HE (3:12 HB 1323 ■ 14 P4 Copy 1 Variation in the Rate of Infant Mortality in the United States Birth Registration Area RAYMOND PEARL. Ph.D. BALTIMORE Reprinted from the Transactions of the Eleventh Annual Meeting of the American Child Hygiene Association, St. Louis, October 11-13, 1920. »'l \K^ xZ'^'^ S^ ^^ VARIATION IN THE RATE OF INFANT MORTALITY IN THE UNITED STATES BIRTH REGISTRATION AREA^ RAYMOND PEARL, Ph.D., BalUmore Until recently it has been impossible to discuss on any accu- rate or satisfactory basis the infant mortality of any considerable portion of the United States. This difficulty has arisen from the fact that except in a few localities, notably some of the New Eng- land States, there has been in the past no adequate system of birth registration. The most accurate practical method of stating the force of infant mortality is to relate the number of deaths of infants under one year of age in a given time unit to the number born in the same time unit. Consequently, one needs accurate birth statistics before infant mortality can be adequately discussed. It is a matter of great satisfaction to everyone interested in the subject of infant mortality that at last there is well established a Birth Registration Area for the United States, and four annual reports on Birth Statistics of this area have been issued to date by the Census Bureau. We are well embarked now on the policy of adequate birth statistics for the country and unquestionably with- in a comparatively few years the Birth Registration Area will cover the major portion of the country as the Death Registration Area now does. In the short period since the Birth Registration Area has been established its growth in extent has been gratifyingly rapid. The first report on birth statistics for the year 1915 com- prised data from an area including approximately 31 per cent of the population of the country. The 1918 birth statistics report gives data from an area including 53 per cent of the population. This furnishes a sufficient volume of material so that one may begin the mathematical analysis of some of the problems of infant mortality with some assurance of reaching valid conclusions. » Papers from the Department of Biometry and Vital Statistics, School of Hygiene and Public Health, Johns Hopkins University, No. 18. This paper, which is a preliminary and condensed abstract of a much more detailed investiga- tion of the subject, shortly to be published elsewhere. The purpose of the present paper is a modest one. It aims simply to present briefly some of the facts of variation in rate of infant mortality in different geographic or demographic units of the population. The first step in the solution of any problem is obviously a clear definition of the problem itself. We shall see, as we pass from city to city, town to town, or rural county to rural county, that the rate of infant mortality varies greatly. In a hypo- thetical commonwealth where the most perfect administrative con- trol over infant mortality possible or conceivable had been attained, this variation would largely disappear, the only residue of diversity between communities in respect of infant mortality being such as arose purely by the operation of chance, that is, from random sampling, or from uncontrollable environmental factors, such as climate or soil. Now, with the actually existing condition of vari- ation between different communities in respect of infant mortality, it is obvious that there must be particulate and presumably deter- minable reasons for each particular difference which exists. Oper- ating on a basis largely of empiricism and a 'priori reasoning, efforts to reduce infant mortality have in the past been attended with considerable success. Also, with the advance of general sanitation the death rate under one year of age has fallen enormously. Greenwood ' quotes some interesting figures on the point from Farr, which we may well reproduce here to show how enormous has been the Improvement. Period 1730-49 1750-69 1770-89 1790-1809 1810-29 Percentage Deaths under 5 years.... 74.5 63.0 51.5 41.3 31.8 But after such a decline as these figures indicate the continu- ation of the business offers a difficult problem to the administra- tive official, whose procedures are grounded essentially only on the two pedestals of what he thinks has worked in the past and what he believes logically ought to work. The easy part of the conflict has happened and is in the past. To continue the good fight with the same relative measure of success, one presently must needs know more precisely than is now known the pattern of the causal nexus which controls and determines the rate of infant mortality. And it is real knowledge, not a priori logic, that is wanted. Let a ' Greenwood, M. Infant mortality and its administrative control. Eugenics Rev. Oct. 