CHAPTER IX
Tabulating Procedures, 1932-57
At the beginning of the 25-year period from 1932 to 1957, the tabulating equipment in use for medical statistics in the Surgeon General`s Office was practically the same as that installed during World War I. The 45-column cards were punched and verified on manually operated machines, and all sorting was done by means of the vertical sorter. This machine, operating at 150 cards per minute, arranged the cards into sequence by diagnosis. The cards for each diagnosis were then subdivided by various classifications, such as new cases and readmissions, area of admission, month of admission, rank, sex, race, and type of disposition. The cards in each subdivision were then processed through the nonprinting tabulator and the results posted to a card which eventually contained all the information pertinent to that diagnosis. Similar sorts and counts were made to get data for posting cards representing specific causative agents, locations, surgical operations, or other items.
Data on posting cards for individual diagnoses were combined into the categories of the International List, and the data were posted to worksheets, with the information arranged in the form desired for publication. The count and aggregate days on these worksheets were divided by the appropriate strengths, and the resulting rates were posted to the final worksheet which was then ready for the printer.
Up to 1936, the tabulating equipment used for medical statistics in the Surgeon General`s Office remained the same. In 1936, however, the Vital Statistics Division installed machines using 80-column tabulating cards instead of 45-column cards and incorporated many improvements over the equipment used for the previous two decades. The punches were electrically
166
driven, with provision for automatic duplication of common information, and both the punching and verifying machines had automatic card feed and card ejection. The sorter was of the horizontal type first used in the late 1920`s, with speed increased to 400 cards per minute as compared to 150 cards per minute, and provided counts of each position punched in the column being sorted. In addition, a tabulating machine was installed which automatically printed a maximum of five 10-digit totals for each sort group. These totals were copied onto posting cards from which final tables were prepared.
To accommodate to 80-column card equipment, it was necessary to revise the punchcard forms, and a number of new forms were adopted. Dr. Hollerith originally used a round perforation in the card. The size of the hole punched limited the number of columns available on the card to 45. When a punch with a small rectangular head was patented, it became possible to increase the number of columns on the card to 80, which resulted in an increase in the potentials of the card.
Two forms, designated medical card and traumatism card, had been used for individual medical records reporting admissions for disease and injury, respectively. A veterinary card of similar design had been used for individual medical records of sick and wounded animals. Other forms were designed now, and the functions of these are described by their titles, such as, "Meat, Dairy, and Forage Inspection` and `Physical Examination for Flying.` These new cards were of the dual type, which provided for writing the codes on the punchcard itself and then punching these codes into the same card.
The first four of the card forms just mentioned remained in continuous use through the year 1943. At one time, the files of punched cards for individual medical records were complete for the period 1919-40, but these did not survive the competition for filing space during World War II. As might be expected, the card volume, which had been 117,000 for 1936 and 138,000 for 1940, jumped to 765,000 for 1941, 2,500,000 for 1942 and over 6,000,000 for 1943.
In anticipation of the large volume of individual medical records which would be generated by the extended operations in the European Theater of Operations, U. S. Army, a separate machine records unit was set up in England in 1942 in the Office of the Chief Surgeon of the theater. This arrangement
167
was for the purpose of providing the Chief Surgeon, European theater, with statistical data on a priority basis, after which the punchcards were forwarded to the Office of the Surgeon General in Washington. More than a million punched cards were prepared in the Chief Surgeon`s Office, European theater, and transmitted for the period 1942-44.
The punchcard forms and codes used, in what came to be called the London office, differed in many respects from the ones used in Washington. Although the London office initially adopted the Washington codes for diagnosis, anatomical location, surgical operation, and certain other items, a different card form was used, and in the middle of 1943 an independent system of codes for most items was instituted. This introduced many difficulties into the consolidation of London cards with the Washington cards for the tabulation of data for the European theater and for the total Army. For 1944 data, the London and Washington offices used the same card forms and codes, with some exceptions for specific fields which had little applicability to dispositions overseas. After 1944, individual medical records for all theaters were sent directly to the Washington office for the preparation of statistical punchcards.
At the beginning of World War II, a second tabulating installation was set up in Washington within the Surgeon General`s Office. The Supply Division had adopted electric accounting machine methods for use in stock accounting in 1941 and soon extended the application to procurement accounting, the International Aid Program, and other jobs connected with supply. The installation quickly grew to include 6 alphabetical tabulators, 10 horizontal sorters, 4 reproducing summary punches, 3 multiplying punches, 2 collators, an interpreter, and a battery of key punching and verifying machines. Two officers exercised administrative and technical direction, while eleven technical sergeants served as procedure writers, operators, and supervisors of the civilian force. The unit operated on a two-shift basis, with a third shift added from time to time.
