Assessment of drug involvement: applications to a sample of physicians in training

The investigators examined survey data of lifetime and recent drug use in national samples of 2036 senior medical students and 1772 resident physicians to test whether patterns of lifetime drug use could be characterized adequately by a single underlying dimension of 'drug involvement'. The data analysis was based on a two parameter normal item response theory (IRT) model using the marginal maximum likelihood estimation method. The results showed that a single latent dimension of 'drug involvement' characterized individual drug use differences on the following measures: substances ever used, substances used in the previous year or previous month, and the sequential order of first use for each substance. The dimension was equivalent for students and resident physicians, and for both genders. Those who professed 'no religion' tended to be more drug involved. Physicians-in-training with a higher drug involvement score based on lifetime use were more likely: (a) to have used higher-ranking drugs (such as LSD and prescription opiates) in the past year; and (b) to have used a greater number of different drugs during the past month. Subjects first began to use each of the substances in a relatively invariant sequence corresponding to that predicted by the model. The implications of this model for evaluating the drug use histories of physicians-in-training, for identifying subgroups at greater risk for continued drug involvement after the beginning of medical training, and for further psychological, biological, and sociocultural research on the nature of 'drug involvement' are discussed.