1912, pp. (of reprint) 1-23. QiSt Autiioi V.fM -0 1922 single example illustrate. It has been maintained that excessive infant mortality is primarily the resultant of the so-called "degrad- ing influences" of poverty, and such a contention stirs a warmly sentimental feeling of agreement in the minds of the well-meaning public zealous to do good. This relationship obviously ought to be true, therefore to a too-common type of mind it must be and is true. But Greenwood and Brown' in what may fairly be regarded the most thoroughly sound, critical and penetrating contribution which has yet been made to the problem of infant mortality are unable "to demonstrate any unambiguous association between pov- erty . . . and the death rate of infants." The plain fact is that before control or ameliorative measures can be applied with the maximum of efficient economy to the gen- eral public health problem of infant mortality we must know a great deal more than we now do about the factors which induce spatial and temporal differences in the rate of that mortality. But first we must get an adequate conception of the magnitude and character of the differences themselves. Let us, therefore, turn to the examination of the facts regarding variation in infant mortal- ity in the United States Birth Registration Area. Variation Data In this work we have studied the variation in the rate of infant mortality (deaths per thousand births) for the following groups: 1. Total population in cities of population of 25,000 or over In 1910. 2. Total population in cities of under 25,000 population in 1910. 3. Total population in rural counties of registration states. 4. White population in cities of population of 25,000 or over in 1910. 5. White population in cities of under 25,000 population. 6. White population in rural counties of registration states. 7. Colored population in cities of population of 25,000 or over in 1910. 8. Colored population in cities of under 25,000 population. 9. Colored population in rural counties of registration states. In order to make possible a better appreciation of the nature of the frequency distributions Figure 1 has been prepared. This shows for the year 1918 the frequency polygons for the total popu- » Greenwood. M. and Brown, J. W. An examination of some factors Influenc- ing the rate of infant mortality. Jour. Hyg. Vol. XII. pp. 5-45, 1912. lation of (a) cities of 25,000 and over, (b) cities of under 25,000, and (c) rural counties. so aio f30 £50 erO ■1.43 ;i.42 ■1.95 ■2.95 : .90 ■6.28 7.01 ■8.10 ■1.80 ^2.06 1 In 1910. 2 In concrete units, I. e., rate of deaths under 1 per 1,000 births. -f .3148 — .0786 -I-.2455 -I-.3237 -)-.1934 -i-.2217 4-. 4840 -1-.5625 -I-.3204 -[-.3536 -I-.2833 -I-.4328 -f-.1799 -I-.2802 -I-.4984 -f .5819 In Table I are presented the chief physical constants' of the distributions of variation in infant mortality. These constants have been determined by the method of moments from the original raw data." * For a very brief and summarized Introduction to the modern mathematical treatment of frequency curves see Pearson, K., "Tables for Statisticians and Biometricians," 1914, pp. Ix to Ixx. References to the basic literature on the subject will be found there. "I am greatly Indebted to my assistant, Mrs. Charmlan Howell, for aid In the arithmetical work of this paper. The constants tabled are: 1. The arithmetic mean. 2. The median. This nleasures the value above and below which exactly half of the variates occur. 3. The standard deviation. This constant measures In absolute units the degree of "scatter" or variation exhibited by the distribution. 4. The skewness. This constant measures the degree of asymmetry of a frequency distribution. If a distribution is perfectly sym- metrical on both sides of the niean so that If folded over upon the mean as an axis the two limbs would exactly coincide, the value of the skewness is zero. From the data presented in Table I. the following points are to be noted : 1. There is no certainly significant decline in the mean value of the rate of infant mortality during the four years covered by these statistics in any of the demographic units considered. 2. In 1918 there was a general tendency towards an increase in the mean rate of mortality over that which obtained in 1917. This increase is unquestionably to be attributed to the influenza epidemic of the autumn and winter of 1918. A careful examina- tion of the rates by months will convince one that the mortality of infants increased very materially during the period of the epi- demic. Whether this increased number of deaths was truly to be charged to influenza does not concern us here. The important fact is that the rate of infant mortality markedly increased coincidently with the existence of the epidemic. It is noteworthy that this in- crease in the infant mortality rate in 1918 is practically confined entirely to the cities. The rural counties, whether for white or col- ored or total population, show little or no change in 1918 as com- pared with 1917. 