Prediction of drug abuse recurrence: a Swedish National Study.

Abstract

Relapse from drug abuse (DA) is common, but has rarely been studied in general population samples using a wide range of objective predictors. Using nationwide registries, we ascertained 44 523 subjects first registered for DA between the ages of 15 and 40 in 1998 to 2004 and followed for 8 years. We predicted relapse in subjects defined as a second DA registration. We also predicted DA relapse in relative pairs concordant for DA but discordant for relapse. In multivariate regression analyses, the strongest predictors for relapse were prior criminal behavior, male sex, being on social welfare, low school achievement, prior alcoholism, and a high-risk father. A risk index trained from these analyses on random split-halves demonstrated a risk ratio of 1.11 [95% confidence intervals (CIs) 1.10-1.11] per decile and an ROC value of 0.70 (0.69-0.71). Co-relative analyses indicated that a modest proportion of this association was causal, with the remainder arising from familial confounders. A developmental structural equation model revealed a complex interviewing of risk pathways to DA with three key mediational hubs: low educational attainment, early age at first registration, and being on social welfare. In a general population sample, using objective registry information, DA relapse is substantially predictable. However, the identified risk factors may not be valid targets for interventions because many index familial risk and may not impact causally on probability of relapse. Risk for DA relapse may reflect an inter-weaving, over developmental time, of genetic-temperamental vulnerability, indices of externalizing behaviors and social factors reflecting deprivation.