, the time costs to obtain deferrable or more discretionary care may in some cases be higher for lower income persons than for persons of high income whose employ- ment and other advantages may permit them more leisure or personal time to seek 23 nonemergency health care. On the other hand, there is the more likely alternative possibility that lower income persons may generally place a lower value on their time than persons of greater income. This latter group includes self-employed and professional people with generally greater opportunity costs of seeking care, i.e., whose time may be more valuable to them. However, there is evidence that having a regular source of care also greatly increases utilization and lower income groups more frequently lack this regular source of care.10 Demand for physicians services by low income persons might be further moderated by the availability of, and possibly easier access to, certain health facilities such as hospital outpa- tient clinics or neighborhood health centers which in some instances may be an even less ex- pensive source of health care under a program such as CHIP than at present. Since CHIP would not require physicians to ac- cept a fee schedule as full payment for services under the Employee Plan (contrary to the other subplans), higher income persons covered under this plan might become preferred patients because the physician could charge them more, even though the additional charges would be borne by the patient. It has been suggested that in view of the situation faced by Medicare reci— pients whose doctors refuse to accept Medicare reimbursement as payment in full (i.e., refuse to accept “assignment"l, some physicians might decline to treat Assisted Plan or Medicare patients—in either case resulting in discriminative care favoring persons of higher in- come.“ Finally, there might be subtle or not-so-subtle pressures acting to displace the lower income per- son's demand for physician services by what might be termed the “worried well" in the more affluent groups. “Well~person” care is particular- ly income-sensitive due to its deferrable nature, and while CHIP excludes routine examinations from reimbursable services, at the extreme there ‘0 Aday, Lu Ann. Economic and Noneconomz'c Barriers to the Use of Needed Medical Services. Paper presented at the Eastern Economics Association meeting, October 25, 1974. Note that any form of NHI may change the experience from which elasticity estimates were derived, by making regular care available to persons who prior to NHI would not have established a source of care. This could lead to higher increased utilization than forecast. “ Cruikshank, Nelson H. Statement (on behalf of the National Council of Senior Citizens) in Hearings before the House Com- mittee on Ways and Means, on the Subject of National Health Insurance, 93rd Congress, 2nd Session, June 28,1974. is little that could be done to prevent some pa- tients from successfully filing claims for a variety of optional diagnostic marginal conditions. If de- mand of this form were generated by CHIP and tolerated or encouraged by physicians, then under conditions of relatively inelastic supply (whether through insufficient increases in numerical manpower or in productivity expan- sion) the socioeconomically less advantaged in- come classes might be expected to suffer most through difficulties in competing for the physi- cian’s limited time. In the short-run, this might happen with supply constraints whether or not genuinely abusive demand arose. simply because the traditional entry point to care for the upper income classes is through the office-based primary physician where strong contacts may already be in existence, whereas lower income classes more often may not have a family physi- cian or are used to seeking care directly through clinics or on an emergency basis, i.e., without an appointment.” When supplementary insurance influences are permitted to operate in the model, the pattern of demand for medical services among the income groups is seen to change little from that registered in the initial estimates. Table 9, show- ing the effects of a uniform 25 percent reduction in CHIP coinsurance, when contrasted with the initial estimates shown in Table 5, reveals that changes in demand of the magnitude of 2—4 per- cent in individual income classes would be ex— pected if supplementation were to occur in this fashion. These results suggest that demand pressures arising from additional coverage for ambulatory care may not be appreciably greater than those foreseen in the basic CHIP impact, and that furthermore, supplementation of this form would not act to change significantly the outlook described above for possible patterns of medical office demand among the income classes under CHIP. When supplemented coinsurance reductions are scaled to high income classes (Table 8 shows the average 25-percent reduction series), results are essentially no different from those estimated using the uniform 25‘percent coinsurance reduc- tion assumption. Again, this partly reflects the close relationship existing between the income» scaled CHIP coinsurance rates and coverage at ‘2 These relationships could be further complicated by changes in time costs following market reequilibration in the face of limited supply response. 24 baseline, and therefore the small differences in price change between income groups before and after CHIP. Another factor is the large propor- tion of the total population in the model‘s top in— come class, with the result that only a small reduction in the CHIP coinsurance for this class is required in order to achieve a high-income- weighted average 25-percent coinsurance redue tion. When supplementation is biased toward the low income groups, as seen in Table 7, demand climbs to around 35 percent for the lowest three income groups and up to about 32 percent for the $7,500-$10,000 class.13 Though it is considered unlikely that supplemental insurance would be distributed in this manner or to this extreme, the results do serve to suggest that the existence of supplementation biased toward either income ex- treme would not act to strongly skew demand in that direction. C. Community Pharmacy Services As may be seen from Table 5, differences in de- mand across income groups estimated for com- munity pharmacy services are modest, with only slightly more than 2 percent difference in demand between the extreme income groups. This, as in the case of medical office services, is a function of the positive relationship between baseline average coinsurance, the CHIP income-scaled coinsurance rates, and the low price elasticity assumed for pharmacy services. It reflects the fact that while private health insurance coverage for pharmaceuticals is sparse at all income levels, the lower income groups do have financial resources for this service through Medicaid, and to some extent through other public assistance programs. Physician prescribing behavior could well be altered under CHIP, with prescription volume rising considerably as physicians recognizd the patient‘s greater ability to pay for them. If the conditions described earlier regarding moderate net demand increases for physician office ser- vices by those of lower income should prevail, then it is probable that somewhat the same phenomenon would occur among income-specific ” Rate changes for the three lowest classes are the same because of the very similar baseline rates of insurance coverage and the necessity to reduce CHIP coinsurance to (nearly) zero to effect the income-scaled average 25—percent coinsurance reduction. net demand changes for pharmacy services. De- mand would probably be higher for higher income persons and not as high among those of lesser means. Were supplemental insurance for pharmacy services obtained in the manner shown by Table 9, which reflects a uniform 25»percent reduction in CHIP coinsurance across all income groups for this service, the extra demand generated would appear to be negligible within all income classes. By and large, this is explained by the relative nonresponsiveness of demand to consumer price changes assumed for the pharmacy sector. Moreover, as is noted above, there is a close positive relationship between the CHIP coin- surance levels and baseline coinsurance levels for this service. Both of these factors cause the dif- ferences occurring between income groups to be small even when supplementation is introduced. For comparison to the other income-scaled coin- surance reduction series, the CHIP coinsurance liabilities for pharmaceuticals were reduced by an average 25-percent toward either income ex- treme. These results, depicted in Table 7 for the low income - scaled reduction and in Table 8 for the high income-scaled version, reveal little income-directionality effects, and scant dif» ferences from the uniform 25-percent coinsurance reduction series—or from the basic results for that matter. Results from the 15—percent coin- surance reduction series— the reduction assumed in the “hybrid" supplemental insurance series— are not essentially different from those detailed here for the 25- percent coinsurance reduction, whether uniform or income-scaled. D. Dental Services Estimated demand changes for dental services recorded across income groups are less obvious because of the nature of the coverage specified for CHIP. The proposal states that coverage shall be limited to dental services for children under 13 years of age. Because demand change computed exclusively for the under-13 group does not repre- sent change for the entire population, the dif- ferences in their utilization levels before and after introduction of CHIP had to be expressed as percentages of the total population base. Dental services demand change shown for each income group across this subpopulation is therefore not 25 only a function of coinsurance rates, but also varies directly with the numbers of children per income group. While in these models the lower CHIP coinsurance rates always exert an influence toward greater demand change in the lower in- come groups, there is an offsetting tendency toward a higher proportion of children in the up- per income groups. Therefore, the results shown in Table 5 for the initial CHIP impact indicate greater demand changes among those of higher income—on the order of about 13 percent bet— ween the lowest and highest income groups. Under conditions of a uniform 25—percent reduction in CHIP coinsurance, demand for den- tal services among the income groups changes very little, as is seen from Table 9—less than 1 percent in any income class from the results registered for the initial CHIP impact (Table 5). This is the case also when the average 25-percent reduction is biased toward the higher income groups. Somewhat greater differences among in- come groups are found when the coinsurance reductions are made to favor the lower income groups (see Tables 7 and 8). As has been noted in the study, price elasticity of demand was held constant across income groups for all health service areas. If, in the case of dental services, demand is more elastic in the lower than the higher income classes, a possibly significant bias may have been introduced by the constancy assumption. Intuitively, demand is sensed to be more income-elastic for dental ser- vices than for the other health services con- sidered, but there is presently little evidence to document this view.” Since even the “higher" price elasticity estimates used in this analysis are quite conservative when compared with the few other available measures in the literature (see Techical Appendix), it is safe to say that the use of income—specific elasticities of demand for den~ tal services could yield higher overall demand change estimates than those calculated here, with much sharper differences among income classes expected. " Income-specific demand elasticity coefficients for health services simply do not exist. One notable effort being under- taken in this area is the RAND Health Insurance Experiment, which is expected to provide a significant contribution to pre- sent knowledge of the interactions of price and demand for health services. (see Inquiry, op. cit). VII. Estimates of the Manpower Impact Of CHIP Before attempting to estimate the manpower impact of CHIP, it is useful to review some of the major assumptions and problems associated with these estimates. In the first place, the estimates were obtained by applying the 1970 mix of health manpower by health service area to the projected increases in demand for each of the services (see Table 10). The manpower figures assume that when utilization of a particular service rises a cer- tain percentage, then a similar rise occurs in the numbers of persons employed in that service. The application of a 1970 manpower mix also means that no basic changes in the system of health care delivery were allowed for in the model; services would be produced by the same manpower in the same proportions as in the base year. Such an assumption discounts the possibility that by 1976 there could be a greater dependency on auxiliary manpower, even if greatly increased demands were placed on relatively fixed physician, dentist, and pharmacist supplies.15 A second shortcoming in the estimates derives from the model's inability to account for changes '5 It is of importance to recall in moving directly from quan- tity of services demanded to health manpower requirements the implicit assumption that productivity factors, including greater use of traditional aides, the use of expanded function aides, increased organizational and technological efficiency and increases in the primary provider's hours of work, do not appreciably alter the requirements for manpower by 1976. in the market equilibrium, whereby demand in- creases might be limited through the introduction of higher prices or other rationing devices. If the supply of resources is not sufficiently elastic to satisfy the demand for services which might be initially generated, the introduction of higher costs for the acquisition of care would check the demand increases. This would lead to an apparent lessening in manpower requirement. The figures generated by the model estimate the manpower that would be required in the absence of such higher costs— perhaps a more desirable target than requirements based on a suppressed level of demand. Finally, the inability of the model to account for interrelationships between inpatient and outpa- tient utilization causes some unrealistic estimates to be obtained for certain manpower types. It is likely that a very high proportion of hospital pa- tients have significant health needs that require treatment. Even if these persons are discouraged from utilizing inpatient facilities, therefore, they would probably not be forced entirely out of the health care market. It is also likely that the same types of health manpower would be utilized in the treatment whether it took place in a inpatient or an outpatient setting. The projections from the model that show possible decline in registered nurse and allied health manpower requirements are therefore thought to be highly unrealistic. Table 10 Estimates of health manpower requirements under CHIP by type and health service category: 1970 (in 1,000’5) Health service category Manpower Medical Short-term Pharmacy Dental All other health type office hospital services office services and manpower activities Physicians (M.D. 8; 13.0. ............ 152.2 118.2 - 52.3 Dentists ............. - 2.3 ‘ 85.1 1 4.8 Registered nurses ..... 133.5 424.0 165.5 Registered pharmacists ....... - 1 1 .4 106.5 11.5 Allied health manpower ........ 