- Email: [email protected]
TOWARDS THE DEVELOPMENT OF ELECTRONIC DATA-PROCESSING SYSTEMS FOR MEDICAL RECORDS
1176
Special
Articles
limited information, such as postmortem findings (Smith and Melton 1964, Bahn et al. 1965) and laboratory data (Jurgens and Rosevere 1964, Smith and Melton 1963). Baird and Garfunkel (1965) successfully recorded clinical summaries and laboratory data, compiled chronologically, but the programme had to be discontinued because of in-put errors, cost, and lack of data during machine processing. We describe here a pilot study of a method to automate the keeping of medical records and discuss some of the limitations and difficulties of the method.
TOWARDS THE DEVELOPMENT OF
ELECTRONIC DATA-PROCESSING SYSTEMS FOR MEDICAL RECORDS A. E. BENNETT M.B. Lond., D.I.H. LECTURER
W. W. HOLLAND M.D., B.Sc. Lond.
BACKGROUND TO PRESENT STUDY
The initial impetus was a growing dissatisfaction with the method of keeping case-notes. This had resulted in individual medical and surgical departments appointing special staff to maintain their own records, thus increasing the already considerable number of clerical staff engaged in this work. Also, the diagnostic index is at present maintained only for inpatients, which makes identification of cases for research very difficult. A preliminary study of the handling and content of inpatient records (Holland et al. 1964) indicated three main areas in the record system in which a computer could be useful. Firstly, it could reduce the time between discharge of an inpatient and the final filing of the notes. At present this takes more than three weeks, and the notes are handled nine times. Secondly, it could readily provide medical and descriptive data on a patient in the form of a print-out. Thirdly, it could automate the patient diagnostic index and supply accumulated data for medical and administrative purposes.
READER
DEPARTMENT OF CLINICAL EPIDEMIOLOGY AND SOCIAL ST.
THOMAS’S
HOSPITAL MEDICAL SCHOOL,
MEDICINE, LONDON, S.E.1
ELECTRONIC computers have often been acclaimed as the solution to the problems of medical record keeping. Such a view is too naive, for there are still many difficulties to be overcome. Much work has already been done in this area, especially in America, and the main objective has usually been to put into a computer-translatable form as much of the medical record as possible (Korein et al. 1963, Schental et al. 1960, 1961, 1963). ’But the amount to be processed for each patient is vast, and some workers have used only
LIMITING
CONCEPTS
One of the difficulties in devising a method of data processing for medical records is that the case-notes are required - not only for patient care but also for research and administration, and the requirements are not necessarily the same in all instances. Another problem is that there is considerable inter-observer and intra-observer variation in taking clinical histories and assessing physical signs. - Cochrane et al. (1951) found that four _
_ experienced physicians, interviewing
com-
parable populations of coalminers in
a
pneumoconiosis
survey, showed considerable variation in history-taking. Four symptoms were studied, and in threecough, sputum, and pain in the chestthere was a significant difference in the : answers recorded by the observers. Fairbairn et al. (1959) found a similar discrepancy in questionnaires for chronic bronchitis. In a study of the observation of respiratory physical signs, the results of nine observers fell midway between chance and total agreement (Smyllie et al.
Fig. I-General-medical
summary
sheet,
to be
completed for
all
patients.
1965). Blood-pressure measurement is also subject to observer bias (Holland 1963, Rose et al. 1964). The recording of all possible information assumes that all information is equally valuable. This appears to be untrue. Further, not all the information is needed for research or administrative We decided, therefore, to purposes. concentrate on the development of patientsummaries. For medical care the total record must be retained, and duplication should be expected and accepted in the early stages of developing any new system.
1177
Fig. 2-Flow
chart of admission procedure.
Only uniform data of practical use should be included in the summary, and the recording of such data should remain the responsibility of the medical staff. Any forms or procedures should be simple enough to be completed by a doctor without special training. The same data should be recorded for both inpatients and outpatients. METHOD
Information
on
each
patient was coded on special forms,
Fig. 4-Example of a one-page computer print-out of blood-pressure, and of weight and blood-pressure.
data using
an
Fig. 3-Flow chart of discharge procedure.
which were fed into a ’Lector ’ photoelectric reading device. The recorded data were then fed directly into the computer. This method was used in St. Thomas’s Hospital for six months (Holland 1965). Two firms cooperated in the experiment-one medical and one surgical. One Lector form, the general-medical
XTAB programme.
