AN IN-DEPTH EXAMINATION OF THE 1980 DECENNIAL CENSUS EMPLOYMENT DATA FOR HEALTH OCCUPATIONS. Executive Summary + li oy fruh) 0.8, DEPOSITORY SEP 11 1984 '''' AN IN-DEPTH EXAMINATION OF THE 1980 DECENNIAL CENSUS EMPLOYMENT DATA FOR HEALTH OCCUPATIONS Executive Summary U.S. DEPARTMENT OF HEALTH & HUMAN SERVICES Public Health Service Health Resources and Services Administration Bureau of Health Professions '' ''Introduction l pf x ruUB The availability of and access to accurate, timely, and reliable data are. essential to the Bureau of Health Professions (BHPr) in carrying out its functions of monitoring and analyzing developments relating to health personnel, and projecting the supply of and requirements for health personnel. A prime potential source of comprehensive data on persons in health occupations (and all occupations) is the 1980 Decennial Census of Population and Housing conducted by the Bureau of the Census. The American Institutes for Research (AIR), 1790 Arastradero Road, Palo Alto, California conducted an evaluation of the 1980 census data on health occupations under Contract No. 232-82-0017 for the Bureau of Health Professions. The objective of this investigation was to analyze and evaluate the data on health occupations to determine whether they are sufficiently reliable to be useful to BHPr and to determine what specific qualifications or statistical adjustments should accompany their use. The 5 percent Public Use Microdata Sample (PUMSA) from the Census was used to examine data for physicians, dentists, optometrists, pharmacists, podiatrists, veterinarians, physician assistants, registered nurses, licensed practical nurses, nurses aides, orderlies and attendants and personnel in ten allied health occupations. An extensive review of other sources of data on the supply and characteristics of health personnel was also undertaken in order to provide a basis for comparison with and evaluation of Census data. ''This summary gives an overview of the project which is described in detail in the final report for the project "An In-Depth Examination of the 1980 Decennial Census Employment Data for Health Occupations". Key project personnel from AIR included Frances Stancavage, Project Director, and Dr. Sandra Wilson-Pessano, Principal Investigator. A list of persons within various Divisions and Offices of the Bureau of Health Professions and the Bureau of the Census, as well as persons in other Federal agencies, professional associations and researchers in various settings who provided information and assistance to this project is provided in the final report. Those interested in further information should contact Ms. Ernell Spratley (301 -443-6662), John Drabek, Ph.D. (301 - 443-6662) or Mr. Howard V. Stambler (301-443-6936) at the Office of Data Analysis and Management, Bureau of Health Professions, Room 8-41, 5600 Fishers Lane, Rockville, Maryland 20857. ''Methodology The evaluation of 1980 decennial census data on health occupations required several major analytic tasks. First, the procedures of the Bureau of Census were examined in detail in order to understand the ways in which health occupations were classified during the 1980 census and also to examine the potential sources of sampling and non-sampling errors that effect the accuracy of these data. Second, an analysis of individual census records for health personnel was carried out in order to evaluate the numbers of persons and the characteristics of respondents classified as health personnel by the 1980 census. These analyses permitted the development of refined census counts for some occupational categories by screening or reclassifying individuals who had less educational Preparation than required for the occupation into which they had been classified. For example, persons who were classified as physicians and reported that they had completed fewer than 6 years of college were excluded from the refined counts of numbers of persons in that profession. By means of these analyses, demographic profiles were generated for personnel in the various health occupations at a level of detail not otherwise available from aggregate census data or (in many instances) from any other data source. The third task was an extensive review of data on the supply and characteristics of health personnel from sources other than the census in order to provide a basis for comparison with and evaluation of census data. ''The fourth major analytic task was an evaluation of the accuracy of census data on health personnel based on comparing these data with non-census counts of persons in these same health occupations. Both gross comparisons and, whenever possible, comparisons of the demographic, educational and economic characteristics of specific groups of personnel were carried out. The Comprehensive Report provides a detailed discussion of the methodology used and the results from each of these tasks. ''Findings Matching Census Categories and Occupational Groups of Interest to Health Professions Analysts The occupational classification scheme used by the 1980 Census provides estimates for each of six independent practitioner occupations, one mid-level practitioner occupation (physician assistants), three levels of nursing personnel (RNs, LPNs, and nursing aides, orderlies and attendants) and 10 allied health occupations at the therapist or technologist/technician level. However, correspondence