The cyclical nature of the Supply Division jobs created valleys as well as peaks in the workload. With two machine records units occupying almost adjacent quarters, it was logical that they should be combined. Accordingly, at the end of 1943, the Supply Division installation took over the machines and personnel of the Medical Statistics unit, and tabulations for the
168
Medical Statistics Division were furnished on a service bureau basis.
The tabulation of medical statistics became increasingly complex for World War II. The data for 1942 and 1943 were on three different punchcard forms, and new ones were designed for 1944 and 1945,
in response to demands for additional kinds of information and new approaches in analysis. London cards had to be checked with Washington cards to eliminate duplication for cases which were later evacuated to the United States or which resulted in death. The 1943 London cards used two code systems, both of which differed from the codes used in Washington. The 1944 cards represented a wholesale revision of the forms and most of the medical and nonmedical codes. Sampling had been introduced in 1943 for selected types of cases, so that provision had to be made for making the separate categories of cases comparable before combining them, and the sampling pattern was slightly different for each year. The analysis of the data was aimed at following the admissions in a given year to their eventual disposition, which may have been several years later, involving different card forms, different codes, and different sampling coverages.
With the cessation of hostilities, the tabulating requirements for supply functions dropped sharply, and the machine records unit was transferred to the Medical Statistics Division. About half the equipment was canceled despite the continuing workload generated by the wartime strength of the Army.
The postwar years of 1946-49 brought repeated pressures for the reduction of machine rentals and personnel ceilings. Key civilian positions to replace those of technical sergeants were classified according to standards which appeared to be based largely on the number of persons supervised and the number of subject matter fields involved. The resulting grade structure was so low that it was difficult to hire competent personnel, and those who were able to fill the jobs satisfactorily soon found employment elsewhere at a higher grade. It was not until the middle 1950`s that the job classifications finally reached a level more nearly commensurate with the technical requirements of the positions.
In 1950, the ESM (Electronic Statistical Machine (IBM 101)) was added to the installation. This machine is primarily a sorting and counting machine similar to the original Hollerith
169
selective counter. It has a capacity for 60 independent card counts and 2 totals of additive information such as days lost per case. There are 60 "recode selectors" which derive their name from their ability to group many separate codes into a single category for sorting or counting purposes. Thus, the cards sorting to pocket 2 or counting in number 57, for example, can represent the combination of several "station" codes comprising a given theater or may represent a certain combination of codes for rank, sex, race, length of service, diagnosis, and type of disposition. By wiring the machine to set up only the acceptable combinations of codes, it is used to edit the cards for consistency of punching in related terms. The machine was found especially useful in setting up initial controls on a variety of categories to which later tabulations could be checked. Its flexibility in sorting facilitated making distributions involving combinations of codes or in selecting a specified category of cases for special study. The capacity for counting 60 separate items of information for a single sort group permitted the cross tabulation of one item with several other items, simultaneously. A second ESM and a summary punch attachment were on rental for a short time, but these were discontinued because it proved more feasible to make the large-scale summarizations on the alphabetical tabulators.
The tabulating procedures varied from year to year in terms of the card forms and codes used, the sampling coverage, and the particular tabular presentations to be made, but there was a broad pattern which was characteristic of each of the years since 1941. The description which follows generalizes the current procedures, with an indication of the principal variations which may be of interest.
The first stage in the tabulating process is a consistency check of the cards. Immediately after the punching and verifying operations, while the transcription slips and the punched cards are still associated in packs, the cards are passed through the ESM. The number of cards for diagnoses, representing the principal cause of admission for each of the cases, is checked to the number of transcription slips from which they were punched. The total number of cards for all diagnoses is checked to a control total generated from the first card for each case. As a by?product of the same pass of cards through the ESM, cards are rejected if they have unacceptable codes in specific items or in
170
combinations of items. These cards are compared with the transcription slips, and if the inconsistency is not due to a punching error the transcription slip is returned to the coding section for correction. Before the ESM was available, consistency checks were made by sorting successively on each of the major items, which was found feasible only for large accumulations of cards and was cumbersome for checks involving combinations of items.