241.7 1,910.6 0.1 174.4 1,364.6 27 They appear because the decline in hospital inpa- tient utilization that is projected to occur in the absence of substantial supplementation is not properly connected to a possible rise in outpa- tient utilization. With the probable high sup- plementation of inpatient services and the pro- bable substitution of outpatient services. a decline in requirements for those manpower types that are based largely in institutional set— tings is not anticipated. As part of the supplemen- tation assumptions. therefore, it was assumed that the population might not be willing to accept lower coverage for hospitalization services. When this assumption was evaluated, the projected re- quirement estimates rose substantially from the basic estimates for all types of manpower except dentists. A further limitation on the manpower estimates derives from the large numbers of diversely specialized manpower who are treated as part of a basic, homogeneous group. Although the physician requirements estimates do not imp- ly an overall shortage problem in the absence of large amounts of supplementation, there may still be sizable and troublesome demand shifts within specific physician specialties. The diversity within the allied health classification also prevents any estimates of demand increases for ‘any particular occupation from being made within that category. If there were no major supplementation of the CHIP provisions, with the possible exception of dental manpower there would appear to be little cause for real concern about the ability of ex- isting resources to satisfy the additional demand which CHIP might generate. In the initial basic estimates using the “higher" elasticity assump- tion and the semilog model (Table 11 and Figure 5), the most substantial requirements increases are predicted for pharmacists (8 percent), and dentists (19 percent). There is a very small an- ticipated increase in the need for physicians and ‘no projected increase for either registered nurses or allied manpower. Overall, then, the strain on the existing health manpower supply would probably be negligible, except for dentists which could experience considerable increases in demand. Once the possibility of supplemental coverage is acknowledged, however, the estimated changes in manpower requirements present a different picture. In the “hybrid” supplementation series estimates (Table 11 and Figure 8), under the semilog model, “higher" price elasticity assump- tions, requirements estimates for all manpower types except dentist and pharmacists are raised markedly. The physician estimate rises from a modest 2-percent increase in the basic series (Table 11) to a substantial 9-percent increase; the registered nurse estimate changes from an 8- percent drop in demand to a l-percent increase. The allied health manpower estimate similarly goes from an 8-percent drop to no change in de- mand. Additional supplementation would elevate the manpower requirements estimates even higher. It is thus apparent that moderate levels of Table 1 1 Estimates of the percentage change in overall requirements for health manpower due to the introduction of CHIP, by elasticity assumption, supplementation assumption and manpower type 1/ Figures reflect any estimated decrease in demand for short-term hospital services Elasticity assumption Allied Supplementation Physicians Registered Registered health assumption (M.D. 8: D.0.) Dentists nurses pharmacists manpower Lower Higher Lower Higher Lower Higher Lower Higher Lower Higher No supplementation ...... 0.7 1.5 2.0 18.6 -3.5 -8.0 3.5 7.6 -3.8 -'7.7 Uniform 25% reduction . . . 3.0 6.6 2.4 21.6 -0.9 -2.2 4.2 9.3 -1.6 -2.9 Uniform 40% reduction . . . 4.3 9.8 2.6 23.6 0.7 1.6 4.7 10.4 -0.3 0.3 Hybrid series ............ 3.8 8.6 2.3 20.6 0.5 1.1 4.3 9.4 -().4 0.0 1/ All estimates are from the semilog model. 28 supplementation could seriously affect the supply of health manpower needed to satisfy the levels of demand that would be generated after the in- troduction of comprehensive NHI. For either set of elasticity assumptions, and regardless of the amount of supplementation, similar requirements estimates are obtained under the linear and semilog curve assumption. The comparable doublelog curve estimates ap- pear to be unrealistic, especially where sup- plementation appears to bring the costs of acquir— ing services near zero. Since the model fails to in- corporate nonmoney costs, the doublelog curve (which allows demand to rise indefinitely as money price approaches zero) produces par— ticularly untenable results. In actuality, the ra- tioning costs, such as such as time, which would be generatdd under a high demand load would keep the total costs of obtaining care well above the zero costs assumed by the doublelog curve. One estimate, for example, predicted that a 30- percent increase in waiting time could reduce out- patient utilization by 10 percent. The selection of different sets of elasticity assumptions produce major differences in the demand estimates. Under both the linear and semilog curve assump— tions, use of the set of higher elasticity assump» tions causes the projected requirements estimates to more than double for each manpower category when compared with results using the lower elasticity assumptions. A. Physicians In the first set of estimates, which were run without adjustments for supplementation or shifts in demand between services, the projected increases in physician requirements were minimal (around 1 or 2 percent). However, once modifications were made to adjust for supplemen- tary insurance, the estimates rose substantially. In the hybrid supplementation series (Table 11 and Figure 8), where the estimated overall decrease in hospital utilization is only 5 percent, and 25 percent supplementation for medical office services is assumed, the semilog, higher elasticity assumption leads to projected requirements in- creases of nearly 9 percent for physicians. Relatedly, if it were assumed that no loss in physician requirements would result from the projected declines in hospital utilization (because of demand shifts and supplementation) estimated increases in physician requirements would be raised to more than 8 percent with no other sup- plementation and over 12 percent with uniform 40 percent supplementation. (See Table 12 and Figures 5 and 7.) B. Dentists The limitation on coverage of dental services by age under CHIP helps to restrain the potential impact on requirements for dental manpower. Table 1 2 Fstimates of the percentage change in overall requirements for health manpower due to the introduction of CHIP, by elasticity assum tion, supplementation assumption and manpower type Fstimates assume no decrease in demand for short-term hospital services Elasticity assumption . Physicians Registered Registered Allied Supplementation (M.D. & D.O.) Dentists nurses pharmacists health assumption manpower Lower Higher Lower Higher Lower Higher Lower Higher Lower Higher No supplementation ..... 3.8 8.6 2.2 19.1 1.5 3.4 4.2 9.3 0.7 2.3 Uniform 25% reduction... 4.7 10.7 2.5 21.9 1.9 4.5 4.7 10.3 0.9 3.0 Uniform 40% reduction... 5.3 12.1 2.6 23.7 2.2 5.2 5.0 11.0 1.1 3.5 Hybrid series ........... 4.8 10.9 2.4 20.8 2.0 4.8 4.5 10.0 1.0 3.2 All estimates are from the semilog model. However. if demand is relatively elastic. re- quirements would probably rise considerably under CHIP. In the initial basic estimates, which are based upon no supplementation and an elasticity coefficient of -1.0, increases of about 19 percent are projected for dentist requirements (Table 11). Assuming relatively inelastic demand for dental services, the estimated increase in manpower requirement would be only about 2 percent. If the pattern of supplementation under CHIP were to be expected to parallel existing patterns of coverage then the existing low levels of dental coverage would mean that supplementa— tion would do little to further increase dental manpower requirements. In the “hybrid" series,which forecasts the effects of CHIP assum- ing 15- percent supplementation of dental and pharamacy services and higher supplementation of medical services, the estimates show a 21— per- cent increase under the semilog curve with higher elasticity, and a 2—percent increase when inelastic demand is assumed. If demand for dental services is, in fact, elastic (elasticity coefficient of unity or greater) there could be a significant in» crease in the number of dentists required, ac— cording to those estimates. Some of the increased demand could probably be accommodated through greater use of allied health manpower, who are projected to experience only small in— creases in demand requirements, but the poten— tial for strain in the supply of dental manpower appears to exist. C. Registered Nurses None of the initial estimates for RN re- quirements predict large increases. Because such a high percentage of active nurses are employed in hospital settings (nearly 60 percent), the pro— jected declines in hospital utilization that would occur without substantial supplementation hold down the projected nurse requirements estimates. Even if it were assumed that the nurses who are presently employed in the treat- ment of hospital inpatients would continue to treat them in the same numbers either on an inpa- tient or outpatient basis, the basic higher elastici- ty semilog estimates (Table 12 and Figure 5) show an increase in requirements of under 4 percent. However, supplementation elevates the estimates only to around 4 or 5 percent. Overall, under current staffing and employment patterns there would seem to be little reason for concern 30 about the adequacy of the numerical supply of registered nurses overall. Given the large number of trained nurses who are thought to be professionally inactive, it would be anticipated that the available supply could adjust with little problem to any demand change which might arise under a National Health Insurance plan like CHIP. D. Phai macists Since the observed elasticity of demand (n = -0.07) for pharmacy services showed demand to be relatively insensitive to price changes, the model’s projected increases in estimated re- quirements for pharmacists are generally small. Even under different levels of supplementation and an arbitrary demand elasticity value of -0.15 (roughly twice the observed value) the projected increases above baseline range only from 8 to 10 percent under the semilog curve assumption (Table 11). Additional supplementation beyond the CHIP benefits would not appear to have a ma- jor effect in further increasing the requirements for pharmacists. Some of the difference between the lower and higher estimates can be attributed to the projected declines in hospital utilization and the proportion of pharmacists employed in the hospital sector. It is felt that any need for in— creases in the supply of pharmacists suggested by the figures would probably be overstated because of the potential for substituting ancillary manpower in the performance of a number of tasks in the pharmacy service sector which do not require pharmacist training. Furthermore, as in nursing, there are many trained pharmacists who are professionally inactive and who might be in- duced to return to practice if a serious shortage of pharmacists arose. E. Allied Health Manpower The assumptions used to derive initital estimates of manpower requirements probably exert their greatest effect in the case of allied health manpower, because they are highly sen- sitive to the projected decline in hospital utiliza- tion and to the model's inability to accept a chang- ing manpower mix. Even if no expansion of de— mand were to occur under CHIP, there would pro- bably be an increase in the requirements for allied health manpower with the expected con- tinuation of the growth in task delegation and in health maintenance organizations. The mean— ingfulness of the allied manpower figures is also limited by the broad range of occupations and skills contained in that category; there would pro- bably be wide variations in the impacts of CHIP on the requirements for health manpower among the different types of allied health personnel. The general viewpoint that National Health In— surance, per se, would not create much of an add- ed burden on the supply of allied manpower is lit- tle changed when adjustments for supplementa— tion and for lower inpatient hospital coverage are incorporated into the model. When either the semilog or linear demand curves are used, neither high elasticity nor large amounts of supplementa- tion indicate prospects of vast increases in addi- tional requirements for allied health manpower. In the “hybrid” supplementation series which assumes no net decline in demand for short-term hospital services (Table 12 and Figure 8), pro- jected increases in manpower requirements are estimated at about 3 percent for both the semilog curve and the linear curve assumption. The relatively small projected impact on this man- power source would probably enable greater utilization of allied health personnel to provide services in areas such as medical offices and out- patient clinics where greater demand impacts are expected and where larger increases in re- quirements for other personnel are anticipated. Since some of the projected increases in re- quirements for physicians, dentists, and phar- macists could probably be accommodated by greater utilization of allied health personnel and registered nurses, the manpower requirements 31 for the allied and RN manpower sectors could well exceed the projected estimates. As has been indicated, in order to account for possible distortions caused by the inability of the model to place those persons removed from the hospital setting back into the health delivery system, one series of the manpower requirements estimates were further adjusted to reflect a con- tinued demand for the services of the health man- power who would ordinarily have been serving those persons removed by the model from the hospital setting. The modification was based on the assumption that there would be no overall decline in the number of persons serving the inpa— tient hospital setting. In this way, the negative impact of the projected decline in hospitalization was eliminated. The effect of this modification is most noticeable for those types of health man- power (nurses and allied health manpower) which are employed most regularly in the hospital set- ting. Overall, it provides what are considered to be more realistic projections of future re- quirements for these personnel, resulting in positive projected requirements for all categories of health manpower. As is seen in Table 12 and Figures 5 through 8 (unshaded portions), the estimated increases in requirements for registered nurses and allied personnel remain small, generally under 5 percent. Since a signifi- cant proportion of physicians are employed in in— stitutional settings, this modification has some impact on the projected increases in physician re- quirements, although they continue to lie predominantly in a moderate range. The modifica- tion has little effect on the estimate for phar- macists and dentists, who work largely outside of the hospital setting. FIGURE 5 ESTIMATED PERCENT CHANGES IN REQUIREMENTS FOR HEALTH MANPOWER UPON THE INTRODUCTION OFCHIP, ASSUMING N0 SUPPLEMENTARY INSURANCE. RESULTS ARE BASED ON THE SEMILOG MODEL AND ALTERNATE ELASTICITY ASSUMPTIONS. ‘ m LOWER ELASTICITY ASSUMPTION HIGHER ELASTICITY ASSUMPTION PERCENT CHANGE 2O -'- PHYSICIANS DENTISTS REGISTERED PHARMACISTS ALLIED HEALTH REGISTERED MANPOWER —_ NURSES 1°‘—— SHADED AREA INDICATES CHANGE IN REQUIREMENTS REFLECTING ANY ESTIMATED DECREASE IN DEMAND FOR SHORT-TERM HOSPITAL SERVICES UNSHADED AREA INDICATES CHANGE IN REQUIREMENTS THAT ASSUME NO DECREASE 1N DEMAND FOR SHORT- TERM HOSPITAL SERVICES 32 PERCF' 1T . CHANGE FIGURE 6 ESTIMATED PERCENT CHANGES IN REQUIREMENTS FOR HEALTH MANPOWER UPON THE INTRODUCTION OF CHIP, ASSUMING A UNIFORM 25% REDUCTION IN CHIP COINSURANCE DUE TO SUPPLEMENTATION. RESULTS ARE BASED ON THE SEMILOG MODEL AND ALTERNATE ELASTICITY ASSUMPTIONS. I m LOWER ELASTICITY ASSUMPTION HIGHER ELASTICITY ASSUMPTION PHYSICIANS DENTISTS REGISTERED PHARMACISTS REGISTERED NURSES SHADED AREA INDICATES CHANGE IN REQUIREMENTS REFLECTING ANY ESTIMATED DECREASE IN DEMAND FOR SHORT-TERM HOSPITAL SERVICES UNSHADED AREA INDICATES CHANGE IN REQUIREMENTS THAT ASSUME NO DECREASE IN DEMAND FOR SHORT- TERM HOSPITAL SERVICES 33 ALLIED HEALTH MANPOWER FIGURE 7 ESTIMATED PERCENT CHANGES IN REQUIREMENTS FOR HEALTH MANPOWER UPON THE INTRODUCTION OF CHIP, ASSUMING A UNIFORM 40% REDUCTION IN CHIP COINSURANCE DUE TO SUPPLEMENTATION. RESULTS ARE BASED ON THE SEMILOG MODEL AND ALTERNATE ELASTICITY ASSUMPTIONS. I m LOWER ELASTICITY ASSUMPTION HIGHER ELASTICITY ASSUMPTION PERCENT CHANGE 25 PHYSICIANS DENTISTS REGISTERED REGISTERED NURSES PHARMACISTS ALLIED HEALTH MANPOWER SHADED AREA INDICATES CHANGE IN REQUIREMENTS REFLECTING ANY ESTIMATED DECREASE IN DEMAND FOR SHORT-TERM HOSPITAL SERVICES UNSHADED AREA INDICATES CHANGE IN REQUTREMENTS THAT ASSUME NO DECREASE IN DEMAND FOR SHORT- TERM HOSPITAL SERVICES 34 FIGURE 8 ESTIMATED PERCENT CHANGES IN REQUIREMENTS FOR HEALTH MANPOWER UPON THE INTRODUCTION OF CHIP. ASSUMING A “HYBRID” COINSURANCE REDUCTION (VARIABLE SUPPLEMENTATION - SEE TEXT). RESULTS ARE BASED ON THE SEMILOG MODEL AND ALTERNATE ELASTICITY ASSUMPTIONS. m LOWER ELASTICITY ASSUMPTION PERCENT CHANGE 25-r—- HIGHER ELASTICITY ASSUMPTION —l..