Benefits and pitfalls of pooling datasets from comparable observational studies: combining US and Dutch nursing home studies

Abstract

Different research groups sometimes carry out comparable studies. Combining the data can make it possible to address additional research questions, particularly for small observational studies such as those frequently seen in palliative care research. We present a systematic approach to pool individual subject data from observational studies that addresses differences in research design, illustrating the approach with two prospective observational studies on treatment and outcomes of lower respiratory tract infection in US and Dutch nursing home residents. Benefits of pooling individual subject data include enhanced statistical power, the ability to compare outcomes and validate models across sites or settings, and opportunities to develop new measures. In our pooled dataset, we were able to evaluate treatments and end-of-life decisions for comparable patients across settings, which suggested opportunities to improve care. In addition, greater variation in participants and treatments in the combined dataset allowed for subgroup analyses and interaction hypotheses, but required more complex analytic methods. Pitfalls included the large amount of time required for equating study procedures and variables and the need for additional funding.