Summary tabulations of individual medical records are made on the basis of a calendar year. Since the records are normally submitted upon disposition of the patient, they do not account for all cases treated in a calendar year until records for the cases remaining under care on 31 December have been received. Many of these cases are of short duration, and final dispositions are received during January and February of the succeeding year. Punchcards for these cases are prepared, not only for the calendar year of disposition but for the preceding year as well. At the end of the February report month, a `remaining card` is submitted for each case which was in a medical treatment installation on the preceding 31 December and which did not have a final disposition in January or February. When these remaining cards have been punched, the tabulating file for the preceding year can be considered complete. It then consists of all cases treated in a calendar year, including CRO (carded for record only) cases as well as cases excused from duty, cases admitted in the specified calendar year plus cases carried over from the previous year, and cases having a final disposition in the specified year plus the cases carried over from that year to the subsequent year.
The punched cards are sorted into diagnosis sequence and are then put through the ESM to produce a report establishing control totals. The gross number of cards for all diagnoses combined is checked for agreement with the accumulation of counts from the consistency check, immediately after the punching and verifying operations. For each diagnosis, the ESM prints the number of admissions, the total incidence, the number of CRO cases, the census at the beginning and at the end of the year, the frequency of deaths and separations, and other significant counts. Some of these categories are further subdivided by year of admission, year of disposition, or sampling ratio.
171
As the various counts are being made, the cards are sorted into categories by type of personnel and type of case. For example, records for active-duty Army personnel are segregated from records for all other personnel. Further subdivisions are made to select cases of special interest, such as deaths, disability separations, and battle casualties. These are reproduced and established as auxiliary files. For the years 1943-54, in which the nonbattle morbidity files were based on a 20-percent sample, the cards representing 100-percent coverage were sorted on the units digit of service number to select a comparable 20-percent sample to be included in the morbidity file.
Requests for tabulations of individual medical records are submitted by the Statistical Analysis Branch to the Machine Records Branch in the form of table outlines indicating the nature of the data to be shown in the stub and column headings of the table. Sources of data are specified in terms of the card files and item codes to be used. The Machine Records Branch prepares detailed operating instructions for securing the information requested, making the tabulation conform as closely as possible to the format of the final table. A copy of the procedure is sent to the Statistical Analysis Branch.
Tabulations for large-scale summarizations of individual medical records often start with the creation of a `basic summary card` which contains all of the breakdowns used in the complete series of tables. This is especially desirable when the summarization is to include more than one card form or requires extensive grouping of original codes. The detail cards have no space for additional codes, so when a new code structure is desired for an item, such as diagnosis, the cards are summarized on the basis of the original codes; the summary cards are then merged with a master deck of `translation cards` from which a `presentation code` is copied into the basic summary cards.
The expansion of sample data to the estimated 100-percent coverage has been done in several ways. The tabulation may show only 20-percent data, in which case the counts would be multiplied by 5 during the operation of posting the final table. However, it is not always necessary to first multiply the 20-percent data by 5, since averages and percentages would not be affected by using the 20-percent data directly, and rates can be computed by using the reciprocal of 20-percent of Army strength instead of using the actual strength. Another means
172
of expanding 20-percent sample data to the estimated 100-percent coverage is to reproduce the summary cards four times, so that the final tabulation will add to the 20-percent summary data five times. This is a convenient method when the summary cards for 20-percent data are to be combined with other summary cards representing 100 percent of another group of records.
If the basic summary cards contain no summations of additive fields of the detail cards, that is, if all of the data on the summary cards consist of counts of cases, the tabulator can be wired to `count by fives` by adding 5 instead of 1 each time a card passes through the machine. This method can be used even though the detail cards represent more than one sampling coverage. For example, for 1955 cards the tabulator adds 1 when the detail card is a part of the 100-percent sample, adds 2 if the detail card is part of the 50-percent sample, and adds 20 if the detail card is part of the 5-percent sample.
The basic summary cards are tabulated to produce the data from which final tables are posted. When the basic cards are used for a series of tables, it is sometimes expedient to cut a `super summary card` aimed at a particular table. This procedure reduces the volume of cards to be tabulated in the various degrees of consolidation required for table posting and supplemented reference tabulations.
The tabulations are sent to the Statistical Analysis Branch to be reviewed for agreement with control totals and checked for internal consistency within the tabulation. If the tabulations require no additional consolidations, the necessary rates, averages, or percentages are written directly onto the machine run. Data for the final table are posted to worksheets which provide for any additional consolidations and rate computations which may be necessary and show all headings and footnotes to be typed. These worksheets are checked for arithmetic consistency and reviewed for reasonableness before they are typed and sent to the printer. Further tabulation procedures are expected to mechanize the computations of rates and other statistical measures. The goal is to produce tabulated reports which can be used as final copy for tables to be reproduced by offset printing.