— 20-— 15-— E_ 10~P 5__ C [Ill/ll! PHYSICIANS DENTISTS REGISTERED REGISTERED ‘ . —H NURSES PHARMACISTS ALLIED HEALTH MANPOWER T 5__l SHADED AREA INDICATES CHANGE IN REQUIREMENTS REFLECTING ANY ESTIMATED DECREASE IN DEMAND FOR SHORT-TERM HOSPITAL SERVICES UNSHADED AREA INDICATES CHANGE IN REQUIREMENTS THAT ASSUME N0 DECREASE 1N DEMAND FOR SHORT- TERM HOSPITAL SERVICES 35 VIII. The Possible Shift From Inpatient To Outpatient Services One almost certain consequence of CHIP would be a greater increase in demand for office—based physician care, hospital outpatient, and other clinical services, than for the inpatient services of hospitals, since most of the population already has extensive hospital insurance coverage more liberal than CHIP while coverage for physician services is comparatively limited. The likelihood of such a shift is greatly increased by the fact that, overall, the consumer's share of the total cost of short-term hospital care is expected to rise under CHIP. To the extent that persons do not supplement the public plan with additional coverage for hospital care, the higher net price of receiving diagnostic or other services on an inpa- tient basis could very well help to curtail unv necessary use of expensive inpatient facilities. The equalizing of coinsurance levels across ser— vice categories as specified in the CHIP legisla- tion is viewed as intentional—designed to make consumers share the financial liability of the higher relative costs of hospitalization and thus to alter current incentives which sometimes make the most expensive forms of care least cost- ly to the consumer. HEW Secretary Casper Weinberger. in a statement before the Senate Committee on Finance in May, 1974, expressed the hope that a plan like CHIP could “achieve a significant reorientation of the use of services" and that through uniform cost-sharing across ser- vices, demand could be “shifted from acute and in— patient to preventive and ambulatory care." In ef- fect, the CHIP benefits were apparently designed to counter an established trend towards greater protection against the costs of hospitalization. The benefit package is not intended to force persons who may currently be overutilizing hospitals completely out of the health care market, but only to provide them with proper in- centives so that they (or their physician) would choose less expensive forms of care when ap- propriate. The anticipated reduction in hospital utilization would probably be accompanied by a shift towards greater utilization of outpatient services. The probability of such a shift is sup- ported by findings of a 1972 study that a signifi- cant positive relationship existed between price for inpatient services and utilization of outpatient 37 services. With estimates that the cross-price elasticity fell somewhere between 0.85 and 1.46, the authors suggested that a given percentage in- crease in inpatient prices would lead to a similar percentage increase in outpatient visits.l6 Although these figures may not be directly ap- plicable to the MAB model, they do succeed in making a crucial point— namely, that if utilization of inpatient services were driven down by a rise in their cost, much of the suppressed demand might reemerge in outpatient settings, where the price and net cost to the consumer would be lower. According to many observers, utilization of hospital services could be substantially reduced without denying anyone needed care. Estimates of hospital overutilization generally fall within a range of 15 to 25 percent. Programs which have attempted to control abuse have claimed reduc— tion on the order of 15 to 25 percent in the average length of stay,17 and the number of un- necessary admissions has been estimated at 15 to 20 percent.18 If the alteration in incentives which presently favor hospitalization were properly designed, projected decreases in hospital utiliza- tion of 15 to 20 percent (which might occur if there were little supplementation of CHIP benefits) would not appear to be unreasonable. When the effects of different coinsurance rates and deductibles under different demand curve assumptions were calculated, there was an im- plicit assumption that the cross price elasticity of demand among the different types of services was zero; in other words, it was assumed that level of consumption of any service was not af- fected by changes in the price of any other. However, in some cases, health or medical ser- vices can be considered to be complementary; '3 Davis, Karen and Russell. Louise B. “The Substitution of Hospital Outpatient Care for Inpatient Care." The Review of Economics and Statistics. Vol. 2, May, 1972, pp. 109-113. ” Stockton. William. “Useless Admissions Add to High Cost of Hospitalization." Washington Post. April 18. 1974. p A21. ‘8 Denenberg. Herbert S. Statement in Hearings before the House Committee on Public Health and Environment. Commit- tee on Interstate and Foreign Commerce, on National Health Insurance and Health Care. 93rd Congress, lst and 2nd ses- sion,1973—1974. that is, a price-induced decrease in consumption of one service might be accompanied by a con- sumption decrease for another. In many other cases, these services can be considered substitutable (that is, increases in the price of one service would induce increases in the consump- tion of alternative forms of care). Unless there are no complementary or substitution effects, or unless they offset each other, there would be a nonzero price elasticity between the two ser~ vices. If the net difference between the two ef— fects were significant, failure to take account of the impact of changes in price (or levels of in- surance coverage) for one service on consumption in others would produce inaccurate demand estimates. _ It should especially not be expected that higher cost-sharing of inpatient expenditures would result in lower utilization of inpatient services without affecting the utilization of other services. Although persons may be discouraged from ac- quiring inpatient hospital services because of cost, their demand for services may not be com— pletely dissipated; if it were based on real medical need, it would probably be transferred to alter— native forms of care which are less expensive, such as hospital outpatient or medical office ser- vices. Failure to account for this possibility may result in a sizable underestimation in the demand for outpatient services. It is generally felt that the benefits specified in CHIP would be substan- tially supplemented. in a variety of ways, and that reduction in hospital utilization would be of much lower magnitude than initially estimated. In spite of the maximum liabilities under CHIP, consumers would probably retain their high aver- sion to the risks of inpatient costs and choose to secure a higher level of protection against their occurence. The weighting of supplemental in» surance coverage towards expensive services could frustrate the intended impact of CHIP’s uniform benefit program to reduce unnecessary, costly hospital utilization. In addition to supplemental coverage, other factors will affect the size of the shift from inpa- tient to outpatient services. One possibility is that CHIP, by increasing demand pressure on am- bulatory physician care services, might en~ courage physicians to make more extensive use of inpatient services for those who can afford them, even in providing treatment which is generally 38 performed in the medical office now. They may also find it more profitable with respect to the use of their own time to use hospital facilities when the option exists. As a somewhat counterbalancing influence, CHIP‘s offering of an HMO option and growing HMO participation could alleviate pressure on hospital inpatient services. The parallel develop- ment of PSRO's is also an important factor from the standpoint of utilization review committees, which should serve to suppress the volume of un- necessary or prolonged hospital stays. There is some concern aboutthe inflationary impact on hospital inpatient prices which might evolve from a shift towards the outpatient sector. First, because a significant portion of the total costs of providing care are fixed or necessary regardless of the number of patients (such as debt repayment, plant maintenance, utilities, and minimum staffing levels), the average cost per pa— tient would rise as these costs were distributed over a smaller inpatient population. Further— more, as patients with the least acute needs were shifted out of the inpatient setting, the patients who are most expensive to treat would remain. This change would also contribute towards higher average patient expenses. Together, these forces might work towards restraining the de- mand for inpatient services further. The initially high demand increases (estimated between 18 and 26 percent in the semilog, “higher" elasticity estimates) which are projected for the medical office setting do not reflect the possible shift to outpatient services from inpa» tient services which might be induced by CHIP. It would appear unlikely that this sector of the health care delivery system could immediately ac- commodate demand increases much in excess of '20 or 25 percent. Since costs could be expected to rise under the added demand pressures, there .might well be a shifting of some persons back into the hospital inpatient setting, and others might be priced completely out of the market. At the same time, however, if the projected decline in utilization of hospital inpatient services were to occur, there should be resources freed by that sector which could be called upon to alleviate the demand pressures on the outpatient sector and which could help limit the rationing of care necessary to control demand there. IX. The Impact Of Deductibles The demand estimates that have been presented reflect only the average coinsurance rate as the insurance variable, and the effect of deductibles was not taken into account; i.e., the deductible component of the average coinsurance rate was assumed to be zero. The deductible is a stipulated amount which the consumer must pay for a reimbursable health service before the benefits provisions of health insurance policy begin to cover any portion of the expenses. Like coinsurance, the inclusion of deductibles in an in- surance plan is commonly justified on the grounds that they function in suppressing un— necessary utilization of a health service, and reduce the premiums paid by the insured by reducing the liability of the insurer. To the extent that deductibles provisions are active in con- straining demand, the demand estimates which have been developed are subject to upward bias.19 However, there is much disagreement among health economists over the effectiveness of deductibles in deterring utilization. The studies that have been conducted suggest that insurance companies have little idea what effect deductibles have on utilization or cost—sharing.2°,2‘ (See also ‘9 Huang and Shomo (op. cit., pp. 91-2) posit that if the amount of a deductible is very high and expected discretionary health expenditures are so low that the insured consumer is unlikely to exceed it, then he acts as though he is not insured and his marginal coinsurance rate (of 1) has little effect on utilization. On the other hand, with a low deductible, the chances of the consumer meeting it are high and his marginal coinsurance rate closely relates to his consumption of health services. To further explain their reasoning, the authors define a variable E(k), the “expected" coinsurance rate, equal to P1*K+P2, where P1 is the probability of meeting the deductible, P2 is the probability of not meeting the deductible, and P1+ P2= 1. In the case where the consumer is certain to meet the deductible, P1=1 and the expected coinsurance rate. K, will equal the marginal coinsurance rate. Where the consumer is certain not to meet the deductible. P1+ 0 and E(K) will equal to 1, i.e., the consumer behaves as though uninsured. Thus, the closer E(Kl is to K, the less likely the demand estimates are to being overstated. Conversely, the closer ElKl is to 1, the greater the probability that the results would be overstated. 2° Zelten, Robert A. The Impact of Cost Sharing on the Utilization of Health Services. Unpublished paper, Wharton School, Department of Insurance, University of Pennsylvania. 2‘ Hall, Charles P. [et al.]The Effects of Cost-Sharing in the Medicaid Program. Material submitted for the record by Wilbur J. Cohen in his statement (on behalf of the American Public Welfare Association) in Hearings before the House Committee on Ways and Means, on the Subject of National Health Insurance, 93rd Congress, 2nd Session, June 28,1974. 39 section on “possible biases"). Even if they are suc— cessful in inhibiting initial demand, their applica- tion only to first dollar expenditures means that, once the deductible provision is satisfied, their restraint is instantaneously lifted. Some observers22 suggest that a backlog of demand which was initially restrained by the presence of deductibles may be released once the deductible is met; there would be instances where the need for care may have been enlarged by the delay, and there would be others where consumers might purposely increase their consumption of health services in order to make the insurance available so that they could obtain full value for their investment, or to obtain their “money's worth" for the added cost of the deductible payments. If a large portion of the insured population were to react in either of the above fashions, any initial demand savings might not oc- cur or might be largely dissipated by higher ex— penditures from those who do meet the deducti- ble requirement. The basis for an appraisal of the impact of deductibles on the demand rates comes from Rossett and Huang, who estimated the effect on demand for different deductible requirements in relation to coinsurance levels for physician and hospital expenditures.23 The degree of utilization suppression for different deductible amounts ac- cording to their data is shown in Table 13. These estimates, however, are highly generalized. They are not income—specific, while the deductibles pro— visions in the Assisted Plan and Federal Plan for the Aged are. Although hard data are suprisingly meager, there is some empirical evidence that lower income groups are more sensitive to such cost-sharing requirements then are others“,25 22 Jesmer, Shirley and Scharfenberg, R.J. "Problems in Measuring the Effect of Deductibles Upon Hospital Utiliza- tion." Blue Cross Reports, VII, No. 5, Oct. 1968, pp. 1-7. ’3 Huang. Lien-fu and Shomo, Elwood, op. cit, p. 91. Peel, Evelyn and Scharff, J. ”Impact of Cost Sharing on Use of Ambulatory Services Under Medicare, 1969." Social Security Bulletin, Oct. 1973. 2’ Scitovsky, Anne A. Effects of Coinsu'ra‘nce on Medical Care Utilization Paper presented at the 1015t Annual Meeting of the American Public Health Association. San Francisco, Ca... Nov. 4-8, 1973. 24 Table 13 Estimated percent reductions in care demands with various coinsurance rates due to CHIP deductibles Average coinsurance No $50 $100 $1 50 rate deductible1 deductiblel deductible1 deductible2 .25 .......... 0 5 8 10 .20 .......... 0 6 10 12 .10 .......... 0 10 14 1 5 .00 .......... 0 14 18 ~ 1Impact estimates for deductibles of $100 and under are from Huang. Uen-fu 8: Shomo. Elwood (Robert Nathan Associates). Assessment and Evaluation of the Impact afArchelypaI National Health Insurance Plans on US. Heallh Manpower Requirements: DHEW Publication No. (HRA) 75-1. DHEW/HRA/BHRD. July, 1974. 2The impact estimates for the $150 deductible are extrapolated from those for deductibles of $100 and under. Also, the estimates were based upon the price and income structure of 1969 and an Engle‘s curve of 1960.” Since the population’s health in- surance coverage was appreciably lower in 1960 than in 1969, the degree of deductibles impact derived from these data could be overstated. A further problem is raised by not having an equivalent impact estimate for the $150 deducti- ble provision of the Employee Plan; while the $150 deductible might roughly correspond to a $100 amount in 1969 dollars, this was not assum~ ed. Instead, an impact estimate for the $150 deductible was obtained by extrapolation to a value somewhat greater than the impact of the lower deductible amounts. The results obtained from this analysis are shown in Table 14 in which the initial CHIP de- mand estimates (without supplementation) have been adjusted for deductibles effects. These results should only be considered as tentative pending the availability of more reliable informa- tion on the behavior of consumers in response to deductibles. Overall, the estimates will serve to suggest the possible sensitivity of demand to the CHIP deductibles provisions within the context of this particular analysis. Under the “deductibles" assumption, demand for short-term hospital services is found to decrease by about 4 percent according 00 the semilog model (with “higher” price elasticity), while demand for medical office services falls about 2 percent. The influence of deductibles on demand for dental office services is not an- ticipated to be great although the model shows decreases of over 3 percent from the initial CHIP estimates without adjustment for deductibles “ The Engle's curve indicates the behavior of consumption expenditures with changes in income. 