Psychiatric Annals

The Development of Automated Personality Assessment in Medical Practice

Wendell M Swenson, PhD

Abstract

Ever since the pioneer investigations of Sarbin1 and Meehl,2 behavioral scientists have attacked the problem of the use of a mechanistic process of decision making that might improve or even replace the efforts of the skilled clinician in converting psychometric data into behavioral descriptions or predictions. The rapid development of high technology in the area of electronic data processing has made this former fantasy a reality. The motivation for such development has come primarily from the awareness, by psychologists, of the physician's need for clear-cut, understandable information regarding the psychologic character of his patient. The physician himself has long been aware of this need but the lack of time has made a complete and comprehensive psychologic examination impossible.

Because it has been well established that at least 50% of the patients in a physician's everyday practice (at least in a secondary or tertiary medical center environment) have significant psychologic problems, it has been abundantly clear that this aspect of medical practice can never be handled on a day-to-day basis by the resources of psychiatry or clinical psychology alone. Within a large medical center one is continually confronted with the question: Are there other means of obtaining and winnowing those volumes of factual data that the psychiatrist has to sift to select the meaningful salient cues, indicators, or suggestions of what is wrong with the patient? In other words, how can the physician get, in understandable language, specific information regarding at least the superficial mental status ol his patient and do so with a minimal expenditure of effort, time, and money.

The solution to this problem appeared to us to be a relatively simple one - one of a data-organizing and matching operation. In other words, can the experts' test be analyzed and mechanized into a relatively simple operation and subsequently, can a computer program be developed such that, if the answers to this mechanized test are fed into electronic hardware, the program will produce this much-needed psychologic information?

Such was the climate that we found ourselves in at the Mayo Clinic in 1960. In large-scale screening operations involving the selection of patients who need detailed psychiatric examination, a computerized process of both test scoring and interpreting seemed to be the obvious solution. In searching for a psychometric device that could select on a much better than chance basis those individuals needing further psychiatric examination, we found the MMPl to be a natural solution to our problem. The Medical Relations Division of International Business Machines, Inc.. in Rochester, Minnesota, was available to collaborate in just one such venture. A proposal to proceed in this venture was presented at the Second IBM Medical Symposium in the fall of 1960.' To our knowledge, this was the first formal consideration of the automation of structured personality test interpretation. In 1961 we began to use computer technology at the Mayo Clinic to make possible the routine psychometric assessment of medical patients by means of the automation of scoring and interpretation of the MMPI. At that time (here was virtually no literature bearing any information on this problem. Attempts had been made to evaluate MMPl profiles on a mechanical basis.4,5 However, with such mechanical interpretation the process was too lengthy for any type of large-scale use in a screening process.

To overcome these obstacles the standard format of the MMPI was modified to permit its assembly into a booklet of 23 standard IBM cards on which the 550 MMPI statements were printed. Figure 1 shows the original booklet with the general instructions on the booklet cover. Using an electrographic pencil, the patient was able to complete…

Ever since the pioneer investigations of Sarbin1 and Meehl,2 behavioral scientists have attacked the problem of the use of a mechanistic process of decision making that might improve or even replace the efforts of the skilled clinician in converting psychometric data into behavioral descriptions or predictions. The rapid development of high technology in the area of electronic data processing has made this former fantasy a reality. The motivation for such development has come primarily from the awareness, by psychologists, of the physician's need for clear-cut, understandable information regarding the psychologic character of his patient. The physician himself has long been aware of this need but the lack of time has made a complete and comprehensive psychologic examination impossible.

Because it has been well established that at least 50% of the patients in a physician's everyday practice (at least in a secondary or tertiary medical center environment) have significant psychologic problems, it has been abundantly clear that this aspect of medical practice can never be handled on a day-to-day basis by the resources of psychiatry or clinical psychology alone. Within a large medical center one is continually confronted with the question: Are there other means of obtaining and winnowing those volumes of factual data that the psychiatrist has to sift to select the meaningful salient cues, indicators, or suggestions of what is wrong with the patient? In other words, how can the physician get, in understandable language, specific information regarding at least the superficial mental status ol his patient and do so with a minimal expenditure of effort, time, and money.

