Soc. S~i. & Med..
Vol. 12. pp. 45 to 47 Pergamon Press 1978. Printed in Great Britain.
ASSESSMENT OF PSYCHIATRIC ILLNESS SEVERITY BY FAMILY PHYSICIANS* PETER HESBACHER, K A R L RICKELS, ROBERT W . D O W N I N G a n d PAUL STEPANSKY'~
Private Practice Research Group, Department of Psychiatry, University of Pennsylvania, Philadelphia, Pennsylvania Abstract Independent psychiatric illness assessments of family practice patients are compared. Of the three ratings self-ratings, evaluation by the treating physician, and a clinical interview by a research psychiatrist patient and research psychiatrist ratings achieved the highest level of agreement (509'o of the variance). Of the three kinds of raters, treating physicians showed the greatest variability. Least agreement appeared for Practice Two, upper middle class college town people who were relatively high in interpersonal sensitivity complaints. Most agreement appeared for Practice Three, ghetto inner city people who were relatively high in somatic complaints. The rating performance of treating physicians is attributed to both rating instrument and rating context.
INTRODUCTION
Psychiatric illness assessment, particularly during the early stages of emotional illness, has largely become the province of primary care physicians [1-4]. Quite understandably, this development has generated efforts to determine the accuracy with which psychiatric illness is detected in private family practice [5, 6]. Recent studies evaluating psychiatric illness assessment by non-psychiatrists have typically relied upon ratings obtained from the treating physician and the patient [7, 8]. In the present study, the family physician's ability to judge severity of psychiatric illness is assessed through comparison of three independently obtained criteria: ratings by the family physician; self-ratings by the patient; and ratings made by a research psychiatrist during a structured psychiatric interview held at the time of the visit to the physician's office.
area [9]. All seven participating physicians had previously cooperated in research conducted by these investigators. The four practices in which a research psychiatrist also participated were chosen in such a way that respondents represented a wide range of social class for both blacks and whites. Fifty patients in each practice were to be seen by the research psychiatrist. Treating physicians typically sent suitable patients to the psychiatrist, sitting in another office, whenever he was free. However, when the family physician considered the patient to be a "case", he often asked the patient to wait until the psychiatrist was available. A total of 182 patients received a psychiatric interview. Characteristics of these patients and the four practices in which they were seen are given in Table 1.
Measures Three independent assessments of psychiatric illness were obtained. Before seeing the physician, the patient completed the 35-item version of the Hopkins Symptom Checklist (HSCL) [8]. On this measure, patients record their own estimates of symptom severity, present during the past week, on a 4-point scale ranging from "not at all" (1) to "extremely" (4). Responses are summed and divided by the number of answered items to generate a total score ranging from
METHOD
Population Present data were collected during a health survey of seven family practices in the Philadelphia (U.S.A.)
1 to4.
* Data were collected in the offices of private family practitioners in the Philadelphia area and analyzed through support from USPHS Grants MH-08957-8. The authors wish to express their gratitude to Drs. Clark, Rial, Rosenfeld, and Zamostein, members of our Private Practice Research Group who engaged in data collection; and to Mrs. Vlabovich, programmer, for her data handling activities. + Dr. Hesbacher is Assistant Professor of Sociology, Department of Psychiatry, University of Pennsylvania. Dr. Rickels is Professor of Psychiatry and Pharmacology, University of Pennsylvania, Director of the Psychopharmacology Research Unit. Dr. Downing is Associate Professor of Psychology, Department of Psychiatry, University of Pennsylvania. Mr. Stepansky is a Doctoral Candidate in History, Yale University, and Research Assistant, Psychopharmacology Research Unit.
Following a routine office visit with the patient, the treating physician completed an Overall Judgement of Psychopathology (OJP). On the OJP, the physician simply makes a global judgement of the severity of each patient's psychopathology, relative to other patients in his practice, using a 7-point scale ranging from "not present" (1) to "extremely severe" (7) [10]. While a variety of judgments guided clinical conclusions, physicians were asked to assess illness in terms of overall impairment during the past week including the present day. After completing his rating, the physician had the patient see the psychiatrist. The psychiatrist conducted a standard psychiatric evaluation, focusing on the content and severity of 45
46
PETER HESBACHER et al.
