A methodological consideration in the construction of actuarial interpretation systems

Several actuarial interpretive systems have been constructed utilizing medical records as the source of criterion data. The present study compared adjective frequencies tabulated on a 100-item checklist from two samples of behaviorally disturbed children and adolescents employing different data collection procedures. Frequencies for the medical record sample (n = 79) were obtained by raters who reviewed case history materials and completed the checklist. The checklist sample (n = 79) was comprised of subjects matched on sex, age, and race and on whom the checklist was completed following a diagnostic evaluation. Significant differences in adjective frequencies were obtained on more than one-third of the items. When descriptors reflected syndrome-specific behaviors, minimal or no differences were found. However, consistent under-reporting of less diagnostically specific though clinically relevant characteristics was indicated in the medical record sample. This was attributed to the lack of systematic review of potential correlates. The implications for the development of actuarial systems, conduct of retrospective investigations, and construction of empirical classification systems based on medical record data were discussed.