Causal inference with multiple treatments Matched methods using propensity scores are a recommended form of analysis to identify causes and effects from non-clinical data. However, current matched methods are mostly limited to comparing one exposure type versus another. Prominent in this limitation is a difficulty in performing causal inference in the multi-treatment setting. However, causal methods for multiple treatments are often required in an aging population where the number of available therapies or interventions for common ailments such as arthritis or heart disease is growing. We propose to develop a method for a matched comparison of three exposure levels, which allows us to estimate the causal effects of three anti-rheumatic therapies on rheumatoid arthritis outcomes and of three types of caregivers on outcomes in a nursing home data set.
Causal inference with multiple treatments Effective use of causal inference is critical for a comparison of treatment effectiveness. The proposal develops methods to make causal inference in the presence of three treatments and implements a series of simulations to identify when these methods can provide unbiased estimation of treatment effects. Successful implementation of this method will improve understanding of which therapies can best improve rheumatoid arthritis outcomes and which form of hospital utilization provides optimal outcomes in a nursing home population.