Table 1. Description of the practices Practice One (N = 42)
Practice Two (N = 42)
Practice Three (N = 47)
Practice Four (N = 51)
Suburban, White, Middle class
Surburban, White, Upper class
Inner city, Black, Lower class
Urban, White, Working class
99 65
98 70
1
65
99 65
73 37 Class III (lower end)
69 42 Class II (lower end)
44 46 Class V (middle)
7l 50 Class IV (middle)
Characteristics
Race: White °/o Sex: Female ~!/,~ Marital status: Married % Age: Median years Social class: Median Position
psychiatric disturbance, during which he completed the Psychiatric Interview (PI) [11]. The PI yields a subjective global severity rating, ranging from "normal" (1) to "severe" (6), with the dividing line establishing clinical significance falling between the ratings of"mild" (3) and "moderate" (4). The PI also provides an objective rating of severity, derived from a weighted combination of ten symptom groups (determined from the patient's reported experience during the previous week), and twelve additional abnormalities (manifested during the interview and rated after the patient had left), assessed on a scale ranging from "not present" (0) to "severe" (4). As each symptom group received a weight of 1 and each manifest abnor-
mality received a weight of 2, the total objective score may range from 0 to 136.* Data analysis
Analyses comparing the three independent assessments of psychiatric illness utilized the Pearson product moment correlation coefficient (r) [12]. As the Psychiatric Interview produced two ratings of total illness, four ratings were actually compared. Correlational analyses on the four ratings were performed for each practice individually and for all four practices combined. RESULTS AND DISCUSSION
* The total objective score of patients in the present study sample actually ranged from 0 to 53, and patients were divided into the following nine groups: 0-5 = 0; 6-11 = 1; 12-17 = 2; 18-23 = 3; 24-29 = 4; 30-35 = 5; 36-41 = 6; 42-47 = 7; and 48-53 = 8. Analyses treated this measure as having nine scale positions ranging from 0 t o 8.
Results are given in Table 2. Mean scores for the three rating instruments, as well as coefficients of correlation for each pair of ratings, are provided for the four practices individually and combined. As two psychiatrist ratings were employed, a total of five comparisons resulted.
Table 2. Psychiatric illness severity: inter-rater reliability Mean ratings
Practice One Two Three Four Combined
N
Patient HSCL total (1-4)
42 42 47 51 182
1.45 1.33 1.71 1.52 1.51
Psychiatrist PI PI Subjective Objective (1-6) (0-8) 1.95 1.76 2.89 2.21 2.21
,
1.33 0.83 2.68 1.61 1.61
Physician OJP (1 7) 2.07 2.31 3.49 2.88 2.68
Correlations
Practice One Two Three Four Combined
Patient Psychiatrist Subjective Objective 0.68 0.57 0.65 0.70 0.70
0.74 0.51 0.75 0.74 0.77
PatientPhysician 0.45 0.28 0.68 0.43 0.54
Physician Psychiatrist Subjective Objective 0.60 0.47 0.70 0.61 0.61
0.65 0.42 0.65 0.62 0.61
Assessment of psychiatric illness severity by family physicians Looking first at data for the four practices combined, correlations for all five sets of comparisons were highly significant (p < 0.001) Specifically, the correlations obtained for the patient-psychiatrist subjective and patient-psychiatrist objective comparisons are 0.70 and 0.77, respectively (with the amount of variance accounted for being 49 and 590, respectively). The two physician-psychiatrist comparisons both yielded correlations of 0.61 (with 37% of the variance being accounted for); and the patient-physician correlation was 0.54 (accounting for 29% of the variance). Thus, greatest agreement about psychiatric illness severity is seen between patient and psychiatrist, substantial agreement is shown for physician and psychiatrist, and least agreement is indicated between patient and physician. Considering the different ratings instruments used, and the different contexts in which they were completed, the level of inter-rater agreement emerging from these analyses must be seen as impressive. The correlations obtained are clearly of sufficient magnitude to suggest that the family physicians assessed here are capable of adequately judging the severity of emotional illness within their patients. Even so, it seems important to point out several factors which help to explain the finding of less agreement between physician and patient than between psychiatrist and patient. The treating physician was at a disadvantage relative to the research psychiatrist in terms of both rating instrument and rating context. Thus, the physician-rated OJP calls for a global judgment without specifying