A data-based examination of selection-bias in the evaluation of a job-based alcoholism program

We describe the strategy of constructive confrontation present in most job-based alcoholism programs, and conclude that it is compatible with the major concepts and practices of discipline in the workplace. We review past evaluation studies of these programs, noting the rather remarkable positive outcomes often reported. We then review reasons for both accepting and rejecting these unusual findings. One of the major biases that could produce the favorable findings-selectivity bias-is then examined, using both field data and record data from a large national corporation. Samples from records in the review period 1964 through 1977 generated 377 case histories. The study also used field interviews in the same corporation in a national sample of 19 locations. We interviewed supervisors who had experienced an employee in the alcoholism program (A sample = 154) and a sample of all other supervisors (B sample = 351) who described a nonalcoholic problem employee and how he/she had been managed. The remainder of the report focuses upon two possible sources of selection bias: suspected but not referred cases in the field data, and incomplete, "lost," cases in the record data. In both instances, the selectivity seemed to be quite minimal, insufficient to damage the conclusion that the program intervention had produced significant positive changes in job performance.