The Parametric G-Formula to Compare the Effectiveness of Two Anemia Management Strategies Among Medicare Dialysis Patients

• Activity Number: 353
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Yi Zhang*+ and Mae Thamer
• Companies: Medical Technology and Practice Patterns Institute and Medical Technology and Practice Patterns Institute
• Keywords: Causal Inference ; G-formula ; Claims data ; inverse probability weighting ; dynamic treatment strategies ; time-dependent confounding

Abstract

Background. Conventional statistical approaches for observational studies are not well-equipped to deal with dynamic treatment strategies primarily due to time-dependent confounding. The parametric g-formula can appropriately adjust for this type of confounding. However, its use and feasibility based on claims data are unknown. Objective. To assess the feasibility of applying the g-formula approach to large claims. Methods. We implemented and adapted the parametric g-formula macro to compare the risks and benefits of targeting two hematocrit (Hct) levels (high vs. low) among Medicare dialysis patients. For validation, we emulated Normal Hematocrit Study (NHS), an existing RCT that compared same Hct target strategies, and evaluated statistical efficiency and sensitivity of estimates. Results. The estimates obtained from the g-formula were similar to those reported in the NHS. Results were robust to assumptions about missing Hct and extent of modeled treatment history. Conclusion: The parametric g-formula is a feasible technique to compare dynamic treatment strategies using claims data within a framework in which this method can be generalized for a broader causal inference questions.