Medical Statistics of the United States Army, Calendar Year 1953
APPENDIX
Sources, Definitions, and Methodology
SOURCES
Primary Source
Individual medical records constitute for Army personnel the source of all data relative to admissions, dispositions, and days lost from duty, except for a few instances where other sources are specifically cited. These records are submitted monthly by each Army medical-treatment facility, worldwide, for all dispositions of personnel treated on an excused-from-duty basis, plus certain other cases ?carded for record only.? (See ?CRO cases.?) The records are submitted on the Medical Report Card (DA Form 8-24), the Field Medical Card (DA Form 8-26), or the Emergency Medical Tag (DA Form 8-27). The individual medical record summarizes the salient facts of the case, including the date and place of initial admission and each transfer to another medical-treatment facility, the conditions treated, the circumstances relating to injuries, the surgical operations performed, the length of stay in hospital, quarters, on leave, etc., and the type of final disposition. In addition to the reports submitted for completed cases, an interim report called a ?Remaining Card? is made on cases treated during the year but not disposed of by a particular subsequent date. This makes it possible to complete the experience for the particular year and so to compute rates, etc.
Summary Report Sources
The Morbidity Report (DD Form 442) is a 4- or 5-week report month summary of the admissions and dispositions of Army, Navy, and Air Force personnel, submitted by each medical-treatment facility. It includes a distribution of cases by diagnostic class and selected diagnoses, and reports the average number of persons regularly provided their primary medical care by the facility. Data from this report are used in the present volume in a number of instances, as in data on non-Army personnel, comparison data on selected infectious and parasitic diseases, etc.
The Beds and Patients Report (DD Form 443) furnishes for each Army hospital and infirmary a monthly (report month) census of patients, by type, itemizes the flow of patients, and indicates the available bed capacity and the extent of its utilization.
The Outpatient Report (DD Form 444) is prepared for each 4- or 5-week report month to provide data on medical care furnished to outpatients, that is, to cases treated without being excused from duty. The report shows, by category of personnel, data on outpatient visits and treatments and the frequency of various types of physical examinations and immuni?zations.
The Dental Service Report (DD Form 477) is prepared monthly by each Army installation or unit which provides its own dental service. The report shows, by type of personnel, dental operations and treatments; examinations completed, personnel data, etc.
The Report of Veterinary Meat and Dairy Hygiene Inspection (DD Form 8-134) is prepared monthly by all Army installations or units with meat and dairy inspection activities.
101
Information on Army-owned animals, including admissions, days lost, and deaths, is derived from the Veterinary Report of Sick and Wounded Animals (DD Form 8-129).
The source of Selected Service data is cited at the end of ?Processing of Selective Service Registrants (1950-53).? Sources of Army Medical Service personnel data are cited on source tables 23 and 25.
DEFINITIONS
Admissions, in general, refers to the number of persons admitted to treatment on an excused-from-duty basis and not returned to duty prior to 2400 hours the same day. However, in this annual report, the nonbattle CRO (carded-for-record only) cases have been included with the admissions. Where CRO cases are known to be an important factor in evaluating admission data, their effect is discussed in the text.
CRO cases are cases ?carded for record only,? as distinguished from cases carded because they are treated on an excused from duty basis. A CRO individual medical record is prepared whenever any one of the following kinds of cases occurs among persons not excused from duty: disability separations, deaths, battle casualties, cases evacuated in patient status from an overseas area to continental United States, pregnancy cases, and new cases of venereal disease and nongonococcic urethritis. A CRO report is also made for those administrative separations (inaptitude, unsuitability, etc.) that have a medical basis, and, occasionally, for outpatients having a condition that, in the opinion of the attending medical officer, might result in a later claim upon the Government.
Incidence for a given diagnosis refers to the total number of new cases reported for the condition, including secondary diagnoses as well as causes of admission. CRO cases are included. This does not ordinarily reflect the true total incidence of the condition since outpatient cases, except those carded for record only, are excluded. Note, however, that all outpatient battle casualties, venereal disease, and nongonococcic urethritis cases are required to be carded for record only.
