Regression modeling strategies: an illustrative case study from medical rehabilitation outcomes research.

The practice of outcomes research is growing in all segments of the health care industry, yet few practitioners and researchers are prepared to deal with the completion of statistical analyses that characterize the new focus on results. This article discusses basic model formulation and interpretation. It also encourages the use of statistical models that study the simultaneous effects of many variables on an outcome and gives examples of relationships among variables that are not simple and linear. The methods are illustrated with a dataset consisting of stroke rehabilitation inpatients discharged during a 3-year period with an admission date that is within 1 year after stroke.