3. There is no unequivocal difference in the mean rates of infant mortality in the larger as compared with the smaller cities. Considering the largest differences in mean rates for total popu- lations in cities of 25,000 and over as compared with cities of under 25,000 there is no difference which is as much as even three times its probable error. 4. The mean rates of infant mortality are notably smaller in the rural than in the urban areas. This fact has, of course, long been well known. The first writer on vital statistics, in the sense in which we now understand that subject. Captain John Graunt, more than 250 years ago pointed out that rural communities ex- hibited generally a lower rate of mortality than urban communi- ties. The difference between urban and rural rates of infant mor- tality is reflected just as clearly in the high absolute rates of the colored population as it is in the lower rates of the white population. 5. The mean rates of infant mortality are, roughly speaking, something like twice as high for the colored population as for the white population in each of the demographic units considered, and at all times. This again is a fact in general well known, but here we have precise figures on the point, with probable errors, which show definitely how tremendously poorer the negro baby's chances of surviving the first year of life are than the white baby's. 6. The cities of over 25,000 exhibit distinctly less variation in respect of infant mortality than do either the smaller cities (under 25,000) or the rural counties. The smaller cities and the rural counties exhibit about the same degree of variation relative to their means, but absolutely, in terms of standard deviation, the rural counties show less variability than the cities under 25,000. The colored distributions exhibit a much higher degree of variation in respect of infant mortality however measured, whether absolute or relative, than do the white populations. In general, it may fairly be assumed that the greater the variation exhibited by a given class of the community in respect of infant mortality, the greater the chance of effective control and reduction of the average infant mortality by administrative measures. There can be no question that there is no field which offers so great opportunities in this direction as the colored population. 7. The skewness is seen to be positive in sign in every case but one. In that case (1916, cities over 25,000 total) the skewness is not significant in comparison with its probable error. With this excep- tion the curves tend to tail off more gradually and farther towards the right end than towards the left end of the range. In other words, the rate of infant mortality in these different American demo- graphic units tends generally to distribute itself in a substantially unsynmietrical fashion about the mean, extremely high rates oc- curring more frequently than correspondingly low rates. This fact might perhaps be taken to indicate that the task confronting the administrative control of infant mortality in the United States, and yet to be accomplished, is even greater than what has already been accomplished in the past, great and worthy of commendation as that is. Data on the Idmitatioiis to Administrative Control of Infant Mortality. We have seen that there is a high degree of variation in the rate of infant mortality as we pass from community to community. Some communities have infant mortality rates several times higher than those prevailing in other communities of the same size. This creates the presumption at once that proper administrative activity might reduce the rates of these abnormally high communities to a level commensurate with those found in the lower group. It is the purpose of this section of the paper to examine this presumption critically. At the outstart it is evident that there are some causes of in- fant mortality which are, in their very nature, beyond hope of effective practical human control. Thus, children born with marked congenital hydrocephalus will presently die, in spite of anything the health officer can do, no matter how active and intelligent he may be. There are other causes of death falling in essentially the same category in this respect. Not as any final or dogmatic settlement of the matter, but rather as a tentative first approximation made for the purpose of seeing whether any suggestive lead may appear, I have ventured to attempt to classify the principal causes of mortality in the first year of life into two groups. The first of these groups aims to in- clude those important causes of infant mortality which are