Battle casualties. The data on battle casualties in this report are limited primarily to 1953 experience, and are included in order to supplement other 1953 data on injuries, noneffective days, disability separations, and deaths. A later, more detailed report will present data on battle casualties for the entire 1950-53 period. Battle casualties who were killed outright, or who died before reaching the first medical treatment facility, are classified as killed in action (KIA). Others sustaining wounds or injuries related to combat are termed wounded or injured in action (WIA). ?Died of wounds? (DOW) cases include only those WIA cases who died after reaching a medical facility. WIA cases who return to duty before 2400 hours on the days they are wounded are considered ?carded for record only? (CRO). Rates for WIA cases are presented both with CRO cases included and with CRO cases excluded. Data on the ?missing in action? and on the ?captured? are not included in medical service reports.
Rates of admission, incidence, and separation are generally presented in this report in terms of ?number per 1,000 average strength per year.? Death rates, because of their lowness for most categories, are presented as number per 100,000 average strength per year. The rate is computed by dividing the number of admissions (or deaths, separations, etc.) by the average strength for the period covered. When annual rates are computed for periods of less than one year, the experience during the observed portion of the year is multiplied by an appropriate factor so as to indicate what the annual experience would have been had the same relative level prevailed throughout the rest of the year.
Days lost from duty because of illness or injury are expressed two ways: (a) average days per case and (b) the noneffective rate. The noneffective rate is the average number of persons who are excused from duty each day, per 1,000 average strength. This rate is computed by accumulating the total days lost during the year (by cases carried over from the
102
previous year and readmissions as well as by new cases during the current year) and dividing by 365 times the average strength (in thousands).
Average strength. The rates described above are based on average strengths provided each report month on DD Form 442 (Morbidity Report). The strength reported is the average daily number of persons entitled to receive primary (dispensary type) medical care from the medical facility making the report. The average strengths for different installations are combined to get average strengths by broad areas for a given report month. The average strength for the year is computed by weighting appropriately and combining the average strength for each report month. To obtain certain breakdowns of the average strength, such as by age, length of service, etc., the average strengths obtained from DD Form 442 are distributed in the proportions shown by data available from The Adjutant General?s Office.
Average duration. ?Average duration,? as presented in this report, may be defined as the average number of days excused from duty per case admitted for a disease or injury not previously recorded. The days include the total elapsed time from the date of admission to the date of disposition, plus the days lost on ?readmissions? that are considered as continuations of the original case. The data on average durations have been based upon dispositions during 1953, including cases admitted in prior years, in order to impose no limit on the maximum duration for individual cases. CRO cases are included unless otherwise specified; most of these cases are venereal disease cases, and durations excluding CRO cases are also shown for this category.
In arithmetic terms the average durations are computed as follows:
Average duration = Days lost by new admissions disposed of in 1953 + Days lost by readmissions disposed of in 1953/Number of new admissions disposed of in 1953
Averages by diagnosis are based on the admission diagnosis. In order to take some account of the effect of secondary diagnoses, a separate average is presented for the single-diagnosis cases.
Race. The category ?white? includes all cases not specifically stated to be Negro. Of the total number of individual medical records received, about one-half of one percent were for ?other races? (other than Caucasian or Negroid) and one-fourth of one percent were classified as unknown race. Including both of these categories in the group characterized as white has no appreciable effect on the resultant rates.
EPTS is an abbreviation for ?existing prior to service.? The EPTS condition is considered to be incurred ?not in line of duty? unless qualified as ?EPTS, aggravated by service.?
Evacuee cases are those which are evacuated in a patient status from overseas to continental United States.
METHODOLOGY
Individual medical records are received and processed by the Medical Statistics Division of the Office of The Surgeon General. The large volume of these records makes it feasible to process only a sample of the total records received. This section of the report describes briefly the method of sampling, and its effects upon the data presented.