The solution to this problem appeared to us to be a relatively simple one - one of a data-organizing and matching operation. In other words, can the experts' test be analyzed and mechanized into a relatively simple operation and subsequently, can a computer program be developed such that, if the answers to this mechanized test are fed into electronic hardware, the program will produce this much-needed psychologic information?

Such was the climate that we found ourselves in at the Mayo Clinic in 1960. In large-scale screening operations involving the selection of patients who need detailed psychiatric examination, a computerized process of both test scoring and interpreting seemed to be the obvious solution. In searching for a psychometric device that could select on a much better than chance basis those individuals needing further psychiatric examination, we found the MMPl to be a natural solution to our problem. The Medical Relations Division of International Business Machines, Inc.. in Rochester, Minnesota, was available to collaborate in just one such venture. A proposal to proceed in this venture was presented at the Second IBM Medical Symposium in the fall of 1960.' To our knowledge, this was the first formal consideration of the automation of structured personality test interpretation. In 1961 we began to use computer technology at the Mayo Clinic to make possible the routine psychometric assessment of medical patients by means of the automation of scoring and interpretation of the MMPI. At that time (here was virtually no literature bearing any information on this problem. Attempts had been made to evaluate MMPl profiles on a mechanical basis.4,5 However, with such mechanical interpretation the process was too lengthy for any type of large-scale use in a screening process.

To overcome these obstacles the standard format of the MMPI was modified to permit its assembly into a booklet of 23 standard IBM cards on which the 550 MMPI statements were printed. Figure 1 shows the original booklet with the general instructions on the booklet cover. Using an electrographic pencil, the patient was able to complete this test without the benefit of a desk or lapboard. as part of the initial medical examination. This usually was accomplished while the patient was waiting to be given a schedule of laboratory tests and other appointments. When the test booklet was completed, it was delivered to the computer center where certain identifying data were punched into the facecard and the deck of IBM cards was scored by the then current IBM 7040 system.

Figure 1. MMPI test booklet and general instructions. (From Swenson WM. Pearson JS: Psychiatry - Psychiatric screening. J Chronic Dis 1966; 19:497-507 By permission of Pergamon Press.)

Figure 1. MMPI test booklet and general instructions. (From Swenson WM. Pearson JS: Psychiatry - Psychiatric screening. J Chronic Dis 1966; 19:497-507 By permission of Pergamon Press.)

Table

TABLE 1DESCRIPTOR LIBRARY FOR DEPRESSION SCALE

TABLE 1

DESCRIPTOR LIBRARY FOR DEPRESSION SCALE

Because of the consistency with which certain statements appeared in psychologists' individual interpretations of the MMPI profile, it was relatively simple to develop a "statement library" to describe in nonpsychiatric terminology most of the basic combinations of scale elevations and relationships in the MMPI test profile. This library originally consisted of about 75 behavioral descriptive statements. For example. Table 1 lists the original live statements for the depression scale. We were well aware that it was impossible to simply describe the various scales as completely separate entities, so we found it necessary to make certain adjustments or "branching operations" to take into consideration various combinations of scale elevations. The IBM 7040 computer then selected a certain number of the specific behavioral statements from the computer's memory and printed them in order of significance of deviation. The original resulting printout is shown in Figure 2.

The format used carried no diagnostic labels or psychiatric jargon. Emotion-laden words such as "schizophrenia" or "psychopathic personality" were carefully avoided. This information was usually in the medical history approximately 12 hours after the patient had completed the test and it was available to the medical consultant before his final meeting with the patient. In our medical center, the test results were used not only as a screening device but also as an indication of the salient personality characteristics of each patient on a "thumbnail sketch" basis. If certain aspects of the profile were sufficiently deviant, the interpretation was prefaced by the statement "consider psychiatrie evaluation."

