Patients are the people who should report on their health and experience. When patients are unable to do this due to illness or impairment, proxies such as relatives may be asked to report for them. This ensures that the sickest and most vulnerable patients are not excluded from measurements of health and experience. However, patients who are not able to report may differ from patients who can in important ways. Proxy responses may also differ from what a patient would have said if they could report. If these issues are not accounted for when analyzing data, the results may be misleading.
The use of proxies means that health and care experience can be collected from patients whose data would otherwise be missing. However, the use of proxies also leads to analytic challenges that can adversely affect statistical analysis if they are not addressed. This proposal will provide researchers with guidelines for choosing the most appropriate statistical method to adjust for proxy responses in cross-sectional and longitudinal studies. New methods that address some of the limitations of current methods when they are applied to proxy responses will also be developed.
The team will develop statistical methods that can address multiple types of patient-reported outcomes (e.g., physical, emotional/psychosocial, etc.) simultaneously. The team will use extensive simulation analyses based on realistic scenarios to examine current as well as newly proposed procedures for adjusting for proxy-reported data in cross-sectional and longitudinal studies. These simulations will allow optimal methods that should be used by investigators for their own questions and datasets to be identified.
The team will provide guidelines for statistical procedures aimed at adjusting for proxy-reported data in cross-sectional and longitudinal data sets. The team will also provide software and tutorials implementing and describing the possible methods so that they can be used by other researchers. In addition, the proposed methods will be applied to three important health questions. The first application will estimate physical and mental health outcomes for older cancer survivors that are used in public reports and ratings. The second and third applications will examine the effects of palliative care on quality of life in patients with Parkinson’s disease and related conditions.