either (a) actually now effectively controlled by the efforts of health officials, either directly, or indirectly through general sanitary and hygienic improvements, or (b) are obviously capable theoretically of control and amelioration if sufficient pains be taken. The sec- ond group aims to include those causes of infant deaths which are either (a) in the nature of the case, out of the range of effective practical, direct control or amelioration, or (b) are not in fact now controlled in any appreciable degree. Let us see what such a classi- fication, to a first approximation, looks like. Tentative Olassiflcation of Principal Causes of Infant Mortality. A. Causes of death actually now well B. Causes of death not controlled controlled, or capable theo- Tuberculosis of the lungs retically of direct control in Tuberculous meningitis greater or less degree: Other forms of tuberculosis Measles Syphilis Scarlet fever Organic diseases of the heart Whooping cough Malformations Diphtheria and croup Premature birth Dysentery Congenital debility Erysipelas Injuries at birth Tetanus Meningitis Convulsions Acute bronchitis Pneumonia Bronchopneumonia Diseases of the stomach Diarrhea and enteritis External causes One realizes that it is a bold thing even to set down such a classification as the above. It is certain to stir up the rancor of extremists in both directions. But extremists are nearly always wrong. Calm and unprejudiced persons will admit that some such classification as that here attempted is possible. Perhaps some further discussion of this classification may make clearer its point of view, and may win at least that measure of agreement with it which will at least permit the consideration of the discussion of its consequences which follows. Taking column A first, presumably no competent health official would deny that the first diseases in the list (scarlet fever, whoop- ing cough, diphtheria and croup, and dysentery) have been, can be, and are in greater or less degree effectively controlled in respect both of their incidence and their mortality. With this same group clearly belongs also diarrhea and enteritis, and convulsions, on the justifiable assumption that in the vast majority of cases convul- sions in infants are consequent upon violent enteric infections, which clearly belong in the controllable class. Diseases of the stomach, as causes of death under one year of age, again in the vast majority of cases undoubtedly mean infection — filth diseases in short — which come in the same category, so far as concerns con- trol, as diarrhea and enteritis. Regarding the rest of the diseases in the A group (erysipelas, tetanus, meningitis, acute bronchitis, pneumonia, bronchopneumonia, and external causes) the point of view of which led to their inclusion here is as follows : If the envir- onmental conditions surrounding the infant in the community and in the home, and the care given it, were made as favorable as they might be made, and actually are in the homes of the hygienically intelligent well-to-do, the death rate from each of these causes would be enormously reduced relatively in comparison with what it actually is. As a matter of fact visiting child welfare nurses are doing a mighty work in just this direction in many communi- ties. They teach parents how to care for their infants, protect them from these infections, and nurse them to a non-fatal issue in many cases if they do get infected." Now for the B column. The first three items are the various forms of tuberculosis. The fanatic will no doubt promptly assert that nothing is so easily and readily controllable as these. But let us make haste slowly and remember certain things: First, that we are here talking about deaths under one year of age, that is fatal tuberculosis in the first months of life; and second, that our classification premises, in specific and stated terms, direct control, that is control through agencies acting directly upon the infant or his environment. Theoretically it is possible to reduce materially the mortality under 1 from tuberculosis. If every child born to tuberculous parents was instantly and ruthlessly removed from the home from the moment of birth, and reared in an environ- ment where no contact with tubercle bacilli was possible, unques- tionably enormously fewer infants would develop tuberculosis in the first year of life than now do. A recent paper by Bernard and Debre' furnished an instructive example, showing in a single case how a child removed at II/2 months from its tuberculous mother, threw off completely its own tuberculosis. But practically it is perfectly clear that neither in the past has there been, nor in the present is there, nor probably for some time in the future will there be rigid isolation of offspring from tuberculous parents to an ®No "6ttetj;who knows at first hand what child- welfare public health nursing is actually accomplishing in these directions will question the putting of these diseases in the controllable column. Their mortality rate can be materially reduced if communities will take the trouble to go intelligently about it. ^Bernard L,., and Debre, R. Bull. Soc. Med. des Hosp. T. 44, p. 1658. 