It illustrates the assessment of
pulse-rate and
Plot 1 is the relation of pulse-rate (Y-axis) to systolic blood-pressure (X-axis) in 623 men aged 40-59 years. Plots 2 and 3 are the relations, with the 623 men separated into age-groups 40-49 years (plot 2) and 50-59 years (plot 3). Plot 4 is the relation of weight in kilograms (Y-axis) to systolic blood-pressure (X-axis) in 620 men aged 40-59 years. Printed below each plot are the regression equation, correlation coefficient, and standard deviations. The points P on the plots mark the regression line. The asterisks are single observations, A represents 10 observations, B represents 11 observations, and so on. same
1178
sheet (fig. 1), was used for all patients. This contained the patient’s identifying details and some general management, diagnostic, and operative details. Space was allotted for recording penicillin allergy, past immunisations, and certain chronic diseases. In addition to the generalmedical sheet, each firm was allowed to include specialised forms containing information of particular interest. Thus the surgical firm, which was concerned mainly with genitourinary surgery, produced a form for cystoscopy findings and another for details of special investigations. The department of obstetrics and gynxcology also produced forms recording information on each delivery. Since the Lector deals only with numerical information, there had to be another method for collecting the basic uncodable details of the patient, such as name, address, general practitioner, &c. These could be collected at a teleprinter station, either at the time of a patient’s recom-
mendation for admission or at the actual admission. For the experimental period this information was recorded on admission forms, then punched into paper tape. Flow charts of the proposed system are shown in figs. 2 and 3. The forms, procedures, and methods devised for the preparation and handling of standard case-summaries proved workable. The summary forms were easily completed by the house-officers. The ability of doctors to use the International Statistical Code for Diagnostic and Operation Coding was tested and found to be accurate. Any disagreements found in comparison with the coding clerk were minor and most often arose when the doctor had a more detailed knowledge of the patient and therefore coded more accurately.
acquisition of new skills; whereby the volume of paper work involved in coding, transmitting, and storing data will be lessened. This would involve the use of a number of input stations with visual-display units and perhaps teleprinters. The information fed in could be seen visually and the computer would be able to question or reject incomplete data and indicate what additional information is required. Few programmes have yet been written for dealing with this type of data, and fewer still for the analysis. The BIMD series developed by the University of California, Los Angeles, for use on I.B.M. 7090 and 7094 machines is already generally available; but of perhaps greater use to medical workers are the programmes developed by Massey (1964), which enable one to see the nature of association between two or more factors, as well as provide a numerical and statistical measure. An example of a print-out of data is given in fig. 4. This sort of analysis is essential, and these programmes allow the data to be analysed easily and quickly. Few, if any, of the computer firms are prepared at the moment to put their resources into developing the necessary aids for the analysis of medical data, and it is difficult for untrained medical personnel to work alone. We are grateful for the help given by Mr. A. J. Goldman in these pilot trials; to the registrars and housemen of St. Thomas’s Hospital for the additional work they have undertaken; and to Prof. F. Massey, of the University of California, Los Angeles, for his help and use of his XTAB programmes. The studies were supported partly by the endowment fund of St. Thomas’s Hospital and partly by a grant-inaid from the Ministry of Health. We are also indebted to I.C.T., Ltd., for help with some of the development, and to the I.B.M. Endowed Research Time Scheme for analyses with XTAB and BIMD programmes.
DIFFICULTIES ENCOUNTERED
REFERENCES
encountered was the difficulty problems of interesting medical personnel in the procedures. Unfortunately the lessons propounded by Cochrane et al. in 1951 have not yet permeated the main-stream of medical thought. Most physicians still believe that what is written is the truth and must therefore be recorded. It is difficult to persuade them that it is necessary to define exactly what information is of value. The direct practical applications of such a system have been indicated. The potential for medical research is considerable. Technically there is no problem. But in order to justify the use of a computerised system it is essential that what is recorded must be used. Recording all information for possible use in years to come is unrealistic : looking for correlations among twenty recorded items involves very many possible combinations. There are 190 possible combinations for any 2 factors, 1140 for 3 factors, 4845 for 4 factors, and so on. It would take many years to sort out which correlations are real and which are false, despite statistical tests of significance. We found it difficult to persuade the house-officers to complete the records, for this added to their already heavy work-load. The introduction of a grade of clerk to record what the doctor directs does not really solve the problem. The doctor must have time to tell him what to write down, and it might be just as easy for him to write it down himself in some suitable form.
