Proactive safety surveillance.

Growth in health information systems presents opportunities to enhance postmarketing safety surveillance of medical products. Spontaneous suspected side effect reports provide the foundation, but we need to 'proactively' improve their quality and our strategies to seek signals. In our more familiar 'reactive' mode, we examine hypotheses from inquiries or publicity. Such responsive evaluations remain essential but may miss latent information on unsuspected risks. Efficient techniques to disclose hidden clusters and associations may emerge through adaptation of approaches from industrial quality control and other disciplines. Data-driven techniques like exploratory analysis, control charts, and time series modeling may help in sifting through accumulated data and in screening consecutive submissions to discern hints of new product hazards or of more specific understanding about previously identified potential side effects. We also need to cultivate non-spontaneous data for hypothesis generation as well as testing, the systematic epidemiologic evaluation of questions and concerns. This hypothesis testing function will assume greater importance if proactive safety surveillance methods yield larger numbers of putatively positive findings. Whether from spontaneous reports or other sources, signals that could have arisen by chance alone usually represent only clues to potential hazards until or unless they can be verified through independent studies.