Are finite population corrections appropriate when profiling institutions?

Over the last two decades, there has been growing interest in comparative performance evaluation of health institutions. Small hospitals and small nursing home facilities might have too few eligible cases for sufficiently precise comparisons in a single year (or other standard reporting period). An analyst might be tempted to use the smaller standard error estimates obtained under the finite population sampling model (FPSM), rather than those obtained under a superpopulation model. The authors argue that the FPSM does not properly represent the reliability of quality profiling information in such situations. The authors consider alternative methods for obtaining adequately precise profiles for small institutions, including moving averages and hierarchical models. (December 2006)