ABSTRACT
This article explores the possibility of randomly assigning groups (or clusters) of individuals to a program or a control group to estimate the impacts of programs designed to affect whole groups. This cluster assignment approach maintains the primary strength of random assignment--the provision of unbiased impact estimates--but has less statistical power than random assignment of individuals, which usually is not possible for programs focused on whole groups. To explore the statistical implications of cluster assignment, the authors (a) outline the issues involved, (b) present an analytic framework for studying these issues, and (c) apply this framework to assess the potential for using the approach to evaluate education programs targeted on whole schools. The findings suggest that cluster assignment of schools holds some promise for estimating the impacts of education programs when it is possible to control for the average performance of past student cohorts or the past performance of individual students.
Subject(s)
Cluster Analysis , Data Interpretation, Statistical , Education/standards , Educational Measurement/methods , Models, Statistical , Program Evaluation/methods , Randomized Controlled Trials as Topic , Bias , Child , Effect Modifier, Epidemiologic , Humans , Longitudinal Studies , New York , Reproducibility of Results , Research DesignABSTRACT
BACKGROUND: This study hypothesized that an office-based training program in clinical breast examination (CBE) would improve the lump-detection skills of primary care physicians. METHODS: A one-group pretest-posttest design was utilized, without outcome measurements taken prior to instruction, immediately afterwards, and six months later. Fifty physicians from 117 practices that had been randomly selected and met program eligibility criteria elected to participate. The intervention was a one-hour office-based CBE training program based on the MammaCare method, a standardized approach to teaching the detection of breast lumps using silicone breast models. RESULTS: There were five lumps in the silicone model. The mean number of correct lump detections increased significantly, from 0.66 before to 3.2 after instruction, and this gain was maintained at six-month follow-up. The mean number of false positives decreased from 2.9 before to 1.16 after instruction, increasing slightly to 1.6 at six-month follow-up. Ninety-six percent of the physicians reported that they had modified their methods of CBE as a result of the training. CONCLUSIONS: This office-based instruction using an academic detailing model improved the participating physicians' abilities to correctly detect lumps in a silicone breast model and was received favorably.