between census Classifications and occupational groups of interest to health professions analysts is affected by the following factors: ° Census classification is based on self-reported job title, job description and industry, while health professions analysts rely heavily on licenses, certifications and specialized training to distinguish among health occupations. ° Some respondents have been misclassified, generally into occupations of higher status, Probably on the basis of ambiguous job descriptions. ° Some teachers, administrators, and researchers are classified into functional groups by the census, and thus the health occupations they belong to cannot be identified. For occupations with clear educational requirements, screening against minimum education criteria (adjusted by age if necessary) was used to minimize the first source of error. However, for the purposes of this project the usefulness of the educational attainment data collected by the Census was hampered by the following: (1) No data were collected on ''the content of the education received or on the degrees received. Although the number of years of college education was collected, it was not possible to determine how the years of education reported related to the individual's training for his reported health occupation. 2) No data were collected on vocational education received outside of a college setting. This is a major disadvantage since some nursing and much allied health education occurs ina non-college setting. Screening each of the independent and mid-level practitioner occupations (except nurses), dietitians, and four types of therapists resulted in the elimination of 5% to 50% of the respondents in these occupational categories. (A minimum education screening criterion was also applied to RNs, but it probably was not effective for removing most misclassified respondents.) The higher rejection rates characterized occupations with definitional as well as misclassification problems. Correspondence Between Census and Other Estimates Adjusted Census estimates based on the educational and/or industry screening were compared with estimates from other sources as illustrated in Table 1. Due to variability in the quality and suitability of the other sources, conclusive comparisons were only possible for the independent and mid-level practitioners, including RNs. These comparisons revealed census undercounts of up to 23 percent for the various independent practitioner occupations (except nurses) largely attributable to the classification of sizable segments of persons in these professions as teachers, managers, etc. Adjusted Census estimates for dentists and optometrists were virtually identical to other ''Table 1 Agreement Between Health Occupations Supply Estimates Based on 1980 Decennial Census and Other Sources (1) (2) (3) (4) BHPr AIR or Other % Difference Census Adjusted Non-Census Between Occupation Estimate 1/ Estimate 2/ Estimate (2). and (3) 3/ Phy si cians 435,600 408 , 300 430,200 <5 Dentists 127,600 118,100 118,300 0 Optometrist 25,100 21,700 21,300 0 Pharmacists 148 , 200 125,500 138,400 -9 Podiatrists 7,500 6,700 8,800 -24 Veterinarians 34,800 31,600 34,200 -8 Physician Assistants 6,200 7,700 -19 29,900 8,400 +9 Registered Nurses 1,316,300 1,297,300 1,227,900 +6 Licensed Practical Nurses 450,000 450,000 549 ,300 -18 Nursing Aides, Orderlies & Attendants 1,423,600 1,423,600 NONE 4/ -- Clinical Laboratory Technologists & Technicians 247 , 800 247 , 800 NONE 4/ - Dental Hygienists 48 ,500 48 ,500 44,700 +9 Dental Assistants 164,200 164,200 133,400 +23 Dental Laboratory Technicians N/A 5/ 44,500 33,300 +34 Dietitians 67,700 32,500 23,900 +36 Medical Record Technologists & Technicians 15,400 15,400 Medical Record Clerks N/A 5/ 50,700 TOTAL MEDICAL RECORD PERSONNEL 52,300 NONE 4/ ial Occupational Therapists 17,900 14,400 17,800 -19 Physical Therapists 44,100 32,200 32,300 0 Radiological Technologists & Technicians 96 , 800 96 ,800 Radiographers N/A 5/ 14,800 TOTAL RADIOLOGY SERVICES PERSONNEL 111,700 86 ,000 +30 Respiratory Technologists & Technicians 50,500 50,500 NONE 4/ -- Speech Therapists & Audiologists 42,000 39, 800 42,100 -5 1/ Based on cases with non-allocated occupation, age, and education on the 5% PUMSA file. 2/ Downward adjustment based on education and/or industry screening when possible. managers by the census. 3/ Calculated as a percentage of the non-census estimate. 4/ For this occupation, no acceptable estimates for the total 1980 supply were available. Does not include any upward adjustments for persons who are in these occupations but are classified as teachers, researchers, or Comparisons of supply within the hospital sector were carried out, however, and are reported in the comprehensive report. 