Actuarial versus Clinical Prediction in Psychopathology

¹ Research Professor of Medical Psychology, University of Missouri, Columbia, Missouri

The accuracy of clinical and actuarial prediction has been examined in relation to a variety of psychopathological criteria. Studies have been cited in which actuarial and clinical predictions have been made from objective and projective psychological test data as well as on the basis of a variety of clinically judged attributes and characteristics of patients. Fourteen such studies have been cited, although over 50 have been reported; the others dealt with predicting such criteria as grades in school (Doleys and Renzaglia, 1963; Melton, 1952; Pierson, 1958; Rosen and Van Horn, 1961; Sarbin, 1943; Watley and Vance, 1964), success in specialized training (Bloom and Brundage, 1947; Bobbitt and Newman, 1944; Cliff, 1958; Conrad and Satter, 1947; Dunlap and Wantman, 1944; Gregory, 1956; Kelly and Fiske, 1951; Lepley and Hadley, 1947; Parrish, Klieger and Drucker, 1957; Truesdell and Bath, 1956), and parole violation (Borden, 1928; Burgess, 1941; Glaser, 1954; Glaser and Hangren, 1958; Schneider, Lagrone, Glueck and Glueck, 1944). In 13 of the 14 studies the accuracy of the actuarial predictions was found to exceed or at least equal the accuracy of clinical predictions. The second of Lindzey's two studies was the only one in which clinicians' predictions were found to be more accurate than actuarial predictions. Only rarely does one encounter such consistent findings in any area of human behaviour.

This comparison of clinical and actuarial prediction, together with a number of related issues, justifies the advice offered years ago by Meehl; those of us concerned with making predictions in areas of psychopathology should re-examine our activities much more carefully than we have done so far. There seem to be several inescapable constraints on our predictive functions that may be more effectively accommodated if we attempt to answer the following questions.

1. What are the specific patient-characteristics or behaviours we wish to predict.

2. Which of these can be established as reliable criteria?

3. How accurately can these attributes be predicted clinically and actuarially?

4. What is the cost of predicting these attributes by either method?

Even though these questions and the relevant data are beginning to have an impact on the analysis of prediction in the United States, serious reservations about the prediction venture itself remain. One of the most important sources of dissatisfaction is the modest absolute level of predictive accuracy that can be achieved by any method. Continued open and critical examination of various methods of making predictions about psychopathological behaviour is certainly needed.

This article has been cited by other articles:

				S. Aegisdottir, M. J. White, P. M. Spengler, A. S. Maugherman, L. A. Anderson, R. S. Cook, C. N. Nichols, G. K. Lampropoulos, B. S. Walker, G. Cohen, et al. The Meta-Analysis of Clinical Judgment Project: Fifty-Six Years of Accumulated Research on Clinical Versus Statistical Prediction The Counseling Psychologist, May 1, 2006; 34(3): 341 - 382. [Abstract] [PDF]

				J. D Hamilton Do we under utilise actuarial judgement and decision analysis? Evid. Based Ment. Health, November 1, 2001; 4(4): 102 - 103. [Full Text] [PDF]