1920. Rev. in Jour. Am. Med. Assoc. March 19, 1921, p. 824. See also paper In the pres- •ent volume of these Transactions, on "Prevalence and Management of Tubercu- losis in Infancy," by T. C. Hempelmann, p. extent or degree suflficient to influence the infant mortality from tuberculosis, in the entire Registration Area of the United States, by as much as one unit of the death rate. The mortality from tuberculosis under 1 year of age has to be sure declined during the past 40 or 50 years but no more rapidly than the general curve of tubercular mortality at all ages. But many persons fail to find any evidence that control measures have had anything to do in bringing down the general tuberculosis death rate. In this connec- tion a recent paper by Given' on the influence of administrative or control measures upon the course of tuberculous mortality in gen- eral is interesting. He says : "Statistics show us that, in spite of all that has been said and done for the prevention of tuberculosis, our efforts in regard to pulmonary tuberculosis have not been attended with the anticipated success. The decline in mortality from this cause dates from 1838, and has continued steadily ever since down to 1913. Koch's discovery of the tubercle bacillus in 1882 does not appear to have affected it in any way." I know of no evidence that anything now being done is sen- sibly influencing the rate of mortality from tuberculosis in infants under 1. Some individual physicians may have been particularly successful with a small number of tuberculous babies under his care. But statistically it means little in the toll of 2,501 deaths under 1 which tuberculosis is recorded to have taken in the Regis- tration Area of the United States in 1918. Actually if the truth were known the total would be much larger even than this. About fatal congenital organic diseases of the heart, congeni- tal malformations grave enough to be fatal in the first year of life, and fatal congenital debility there will probably be no dispute. The mortality from fatal congenital syphilis is again, like tubercu- losis, theoretically controllable." But practically and in fact, is it controlled? The writer feels extremely doubtful about it. Regarding premature birth, and injuries at birth, much the same reasoning applies, but with the additional consideration that presumably intelligent prenatal education of the mothers and improvement of prenatal environmental conditions would reduce * Given, D. H. C. Some deductions from the statistics on the prevention of pulmonary tuberculosis. Bull. Med. Jour. Feb. 12, 1921, pp. 225-226. » Cf. Jeans, P. C. Syphilis and Infant Mortality. Trans. Am. Assoc, for Study and Prev. Inf. Mortality. Vol. IX, p. 155. these mortality rates in some unknown, but probably not large degree. Actually, however, there is no tangible evidence that these causes of death are in effect administratively controlled in any appreciable degree in this country at this time. Finally it should be said that one occasionally important cause of infant mortality is omitted entirely from the classification. This is influenza. The reason for the omission is simply that the statis- tical discussion which follows is based upon 1918 mortality figures TABLE II — Showing the deaths under one year of age per 1,000 living births for (A) controllable, and (B) non-controlled causes of death in certain American cities of 100,000 population or over in 1910. City Births In 1918 Deaths under one year A. From causes control- lable in some degree A. Rate of control- lable deatbs B. From B. Bate of causes not con- not con- trolled trolled deaths Rate per 1,000 births from all causes Bridgeport New Haven Washington Indianapolis Louisville Baltimore Boston Cambridge Fall River Lowell Worcester Detroit Grand Rapids .... Minneapolis St. Paul Albany Buffalo Bronx Borough . . . Brooklyn Borough . Manhattan Borough Queens Borough . . Richmond Borough Rochester Syracuse Cincinnati Cleveland Columbus Dayton Toledo Philadelphia Pittsburgh Scranton Providence Richmond, Va. . . . Seattle Spokane Milwaukee 4,910 4,869 8,162 6,196 4,368 15,143 20,062 2,672 3,646 3,286 6,238 27,036 2,836 8,704 5,155 2,153 13,989 16,763 49,515 59,227 9,467 2,677 6,855 4,352 7,913 20,699 4,464 3,282 5,524 43,408 15,875 3,139 6,384 3,840 5,910 2,194 11,090 226 46 190 39 399 49 270 44 239 55 1,225 81 1,092 54 144 54 403 111 302 92 212 40 1,296 48 110 39 198 23 160 31 96 46 866 62 496 30 2,232 45 2,855 48 389 41 113 42 283 41 265 