Project Grant R01GM169395

The National Institute of General Medical Sciences awarded the University of Wisconsin–Madison $310,405 on September 5, 2026, under the Biomedical Research and Research Training program (CFDA 93.859) for development of causal inference methods for complex biomedical data.

The project develops distributional balancing methods to support causal inference in critical care settings where standard assumptions—no unmeasured confounding, single-time exposures, and ideal measurement—rarely hold. The work addresses three specific gaps: Aim 1 develops proximal causal learning methods using treatment- and outcome-confounding proxies to handle unmeasured confounding; Aim 2 creates sequential distributional balancing methods for longitudinal treatment decisions, replacing unstable inverse-probability products with history-aware balancing weights; Aim 3 integrates distributional weighting with spline-expanded neural networks for continuous dose-response estimation. Deliverables include identification results, estimators, theory, diagnostics, and software tools. All three aims incorporate sensitivity analyses and formal diagnostics for balance, overlap, and support assumptions.

Performance runs through June 30, 2030, with work conducted in Madison, Wisconsin. This is assistance type Project Grant, classified under NIH's standard R01 mechanism for investigator-initiated research.