PRELIMINARY EVALUATION OF COMPETING SCREENING TESTS FOR MAJOR DEPRESSION AND SUBSTANCE ABUSE AND DEPENDENCE IN AN ALCOHOLIC POPULATION: AN APPLICATION OF RECEIVER OPERATING CHARACTERISTIC (ROC) METHODOLOGY
Abstract (summary)
The major objective of this research was to demonstrate ROC analysis as a more appropriate methodology which would serve as a standard in the evaluation of screening tests in psychiatric epidemiology. The specific application aimed to separately evaluate five competing depression (e.g., Center for Epidemiological Studies Depression Scale; Beck Depression Inventory) and two drug abuse Language of Addiction-Drug Version (LAD-D) Scale; Reasons for Drug Use (RDU) Scale tests in a random sample of an alcoholic population using various diagnostic standards e.g., Schedule for Affective Disorders and Schizophrenia (SADS); Diagnostic Interview Schedule (DIS) . Although measures of accuracy were independent of prevalence and the pre-established criterion values of the screening tests, they were found to be quite dependent on the spectrum of major depression and substance abuse and/or dependence (A/D) in the alcoholic sample. Broadening the clinical and co-morbid spectrum in the comparative groups, by including patients with a past history of major depression or substance A/D, resulted in a diminution in the accuracy of each screening test. It was suggested that the (a) heavy loading of clinical features and minor affective disorder in the comparative groups, (b) extensive co-morbidity in the entire alcoholic sample, (c) sensitivity to depression "reactions", (d) long-standing problems in nosological differentiation and (e) possible interrelationships between the essential and associated features of major depression and substance A/D were primarily responsible for the depression screening test's overall poor performance. While the level 2 efficacy of the depression screening tests were indistinguishable, the drug abuse tests were found to offer superior discriminability in differentially detecting substance A/D. The RDU was particularly well suited to detect current SADS summary diagnoses of substance A/D and specific DIS diagnoses of Cannabis, Amphetamine, Barbiturate/Tranquilizer and Cocaine A/D, and the LAD-D demonstrated superior performance in detecting Opioid A/D. Neither drug abuse test appeared to have greater detectability for DIS summary diagnoses of substance A/D. The major implications of the study pointed to (a) the danger of comparing accuracy measures derived from "equivalent" samples or independent investigations and (b) depression and/or anxiety as dimensions of psychopathology which might transverse categorical descriptions in psychiatric nomenclatures.