40 Table 14 Btimated percentage change in health care demand resulting from CHIP, by income and health service category assuming decreases in demand due to deductibles provisions1 No supplementation assumed Health service category Income Medical Short-term Pharmacy Dental office hospital services office All incomes ..... 16.7 -23.0 10.4 19.5 Under $2,500 .... 25.4 -1.1 13.4 11.3 $2,500-$4.999. . .. 18.8 -12.3 11.6 12.7 $5,000-$7,499. . .. 16.0 23.0 10.5 16.3 $7,500-$9,999. . ,. 14.7 -24.5 10.1 18.1 $10,000 and over. . 15.4 -27.4 10.0 22.1 1 . , . . . , Estimates are from the semilog model, “higher” elastrcrty series. (Table 5). This is because of the tendency toward more use of dental services among those of higher income and the inverse relationship between coin- surance levels and the degree of impact of the deductible. Pharmacy services are found to res- pond little to the deductibles adjustment as was anticipated because of the smaller amount of the deductible (up to $50) and because of the relative- ly low elasticity of demand assumed for the ser- vice. In at least one way, the above estimates may be a more accurate reflection of reality than the estimates computed without . adjustment for deductibles effects. This is because of the model's use of baseline gross effective coinsurance rates, i. e., out-of—pocket expenditures as a percentage of total. Since most insured persons are subject to deductibles, their out-of—pocket expenditures pro- bably contain a significant deductibles compo- nent. In essence, then, deductibles and coin- surance are merged into one cost-sharing feature. Therefore, the baseline coinsurance rates pro- bably overstate the average coinsurance rate that would exist in the absence of deductible ex- penditures. Conversely, the CHIP-specified coinsurance rates, when assumed to be the average coin- surance, understate the proportion of costs that the population would be meeting out-of-pocket because their marginal coinsurance rate would be 100 percent while their deductibles requirements were being satisfied. To make the process com- parable, therefore, the adjustment made for deductibles effects can be viewed as reflecting the fact that consumers would be paying 100 per- cent coinsurance (i.e., a marginal coinsurance rate of 1.0) on their expenditures until the deductibles were met and the CHIP-specified coinsurance rates only on those expenditures which exceed the deductible limit. There is another feature of the CHIP plan which may lower somewhat the overall propor- tion of expenditures which the population would pay directly—specifically the maximum liability. 41 This provision which is not a common property of current insurance plans (there are usually established limits on the carrier’s liability) means that, once the limit is reached, effective marginal coinsurance drops immediately to zero, and the money cost of acquiring care also (theoretically) drops to zero. Although it would be expected that only a small proportion of the population would incur such large expenditures”, those who do are apt to have profound health needs and may incur very large expenditures in relation to their numbers. Their probable occurrence may thus lower overall effective coinsurance rates and somewhat offset the possible distortions in the estimates introduced by the failure to remove deductibles expenditures from the pre-CHIP average coinsurance estimates. 2’ For example, under a coinsurance rate of 25 percent in the Employee Plan, a family would have to incur costs exceeding $6,000 in total expenses for reimbursable health services during a year in order to meet the $1,500 maximum family liability stated in the plan. X. Conclusions In the interpretation of the results presented in this report, it is extremely important to remember that the estimates are at best general approximations. This is particularly true for the estimates of gross effective coinsurance in 1976 before the introduction of CHIP and for the estimates of elasticity. Still, as general approx- imations, the estimates do provide a perspective on the probable manpower effects of an NHI plan such as CHIP. If the semilog assumption of de- mand curve shape is correct, then given the realization of extensive supplemental insurance purchases, the probable demand pressure on the physician supply would appear to be moderately heavy with that on the dentist supply somewhat more substantial. The demand pressure on R.N.’s and allied manpower would be small. Even if the demand curve should be noticeably more conser- vative than assumed in the above estimates, i.e., should it be linear, then the manpower require- ment changes would still approach those generated from the semilog model. This would result in slightly less demand pressure on the physician manpower supply and little appreciable effect upon other types of manpower except den- tists. Results from the doublelog model are con- sidered to be unreliable because of the low coin- surance levels that are associated with this analysis, which in turn lead to artificially high estimates.“ The key factor in evaluating the effects of these demand pressures is the shape of the health ser— vices supply curve, which reflects the respon- siveness of the quantity of services supplied to changes in prices. In this context, the clear distinction must be kept in mind of the difference between the possible cost effects of CHIP and the possible manpower effects. The introduction of CHIP (or any national health insurance plan for that matter) will tend to restructure the health care market and redistribute care among the population, as cost barriers to care are lowered 2‘ The unreasonableness of a doublelog curve shape assump- tion is largely that demand is not infinite at zero cost. Since potential dental care is much greater in relation to dental care received than in the case for other care, it is possible that the dental care demand curve more closely resembles the doublelog curve than do the demand curves of other types of care. 43 for part of the population. The cost effect of such plans thus would be determined by the interac~ tion of increased demand for care with the supply curve, with manpower effects determined in the same way. If, for example, the care supply within the existing labor force is totally elastic and thereby able to provide all of the additional care demanded (largely through increased productivi~ ty), there would be no numerical manpower effect and no major increase in prices, although expen— ditures would increase to pay for the greater amount of care provided. But if the care supply is not perfectly elastic, even moderate demand in- creases will cause increases in prices. If prices do not rise to market clearing levels, the health care market would contain more demand than supply, and some combination of methods of allocation would have to ration the existing supply among those demanding care. In addition to price, there are two other major types of care allocation that appear to be effec- tive in the present market. The first type of care allocation is through the time-cost of care. Many persons may be deterred from seeking care in marginal medical situations by the presence of queues and waits for appointments. When there is more demand than supply at a set price, in- creasing queues and rising appointment waits may automatically provide a form of time-cost ra- tioning, with the extent of queueing dependent upon the degree of shortfall. The last major type of allocation is “quota" or “need" allocation, in which providers themselves ration their time to those whose care needs they feel are more impor— tant, and, perhaps make greater use of referral mechanisms. When faced with more patients than they can treat, care providers are thought to tend to limit their case load to the more critical needs. Empirical evidence suggests that with the in- troduction of CHIP, the rationing ability of price would lessen, and allocations on the basis of time- cost and need would assume a greater role in total rationing, especially if the care supply is not elastic to the increased demand. For example, the increases in queueing time which were observed in Montreal, Canada after the introduction of universal health insurance may have led to lower use of physician office services by middle and up per income groups—groups which had been less sensitive to the previous care price rationing— and an increase in use by the low income groups which had been sensitive to care price.29 This sug- gests that the care supply in Canada was not totally elastic to the increased demand for care, and such might well be the case in the United States. The overall manpower impact of CHIP depends upon the distribution of the manpower types among the affected health service areas. Some types of professional work, such as teaching and research, would probably be affected very little by the introduction of CHIP. Of course, the changes in the demands for manpower in par- ticular types of care depend upon the demand change in specific health service settings. Because there is a large number of auxilliary health manpower that are either not employed in the health field or are employed only part-time, it is anticipated that the supply of these types of manpower could possibly assist in the response to increases in market demand for health services. However, there are many crucial health man- power types—such as particular physician specialists (e.g., psychiatrists), dentists, and op- tometrists—whose members are mostly active and working full—time. Because of the long train- ing period required to enter these professions, short-run increases in the supply of their services will necessarily be based on increased productivi— ty. Since data on price increases and visit output per practitioner during the period of the introduc- tion of Medicare—Medicaid show little productivi- ty increase response of physicians to demand in— creases (this is somewhat supported by the Cana— dian experience), there is reason to question whether the productivity response by these care providers will be completely adequate. There will likely be some increase both in the use of tradi- tional aides and in the practitioner's hours of work, but there is little basis for confidently predicting the degree of this response or its abili- ty to meet increased demand. Possibly policy decisions fostering the development and efficien- cy provisions of HMO’s could prove useful. In the very short term, there will inevitably be price inflation with the introduction of CHIP, as well as probable increases in nonprice rationing. The size of these cost changes will in part depend on the speed (and patterns of care supply) with 29 Enterline, P.E., etaL, op.c1't. 44 which patterns of care demand change. Even if the more liberal estimates of the demand increase should obtain, the overall impact of CHIP on man- power requirements should not be overwhelming. In the short~term, the physician supply may be faced with an increase of about 7-12 percent in re- quirements (under the higher elasticity assump- tion according to the semilog model), and this would be reduced by whatever immediate supply response can occur through changes in produc- tivity. The size of the increase can be largely at— tributed to the anticipated expansion in demand for ambulatory care services, regardless of changes in inpatient utilization. In the dental care sector, there would be less potential for market dislocation if the elasticity and curve assumptions used in generating the 19 to 24— percent increase in dental manpower estimates tended to overstate the responsiveness of consumers to price changes for this service. In this latter area, however, there is a much greater possible variation in the demand impact of CHIP because so little is known about the elasticity of dental care demands to price. Thus, there is a real possibility that the increase in den- tal care demand could be substantially greater than the estimated amount, and (if this is the case), the market dislocation could be severe. One protection against such dislocation might be a temporary ceiling on, or controls of, utilization in this area. In the longer time-span, the picture changes sharply. Manpower supply, expanded function aide roles, HMO and other care organization possibilities, and distributional problems all become more amenable to policy action. In this time frame, it would be possible to address more adequately the basic issues—the social cost of care prices, the appropriate forms of care alloca- tion and rationing among different parts of the population, and the acceptable levels of care that people need and/or expect. In summarizing, it must be said that sufficient understanding of the economic interactions of the health care system needed to accurately predict the effects of CHIP does not yet exist. Since any intervention in the health care system of the order of CHIP (or any such plan) is almost certain to change the existing equilibrium of the system, this inability to predict the course and degree of the changes argues for considerable caution in these innovations wherever and whenever this is possible. On the basis of the rough quantifications presented above, however, it appears that the im- plementation of an NHI plan on the order of CHIP will not create major manpower problems. The exception is the area of dental care where possi- ble high elasticity of demand to price could lead to large increases in care demands, and in the area of physician manpower where large increases in 45 demand for medical office services could arise if widespread purchases of private supplemental health insurance were to occur. Finally, given the likely continued government intervention in the health care system, it is very important that a greate? understanding of the interactions of the health care system be developed. Xl. Development of the Methodology The following sections describe the methodologies and technical considerations underlying 1) the projections of the population, and 2) the estimates of changes in the demand for health services consequent to CHIP, including the price elasticity assumptions and the estimation of gross effective coinsurance. The basic structure of the model is depicted in Figure 1. A. Projections of the Population The first stage in the overall procedure was the development of projections of the population in 1976 (by income), against which the provisions of the CHIP legislation would apply. The projec- tions needed for this purpose differed from the usual population projections by income. Normal- ly, income distibution projections of the popula— tion are based upon constant dollars in some form.30 In the present analysis, however. it was necessary to project the population by income as measured in 1976 current dollars, since a family’s (or an individual‘s) income at the inception of CHIP will determine costs and benefits. Thus, where normally the effects of inflation are remov- ed in such analyses, it was necessary to use cur- rent dollars so as not to disturb the relative im- pact of costs to enrollees over various family groupings. To accomplish this, the current population had to be projected on the basis of assumptions as to income growth among popula- tion groups, and in the level of demographic detail needed for analysis of the health service re- quirements affected by the bill. Since care utiliza- tion rates had already been determined for cer- tain age, sex, and family income-specific categories (as discussed below) in conjuction with other work in the Bureau, these categories were retained for the population projection. To simplify the analysis and permit application of the family income-based utilization rates, it 3° Mr. Roger Harriot of the Bureau of the Census has prepared one such unpublished analysis, entitled “Projections in 1969 Constant Dollars Assuming a 0.25% Reduction of Hours Worked per Year" (May 1971). Another source is the Na- tional Planning Association which prepared such a projection after the 1960 Census and was expected to conduct a similar study after the 4th count of the 1970 Census became available. The NPA estimates are prepared by state. 