Home List of Titles Analysing longitudinal changes in health-related quality of life: how do you account for deaths and other missing data?
Please use this identifier to cite or link to this item: http://hdl.handle.net/1959.3/234304
- Analysing longitudinal changes in health-related quality of life: how do you account for deaths and other missing data?
- Bowe, S.; Young, A.; Sibbritt, D.
- Background: Many longitudinal studies that measure changes in health-related quality of life over time use the Medical Outcomes Study Short-Form 36 (SF-36). The SF-36 covers the major aspects of health, has been validated with adults of all ages and has been used extensively in clinical trials. A limitation of the SF-36 is that it only considers morbidity and not mortality. Aim: To apply a method of transforming the SF-36 to incorporate death; to impute data missing for other reasons, and to assess the impact on study findings using these methods. Methods available to impute data for longitudinal studies will also be discussed. Methods: The focus of this paper is the older cohort of women in the Australian Longitudinal Study on Women's health. Survey 1 was conducted in 1996 (aged 70-75 years, n=12432), with three yearly follow-up. Women were classified as having diabetes based on their self-report of the diagnosis by a doctor. Deaths were ascertained by linkage to the National Death Index. Changes in the Physical Health Component Summary score (PCS) of the SF-36 were measured for women with and without diabetes. The PCS was then transformed to the probability of being healthy at the next survey, taking the value of zero after death. Multiple imputation was used to replace other missing data. Results: Our findings suggest that the SF-36 transformation was valid. The transformation enabled us to detect changes in health in the cohort which were not apparent when only analysing survivors who had complete follow-up data. After then imputing values which were missing for other reasons, women with diabetes were shown to have a greater decline in health over time than other women their age. Conclusion: This study has shown the importance of accounting for death and missing data when studying longitudinal changes in health.
- Publication type
- Conference paper
- Paper presented at the 15th Australasian Epidemiological Association Annual Conference (AEA 2006), Melbourne, Victoria, Australia, 18-19 September 2006
- Publication year
- Diabetes; Health services; Mortality; Women's health
- Australasian Epidemiological Association
- Publisher URL
- Copyright © 2006.