Robust Learning Approaches

Robust Learning Approaches for Assessing Effects and Effect Heterogeneity of Real World Antipsychotic Treatment Regimes in Elderly Persons with Schizophrenia

5R01MH130213-02

12/2022-10/2027

We will develop statistical approaches to extract scientifically robust and valid causal evidence of the effectiveness and adverse outcomes of antipsychotic drugs used by elderly adults with schizophrenia using large, observational, longitudinal databases. Our proposal has high public health relevance because (1) we focus on adults with illnesses associated with a heavy disease burden for whom drug treatments are a critical and often lifetime treatment component; (2) we assess the extent to which patient race/ethinicity and social contextual factors known to influence health behaviors may moderate outcomes; (3) we expand causal inference methodology to characterize the outcome effects of drug exposure and drug regimens, thus providing valuable information to optimize outcomes of long and complex exposures typical of usual care settings; and (4) we develop generalizable approaches to target parameters of general scientific interest.

Publications

2025

Max Rubinstein, Denis Agniel, Larry Han, Marcela Horvitz-Lennon, and Sharon-Lise Normand. 2025. “Bounding Causal Effects With an Unknown Mixture of Informative and Non-Informative Censoring”
Max Rubinstein, Denis Agniel, Larry Han, Marcela Horvitz-Lennon, and Sharon-Lise Normand. 2025. “Bounding Causal Effects With an Unknown Mixture of Informative and Non-Informative Censoring”

Presentations

Presentations and papers from Year 4: 

  1. Yige Li presented “A covariate balancing approach for dynamic treatment regimes,” at the 2026 American Causal Inference Conference (ACIC) organized by the Society for Causal Inference (SCI) in Salt Lake City, Utah, May 2026.
    1. Authors: Yige Li, Ben Buzzere, Marcela Horvitz-Lennon, Sharon-Lise Normand.
  2. Max Rubinstein presented “Bounding causal effects with an unknown mixture of informative and non-informative missingness” (poster) at the 2026 American Causal Inference Conference (ACIC) organized by the Society for Causal Inference (SCI) in Salt Lake City, Utah, May 2026.
    1. Authors: Max Rubinstein, Denis Agniel, Larry Han Marcela Horvitz-Lennon, Sharon-Lise Normand.
  3. Max Rubinstein paper, “Bounding causal effects with an unknown mixture of informative and non-informative missingness,” was accepted for publication in the Journal of causal inference, PMCID: PMC13455597.
    1. https://www.degruyterbrill.com/document/doi/10.1515/jci-2025-0028/html?srsltid=AfmBOorKUoj4tCm6qYBh5OOOdfy1alz3jkpr3Tw-HPsQHS7AeFC7aVHL#articleAbstractView
  4. Larry Han’s paper, “FACTOR: Fairness-Aligned Conformal Transport for Optimal Regions,” was accepted for poster presentation at Statistics and Trustworthy AI for Cross (X)-Domain Acceleration (STAI-X) in Cambridge, Massachusetts, August 2026. The paper has also been accepted for publication in the Proceedings of Machine Learning Research, Volume 335.
    1. Authors: Chenyin Gao and Larry Han.
  5. Abstract accepted to be included in the session titled “COSTS AND EFFECTIVENESS OF INTERVENTIONS TO TREAT MENTAL DISORDERS AND ACCESS INEQUALITIES” in the upcoming 26th WPA World Congress of Psychiatry in Stockholm, Sweden on September 26, 2026. 
  6. Marcela Horvitz-Lennon paper “Social Disadvantage and Risk of Death Among Seriously Ill Elderly Medicare Beneficiaries: Examining Causal Effects and Variation by Race and Ethnicity” is currently under review at JAMA-HF. 

Presentations during Year 3: 

  1. Marcela Horvitz-Lennon. Leveraging data science to assess the role of social disadvantage on health outcomes. Seminar on Biomedical Data Science in the Era of Generative AI, Cambridge, MA (October 2024). 
  2. Marcela Horvtiz-Lennon. Race and ethnicity and the effect of social disadvantage on mortality among elderly Medicare beneficiaries. Seventeenth Workshop on Costs and Assessment in Psychiatry, Mental Health Outcomes, Services, Economics, Policy Research, Venice, Italy (March 2025). 
  3. Sharon-Lise Normand. Causal inference and evidence on drug outcomes among elderly schizophrenia patients. International Day of Women in Statistics and Data Science, Virtual Meeting (October 2024).

Presentations during Year 2

  1. Max Rubinstein. Mixture censoring: a sensitivity framework for causal analyses with censored outcomes. Acadmey Health, Washington DC, July 2024. Late breaking poster.
  2. Yunzhe Qian. Enhancing causal inference for multi-valued treatments: A confounder balanced deep learning instrumental variable approach. New England Student Research Symposium on Data Science and Statistics, Boston University, April 20, 2024. Oral presentation.
See also:

Study Team