5/ This occupation is combined with non-health occupations in the same occupational category. estimates are therefore not available. Unadjusted census NOTE: See the comprehensive report, Section VI-Conclusions and Recommendations for a more detailed summary of census/non-census comparisons. to which they measure the same population being estimated by the census. assumed, therefore, that census data are unreliable for occupations exhibiting a large discrepancy in column (4). On the other hand, a small discrepancy does not, by itself, confirm the reliability of census data. Non-census estimates are highly variable in reliability and in the extent It should not necessarily be ''estimates, implying very little classification of persons in these occupations into the other categories. The most notable census undercount among the independent practitioner occupations was for podiatrists. This undercount appears to result not only from the classification of some personnel as teachers and managers, but other classification problems as well. Physician assistants and RNs, on the other hand, were overestimated by the census as a consequence of definitional problems not solved by adjustments. Correspondence between demographic data from the census and other sources was best for those occupations that had fairly good agreement on the total estimates. Correspondence predictably decreased as the census estimate deviated from the estimate from other sources due to undercounting or overcounting. For the remaining nursing and allied health occupations, census/other comparisons were attempted, but were not definitive due to questions concerning the accuracy of the non-census data. Insofar as they could be evaluated, the only systematic biases observed in census data were those associated with upward misclassification. These biases, which were reduced for some occupations in the adjusted census estimates produced by the project, would have resulted in overestimates of the proportions of female and minority personnel and underestimates of the education and income levels of affected occupations. Recommendations for the Use of 1980 Census Data on Health Occupations Total supply estimates based on census data are especially of interest for the allied health occupations, since most of these occupations lack reliable alternative supply estimates and data on characteristics. While 8 ''this situation increases the attractiveness of the census data, it also means that these estimates cannot be validated against external criteria. The Comprehensive Report provides information to guide the user of census-based estimates for each health occupation. These occupation-specific discussions include adjustments based on educational and/or industry screening, consideration of special definitional problems, and comparisons to available non-census sources. The use of demographic, educational and employment data from the 1980 census is subject to the same cautions and qualifications as the total supply estimates. Predictably, these distributions appear to be most accurate for the occupations in which total supply estimates are also accurate. For most of the allied health occupations, as noted, the degree of accuracy is not known. For the user interested in conducting special analyses of occupational subpopulations using the person-level PUMSA files, these factors must be considered: sampling error, random coding errors, the systematic exclusion of teachers and administrators, specific definitional problems in some occupations, and (suspected) upward bias affecting misclassified respondents. All of these problems are magnified when analyses focus on very small groups, including subnational geographic units. Most health occupations are relatively small and also have highly skewed demographic distributions, with non-whites and one sex comprising very small fractions of the total. It is recommended that special subgroup analyses be restricted to numerically larger occupations, to subgroups that constitute sizable fractions (20% or more) within the occupations, and to subgroups less likely to be affected by systematic bias resulting 9 ''from ‘upward misclassifi¢ation “(if this ‘can be determined). “Such analyses should be approached’ cautiously for occupations where the validity of °°" gross supply estimates’ and’ demographic distributions have not’ been °° *~ demonstrated. rea OI 259 i sjemites bsesad-2uzne: The census’ data Could be more useful ‘if ‘changes were made in’ the’! *° educational attainment data collected ‘in future decennial censuses?! The: following additions to the questions on educational attainment are recommended: (1) An item 6H dégreé(s) ‘received (for those ‘reporting °°! college’ education)) and (2)° An item’ Gn’ vocational training? received /° "°° outsidé a college setting. “The addition of thesé items would be* helpfwi® in eliminating some Of the ‘ambiguity whith exists in’ the educational’ !°°* attainment’ reported by census respondents. ‘As @ result® ‘these’ data would be more useful not only in defining the occtipational classification of >” health personnel but those in other industries as well. BeS2Vvisiit b§I9Gea DiLsaauvomoo ie § i, res &° Sita or POLY np 1O7r PML iagwniss sDSsIsbiLenos + ~~ fey mg - = on hi = see ia ad sa ido Bi {it f L 103 JeinimDos fis 195693 IOs f cd jeo ~esir idgqsiposp Isnoitandue ibuloni ,aquoip iisme visv Li 2 Yidp sved Ig 5 fisme yvlevitsis1 916 enotissquooo ii si € One Ne bai6n esstinw-non djiw ,2 Ifastarb seylsns quoipdue Isiosqe ted? bebasmnoos1 ei FI -l6303 eds Yo anoljos12 J6a3 aquoipdus oF ,anolissaquoce iuspisl ylisolismun o3 betolisesi: ad OJ Bae ,ano:! 200 ede aintiw (910m 310 808) enotsosit sildesia gstusisenoo sepfdtfrimay sgetd orden +4 nxn a ala ao a ane a mn wrens pitjluesy esid oisgsmeseye yd besoslis sd ot eeal aquoipdue '''' puplic HEALTH LIBRARY : UL 8 1985 BHPr U.S. DEPARTMENT OF HEALTH & HUMAN SERVICES Public Health Service Health Resources and Services Administration Bureau of Health Professions ''GENERAL LIBRARY - U.C. BERKELEY NL B0008293ba ''