47 was decided to ignore the distinctions made in the CHIP legislation between family income and in- dividual income for the under-age 65 population. Those under age 65 with yearly incomes above $7,500 were treated according to the provisions of the Employee Health Insurance Plan, while most of those below $7,500 were placed in the Assisted Plan. As is discussed at length below, these simplifying assumptions are not believed to have an adverse effect on the results. The population aged 65 and over was handled as one—member family units in the model, in accordance with the legislative stipulations of the Federal Plan for the Aged. Because published Census reports do not con- tain estimates of the number of persons by age and family income, the population distribution us- ed in this analysis was estimated from the Census Public Use Sample.“1 The only data available in published form were personal income of the in- dividual by age, which would not be compatible with the utilization rates in the model based on family income. (As explained below, the actual tabulations were made in finer income detail than is shown here in order to facilitate the projection process.) It should be noted, however, that the in- come estimates used are based upon a 1-in-100 sample and have greater statistical variance than do normal Census estimates.32 Furthermore, the specific method used for projection of the popula- tion was less precise than it might have been, because of the limited time and resources allowed for the analysis and because the possible gain in accuracy arising from use of a more complex method did not appear to warrant the substantial- ly greater effort that would have been involved.” "‘ The specific Public Use Sample file used was the Spercent sample State file. 32 Income groups are generally within 2 standard error units of the published 20-percent sample data, of which the 5 percent sample is a subset; somewhat more variability is found in the lower than in the higher income groups, as would be expected. “a One such method would be to apply projected population totals by age and sex and income. Having an internal distribu- tion by age. sex, and total money income for 1970, the internal distribution for 1976 could be estimated by use of deter. minants. This mathematical technique essentially maintains the closest relationship possible between the grand total population (by sex) at 1976. the two appropriate rim subtotals (age and income), and a given cell in 1970 and the same four components in the target year. Table 15a Estimates of the population by age, sex, and income: 1976 Family income Age and sex All Under $2,500- $5,000- $7,500- $10,000 incomes $2,500 $4,999 $7,499 $9,999 & over All persons under age 65 (in 1,000’s) ........... 195,805 14,606 10,470 12,653 27,349 168,313 Males ..'. ............ 97,363 7,709 4,620 5,752 7,795 71,268 Under 14 ........... 26,495 1,380 1,477 1,812 2,565 19,261 14-24 .............. 22,795 3,539 1,302 1,525 1,873 14,556 25-44 .............. 27,455 1,251 844 1,252 2,277 21,831 45-64 .............. 20,618 1,200 977 1,163 1,638 15,620 Females ............. 98,442 6,897 5,850 6,901 9,098 69,696 Under 14 ........... 25,471 1,318 1,440 1,771 2,461 18,481 14-24 .............. 22,127 2,390 1,339 1,649 2,106 14,643 25-44 .............. 27,975 1,298 1,264 1,584 2,391 21,438 45-64 .............. 22,869 1,891 1,807 1,897 2,140 15,134 Given a 1970 distribution of the population by age, sex, and family income, two simplifying assumptions had to be made. First, it was assum- ed that all persons would be uniformly affected by “inflation” and by actual growth in real per- sonal income. This means, for example, if a 34— percent inflation rate and an 18-percent real in— come growth rate were to occur over the 6-year period, that these rates would apply uniformly to each individual income group, rather than having more of the growth occur in one income group than another. Using this assumption, it was possible to pro— ject the growth of personal income in the general economy and then to apply a single rate for the future 6—year period to a finely stratified estimate of the 1970 population by income.“ With the in— come distribution of the 1970 population classified in $100 intervals, an income adjustment factor (see below) was applied and the specific overall income group location of each detailed in- terval determined.35 The second simplifying assumption was that in- creases in the population in a specific age-sex category would be proportional across all income 3‘ However. it is recognized that low income persons are hit hardest by inflation, with the sharp rise in costs for basic necessities taking a bigger percentage of low and moderate incomes. “5 For example, if the adjustment factor were 4.9 percent per year (about 35 percent for the 1970-76 period). all persons earn- ing at least $3,700 but less than $5,600 in 1970 would be pro- jected as earning $5,000 to $7,499 in 1976. 48 groups, i.e., that the population increase would not occur at a greater rate in one income group than in another. Under this assumption, it was possible to increase all of the income groups in a specific age—sex category at the ratio of their 1976 population to their 1970 population. Together, these two steps resulted in the projection of the 1976 population as shown in Table 15. Deriving the 1976 population for the aged 65 and over groups presented different problems. According to the legislation, all persons over 65 enrolled in the Federal Plan for the Aged would be treated as individuals. In essence, then, an in- dividual's cost~sharing requirements would be determined by his personal income, regardless of the financial status of his spouse or family. Pro- jected individual incomes for 1976 formed the basis for estimating the number of individuals over 65 who would be subject to the different levels of coinsurance and deductibles found in CHIP. The assumption was made that the per- sonal income of the elderly would expand at the same rate as that assumed for persons under 65. Although Social Security benefits have been ex- tended significantly in recent years, many elderly persons live on fixed incomes, and the income of other older persons increases only in line with the cost—of-living, which is usually less than the average increase in income for the overall popula- tion. Therefore, there could be a downward bias in the demand estimates caused by the unrealistically high income growth rate assumed for the older population. Table 15b Estimates of the population in 1976 by age, sex. and income: 1976 Individual income Age and sex All Under $1,750 $3,500 85.250 incomes $1.750 $3,499 $5,249 8.- over 1 All persons age 65 and over (in 1,000‘3) ............... 22,227 7,681 6.501 4,566 3.479 Males ................... 9,158 1,539 2,802 3.233 1.584 Females ................. 13,069 6,142 3,699 1,333 1,895 The personal income adjustment factor used was obtained by computing the average annual change in per capita personal income between 1967 and 1974.38 This measure was chosen as be- ing the most reasonable index of changes in real income combined with income adjustments caus- ed by general increases in the cost-of-living. Specifically, the rate of change (geometric mean) in per capita personal income during the 1967-74 period was computed, yielding an average yearly increase of 7.8 percent; this was assumed to be the most reasonable approximation of the pro- jected trend in personal income growth to the target year. As the basic income assumption in this analysis, this rate of change was projected to continue through 1976. The age-sex category in- flation was based upon census projections.37 The resulting estimates of the population by age, sex and income are shown in Table 15. Still another adjustment should properly have been made in the population distribution by in- come, however, because this distribution (in cur- rent dollars of 1976) does not correspond directly to the distribution available for utilization rates. The problem is that current-dollar income groups contain dollar inflation for the 1970 to 1976 period—which causes some of the population in the $10,000 and over category not to have the full purchasing power of a 1970 family with a $10,000 income. This means that the CHIP coinsurance and deductibles would apply to the then current income of 1976 and thus would be used in estimating the demand increase, while the base 3“ US. Department of Commerce, Bureau of Economic Analysis, unpublished data on per capita personal income, 1967-74. “7 Age-sex population estimates for 1976 (US. Bureau of the Census, Current Population Reports. Population estimates and projections, (Series E), Series P25, No. 470, Nov. 1971), divided by the total population estimate from the Public Use Sample, were used as inflators for respective age-sex population categories. 49 utilization rates used in the computation of the demand increase are still those of 1970, which cor- respond to a constant dollar income. Ideally, the deductibles and coinsurance should be based upon current dollar family income groups, and the base utilization rates upon constant dollar family income groups. Models of this situation would need twice as many income groups for the age and sex distribu- tion as in the current model. Because it would have been necessary to substantially modify the model to handle this, and since the models of the demand increase used only the utilization rate as a base, the problem was not adjusted for in the analysis. The effect is believed to be minor— perhaps an overstatement of demand changes due to CHIP on the order of 1 or 2 percent.38 As was noted earlier, in order to simplify the analysis, except in the case of the over aged 65 population the distinction between individual and family incomes was eliminated. There was major concern, however, as to whether this simplifica- tion would cause significant distortions in the results because of the different cost-sharing pro- visions for individual and family units in both the Assisted and Employee Health Insurance Plans contained in CHIP. Classifying all persons as members of family units would result in an overestimation of utilization to the extent that the smaller cost—sharing figures were applied to individuals. If the individuals were assessed the appropriate higher costs, they could be expected to reduce their utilization of services. Since a large number of the unrelated individuals in the population are over 65, however, unrelated in- dividuals comprise only about 5.8 percent of the 3“ A test of this bias in an earlier version of this study, using the semilog model, showed that there were no effective dif ferences in the percentage changes in demand for any service category when constant dollar income adjustments were made to the population rather than current dollar adjustments. total under»65 population. Furthermore, there are a number of income ranges (over $7,500, under $1,750, $2,500-3,499, and $5,000-5,249) where the deductible and coinsurance requirements are the same for persons‘ in both family and individual units. Therefore, the proportion of the under-65 population in the model that is subjected to inap- propriate levels of cost-sharing because of this simplification is probably considerably less than 5 percent. For the under~65 population, assignment of per- sons to the Employee and Assisted Health In- surance Plans by family income seemed ap- propriate because the legislation allows low- income employees to participate in the Assisted Plan. Since families with incomes below $5,000 would enjoy lower cost-sharing requirements under the Assisted Plan, it was assumed that all persons below this level would elect to par- ticipate in the Assisted Plan. The lower deduc- tibles and coinsurance rates of the Assisted Plan have therefore been applied to all persons under 65 with incomes under $5,000”, regardless of their employment status. For employed families with incomes of $5,000-7,500 (individuals earning $3,500-5,250), the coinsurance rates and deduc» tibles of the Employee Plan would likely apply, whereas very high medical risks and non working persons in this income group would be expected to enroll in the Assisted Plan. For persons with incomes of $7,500 or more, the same coinsurance 3” Ordinarily, it would be financially advantageous for lower income persons (i.e., those having incomes below $5,000 in 1976) to enroll in the Assisted Plan when they are eligible and . without better coverage, instead of in the Employee Plan. As is noted, the present analysis assumes this to be the case. However. it is possible that a significant number of employed families with incomes between $5,000 and $7,500 income ($3,500 and $5,200 for individuals) would not have a more at- tractive health insurance option than theEmployee Plan under CHIP. The Employee Plan contains some regressive characteristics which might militate against these lower- middle income persons obtaining coverage in the first place, or using their benefits once coverage is obtained. Regressive premium impacts in the Employee Plan (since a fixed dollar premium amounts to a greater proportion of lower incomes than of higher incomes) would possibly induce some employers, especially small and low wage employers, to discourage their employees from obtaining coverage or to favor workers who elect to forego coverage. Furthermore, the regressive nature of the uniform cost-sharing provisions would probably act more forcibly to retard demand among lower-income persons en- rolled in the plan than can be predicted for the other plans or higher income groups. The failure. to account for these possibilities may have resulted in overestimating demand. but the amount or error is probably small. 50 rates and deductibles are applicable whether they are enrolled in the Employee or Assisted Plan. Using family income as the primary basis for assessing cost—sharing requirements means that nearly 80 percent of the population under 65 is subject to the highest coinsurance and deductible rates. Recent estimates by the Social Security Administration of enrollments under the various CHIP subplans placed approximately 73 percent of the under-65 population in the Employee Plan.4O Compatibility between the figures would require that over 25 percent of the population which SSA places in the Assisted Plan be subjected to the highest cost-sharing requirements. Because it is not expected that much of the population eligible for the Assisted Plan would fall within the highest income groups (expecially because of the income criteria for eligibility). the adopted methodology probably subjects a higher propor- tion of the population to the highest cost- sharing requirements than would occur using SSA estimates. However, since income determines eligibility for the lower “assisted” rates, estimates based on projected family incomes seem appropriate. Some of the differences can be attributed to the different years of the estimates; SSA estimates are based on January 1975 incomes while the MAB estimates are based on projected 1976 in- comes. Since the CHIP legislation's $10,000 in- come base is expressed in current dollars, it is an- ticipated that some persons would move into the highest income category (through “inflation" as well as changes in real wages) between 1975 and 1976. B. Price Elasticity Assumptions Economic theory holds that when the cost of a good increases consumption normally decreases because other uses of resources become more at- tractive. However, the amount of any goods or services actually consumed is not simply a func- tion of price, but is the result of the complex in- teractions of costs and supply and demand. In regard to health care, many additional factors im- pinge on the relationship between price and con- “ Estimated Health Expenditures Under Selected National Health Insurance Bills: A Report to the Congress. USDHEW, Social Security Administration. July, 1974. sumption. First, as has been pointed out, the dollar price charged the consumer by the pro- vider is not the only cost which an individual faces when seeking health care. There are also the secondary dollar costs, such as costs of travel and loss of income, as well as the nondollar costs, such as the time spent in travel, time spent in the office queue, and other psychological disincen- tives like rude or hurried treatment, or fear of pain or of what the diagnosis might be. Second, the demand for health care is sometimes very inelastic; i.e., there are times when the cost of care is of little concern. Neither loss of money nor loss of time is important to a very sick person, and to some degree this effect extends over most levels of illness. The extent to which an individual weighs financial considera- tions in his decision whether or not to obtain care is related to his perception of how unwell he is. This perception, in turn, is probably related to level of education and cultural and ethnic factors, among other things. The third consideration is that the reaction of an individual to cost concerns is dependent upon his perception of his own resources, e.g., income, insurance coverage, eilgibility for free care, 0c- cupation leave provisions, and kind and quality of health care available to him. While the above considerations are material to any general analysis of the economics of health care demand, their importance in this particular analysis is even more critical. The problem is one of generalizing gross estimates of a specific population’s responsiveness to price to the general population. To the extent that measures of price elasticity based on limited numbers of observations may differ from the characteristics of the population to which they are applied, any conclusions drawn from the analysis may be inac- curate.