How to use multiple logistic regression in retrospective database analyses

Accurately measuring the effectiveness of medical or pharmacological interventions is essential in economic evaluations. Many economic evaluations are performed retrospectively using medical and pharmacy claims databases. A challenge to using this approach is that variation in treatment patterns prevents simple comparisons between patient groups. Multiple logistic regression is a powerful technique that can overcome this challenge and is suited for evaluating dichotomous outcomes (eg, disease/no disease). An example is presented that applies multiple logistic regression to determine the impact of different drug therapies on hospitalization risk for heart failure among patient with congestive heart failure.