It was not possible to cover all areas of research on clinical judgment and mechanical prediction in this chapter. Most notably, little was said about the validity of judgments made by mental health professionals (e.g., the reliability of diagnoses, the validity of descriptions of personality). An entire book on these topics has been written (Garb, 1998). However, conclusions from key areas of research were described. First, many automated assessment programs for interpreting psychological test results are not validated. Second, although there are reasons to believe that statistical prediction rules will transform psychological assessment, present-day rules are of limited value. Finally, the value of training in psychology and other mental health fields is supported, but research illustrates the difficulty of learning from clinical experience. These last results highlight the importance of continuing education, although continuing education may be of limited value unless it capitalizes on the findings of empirical research.
It is likely that clinical experience is valuable under certain circumstances. Experienced mental health professionals may be more adept at structuring judgment tasks (Brammer, 2002). In virtually of the studies that have been done, the tasks were already structured for clinicians: They were told what judgments to make and they were given information. However, in clinical practice, supervision can be helpful because questions are raised about what judgments and decisions need to be made (Do you think she is suicidal? Has the client ever had a manic episode?). Similarly, supervision can be helpful because supervisors provide direction on what information should be collected. Just the same, although experience may be helpful under certain circumstances, it does not seem to be useful for helping clinicians evaluate the validity of an assessment instrument. Nor does it seem to help clinicians make more valid judgments than graduate students when those judgments are made for a structured task.
A number of recommendations can be made for improving the way that judgments and decisions are made. The recommendations are made for both practicing clinicians and research investigators. First, mental health professionals should not use automated assessment programs to interpret test results unless they are appropriately validated. Second, as discussed earlier, new methods for building and validating statistical prediction rules need to be utilized. Data need to be collected for judgment tasks that have not yet been studied. Also, new analyses, including neural network models and multivariate taxometric analyses, should be used to build statistical rules (Marshall & English, 2000; Price, Spitznagel, Downey, Meyer, & Risk, 2000; N. G. Waller & Meehl, 1998). Third, mental health professionals need to become familiar with the research literature on clinical judgment. By becoming familiar with the results of studies on the validity of judgments made by mental health professionals, they can avoid making judgments for tasks that are surprisingly difficult and for which they are unlikely to be accurate. Fourth, clinicians should rely more on their notes and less on their memories. Fifth, to decrease confirmatory bias, clinicians should consider alternative hypotheses when making judgments and decisions. Sixth, when deciding whether to use an assessment instrument or treatment method, clinicians should weigh empirical findings more heavily than clinical experiences. That is, they should not use an assessment instrument or treatment method simply because it seems to work. In conclusion, to improve clinical practice dramatically, powerful statistical prediction rules need to be constructed and clinicians need to place less emphasis on their clinical experiences and greater emphasis on scientific findings.
Was this article helpful?