Background: Analyses of longitudinal health-related quality of life data often exclude participants who die, which limits the generalizability of the results. Methods to incorporate death as a valid score in the Medical Outcomes Study Short-Form (SF-36) have been suggested but need to be evaluated in other populations. Objectives: We sought to apply a method of transforming the SF-36 Physical Component Score (PCS) to include death. A transformation to estimate the probability of being "healthy" in 3 years, based on the current PCS value, will be developed and validated. Subjects: Women in the Australian Longitudinal Study on Women's Health (ALSWH), ages 70-75 years at Survey 1 in 1996 (n = 12,432), were followed-up at 3 yearly intervals for 6 years. Results: The transformation derived from the ALSWH data provides evidence that the methodology for transforming the PCS to account for deaths is sound. The 3-year equation provided good estimates of the probability of being healthy in 3 years and the method allowed deaths to be included in an analysis of changes in health over time. Conclusions: For longitudinal studies involving the SF-36 in which subjects have died, we support the recommendation that both the PCS and its transformed value which includes deaths should be analyzed to examine the influence of deaths on the study conclusions. Using study data to derive empirical parameters for the transformations may be appropriate for studies with follow-up intervals of other lengths.