Because of this technique's potential for large-scale use. a research proposal was submitted to and approved by the National Institute of Mental Health. With the support of Research Grant MH-06947. the MMPI was given to all patients in the Sections of Internal Medicine at the Mayo Clinic for a period of approximately 3 years. From the accumulation of these data, a source book was developed summarizing the responses of 50,000 general medical patients over that period.0

The acceptance of the use of such an automated procedure by patients was indeed positive. The results of efforts to administer the MMPI to 1,559 patients from May 22 to Oct. 21. 1961, are shown in Table 2. Almost 90% of the patients completed the test with sufficient validity for interpretation. Only 0.2% overtly refused to cooperate. Because of their medical problems, some patients. 5% of that sample population, were unable to complete the test.

The physicians also reported significant enthusiasm for the availability of the kind of information we were able to give them regarding their patient's psychologicstructure. In September 1965 a questionnaire was submitted to all medical consultants who had used the MMPI for at least 4 months. Table 5 summarizes responses to one question in that questionnaire, fn general, between 69% and 73% of the consultants polled thought that lhe MMPI gave them information that was significant to the total medical care of their patients.

Several different forms of the "printout" have been developed at our institution over tbe past 2 decades of our experience. The current form of the printout (Figure 5) is a double sheet folded in half to fit the patient's clinic history. At the top of the upper half are the patient's name, dale, age, sex, and the clinic section to which the patient is registered (not shown in this example). Beneath that appears the profile code according to the standard Dahlstrom and Welsh7 system. This code gives the scale numbers in order, beginning with the most deviant scale and ending with the least deviant. The scale elevations on the profile itself are shown with asterisks so that the clinician can connect these elevations with a pencil to provide a profile. Both raw scores and T score conversions are shown beneath the profile. The interpretative statements appear on the bottom half of (he printout, generally in order of the most deviant to the least deviant as determined by scale elevation. At the bottom is a series of numbers for certain critical or particularly deviant items in the test. Item numbers from the booklet form of the test are used rather than the item itself to preserve confidentiality.

Figure 2. Form in which computer data are printed out. (From Swenson WM, Rome HP Pearson JS. et al: A totally automated psychological test: Experience in a medical center. JAMA 1965: 191:925-927. Copyright 1965. American Medical Association. By permission.)

Figure 2. Form in which computer data are printed out. (From Swenson WM, Rome HP Pearson JS. et al: A totally automated psychological test: Experience in a medical center. JAMA 1965: 191:925-927. Copyright 1965. American Medical Association. By permission.)

Table

TABLE 2RESULTS OF EFFORTS TO ADMINISTER MMPI TO 1,359 PATIENTS- MAY 22 TO OCTOBER 1, 1961

TABLE 2

RESULTS OF EFFORTS TO ADMINISTER MMPI TO 1,359 PATIENTS- MAY 22 TO OCTOBER 1, 1961

During the original research supported by the grant from the National Institute of Mental Health, the MMPI was administered free of charge. After that, it was given to patients on a fee-for-service basis. Our volume of use has increased modestly over the past two decades. In 1985 we were processing between 1 .200 and 1 .500 decklets per month. The use of the tesi has been on an individualized basis during the past several years as far as the specific departments or individual consultants are concerned. Some internists use the lest routinely and actually have their patients read the printout verbatim. At the other end of the continuum there are some physicians who see little or no benefit from the information given them and rarely, if ever, take advantage of the availability of it. With the development of the Mayo Regional Laboratories for use by physicians outside the Mayo Clinic complex, we have instituted a "mail order" procedure. Interested physicians can obtain the test decklets from us by mail; on request we process the tests and return interpretations on a 24-hour turn-around basis. This use of the test has increased to the level of approximately 250 to 500 per month.