the criteria for that judgment, and has been shown in previous research to have rather low interrater reliability [13]. The psychiatrist-rated PI is a structured evaluation form, providing total scores based on specific symptom ratings, and has been shown to have high inter-rater reliability [14]. Also, the physician performed his global rating following a routine office visit with most visits being relatively brief and not focused exclusively upon emotional symptomatology. In contrast, the psychiatrist completed his evaluation form during a structured, standard psychiatric interview. Particularly in view of these factors, the level of agreement reached between physician and psychiatrist appears quite good. Within this context, it should be mentioned that higher levels of inter-rater agreement have been obtained when family physicians and psychiatrists have used the same evaluation forms to rate patients under the same circumstances [15]. For example, intraclass correlations were computed for groups of family physicians and psychiatrists who performed ratings on the Physician Questionnaire (PQ) [10] and the Hamilton Anxiety Scale (HAS) [16] after viewing a video-taped interview of an anxious patient. The correlations obtained were 0.79 for a group of five family physicians, as compared to 0.77 for a group of 9 psychiatrists, on the PQ, and 0.68 as compared to 0.76 on the HAS. All four correlations were significant at the 0.0001 level. Turning now to data for the four individual practices, Table 2 shows the least substantial correlations
47
for all five comparisons within Practice Two, and the highest correlations for patient-physician and physician psychiatrist comparisons in Practice Three, Interestingly, the mean ratings shown in Table 2 indicate Practice Two patients to be least symptomatic (except on the physician rating) and Practice Three patients to be most symptomatic (according to all three raters). Also, patients from Practice Two, located in a white college town, were found to be relatively high in interpersonal sensitivity complaints, while patients from Practice Three, located in a black ghetto community, were relatively high in somatic complaints. Present data do, then, permit the speculation that type and degree of neurotic symptomatology have an impact on the physician's ability to judge severity of psychiatric illness. REFERENCES
I. Hesbacher P. et al. Entry and exit from treatment: the treatment process from a patient perspective. Int. Pharmacopsychiat. 8, 70, 1973. 2. Locke B. Z. et al. Emotionally disturbed patients under care of private non-psychiatric physicians. In Psychiatric Epidemiology and Mental Health Planning (Edited by Monroe R. R., Klee G. D. and Brody E. B.). American Psychiatric Association Report No, 22, pp. 235-248, 1967. 3. Locke B. Z. and Gardner E. A. Psychiatric disorders among the patients of general practitioners and internists. Publ. Hlth Repts 84, 167, 1969. 4. Leopold R. L. et al. Emotional Disturbance Among Patients of Private Nonpsychiatric Physicians in an Urban Neighborhood. Unpublished Report. University of Pennsylvania, 1971. 5. Goldberg D. P. The Detection of Psychiatric Illness by Questionnaire. Oxford University Press, London, 1972. 6. Hesbacher P. et al. Psychiatric illness detection: a comparison of osteopaths and M.D.s in private family practice. Soc. Sci. & Med. 9, 461, 1975. 7. Whitehorn J. C. Guide to interviewing and clinical personality study. Archs Neurol. Psychiat. 52, 197~ 1944. 8. Derogatis L. R. et al. The Hopkins Symptom Checklist (HSCL): a self-report symptom inventory. Behav. Sci. 19, 1, 1974. 9. Hesbacher P. et al. Psychotropic drug use in family practice. Pharmakopsychiat. Neuropsychopharmacol. 9, 50, 1976. 10. Rickels K. and Howard K. The physician questionnaire: a useful tool in psychiatric drug research. Psychopharmacologia 17, 338, 1970. 11. Goldberg D. P. et al. A standardized psychiatric interview for use in community surveys. Br. J. preventive & soc. Med. 24, 18, 1970. 12. Ferguson G. A. Statistical Analysis in Psychology and Education. McGraw-Hill, New York, 1959. 13. Rickels K. et al. Differential reliability in rating psychopathology and global improvement. J. clin. Psychol. 26, 320, 1970. 14. Goldberg D. P. et al. A comparison of two psychiatric screening tests. Br. J. Psychiat. 129, 61, 1976. 15. Rickels K. et al. The Private Practice Research Group (PPRG): a working model from psychopharmacology for clinical research in family practice. J. clin. Pharmacol. 17, 541, 1977. 16. Hamilton M. The assessment of anxiety states by rating. Br. J. med. Psychol. 32, 50, 1959.