Processing individual medical records. All of the individual medical records received are reviewed for conformance with reporting requirements. For sampling purposes, records for Army personnel are grouped into three broad classifications:
a. Special cases
This group includes all deaths; all disability separations; all battle casualties; all cases recovered from enemy control; all cases evacuated to the continental United States in a patient status; and certain nondisability separations from service, such as separations for inaptitude and separations for nonmedical reasons.
b. Other cases, 20 percent sample
103
c. Other cases, 80 percent sample
The last two categories of cases include all cases that do not fall into any of the ?special? categories. The 20 percent sample is selected on the basis of the units digit of the service number, which identifies each individual in the Army. Since service numbers ending in any particular digit should occur with about the same frequency as those ending in any other, the service numbers ending in any two specific digits should constitute a 20 percent sample of all Army cases.
All of the records in groups a and b are translated into codes which are written on ?transcription slips? and then punched into tabulating cards. In general, these codes cover the following information:
Identification of patient (service number).
Classification of patient (rank, sex, race, age, length of service, branch or service, etc.).Admission data (place and date).
Disposition data (place or type of medical treatment facility, date, nature of disposition made of the case).
Days lost from duty (total in each calendar year, total for completed case, days overseas, bed occupancy days, days on leave, etc., and days in quarters).
All diagnoses reported, together with related information (whether ?new? or ?previously recorded,? the anatomical location, the principal surgical operation for each diagnosis, the causative agent of injuries, and the principal residual condition or ?final result,? if any). Since the punch card form provides space to record only two diagnoses and their related fields, any additional diagnoses are punched into ?trailer cards,? using as many as necessary; the punching in trailer cards for a given case is identical with the ?principal card? for the case except for the diagnostic information. The code used is based on the International Statistical Classification of 1948 as adapted for use by the Army, Navy, and Air Force in SR 40-1025-1 ?Joint Armed Forces Statistical Classification and Basic Diagnostic Nomenclature of Diseases and Injuries.? The principal modification is in the coding of neoplasms and traumatisms, which are classified in much greater diagnostic detail, with the anatomical location involved coded as a separate item.
Data which pertain to battle casualties only (or to deaths only, or to disability separations only), have been tabulated from auxiliary files containing data on a 100 percent (no sampling) basis. All other data (admissions, incidence, noneffective days, average days per case, etc.) are tabulated from the 20 percent file of nonbattle conditions. Note that these 20 percent tabulations include the 20 percent portion of deaths, separations, etc.
The question of the possible sampling errors in the data is discussed below in terms of confidence limits. However, even with the data that have not been sampled (e.g., deaths or separations), there is need for caution in using numbers or rates that are quite small. In attempting to generalize from the observed experience, the magnitude of the numbers involved in that experience will affect the reliability of the data insofar as its applicability to other circumstances is concerned. Thus, for example, the 1953 death rate for the total Army for a condition such as appendicitis (0.2 per 100,000) may be expected to vary relatively more from the Army experience for, say, the 5-year period 1950-54 than will be the case with the 1953 death rate based on a larger number of deaths for a group of conditions such as malignant neoplasms (7.6 per 100,000). Similarly, the death rate shown for the group, infective and parasitic disease, for ages 20-24 are relatively more reliable for use in generalizations than those shown for ages 40-44, since, in this case, the population group at older ages was relatively smaller.
Confidence limits. Card counts from the 20 percent sample have been multiplied by 5 in order to show the total number of cards that presumably would have been obtained if all records had been processed. Consequently, all frequencies based solely on the 20 per?
104
cent sample are multiples of 5. These figures may be somewhat higher or lower than the true figure, and this ?sampling error? exists whether the data are presented as frequencies or converted into rates. Thus, it is useful to know how far the ?true value? might differ from the estimate based on the 20 percent sample, and with what degree of confidence one could expect the true value to be within specified limits.
?Confidence limits? indicate the range within which the ?true value? (derived from 100 percent of the records) can be expected to fall with a specified degree of assurance. The 95 percent and 99 percent degrees of confidence are two of the most commonly used. Using the 95 percent confidence limits, one may say that the chances are 95 out of 100 that the true value lies between the limits shown. Likewise, there is only one chance in 100 that the true value lies outside the 99 percent confidence limits.