Table

TABLE 3ATTITUDE OF IVIAYO CLINIC CONSULTANTS TOWARD GENERAL USE OF MMPI

TABLE 3

ATTITUDE OF IVIAYO CLINIC CONSULTANTS TOWARD GENERAL USE OF MMPI

Figure 3. MMPI report printout currently in use ai Mayo Clinic. (From Colligan RC. Osborne D. Swenson WM. et al: The MMPI A Contemporary Normative Study. New York. Praeger Publishers, 1983.)

Figure 3. MMPI report printout currently in use ai Mayo Clinic. (From Colligan RC. Osborne D. Swenson WM. et al: The MMPI A Contemporary Normative Study. New York. Praeger Publishers, 1983.)

Although we have identified ourselves as pioneers or originators of automated personality assessment, we have by no means remained the only investigators in this application of high-technology computers to personality assessment. The field has developed into a fairly competitive one as evidenced by the mushrooming literature - especially in the last decade. We continue to be alone in the area of providing what some of our colleagues identify as an "overly simplified" MMPI interpretation for physicians. Ours continues to be brief, to the point, and lacking in any kind of psychiatric jargon. The modifications we have made have not affected the length of the printout. It continues to be approximately of the same quantity as well as quality as that shown in Figure 3.

As described elsewhere in this issue, we have recently completed a restandardization of the MMPI that will have some impact on our automated printout. We are in the process of developing age-related profiles which will be available on an optional basis to those who wish to compare the responses of a specific patient to those of others in the same age group. We also have begun to prepare a printout that will report some additional scales that have been developed over the past 2 decades to the point of adding significant information in terms of the patient's salient personality characteristics.

The possibilities of "cookbook analysis" referred to by Meehl several decades ago have certainly become realistic.8 At this time the psychologist probably can best function as a technician in certain areas without the expenditure of a great amount of time, relegating most of the program detail to machine processes. This allows his time to become available for more detailed evaluation of more complex aspects of human behavior. The research spin-off of this venture has been tremendous. With the availability of thousands of MMPI profiles from patients with various types of diagnostic classifications, we have been able to identify a number of disease-specific profiles. These have been and will be reported elsewhere. There no longer is any reason for the general practitioner or the internist or the medical specialist to deal with the medical problems of his patient in the absence of information about the patient's functioning personality.

REFERENCES

1. Sarbin TR: Clinical psychology - Art or science? Psychomeirika 1941: 6:391.

2. Meehl IT-: Clinical Versus Statistical Prediction A Theoretical Analysis and d Review of the Evidence. Minneapolis. University of Minnesota Press, 1954.

3. Swenson WM: A preliminary investigation of the possibilities of application of computer devices to the scoring and interpretation of structured personality tests anil their use in a medical center. International Business Machines Corporation Medical Symposium Proceedings 1960; 2:401-415.

4 Marks PA, Sccman W: Actuarial Description of Abnormal Personality: An Atlas for Use with the MMPI. Baltimore. Williams & Wilkins, 1963.

5. Drake Lt, Getting KR: Ali MMPI Codebook for Counselors. Minneapolis. University of Minnesota Press. 1459.

6. Swenson WM. Pearson IS. Osborne D: An MMPI Source Hook: Haste Item, Scale, and Pattern Data on 50,000 Medical Patients. Minneapolis. University of Minnesota Press, 1975.

7 Dahlstrom WG. Welsh OS: An MMPI Handbook: A Guide to Use in Clinical Practice and Research. Minneapolis. University of Minnesota Press, 1960. pp 18-25.

8. Meehl PE: Wanted- A good cookbook. Am Psychol 1956: 1 1 :265-272.

TABLE 1

DESCRIPTOR LIBRARY FOR DEPRESSION SCALE

TABLE 2

RESULTS OF EFFORTS TO ADMINISTER MMPI TO 1,359 PATIENTS- MAY 22 TO OCTOBER 1, 1961

TABLE 3

ATTITUDE OF IVIAYO CLINIC CONSULTANTS TOWARD GENERAL USE OF MMPI

10.3928/0048-5713-19850901-08

Sign up to receive

Journal E-contents