Bahn, R. C., Schmit, R. W., Young, G. G. (1965) Proc. Staff Meet. Mayo Clin. 39, 835. Baird, H. W., Garfunkel, J. M. (1965) New Engl. J. Med. 272, 1211. Cochrane, A. L., Chapman, P. J., Oldham, P. D. (1951) Lancet, i, 1007. Fairbairn, A. S., Wood, C. H., Fletcher, C. M. (1959) Br. J. prev. soc. Med. 13, 175. Holland, W. W. (1963) in Epidemiology (edited by J. Pemberton); p. 271.
One of the
we
FUTURE DEVELOPMENTS
With the advent of large random-access storage units and time-sharing we can envisage a system whereby the insertion of data can be effected simply and without the
London.
(1965) in Mathematics and Computer Science in Biology and Medicine. London. De Bono, E., Goldman, A. J. (1964) Lancet, i, 819. Jurgens, J. L., Rosevere, J. W. (1964) Proc. Staff Meet. Mayo Clin. 39, 818. Korein, J., Woodbury, M. A., Goodgold, A. L. (1963) J. Am. med. Ass. 186, 132. Massey, F. (1964) Personal communication. Rose, G. A., Holland, W. W., Crowley, E. A. (1964) Lancet, i, 296. Schental, J. E., Sweeney, J. W., Nettleton, W. (1960) J. Am. med. Ass. 173, 6. —
—
(1961) ibid. 178, 267. (1963) ibid. 186, 101. Smith, J. C., Melton, J. (1963) Meth. Inform. Med. 2, 85. (1964) J. Am. med. Ass. 188, 958. Smyllie, H. C., Blendis, L. M., Armitage, P. (1965) Lancet, ii, —
—
—
—
—
—
-
-
412.
"... The idea that very expensive experimental resources be shared at some central institute or laboratories is fully accepted both on a national and international basis, and in the most advanced form of expensive physics we are approaching the time when co-operation on a world scale will be achieved for the construction and working of gigantic apparatus. But perhaps we should be taking a sharper look at whether more could not be accomplished by greater national and international integration in smaller, though still expensive, scientific endeavours. As far as scientific discoveries are concerned it should not matter whether they are made in Oshkosh or Vladivostok.... Instead of spreading our national resources too thinly between too many institutions we should concentrate on ’centres of excellence’ and from firm national bases international ’centres of excellence ’could be formed. We might then hear less of brain drains, and if some of our best people did go to work in other countries, their going would be balanced at least partially by a reverse flow."-Lord FLOREY, P.R.S., anniversary address to Royal Society, Nov. 30, 1965. must
Special
Articles
limited information, such as postmortem findings (Smith and Melton 1964, Bahn et al. 1965) and laboratory data (Jurgens and Rosevere 1964, Smith and Melton 1963). Baird and Garfunkel (1965) successfully recorded clinical summaries and laboratory data, compiled chronologically, but the programme had to be discontinued because of in-put errors, cost, and lack of data during machine processing. We describe here a pilot study of a method to automate the keeping of medical records and discuss some of the limitations and difficulties of the method.
TOWARDS THE DEVELOPMENT OF
ELECTRONIC DATA-PROCESSING SYSTEMS FOR MEDICAL RECORDS A. E. BENNETT M.B. Lond., D.I.H. LECTURER
W. W. HOLLAND M.D., B.Sc. Lond.
BACKGROUND TO PRESENT STUDY
The initial impetus was a growing dissatisfaction with the method of keeping case-notes. This had resulted in individual medical and surgical departments appointing special staff to maintain their own records, thus increasing the already considerable number of clerical staff engaged in this work. Also, the diagnostic index is at present maintained only for inpatients, which makes identification of cases for research very difficult. A preliminary study of the handling and content of inpatient records (Holland et al. 1964) indicated three main areas in the record system in which a computer could be useful. Firstly, it could reduce the time between discharge of an inpatient and the final filing of the notes. At present this takes more than three weeks, and the notes are handled nine times. Secondly, it could readily provide medical and descriptive data on a patient in the form of a print-out. Thirdly, it could automate the patient diagnostic index and supply accumulated data for medical and administrative purposes.
READER
DEPARTMENT OF CLINICAL EPIDEMIOLOGY AND SOCIAL ST.