“ In view of the uncertainty underlying present measures of price elasticity in the health " In regard to this point, it is pertinent to note that elasticities which are estimated from actual market data are in- trinsically predicated on the information contained in a given sample at a particular level of gross effective demand. To apply these elasticities (whether determined from a number of obser- vations at different levels of price and demand. or from a single observation) to portions of the demand curve which have not been “observed", but have merely been produced by statistical extrapolation is to make an implicit assumption about the shape and position of the demand curve, both before and after CHIP. (From a staff paper: Evaluation of CHIP: A Technical Review, prepared by the Economic Analysis Branch, Bureau of Health Services Research and Evaluation, HRA, 1974). 51 care area (as is discussed below), the results presented here should be used with respect for the above considerations. C. Elasticity of Demand Estimates There are two important interrelated factors that are used in the quantification of the degree to which the demand for health services is af- fected by changes in the price of care.“2 The first of these is the measurement of price elasticity, which is the percentage change in service utiliza- tion that is associated with a one—percent increase (or decrease) in a particular service price. The other factor, the specification of the shape of the demand curve, determines how the elasticity at each price relates to the elasticity at all other prices. The concepts of price elasticity and de- mand curve specification and their relationships are illustrated in Figure 9. The points X and Y in Figure 9 represent two hypothetical situations, before and after a change in the price of care. The relationship between X and Y represents the net effect of demand elasticity over a number of points; it is termed “arc elasticity" since it is the measure of elasticity over an arc of the demand curve. However, there is little agreement on the shape of demand curves for health services; hence the measurement of elasticity is not necessarily straightforward since the shape of the curve in part determines the magnitude of elasticity estimates. The lines ‘A’, ‘B', and ‘0' represent three different assumptions as to the nature (or shape) of the demand curve,with ‘A’ illustrating the “linear" assumption, ‘8' the “semilogarithmic” assumption; and ‘C’ the “doublelogarithmic” assumption. As can be seen from Figure 9, there is little dif— ference between the linear, semilog and doublelog assumptions except where high and low prices are involved. At the extremes of the curves, the linear model (a) assumes the least change in one variable compared to the other, while the doublelog model (c) assumes the most change. In fact, the doublelog model assumes that demand approaches infinity as the price becomes very low and that the demand never quite ‘2 The price of health services that is relevant to this report is basically the net price to the consumer; a proxy for this price is the coinsurance rate. Since the lower the coinsurance rate, the lower the net price to the consumer, with no coinsurance there is little reason for the consumer to react to price differentials. Figure 9 An example of elasticity curves relating demand to price for health services Health Service Price 5 — a \ \ 4 _ 3 _ 2 —. Y Y!) \\ \ \ \ ‘\ \ PRICE Z \ \ \ s \ 1 - \ \ \ l l l l I 1 2 3 4 5 52 Health Service Utilization becomes zero, no matter how high the price goes.43 This report provides estimates based on all three demand curve assumptions, but with em- phasis on the semilog model. It might be useful at this point to use Figure 9 to illustrate some additional aspects of demand and elasticity. Curve XY represents the relation ship of price and quantity, ceteris paribus. Thus, if price falls, quantity demanded moves ac— cordingly. This is movement along the demand curve. However, if another population were measured, e.g., one which faced a different time price, or one which viewed its state of health dif- ferently or one which perceived its financial resources differently, that population would like» ly exhibit a different relationship between ser- vice price and utilization, with the entire curve shifting to X'Y’ or X"Y". That is, elasticity refers to a movement along the curve, while demand shift refers to a movement of the curve. There is also a relationship between a constant price, e.g., Z, and the position of the curve. For linear de- mand curves, as the curve moves to the right elasticity falls for any given price. Thus the same shaped demand curve has different elasticity im- plications at a given price depending on where it is positioned. The arc X’ to Y' in the illustration shows what might occur if the population faced a shortage of care which led to longer waits and less service utilization at the specified prices.“ The are X" to Y" shows what might happen if a similar population had higher incomes and felt more able to afford the care prices. For those readers who are mathematically in- clined, Figure 10 presents the equations used to estimate the demand change under the three curve assumptions. These were developed by Dr. Lien-Fu Huang, currently of Howard University, for Robert R. Nathan Associates under a contract with the Bureau of Health Resources Develop- “ In the literature, where the model type is not specified as to the demand curve assumption the doublelog assumption has probably been used. This is because the doublelog assumption leads to a constant value for the elasticity parameter over the entire demand curve while the other assumptions do not. While this assumption simplifies the computations. it should be remembered that the doublelog assumption leads to the most extreme values. “ It has been indicated earlier that insofar as nonmoney costs are relatively important in the consumer‘s perception of the total cost of care, the failure to take account of them in the model becomes important—likely biasing the estimates up— ward as the money price approaches zero. Figure 10 Estimation of the care demand adjustment factor For all models Nij = Uij (post-NHI) Uij (pre-NHI) which is estimated as-- Kij Nlj=[1‘njx)+(anF-)] ij 1. Linear Model: K“ U n. [— 1" P.. 11 . 1] 2. Semilog Model: Nij = e Ki]. n. 3. Doublelog Model: Nij = [P- ] 1" ll Where: Nij is the demand change factor for population group i_and care type j, ij is the health service utilization rate at baseline, njx is the elasticity for care type j under elasticity series x, — K.. is the gross effective coinsurance under NHI which is the coverage of the population group i for health service type j times the coinsuran_ce of the population fort-hat care type, plus the proportion of the population without insurance. Since 100% participation in CHIP is assumed, Kij is equal to the CHIP coinsurance rate, R. is the gross effective coinsurance before NHI, defined as the proportion of total expenditures for a given type of personal health service which the consumer pays directly. ment, Health Resources Administration, HEW.“ Despite the vital role played by elasticity in any measurement of care demand, there are no definitive measures of the level of demand elasticity in health care. Problems in most data collection efforts have raised basic questions about the general applicability of exisiting studies. Furthermore, elasticity estimates from these studies vary widely. For example, Davis and Russell and Martin Feldstein used state- aggregated data to obtain estimates of price elasticity of demand for hospital inpatient ser- ‘5 For the development and a discussion of these models. see Huang, Lien-fu and Shomo, Elwood, op. cit. 53 vices ranging from -0.26 to -0.67. “2‘“ Rossett and Haung measured consumer expenditures for hospital and physician care, obtaining price elasticity estimates of -0.35 to -1.20, depending upon the estimate of the gross effective coin— surance of the population (discussed at length later)“ The methodology and estimates of these studies have all been criticized by Phelps and Newhouse, who feel that the estimates are pro— bably biased upward."9 In a 1972 study, these in- vestigators used data from a Stanford University population to derive price elasticity estimates of -0.15 to -0.30 for medical office visits.50 In this case, however, legitimate questions can be raised as to whether the Stanford population is similar enough to the US. population for the estimate to be appropriate. In addition to the comparability problems, there is a question of whether the doublelog curve shape assumption that appears to have been used in this study has affected the elasticity estimate, since the coinsurance range examined was in the extreme low range of zero to 25 percent. The estimates of Phelps and Newhouse have also been questioned as to their applicability over broad price ranges by Martin Feldstein because of curve shape assumptions. 5‘ ‘° See: Davis, Karen and Russell, Louise B. op. cit. ‘7 Feldstein, Martin. ”Hospital Cost Inflation: A Study of Non-Profit Price Dynamics." American Economic Review, Vol. LXI,N0.5, 1961. ‘8 See: Rosett, Richard N., and Huang. Lien-fu. "The Effect of Health Insurance on the Demand for Medical Care." Journal of Political Economy, Vol. 81, No. 2, Part 1. March/April 1973. “‘ Newhouse, Joseph P., and Phelps, Charles E. On Having Your Cake and Eating it Too: Econometric Problems in Estimating the Demand for Health Services. The Rand Cor~ poration. R-1149-IVC, April 1974. Phelps and Newhouse. in their defense of more conservative elasticity estimates. argue that failure to take explicit account of the role of time costs may have introduced an upward bias in most elasticity measurements. Since the total price of medical care can be represented by two gross components, a money price and time price, in the presence of time prices, as has been pointed out earlier. consumer response to changes in the money price will vary—depending upon the relative value of their time. 5" Phelps, Charles E., and Newhouse, Joseph P. “Effect of Coinsurance: A Multivariate Analysis." Social Security Bulletin, June 1972. 5‘ Feldstein. Martin. Econometric Studies of Quantatative Economics. Ed. M.D. Intrilligator and DA. Kendrick, 1974. North Holland Publishing Co. (see page 406). 54 Although these studies and the literature pro- vide little rationale for setting a single-point elasticity estimate for medical care, informal discussions with health economists indicate that there may be at least a general consensus on the possible range of estimates. This consensus ap- pears to range from a low of -O.14 to a high of 0.50 with a midpoint of about -0.30. It should be noted, however, that the presently-used estimates are lower than would be the case if they were based solely on the literature. In these informal discus- sions, there also appeared to be some agreement that elasticities for hospital care should be somewhat less than those for medical office care, because of the greater incidence of serious illness cared for in hospitals. In the area of dental care, the only available elasticity estimates besides a recent one from Phelps and Newhouse (as noted below) are by Paul Feldstein, who estimates an elasticity to price of 1452, and by Alex Maurizi, who recently published an estimate of —1.753. The dental care elasticity estimates by Feldstein and Maurizi were tested for reasonableness but while strong— ly supported by two secondary checks54 they were not used in the present study. Instead, the more 5’ From: Feldstein, Paul J. Financing Dental Care:An EconomicAnalysis. (mimeo draft, 1973). 5“ Maurizi. Alex. Public Policy in the Dental Care Market. American Enterprise Institute, Washington, D.C.. 1975. 73 pages. 5‘ The two checks used were a comparison with estimates of demand change in prepaid dental care and an examination of past price levels and utilization to determine if the elasticity adjustment explained utilization levels. Findley (in Findley, James J. Is Prepaid Dental Care the Solution'?, mimeo, M.A. thesis at Massachusetts Institute of Technology, June 1973, p. 55) found that under prepayment, dental care expenditures were nearly always two or three times those that would be ex- pected using NCHS actual utilization data for the entire population. Interestingly, Findley found that there was little rise in the proportion of enrollees seeing a dentist in the preceding year. In the other test. Cole and Cohen (Cole, Roger B. and Cohen, 'Lois K. “Dental Manpower: Estimating Resources and Requirements" in Milbank Quarterly, July, 1971, pt. 2.) found that the growth in care utilization in the 1950-1965 period was much less than could be explained in terms of population-specific care rates. If an elasticity of -1.4 is assumed for dental care, the change in dollar prices in this period would suppress demand to the observed 1965 level. Since there appeared to be no significant changes in nonmoney factors influencing the relationship, e.g., queueing time. this implied a point price elasticity value of -1.0 or greater; which tends to support the elasticity values of -1.4 and 17 reported by Feldstein and Maurizi, respectively. recent price elasticity estimate for dental ser- vices reported by Phelps and Newhouse (1974 - see below) was used for the “lower" elasticity series. This estimate, while much more conser» vative than the Feldstein and Maurizi estimates, derives support from its relative similarity to other dental estimates developed in the Phelps- Newhouse study even though based upon in— dependent sets of observations. The “higher" dental price elasticity estimate used in the pre- sent study assumes that demand for dental care is elastic, somewhat arbitrarily set at -1.0 (unity). In the 1974 study just mentioned, Phelps and Newhouse undertook a critical review and analysis of elasticity evidence from both publish- ed and unpublished sources.55 In summarizing their findings, these investigators estimated (for the zero to 25 percent coinsurance range) the elasticity of demand for hospital services to be -0.08; for medical office visits, 014; for dental ser- vices, between -0.07 and -0.16; and for prescrip- tion drugs, 0.07.5“ Consistent with their earlier studies, Phelps and Newhouse show a pattern of low elasticity for hospital services and prescrip- tion drugs and a larger one for physician office visits, with a moderate elasticity range reported for dental services. Their results reflect the assumption of linear demand curves to generate elasticity estimates in the 0.25 to 0.0 coinsurance range; in contrast to the extensive use of constant elasticity demand curves elsewhere in the literature (i.e., doublelog models). From a test of this assumption, Phelps and Newhouse found essentially no difference between the linear form and the semilog curve. The “low" elasticity of demand series selected for use in the present report are drawn from this latter study; the “higher" elasticity series is com- posed of a consensus of estimates from the other sources discussed. The “higher" elasticity coeffi- cient used in the pharmacy demand estimates is 5" Phelps, Charles E. and Newhouse. Joseph P. Coinsurance and the Demand for Medical Services. A report prepared under grant from the Office of Economic Opportunity and the National Center for Health Services Research and Develop- ment. R-964-1-OE0/NC, October, 1974. 5‘ It is recognized that elasticity coefficients computed for a zeroto25 percent coinsurance range are somewhat underestimated when applied to a higher coinsurance range in a variable elasticity model. However, there was no reasonable means ofadjusting these elasticities at the time of the analysis. 