Nonresponse and selection bias in treatment follow-up studies

Large-scale studies of addiction treatment employ two stages: select provid ers, then select patients to follow-up. Nonresponse due to noncooperation o f providers and problems of locating and recruiting patients may bias the r esults. We review selection and attrition biases in previous work and in fo ur major United States treatment studies in the 1990s: DATOS (N = 10, 100 c lients, 96 units). NTIES (N = 6,593/71), SROS (N = 3,047/99), and CALDATA ( N = 3,045/86). We develop a standard approach. break down sampling and attr ition rates, and discuss differences in client, program, and methodology fa ctors. We conclude with some methodological recommendations for future foll ow-up studies of addiction treatment. [Translations are provided in the Int ernational Abstracts Section of this issue.]