Data from the 20 percent sample are not intended to show exact numbers of cases. Very low frequencies of zero, 5, 10, 15, etc., based on the 20 percent sample, obviously have a higher percentage of error than large frequencies. For certain diseases of very low incidence, the Morbidity Report is a useful source for obtaining comparative data, and this source has been utilized in the discussion of these diseases. Certain other data shown in the tables may be based on very low frequencies because the condition has been subdivided or cross-classified by other variables. Whenever low frequencies occur, they should not be read literally, but should be evaluated in terms of the sampling error they may contain. Table XXXVI shows the limits within which the true frequency may be expected to lie, when frequencies of 100 or less are based on a 20 percent sample.
The amount of sampling error in rates based on the 20 percent sample depends on (a) the size of the rate itself, and (b) the size of the population or average strength from which the sample was drawn.
Table XXXVII presents the standard error for specified rates and average strengths, based on the formula for the standard error of a proportion. Values for intermediate rates and strengths not shown in the table may be approximated by interpolation. From these standard error values, approximate 95 percent confidence limits may be computed by adding and subtracting two (more precisely, 1.96) standard errors to the rate based on the 20 percent sample. The approximately 99 percent confidence limits may be computed by adding and subtracting 2.58 standard errors to the rate computed from the 20 percent
Number of cases in 20 percent sample | Frequency shown in tables | Probable range of number in total population | Number of cases in 20 percent sample | Frequency shown in tables | Probably range of number in total population | ||
95 percent confidence limits | 99 percent confidence limits | 95 percent confidence limits | 99 percent confidence limits | ||||
0 | - | 0-17 | 0-24 | 11 | 55 | 30-94 | 24-108 |
1 | 5 | 1-26 | 1-35 | 12 | 60 | 33-100 | 28-115 |
2 | 10 | 2-34 | 2-44 | 13 | 65 | 37-106 | 31-121 |
3 | 15 | 3-41 | 3-51 | 14 | 70 | 41-112 | 35-127 |
4 | 20 | 6-49 | 4-59 | 15 | 75 | 45-118 | 38-133 |
5 | 25 | 9-55 | 6-66 | 16 | 80 | 49-124 | 41-139 |
6 | 30 | 12-62 | 9-74 | 17 | 85 | 52-130 | 45-145 |
7 | 35 | 15-69 | 12-81 | 18 | 90 | 56-136 | 49-151 |
8 | 40 | 19-75 | 15-88 | 19 | 95 | 60-141 | 52-157 |
9 | 45 | 22-82 | 18-95 | 20 | 100 | 64-146 | 56-163 |
10 | 50 | 26-88 | 21-101 | ||||
105
sample. For example, if the rate was 10 per 1,000 average strength, based on the 20 percent sample of 250,000 average strength, the chances are 95 out of 100 that the true rate lies between 9.12 per 1,000 and 10.88 per 1,000 average strength; or, the chances are 99 out of 100 that the true rate lies between 8.86 and 11.14 per 1,000 average strength.
It should be noted that in table XXXVII the standard errors of the rates are approximate values only. Because the sample is from a finite population (total Army, a specific geographic area, sex, race, rank, etc.), a finite population correction might be applied to the listed value of the standard error. Since the sampling fraction is 20 percent, the values corrected for finite sampling would be approximately 0.9 times the values shown in the table. Thus the uncorrected values shown overestimate slightly the standard error. It should also be noted that if the confidence limits are computed as indicated in the footnote to table XXXVII, the limits may not be reliable for the smaller rates shown when based on the smaller strengths. This is due to the fact that the normal approximation to the binomial distribution is being used to obtain confidence limits (although, in fact, the true distribution is hypergeometric). For very small numbers (and therefore for very small rates), confidence limits can be obtained from table XXXVI, which is based on accurate computations of confidence limits for the hypergeometric distribution.
