Abstract
Surveillance of health care performance indicators is of growing interest as drugs need to be monitored even after they passed phase 3 clinical trials and are on the market. Individuals in small subpopulations may still have adverse reaction to it, and even if no such adverse reaction is present, monitoring is the only way public concerns can be settled. We compare methods using large sample approximations and exact distributions for the likelihood ratio and show that the efficient score vector is just as effective tool. Furthermore, we demonstrate that it can be applied even for data structures that are too complicated for the likelihood ratio. All new monitoring schemes are simple and they can be represented by simple graphs.
ACKNOWLEDGMENTS
The authors are grateful to Professor Martin Kulldorff of Harvard Medical School and Harvard Pilgrim Health Care Institute for his interest and encouragement. In addition, we wish to thank the Editor and the referees for their kind and prompt handling of this submission.
Notes
Recommended by Nitis Mukhopadhyay
Color versions of one or more of the figures in the article can be found online at http://www.tandfonline.com/lsqa.