55 -0.15, approximately twice the observed value. These elasticity estimates are shown in Table 16. Table 1 6 Selected price elasticity assumptions by health service category Price elasticity of demand Health service category Lower Higher series series Medical office services . . . -0.14 -0.30 Hospital services ........ -0.08 -0.20 Pharmacy services ....... 007 -0.1 5 Dental services ......... -0.16 1.00 D. Gross Effective Coinsurance Rates This analysis assumes that a major determi- nant of demand for health services under NHI will be the amount of coinsurance in effect at the time of its implementation. Coinsurance, the percentage of the total expense for some type of personal health care which the consumer must pay out of pocket, comes into effect after the 7 deductible level is exceeded. The justification for coinsurance is held to be its role in reducing the liability of the insurer and therefore the premiums paid by the insured, and in preventing health care from becoming a free commodity after deductibles are met, and thus in suppressing un- necessary utilization. Not to be confused with coinsurance rates themselves, gross effective coinsurance refers to the weighted mean of the coinsurance rates of the insured and uninsured populations. Hence, a population with extensive third~party coverage for some type of health service would have a lower average coinsurance coefficient than one with less extensive coverage. The present study defines health insurance as including all sources of third party payment for health services: public assistance programs (Medicare, Medicaid and others) and other types of “free” institutional and noninstitutional care, as well as privately purchased health insurance. Thus, the definition of coinsurance expressed above can be redefined as the proportion of total expenses for personal health services which are not paid through public programs or by private health insurance. _ Evaluating baseline average coinsurance presents special problems for an NHI analysis when viewed in the latter manner. This is because of the uncertainty about the extent to which various state and local public assistance programs would still provide “free" health benefits to persons after implementation of CHIP. Failure to accurately appraise this poten- tial source of added benefits could bias the de- mand estimates: (1) downward, supposing its con- tinuance, by not removing its influence from the total expenses at baseline or by not adjusting the CHIP coinsurance rates downward accordingly; or (2) upward, by its exclusion from estimates of baseline average coinsurance, should these sources of benefits be discontinued after CHIP. This dilemma was resolved for modeling pur- poses by computing a set of “framework" demand estimates which serve as a starting point for subsequent modifications, assuming various degrees of supplementation through public sources of "free” health care or investments in private supplementary health insurance. These framework estimates, as shown in Table 5, were derived from estimated baseline (pre-CHIP) coin- surance levels which include both public and private sources of expenditures for health ser- vices. The baseline estimates of 1976 average coinsurance before CHIP are presented in Table 17. The pre-CHIP coinsurance rates are based upon results from the 1970 Andersen and Andersen Survey of Health Services Utilization and Expenditures conducted by the Center for Health Administration Studies (CHAS), Universi- ty of Chicago“, which provides primary data on the distribution of expenditures by health service according to age and income status.58 These data reflect the experience of 3,765 families, comprised of 11,619 noninstitutionalized individuals in the purchase of health services through out-of—pocket payments, private health insurance, public assistance programs including Medicaid, Medicare, and other sources. 57 For a more extensive description of these data see Andersen, R. et aL Expenditures for Personal Health Ser- vices National Trends and Variations — 1953-1970. DHEW Pub. (HRA) 74-3105, Rockville, Md.: National Center for Health Ser- vices Research and Development. 1973. 5" The actual ratios of out-of—pocket expenditures to total ex» penditures which were used as average coinsurance rates in the model are predicated on rates determined from the 1970 CHAS data by the Office of Research and Statistics, Social Security Adminstration. These average coinsurance rates were provided by SSA on the basis of income for the popula— tion under age 65, and aggregated across income groups for persons aged 65 years and over, for the health service categories specified in the model. Table 17a Estimated average coinsurance before CHIP by age, health service category, and income Family income Age and health service category Under $2,500- $5,000— $7,500- $10,000 $2,500 $4,999 $7,499 $9,999 & over Persons under 65: Medical office ........... 0.47 0.46 0.51 0.52 0.53 Shortterm hospital ...... 0.15 0.14 0.12 0.12 0.10 Pharmacy services ........ 0.58 0.57 0.77 0.80 0.78 Dental office ........... 0.59 0.55 0.83 0.81 0.85 Table 17h Estimated average coinsurance before CHIP by age, health service category, and income Individual income Age and health service category Under $1,750- $3,500— $5,250 $1,750 $3,499 $5,249 & over Persons 65 and over: Medical office .......... 0.35 0.35 0.35 0.35 Short-term hospital ...... 0.05 0.05 0.05 0.05 Pharmacy services ....... 0.65 0.64 0.62 0.63 Dental office ........... 0.88 0.83 0.88 0.85 56 Since expenditures for any given service category represent a composite of expenditures for a variety of specific services, a brief summary of the components included in the average coin- surance rates for each service category variable is worthwhile. Expenditures for short-term hospital services are comprised of charges for hospital inpatient care, including room and board, laboratory fees, drugs, x-rays, operating and delivery room fees, and ”extras". Charges for pathologist, radiologist and anesthesiologist are included, as are expenditures for surgical and nonsurgical care in a hospital emergency room or outpatient department. These expenditures reflect charges billed to the patient separately by the doctor, as well as charges billed directly by the hospital. Insofar as possible, expenditures for long-term nursing home services, psychiatric care, and extended care facilities are excluded. In any event, contaminating effects of a long-term care component in the short—term hospital coin- surance rates should be minor since the survey was of a noninstitutionalized population. Medical office services include charges for out- patient visits in the physician's office, patient’s home, and visits to an ophthalmologist for eye care but not for eyeglasses. Dental services in— clude charges by the dentist for his own office ser- vices as well as for those of his auxiliary person- nel, and for dental appliances. Also included are charges for dental laboratories and manufac— turers connected with the patient's treatment. Pharmacy services includes expenditures for prescription drugs prescribed by a physician or dentist and purchased directly by the patient. Drugs received while in the hospital are ex- cluded. Having obtained estimates of average coin— surance rates in 1970, the next step was to adjust them for application to the projected 1976 popula- tion base. The data used for this adjustment were aggregated estimates of expenditures by service category and source of payment for personal health services for years 1967-1972 from the Social Security Administration.” These expen— ditures were projected to 1976 by the use of linear regression techniques, permitting the ex- pression of direct payments as a percentage of the estimated 1976 totals. ('(mper, Barbara S. et al. “National Health Expenditures. 1929773." Social Security Bulletin, Vol.37, No. 2, Feb. 1974, pp.2~l9. 57 The procedure for adjusting the CHAS data- derived baseline coinsurance rates to 1976 expec- tations generally involved controlling them to yield a weighted average approximating the overall estimated change in coinsurance pro- jected from the SSA expenditure data. Slight dif- ferences between the two projected measures oc- cur because of adjustments for definitional discrepancies among the data sources and because of an assumption of negligible change in hospital and medical services coinsurance for per- sons aged 65 and over between 1970 and 1976; i.e., the assumption that benefits from Medicare and investments in private health insurance will not appreciably increase for the elderly during this period. For these reasons, estimated average coinsurance levels for 1976 as projected from the SSA data are somewhat lower for all services than the average rates used in the model. Some characteristics of the SSA expenditure data used for the baseline average coinsurance projections need mention. First, SSA estimates of expenditures used for the short-term hospital category do not discriminate between short and long-term care. Hence, the validity of this measure of projected change in coinsurance is partly dependent upon the assumption of in- significant changes in the mix of expenditures for short and long-term care by third party sources which would influence direct expenditures. A con— tinuation of faster growth in public funds ex- pended for short-term care than for long-term care from 1970 to 1976, for instance, would result in an upward bias in the 1976 estimated average coinsurance for this sector, assuming no tradeoffs in expenditures from other sources. Finally, projected changes in average coin- surance for dental and pharmacy services were approximated through SSA expenditure data that includes dentists’ services. drugs, and other professional and health services, as data were not available separately for these services. A possible bias is indicated by this aggregation because of evidently greater yearly increases in third party coverage for dental services than for pharmacy services in recent years.“ In addition to the assumptions mentioned earlier, the validity of the preceding methodology for evaluating gross effective coinsurance is con- ” See Mueller, Marjorie Smith. “Private Health Insurance in 1972: Health Care Services, Enrollment and Finance." Social Security Bulletin, Vol.37, No. 2, Feb. 1974. pp. 20-40. tingent upon the assumption that the propor- tionate distribution of expenditures by income levels in 1976 will be similar to that of 1970. Moreover, in the case of CHIP, it was assumed that population NHI coverage would be complete. While this will probably not occur since the in- surance plan is voluntary (as was discussed earlier and is noted again later in the report), there is no adequate basis for a lesser esti- mate. This assumption should be kept in mind as representing a possible upward bias in the de- mand estimates. A possible downward bias in the estimates may be caused by the extensive existence of health in- surance with benefits already beyond those of the CHIP proposal. Particularly in some heavily unionized industries, where existing benefits and coinsurance are more advantageous than those proposed in CHIP, it is unreasonable to expect that coverage would be replaced by the lesser CHIP provisions. To quantify this factor requires much more data on existing insurance coverage by type and income then is currently available, however, and it therefore must be viewed as a bias toward understatement in the demand estimates. Although this bias may be partially off- set by the assumption of total coverage under CHIP, the net result is believed to be an underestimate of demand.“1 It is emphasized that, as the name suggests, the framework demand estimates generated from the ratios of CHIP coinsurance rates to pre-CHIP average coinsurance levels are just the base structure upon which alternative forecasts are built; they do not necessarily reflect any final or conclusive estimates. To provide a series of alter- native demand estimates reflecting a variety of possible occurrences under CHIP, the CHIP coin- surance requirements were modified to suggest various degrees and forms of supplementary in- surance. In all, eight series of supplemented CHIP coinsurance rates were computed. The ad- justments fall into four general categories. First, the rates were uniformily reduced across all in- come groups and health service categories by ar- bitrary factors of 40 and 25 percent. These ar- tificial assumptions might be stated as a given percentage of the population's purchasing first or early dollar insurance coverage to offset the "‘ It is also possible that if employees are now better insured than they would be under CHIP, employers will resist paying the extra difference. if CHIP is effected, to bring the employee’s benefits in line with current benefits. 58 CHIP coinsurance liabilities, or as the “average effects" of such coverage reducing effective coin- surance by that amount across the entire popula- tion.62 Second, the artificial assumption was made that supplemental insurance would be purchased according to the severity of the expenditure risk. Thus, short-term hospital services, being the most expensive form of care, and as discussed earlier which are likely to show an overall in- crease in price to the consumer under the present CHIP requirements, may attract considerable in— vestments in supplemental insurance; for such eventualities a 40-percent reduction in CHIP coin- surance has been initially assumed. Uniform reductions of 25 percent in coinsurance for medical office services and 15 percent for dental and pharmacy services were similarly assumed. The third category is composed of coinsurance reductions scaled according to income. These hypotheses are introduced to suggest the redistributive effects of differentially reducing coinsurance according to income, in terms of sup- plementation’s altering the “demand spread" be- tween income classes. The adjustments were generally computed by maximally reducing coin- surance (to zero if necessary) beginning at either income extreme until a given weighted percen- tage reduction in coinsurance was achieved. Thus, the adjustments, by “loading" coinsurance reductions to a maximum extent toward one in- come extreme or the other, emphasize the possi- ble influence of income-related utilization rates (and population structure) on demand for services under CHIP were supplementation biased in either direction. These income-scaled coinsurance reductions take two forms. The first is that the extent of sup- plementation will be greater in the higher income groups, with less investment at lower income levels. The alternative form hypothesizes that supplementation (possibly through public welfare programs, or through a lowering of coinsurance levels in the final version of the NHI plan) will be greater at lower income levels with no change in the highest income group. In each of these alter- natives, a weighted mean coinsurance reduction “2 There are no known empirical antecedents for these percentage reductions. The 40-percent reduction might be con- strued as somewhat less than half the population reducing gross effective coinsurance to zero. with an “intermediate" assumption of 25 percent reduction in gross effective coin— surance. of 25 percent was computed; the resulting estimates of demand change are shown in Tables 7 and 8.“3 The last general category of supplemental in- surance contingencies includes assumptions of the persistence of public health programs follow— ing the advent of NH1. As has been noted, the CHIP coinsurance rates are not necessarily com- parable to the gross effective coinsurance rates applicable to pre-CHIP demand levels. Unless all expenditures are covered either out—of—pocket or by CHIP, effective coinsurance rates after CHIP will be lower than the CHIP rates specified in the bill. Besides lowering their coinsurance re- quirements by buying supplementary insurance policies, persons might continue to acquire free health care through welfare programs, free in‘ stitutions, or other dispensers of care with no direct charges to the patient. Either of these ac- tions lowers the proportion of expenditures which must be met out-of—pocket and lowers the post—CHIP effective coinsurance levels. As is seen from the above discussion, the availability of free services after CHIP can be considered a form of supplementation of the CHIP benefits. Since they, too, have the effect of lowering the share of total expenditures which must be met out-of—pocket, they can be treated in the same manner as voluntary purchases of addi tional insurance coverage. However, their impact may differ substantially from that of voluntary supplementation, as the distribution of free ser— vices is heavily weighted toward the low-income groups. This last series of supplemented CHIP coin- surance rates is based on the assumption that the “free" care, excluding Medicare and Medicaid, that currently exists would be maintained after CHIP. In effect, then, if no other supplementation existed, the population would be paying no coin- surance on these “free" services and the CHIP coinsurance rates would apply only to the rest of their reimbursable health services consumption. The extent to which free services will be con- tinued after the introduction of CHIP is not known. There are a number of “free" services (such as those oriented towards public health '” Since the doublelog demand model includes the expression: pr where the exponent, n, is always negative (see Figure 10), a “zerodivide'condition would be set up were the con— sumer price (K) under CHIP permitted to drop to zero. Therefore, where applicable, post-CHIP coinsurance was ex- pressed as 0.0001. 