TABLE XXXVII.?APPROXIMATE SAMPLING ERROR IN RATES BASED ON A 20 PERCENT SAMPLE
Rate per 1,000 average strength | Standard error of rate per 1,000 average strength assuming various average strengthsa | ||||||||
1,500,000 | 1,000,000 | 750,000 | 500,000 | 250,000 | 100,000 | 50,000 | 25,000 | 10,000 | |
1 | 0.06 | 0.07 | 0.08 | 0.10 | 0.14 | 0.22 | 0.32 | 0.45 | 0.71 |
2 | .08 | .10 | .12 | .14 | .20 | .32 | .45 | .63 | 1.00 |
3 | .10 | .12 | .14 | .17 | .24 | .39 | .55 | .77 | 1.22 |
4 | .12 | .14 | .16 | .20 | .28 | .45 | .63 | .89 | 1.41 |
5 | .13 | .16 | .18 | .22 | .32 | .50 | .71 | 1.00 | 1.58 |
6 | .14 | .17 | .20 | .24 | .35 | .55 | .77 | 1.09 | 1.73 |
7 | .15 | .19 | .22 | .26 | .37 | .59 | .83 | 1.18 | 1.86 |
8 | .16 | .20 | .23 | .28 | .40 | .63 | .89 | 1.26 | 1.99 |
9 | .17 | .21 | .24 | .30 | .42 | .67 | .94 | 1.34 | 2.11 |
10 | .18 | .22 | .26 | .31 | .44 | .70 | 1.00 | 1.41 | 2.23 |
20 | .26 | .31 | .36 | .44 | .63 | .99 | 1.40 | 1.98 | 3.13 |
30 | .31 | .38 | .44 | .54 | .76 | 1.21 | 1.71 | 2.41 | 3.82 |
40 | .36 | .44 | .51 | .62 | .88 | 1.39 | 1.96 | 2.77 | 4.38 |
50 | .40 | .49 | .56 | .69 | .97 | 1.54 | 2.18 | 3.08 | 4.87 |
60 | .43 | .53 | .61 | .75 | 1.06 | 1.68 | 2.37 | 3.36 | 5.31 |
70 | .47 | .57 | .66 | .81 | 1.14 | 1.80 | 2.55 | 3.61 | 5.71 |
80 | .50 | .61 | .70 | .86 | 1.21 | 1.92 | 2.71 | 3.84 | 6.07 |
90 | .52 | .64 | .74 | .90 | 1.28 | 2.02 | 2.86 | 4.05 | 6.40 |
100 | .55 | .67 | .77 | .95 | 1.34 | 2.12 | 3.00 | 4.24 | 6.71 |
150 | .65 | .80 | .92 | 1.13 | 1.60 | 2.52 | 3.57 | 5.05 | 7.98 |
200 | .73 | .89 | 1.03 | 1.26 | 1.79 | 2.83 | 4.00 | 5.66 | 8.94 |
250 | .79 | .97 | 1.12 | 1.37 | 1.94 | 3.06 | 4.33 | 6.12 | 9.68 |
300 | .84 | 1.02 | 1.18 | 1.45 | 2.05 | 3.24 | 4.58 | 6.48 | 10.25 |
350 | .87 | 1.07 | 1.23 | 1.51 | 2.13 | 3.37 | 4.77 | 6.75 | 10.67 |
400 | .89 | 1.09 | 1.26 | 1.55 | 2.19 | 3.46 | 4.90 | 6.93 | 10.95 |
450 | .91 | 1.11 | 1.28 | 1.57 | 2.22 | 3.52 | 4.97 | 7.04 | 11.12 |
500 | .91 | 1.12 | 1.29 | 1.58 | 2.24 | 3.54 | 5.00 | 7.07 | 11.18 |
aBased on the formula for the standard error (S. E. of a proportion, S. E.= √p(1-p)/n, where p is the proportion observed in the sample (and is thus 1/1,000 times the rate) and n is the size of the sample. Sample rate ?1.96 S. E. includes true rate with 95 percent assurance. Sample rate ?2.58 S.E. includes true rate with 99 percent assurance. Not corrected for finite population or for error in normal approximation; see text.