THOMAS’S
HOSPITAL MEDICAL SCHOOL,
MEDICINE, LONDON, S.E.1
ELECTRONIC computers have often been acclaimed as the solution to the problems of medical record keeping. Such a view is too naive, for there are still many difficulties to be overcome. Much work has already been done in this area, especially in America, and the main objective has usually been to put into a computer-translatable form as much of the medical record as possible (Korein et al. 1963, Schental et al. 1960, 1961, 1963). ’But the amount to be processed for each patient is vast, and some workers have used only
LIMITING
CONCEPTS
One of the difficulties in devising a method of data processing for medical records is that the case-notes are required - not only for patient care but also for research and administration, and the requirements are not necessarily the same in all instances. Another problem is that there is considerable inter-observer and intra-observer variation in taking clinical histories and assessing physical signs. - Cochrane et al. (1951) found that four _
_ experienced physicians, interviewing
com-
parable populations of coalminers in
a
pneumoconiosis
survey, showed considerable variation in history-taking. Four symptoms were studied, and in threecough, sputum, and pain in the chestthere was a significant difference in the : answers recorded by the observers. Fairbairn et al. (1959) found a similar discrepancy in questionnaires for chronic bronchitis. In a study of the observation of respiratory physical signs, the results of nine observers fell midway between chance and total agreement (Smyllie et al.
Fig. I-General-medical
summary
sheet,
to be
completed for
all
patients.
1965). Blood-pressure measurement is also subject to observer bias (Holland 1963, Rose et al. 1964). The recording of all possible information assumes that all information is equally valuable. This appears to be untrue. Further, not all the information is needed for research or administrative We decided, therefore, to purposes. concentrate on the development of patientsummaries. For medical care the total record must be retained, and duplication should be expected and accepted in the early stages of developing any new system.
1177
Fig. 2-Flow
chart of admission procedure.
Only uniform data of practical use should be included in the summary, and the recording of such data should remain the responsibility of the medical staff. Any forms or procedures should be simple enough to be completed by a doctor without special training. The same data should be recorded for both inpatients and outpatients. METHOD
Information
on
each
patient was coded on special forms,
Fig. 4-Example of a one-page computer print-out of blood-pressure, and of weight and blood-pressure.
data using
an
Fig. 3-Flow chart of discharge procedure.
which were fed into a ’Lector ’ photoelectric reading device. The recorded data were then fed directly into the computer. This method was used in St. Thomas’s Hospital for six months (Holland 1965). Two firms cooperated in the experiment-one medical and one surgical. One Lector form, the general-medical
XTAB programme.
It illustrates the assessment of
pulse-rate and
Plot 1 is the relation of pulse-rate (Y-axis) to systolic blood-pressure (X-axis) in 623 men aged 40-59 years. Plots 2 and 3 are the relations, with the 623 men separated into age-groups 40-49 years (plot 2) and 50-59 years (plot 3). Plot 4 is the relation of weight in kilograms (Y-axis) to systolic blood-pressure (X-axis) in 620 men aged 40-59 years. Printed below each plot are the regression equation, correlation coefficient, and standard deviations. The points P on the plots mark the regression line. The asterisks are single observations, A represents 10 observations, B represents 11 observations, and so on. same
1178
sheet (fig. 1), was used for all patients. This contained the patient’s identifying details and some general management, diagnostic, and operative details. Space was allotted for recording penicillin allergy, past immunisations, and certain chronic diseases. In addition to the generalmedical sheet, each firm was allowed to include specialised forms containing information of particular interest. Thus the surgical firm, which was concerned mainly with genitourinary surgery, produced a form for cystoscopy findings and another for details of special investigations. The department of obstetrics and gynxcology also produced forms recording information on each delivery. Since the Lector deals only with numerical information, there had to be another method for collecting the basic uncodable details of the patient, such as name, address, general practitioner, &c. These could be collected at a teleprinter station, either at the time of a patient’s recom-
mendation for admission or at the actual admission. For the experimental period this information was recorded on admission forms, then punched into paper tape. Flow charts of the proposed system are shown in figs. 2 and 3. The forms, procedures, and methods devised for the preparation and handling of standard case-summaries proved workable. The summary forms were easily completed by the house-officers. The ability of doctors to use the International Statistical Code for Diagnostic and Operation Coding was tested and found to be accurate. Any disagreements found in comparison with the coding clerk were minor and most often arose when the doctor had a more detailed knowledge of the patient and therefore coded more accurately.