59 needs) which might continue to be provided because they are not amenable to market provi- sion. Certain organizations could be expected to continue provision of free services if they con sider CHIP coverage to be inadequate in specific areas for certain populations (for example, dental care for adults). On the other hand, after Medicare and Medicaid, many health care pro- viders eliminated their price discrimination policies and introduced charges for services to low-income patients where none had previously existed. If current providers feel that their pa- tients have acquired the ability to pay for their services through CHIP coverage, the availibility of free services would probably diminish. The ex- tent to which they are continued will probably be determined by the degree to which unmet health needs are perceived by local health planning agencies and providers of public health services. There are considerable differences in the amounts of free care available among the dif- ferent types of services. For instance, only a small amount of free drug services would remain if Medicaid funds were removed. It would be an- ticipated that the maintenance of free services would be more significant for physician and hospital inpatient care, where the share of free services may account for between 5 and 10 per— cent of total expenditures and for over 15 percent of the inpatient expenditures of the lowest in- come groups.64 Since utilization of free services of all types is higher among the lower income groups, their continued availability would have the greatest impact on the effective coinsurance rates of the low income groups. Unfortunately, there was no information at hand on the availability of free services by income groups specifically for the over-65 and underA65 populations. Figures were derived, based largely on the CHAS “best estimate" survey results65 which were covered by “Medicaid, Welfare, free institutions" and “other free care" by family in- come groups for the whole population. Since most of the Medicaid programs would be discontinued under CHIP, estimates of the percent of Medicaid funds were subtracted from these figures; the re— mainder then reflected the percent of health ex— penditures provided through “welfare, free in- stitutions,“ and “other free care." 64 Cooper, Barbara 8.. op. cit. 65 A ndersen, R., op. cit. The over-65 population received a higher pro- portion of free drug and dental care than did the population as a whole. Since the utilization of such services is heavily weighted toward the low income groups and since the income distribution of the elderly is most heavily concentrated in the lower income classes, it was assumed that the greater utilization of those services by the elder- ly could be attributed to their lower incomes and that the utilization of these free services within each income group was the same for the over-65 and under-65 populations. The estimated share of expenditures received through free services by family income was therefore applied to both the over-65 and under-65 population. Because Medicare funds comprised such a large proportion of the elderly's expenditures for hospital inpatient and physician services, the elderly received a lower proportion of their ser- vices from free sources than did the general population—even though their income was lower. Since free inpatient and physician services are also utilized to a greater extent by the poor, the differences between the under-65 and over-65 utilizations within each income group are prob- ably greater than the aggregate difference. However, as insurance coverage becomes similar for the two populations under CHIP, their behavior might also become more similar. As a compromise, it was assumed the difference in utilization of free service by the over-65 popula- tion within each income group after CHIP would be equivalent to the current aggregate difference in utilization. 60 There was no attempt made to derive new rates on the utilization of free services for the CHIP income categories, even though the income groups under CHIP were not always comparable to those of the published CHAS data. The cor- respondence between the family income groups for the under aged 65 population was close; but the individual income basis for determining re- quirements for the elderly prevented a direct relationship from being drawn between the two. Because numerous assumptions had to be made in order to derive figures on the utilization of free services by the over-65 and under-65 populations, because data used to derive the figures came from different sources, because the status of free _ services after the inception of CHIP is uncertain, and because the derived utilizations were not based on income groups identical to those in CHIP, the revised coinsurance rates should not be considered as necessarily representative of the post- CHIP situation. However, since free ser- vices do play an important role in the supply of health services to the low-income population, their existence should not be ignored. One possibility is that they would continue to be of about the same magnitude as they currently are. If NHI functions properly, there would be little need for them to expand. The continuation of the existing level of free services could thus be con- sidered a probable maximum, and the actual ef— fect of the availability of free services on total utilization might be expected to fall at or below the estimated demand levels presented under this assumption. XII. Summary Of Biases There are a number of possible sources of bias which may affect the demand and manpower re— quirements estimates, many of which have already been noted. The following section presents a summary of the potential biases of the study. It was in large part prepared by Dr. Allen Dobson of the Office of Program Planning Evalua- tion and Legislation, Health Services Administra— tion (HSA), as part of his comments on an initial draft of this paper. The authors have modified and supplemented his summary. (a) The Voluntary Nature of CHIP The model assumes that everyone will become insured under CHIP. This is probably wrong for two reasons. First, those who are currently enrolled in more favorable programs will not opt for CHIP. This fact suggests that the presented demand estimates are too low, i.e., the average coinsurance rate after CHIP's imposition is incor- rectly estimated and should be lower. Counter to this, however, is the potential number of people who will not utilize CHIP and have no other in- surance. These uncovered persons would lend to upward biases in estimates of the percentage of insured persons and hence in demand estimates. On balance, if CHIP is successful in recruiting members, it is reasonable to assume the first bias to be the strongest. Thus, the overall impact of this bias is to suggest that the low demand estimates should be revised upward. It should be noted, however, that the more liberal NHI policies evaluated by Huang-Shomo (see technical appendix) represent an upper limit to such revi- sion. The impression is that if the CHIP estimates should be revised upward, they would not be raised substantially. To do so would imply that liberal and less liberal plans were associated with essentially the same demand. The reason the liberal plans don't require upward revision (because of the bias under discussion), is that they offer services that are approximately equal to the best that is available today. The net result of assuming universal CHIP membership is to com- press the range of demand estimates. (See Sec- tion d below.) 61 (b) The Effect of Deductibles The literature does not offer a precise evalua- tion of the impact of deductibles on the demand for health services. A paper by R. Zelten of the Wharton School, for instance, presents evidence from numerous surveys and private insurance company files which in no way presents a unified picture of deductible effects.“ Zelten cites a 1962 survey of 65 insurance companies and 3 prepay- ment plans that reported only 42 percent of its respondents feeling that deductibles were effec- tive in curtailing excessive utilization. Another survey cited by Zelten was based on 1971 Blue Cross data and suggested that cost-sharing does contain utilization, but it does not reveal to what extent. Zelten presents other information that suggests that deductibles at times have little im~ pact on utilization. He concludes that coinsurance clauses are probably more effective in reducing utilization that are deductible provisions. Data and assumptions presented in Feldstein, Friedman, and Luft‘57 and Rosett and Huang88 suggest that consumers are sensitive to deducti— ble clauses and at an increasing rate as price elasticity rises. It also appears that deductibles are regressive and that, if sizable, they may reduce the relative demand of lower income groups. Since the demand estimates only partial- ly reflect deductible features, it may be that most estimates are too high on this account. (c) The Assumption of Constant Elasticity Across Income Groups ’ This assumption could again produce results with an upward bias to the extent that low in- come groups are more sensitive to net price; whether it is affected by coinsurance or de- “ Zelten, Robert A., op. cit. “7 Feldstein, Martin S.. Friedman. Bernard 5., and Luft. Harold, “Distributional Aspects of National Hdalth Insurance Benefits and Finance," National Tax Journal, December 1972. “5 Rosett N., and Huang. Lieniu, “The Effects of Health In‘ surance on the Demand for Medical Care," Journal ofPolitical Economy, March/April 1973. ductibles. Another aspect of health utilization by lower income families appears to be that their in- itial utilization rates are fairly high when they enter the system. This could lead to an initial bulge in demand if large numbers of low income families are actually brought into the health care system by NHI legislation. The implication is that shortrun problems may be encountered during the imposition of an NHI plan which brings large numbers of low income families into the health care system. This problem would be a single oc- currence, but to the extent that CHIP's deduct- ibles and coinsurance ratios are effectively lowered, some thought must be given to initial en- try dislocations in the health care market place. Other investigators feel that, in addition to any initial one-time surge in demand, there is a “learn- ing curve" effect where much of the increased de- mand occurs over an extended period of several years."9 (d) The Assumption That Future Populations Con— sume Health Services at the Current Rate This assumption has offsetting components. The trend over the past five decades has been towards higher per capita consumption of health services although at a decreasing rate. There is no apparent reason why this trend will subside. On the other hand, the extrapolation of utilization rates of the insured to remaining segments of the population may overstate responsiveness of de— mand to insurance coverage. Insurance con— sumers are a self~selected group. If segments of the population in generally poorer health pur- chase insurance or if persons who purchase in- surance are those who place a higher value on health care, an extension of current insurance— related utilization rates to the entire population under an assumed universal NHI might upward bias results. The overall magnitude of this bias is probably impossible to estimate except through the use of experimental research designs such as the Rand Health Insurance Study.70 ‘19 Andrans, R. et. a]. Planning a National Health Manpower Policy: A Critique and a Strategy. Paper prepared for OPPE/HSMHA, March 5. 1973, (see page 27). If this is the case. there is reason to suspect the accuracy of many of the estimates of elasticity of demand that were conducted over periods of time of 2 years or less. 7" Newhouse, Joseph P., “A Design for a Health Insurance Experiment".1nquz'ry, Vol. XI, N(. 1 (March 1974), pp. 5-27. .62 (e) The Specification of the Demand Model The demand models under consideration state that demand is a function of price (effective coin- surance rate). Clearly, the demand for health ser- vices is also a function of age, race, sex, income, time costs, etc. To the extent these variables are not explicitly entered into the model, biases could be imparted to demand estimates. Many of these variables, however, were entered into the model indirectly by partitioning the population into separate cohort groups. If demand is estimated within cohort groups and then weighted ag- gregate estimates are formed, the impact of fac- tors such as age and income are taken into ac- count.71 The difficulty is that the relative sizes of the various cohort groups (e.g., a relative increase in the elderly who exhibit high utilization rates) may change over time. To the extent that they do, future populations may exhibit different utiliza- tion patterns than do present ones. Biases of this type are undoubtedly a part of the estimates, but they are thought to be minimal. if) Selection of Demand Curve Shape The use of three demand curve shapes pro- duced sharply divergent estimates. It is suspected that the doublelog model is generally too high, but there is no empirical basis for the selection of any particular demand curve shape over the alternatives. Since the doublelog model results tend to be unreliable at coinsurance levels below about 15 percent, however, it may be less suitable to the present analysis. The truth perhaps lies between the linear and semilog models. The linear model's estimates are typical- ly low, especially for CHIP; these low numbers could be the more accurate estimates, however. (g) The Appropriateness of CHIP Demand Estimates to Other Systems of NHI The CHIP demand estimates presented in this report assume supply flexibility. That is, demand increases are estimated for a world where these demands could be met. For NHI packages associated with small overall changes in demand, this is perhaps not an unreasonable assumption. For extremely liberal NHI plans this is not a valid assumption. If such a plan were adopted, demand 7‘ Where weights are determined by the relative sizes of the various cohort groups. would initially rise only to be thwarted by limited elasticity of supply responses. Prices would rise as would queuing times. A model that accurately predicts effective de- mand for very liberal NHI plans at a minimum must simultaneously account for consumer responses to price changes and increased queuing times as well as known facts concerning the sen- sitivity of consumers to varying levels of coin- surance, deductibles, etc. At present little is known concerning the potential impact of queuing time on consumer behavior or about producer responses to large changes in demand. It is 63 unlikely that the techniques reported in this study would give accurate representations of ac- tual market responses to NHI plans that have the potential of increasing demand far past any levels we have observed. What the model actually can do is present estimates of pressures that might be imposed on the health care system under vary- ing NHI assumptions. A critical assumption is that large estimated demand changes would be associated with significant changes in the health care market. Just what these changes might be, or what form they would take is not predictable from the assumptions of the model reported in this study. 7: at: 51.17.75 vw‘ t. ‘3» BHM@J 0.8. DEPARTMENT OF HEALTH. EDUCATION. AND WELFARE Mflc Health Service Hum Resources Administration hum of Health Manpower . MW Publication No.(l-RA) 77-02