61 326 41 963 47 163 37 109 33 186 34 2,876 6d 1,179 74 263 84 342 54 199 52 93 16 55 25 574 52 224 46 200 41 450 55 269 44 210 48 847 56 984 49 111 42 183 50 180 55 248 47 1,199 44 119 42 358 41 135 26 122 57 653 47 669 40 1,889 38 2,456 41 417 44 139 52 276 40 206 47 404 51 790 38 255 57 143 44 270 49 1,993 46 805 51 141 45 352 55 285 74 218 37 90 41 488 44 100 90 112 93 112 14» 115 107 180 159 97 100 86 - 73 87 116 121 75 90 97 93 106 92 119 104 98 101 87 94 124 139 141 123 147 61 77 106 and inasmuch as that was a year in which the influenza mortality was abnormally heavy owing to the epidemic it was thought that it would be unfair to the general relationships exhibited to include this epidemic mortality. Presumably normal endemic influenza should be in the A group, on the same reasoning as the pneumonias. With so much of explanation as to the point of view of this classification let us examine some of its statistical consequences. These consequences I have tested in a preliminary way upon the birth and death data for certain large cities and the registration states in 1918. There were found to be 37 large cities included in both Birth and Death Registration Areas in that year, and 20 states. For each of these cities and states the births were taken from 1918 Birth Statistics and the deaths under one year of age according to causes from Table II of the 1918 Mortality Statistics. From these data the rates per thousand living births for all class A and all class B diseases were separately calculated. The results are set forth in Tables II and III. TABLE in — Showing the deaths under one year of age per 1,000 living births for (A) controllable, and (B) non-controlled causes of death in twenty registration states. Births in 1918 Deaths under one year state A. Prom causes control- lable in some degree A. Rate of con- trollable deaths B. From causes not con- trolled B. Rate of not con- tiolUi! deaths Rate per 1.000 births from all causes 36,971 64,385 39,117 62,338 16,798 34,113 95,640 91,011 55,941 9,642 242,155 75,525 124,586 220,170 15,499 14,478 7,507 63,062 25,682 60,867 1,755 2,482 1.163 2.325 670 2,531 5,284 3,496 1,317 451 10,897 2,850 5,029 14,506 947 308 258 2.529 544 1,854 47 39 30 1 37 40 74 55 38 24 47 45 38 40 66 61 21 34 40 21 30 1,723 2,520 1,522 2,328 743 1,730 4,324 3,760 2,060 499 10,333 2,319 5,206 10,295 783 474 343 2.448 980 2,334 47 39 39 37 44 51 45 41 37 52 43 31 42 47 51 33 46 39 38 38 107 87 80 Kentucky 93 101 Maryland Massachusetts 140 113 89 71 New Hampshire New York 113 97 102 Ohio 94 Pennsylvania 129 126 64 93 Virginia Washington 103 69 79 In the last column of these tables the gross infant mortality rates from all causes of death have been inserted for comparison and to furnish the basis of certain discussions which will follow. It will be noted that the five boroughs of New York City have been treated as separate cities. This appears to be entirely justifiable, both on grounds of size, and of differentiation, any two of these boroughs being as much differentiated biologically and demo- - graphically as, for example, Minneapolis and St. Paul. The first point which strikes one in examining Tables 11 and III is that in the group of causes of death subject to our classifica- tion (which includes in most cases, as will be seen, something over 90 per cent of all the mortality under one year of age) the con- trollable and uncontrolled causes are responsible for approxi- mately an equal degree of mortality. In other words, it appears that if any degree of justification attaches to the classification here suggested, the infant mortality beyond present control by adminis- trative measures is by no means a negligible fraction of the total infant mortality. On the contrary, it represents a substantial lower limit sensibly below which the health officer, no matter how zealous and intelligent his activities, may not hope to go at the present time. If there is a substantial moiety of the existing infant mortal- ity which is uncontrolled by administrative measures and is essen- tially unaffeected by the present or past application of such meas- ures, we should expect that the rate of mortality represented by this moiety would vary but little from city to city or state to state. As we have seen, the main reason why this part of the total infant mortality is beyond control is because it depends upon fundamental biological factors inherent in the parents and the infants. Clearly if this is so, whatever variation appears in this portion of the total infant mortality rate as we pass from community to community must arise from some combination of two factors; of which the first and less important is pure chance, that is, variation arising from random sampling purely; and of which the second is differ- ing racial and other biological characteristics of the