The topography of relapse in clinical samples

In two clinical samples, alcohol consumption, other drug use, and tobacco use were measured at approximately 6 months following admission of individuals into treatment. Using only the alcohol consumption variables, cluster analyses with several different solutions consistently identified abstinent, moderate, and unremitted groups. With the addition of tobacco and other drug use, analyses identified a largely abstinent group, a drug use group that did not drink, a heavy drinking group that did not use other drugs, and a group using both alcohol and other drugs, indicating the need for broad definitions of relapse. All solutions distinguished clusters of tobacco users and nonusers in remission from alcohol and other drug use, and tobacco users and nonusers tended to be differentiated among those continuing to drink or use other drugs, suggesting that the presence or absence of tobacco use marks different outcome groups. Generally speaking, clustering methods using complete and average linkage as agglomeration measures with cosine similarity as a proximity measure produced the most consistent clusters and the most clinically interpretable results.