B1480
Title: Quantitative bias analysis for provider profiling and practice variation studies
Authors: Rolf Groenwold - Leiden University Medical Center (Netherlands) [presenting]
Abstract: Provider profiling and comparisons of performance indicators across centres essentially focus on understanding the (causal) effect of visiting one institution instead of another. At the same time, practice variation (e.g., in the use of medical treatment across centres) could provide a starting point for studies of the effects of medical treatments. Both instances require clear articulation of the assumptions needed to give the results of the analyses a causal interpretation. Violations of those assumptions (for example, unmeasured confounding or lack of conditional exchangeability) thus impacts provider profiling, just as it would impact observational studies of medical treatments. Unfortunately, commonly used methods are not immune to violations of the assumptions and the impact of that could be investigated using quantitative bias analysis methods. The possible role of quantitative bias analysis will be discussed for provider profiling studies and for observational studies that utilize practice variation.
