Selection bias in an inpatient outcomes monitoring project.
Managed care organizations increasingly tout clinical outcomes assessment as the mechanism by which we will ensure quality and compare providers. The authors report on their experience with a multisite inpatient outcomes monitoring project by comparing patients who accepted (N = 51), refused (N = 36), or were not asked (N = 110) to participate in the project. The patients who were asked to participate had significantly longer inpatient stays compared with the unasked group (11.2 vs 6.9 days). Patients who agreed to participate in the project were more likely to have a bipolar (43.1% vs 19.2%) or any affective disorder (94.1% vs 79.5%), and less likely to have a schizophrenic disorder (2.0% vs 11.6%) than the refused and unasked groups. The project participants also had higher 90-day readmit rates (27.5% vs 9.6%), more readmissions (0.51 vs 0.16), and more education (14.59 vs 13.51 years) than nonparticipating patients. In this preliminary study, patient-related variables were found to influence who the staff asked and who consented to participate in this clinical outcomes monitoring project. The authors distinguish clinical outcomes monitoring from treatment effectiveness research and discuss the need to develop methodologies that deal with nonrepresentative patient sampling and intersite variability in recruitment practices.