Adjusting for Baseline Covariates in Net Benefit Regression: How You Adjust Matters

Background and ObjectiveThe literature has shown that different baseline adjustment approaches lead to different results when examining cost and quality-adjusted life-years. To our knowledge, the concept of baseline adjustment in a net benefit (NB) regression has not been studied. The purpose of the study was to explore the impact of different baseline adjustment approaches in an NB framework on the cost effectiveness of an intervention using person-level data.MethodsThis study used data from a randomized controlled trial that evaluated the effectiveness of a multifactorial falls prevention intervention for older home care clients. The outcome was the number of falls at the 6-month follow-up. The cost variable was the total healthcare costs from a societal perspective. Incremental NB values were estimated using four baseline adjustment approaches: (1) the change in NB is the dependent variable; (2) the NB at follow-up is the dependent variable without adjusting for baseline values; (3) the NB at follow-up is the dependent variable adjusting for baseline NB; and (4) the NB at follow-up is also the dependent variable adjusting for baseline cost and effect separately.ResultsWith adjustment of baseline values (Approach 1, 3, 4), the intervention was not cost effective when compared to usual care. Conversely, without baseline adjustment (Approach 2), the intervention was cost effective if decision-makers’ willingness-to-pay per fall prevented was CAN$10,000 or greater.ConclusionsThis study showed that different baseline adjustment approaches in a cost-effectiveness analysis can lead to different results. Future research is needed to determine the most appropriate adjustment approach in planning economic evaluation using NB regression.