populations of the several communities. We should expect the variation in the death rate from the class B group of causes to show very little vari- ation as compared either with the variation in the rate from class A causes or in the gross infant mortality rate from all causes. This a priori expectation is beautifully realized in the actual statistics. TABI/E IV — Frequency distributions of variation in rates of mortality under one per thousand births for (A) controllable, and (B) non- controlled causes. Cities States Rate A Causes B Causes All Causes A Causes B Causes All Causes 15-24 2 5 9 12 3 2 2 1 i i 16 i 13 6 1 1 1 2 1 9 1 7 6 6 1 1 2 1 2 1 1 i 3 3 7 3 2 2 11 7 25-34 35-44 45-54 55-64 1 65-74 2 75-84 2 85-94 5 95-104 4 105-114 3 115-124 125-134 2 135-144 1 145-154 155-104 165-174 1 175-184 Totals 37 37 37 20 20 20 TABLE V. Variation constants from the distributions of Table FV. (Inni]) Mean Median Standard deviation 49.46 -f- 2.04 47.30 ■+■ .90 107.84 -+- 2.75 42.00 ■+■ 2.17 42.50 -+- .83 97.00 + 3.03 47.08 46.15 102.86 40.71 42.27 95.00 18.37 -f- 1.44 8.09 -H .63 24.78 -+- 1.94 14.41 + 1.54 5.52 -t- .59 20.07 ■+■ 2.14 It is seen that the class B causes of death, which are not prac- tically capable at the present time of administrative control or amelioration, exhibit less than one-half as much variation in the rate of infant mortality for which they are responsible, as we pass from city to city or from state to state, as do the class A causes of death, ivhich are capable of administrative control. This relation is true however the variation is measured. This is a novel result, of interest from several points of view. In the first place, the suggestion lies near at hand that if the class A causes of death, which are controllable, show such great variation relatively as they do, it must mark an approximately equal variability in the zeal, intelligence, and efficiency of the administra- tive health officials of these communities. Anyone at all familiar with the organization of municipal and state health departments in this country will find it extremely interesting to study in detail the entries of Tables II and III noting how the class A (con- trollable) and the "all causes" rates fluctuate up and down, while the class B (non-controlled) rates stay, with a very few excep- tions, so extremely constant. One will observe, with great satis- faction what splendid work is being done in some communities in holding down to a low level the infant death rate from controllable causes. Table II forms a real justification of the faith that is in the public health official of vision. It shows that the infant mortality from controllable causes can be kept down to a low level, and is in some communities. In the following cities (17 out of 37) the rate of infant mortality from the controllable causes of class A is actually lower than the rate from the non-controlled causes (class B) . New Haven Cincinnati Wlashlngton Columbus Worcester Dayton Grand Rapids Toledo Minneapolis Providence Albany Richmond Bronx Borough Seattle Queens Borough Spokane Richmond Borough These cities stand as examples of the fact that a considerable portion of the infant mortality rate can be controlled. Summary. This paper is a first biometric survey of the infant mortality statistics of the recently established Birth Registration Area. It is to be regarded as preliminary to certain analytical studies of the problem of infant mortality now in progress in this laboratory. The chief results of the paper are first to set forth and discuss some of the constants of variation in infant mortality in the differ- ent demographic units. This variation, which is large in amount, markedly and consistently skew in the positive direction, defines and throws into high relief the fundamental public health or admin- istrative problem of infant mortality. Why do the communities having rates of infant mortality higher than the average occupy that position? Is it from causes capable of human control, or from causes beyond the possibility of such control? A special prelimi- nary analysis of the data for cities of over 100,000, and the regis- tration states indicates that causes of death capable of administra- tive control are chiefly responsible for the variation observed in the total infant mortality rate, while those causes of infant deaths which, for fundamental biological reasons, are not sensibly in- fluenced or controlled by administrative measures, are a highly stable and constant factor from community to community, con- tributing little to the observed variability of the total infant mor- tality rate. In absolute terms, however, these causes of death not administratively controlled are responsible for roughly 40 per cent of the total infant mortality in the communities discussed. LIBRARY OF CONGRESS 013 825 380 6 ♦ n