acquisition of new skills; whereby the volume of paper work involved in coding, transmitting, and storing data will be lessened. This would involve the use of a number of input stations with visual-display units and perhaps teleprinters. The information fed in could be seen visually and the computer would be able to question or reject incomplete data and indicate what additional information is required. Few programmes have yet been written for dealing with this type of data, and fewer still for the analysis. The BIMD series developed by the University of California, Los Angeles, for use on I.B.M. 7090 and 7094 machines is already generally available; but of perhaps greater use to medical workers are the programmes developed by Massey (1964), which enable one to see the nature of association between two or more factors, as well as provide a numerical and statistical measure. An example of a print-out of data is given in fig. 4. This sort of analysis is essential, and these programmes allow the data to be analysed easily and quickly. Few, if any, of the computer firms are prepared at the moment to put their resources into developing the necessary aids for the analysis of medical data, and it is difficult for untrained medical personnel to work alone. We are grateful for the help given by Mr. A. J. Goldman in these pilot trials; to the registrars and housemen of St. Thomas’s Hospital for the additional work they have undertaken; and to Prof. F. Massey, of the University of California, Los Angeles, for his help and use of his XTAB programmes. The studies were supported partly by the endowment fund of St. Thomas’s Hospital and partly by a grant-inaid from the Ministry of Health. We are also indebted to I.C.T., Ltd., for help with some of the development, and to the I.B.M. Endowed Research Time Scheme for analyses with XTAB and BIMD programmes.
DIFFICULTIES ENCOUNTERED
REFERENCES
encountered was the difficulty problems of interesting medical personnel in the procedures. Unfortunately the lessons propounded by Cochrane et al. in 1951 have not yet permeated the main-stream of medical thought. Most physicians still believe that what is written is the truth and must therefore be recorded. It is difficult to persuade them that it is necessary to define exactly what information is of value. The direct practical applications of such a system have been indicated. The potential for medical research is considerable. Technically there is no problem. But in order to justify the use of a computerised system it is essential that what is recorded must be used. Recording all information for possible use in years to come is unrealistic : looking for correlations among twenty recorded items involves very many possible combinations. There are 190 possible combinations for any 2 factors, 1140 for 3 factors, 4845 for 4 factors, and so on. It would take many years to sort out which correlations are real and which are false, despite statistical tests of significance. We found it difficult to persuade the house-officers to complete the records, for this added to their already heavy work-load. The introduction of a grade of clerk to record what the doctor directs does not really solve the problem. The doctor must have time to tell him what to write down, and it might be just as easy for him to write it down himself in some suitable form.
Bahn, R. C., Schmit, R. W., Young, G. G. (1965) Proc. Staff Meet. Mayo Clin. 39, 835. Baird, H. W., Garfunkel, J. M. (1965) New Engl. J. Med. 272, 1211. Cochrane, A. L., Chapman, P. J., Oldham, P. D. (1951) Lancet, i, 1007. Fairbairn, A. S., Wood, C. H., Fletcher, C. M. (1959) Br. J. prev. soc. Med. 13, 175. Holland, W. W. (1963) in Epidemiology (edited by J. Pemberton); p. 271.
One of the
we
FUTURE DEVELOPMENTS
With the advent of large random-access storage units and time-sharing we can envisage a system whereby the insertion of data can be effected simply and without the
London.
(1965) in Mathematics and Computer Science in Biology and Medicine. London. De Bono, E., Goldman, A. J. (1964) Lancet, i, 819. Jurgens, J. L., Rosevere, J. W. (1964) Proc. Staff Meet. Mayo Clin. 39, 818. Korein, J., Woodbury, M. A., Goodgold, A. L. (1963) J. Am. med. Ass. 186, 132. Massey, F. (1964) Personal communication. Rose, G. A., Holland, W. W., Crowley, E. A. (1964) Lancet, i, 296. Schental, J. E., Sweeney, J. W., Nettleton, W. (1960) J. Am. med. Ass. 173, 6. —
—
(1961) ibid. 178, 267. (1963) ibid. 186, 101. Smith, J. C., Melton, J. (1963) Meth. Inform. Med. 2, 85. (1964) J. Am. med. Ass. 188, 958. Smyllie, H. C., Blendis, L. M., Armitage, P. (1965) Lancet, ii, —
—
—
—
—
—
-
-
412.
"... The idea that very expensive experimental resources be shared at some central institute or laboratories is fully accepted both on a national and international basis, and in the most advanced form of expensive physics we are approaching the time when co-operation on a world scale will be achieved for the construction and working of gigantic apparatus. But perhaps we should be taking a sharper look at whether more could not be accomplished by greater national and international integration in smaller, though still expensive, scientific endeavours. As far as scientific discoveries are concerned it should not matter whether they are made in Oshkosh or Vladivostok.... Instead of spreading our national resources too thinly between too many institutions we should concentrate on ’centres of excellence’ and from firm national bases international ’centres of excellence ’could be formed. We might then hear less of brain drains, and if some of our best people did go to work in other countries, their going would be balanced at least partially by a reverse flow."-Lord FLOREY, P.R.S., anniversary address to Royal Society, Nov. 30, 1965. must