Research & Employment Opportunities

Statistical Evaluation of a Care-Management Program’s Effect on Hospital Admissions

Project Overview

A collaboration with Senscio Systems, a Massachusetts based health analytics company, we are studying whether a care-management intervention reduces all-cause hospital admissions among health plan members with chronic conditions (e.g., CHF, COPD, diabetes, atrial fibrillation, kidney disease). The study compares members enrolled early with a matched comparison cohort enrolled later, so the design mimics a randomized trial even though it is observational. Because assignment was not random, establishing a causal effect requires careful statistical work: matching and weighting methods to balance the two groups, zero-inflated count models to handle admissions data where most patients have zero events, and simulation-based power analysis to plan sample size and detectable effect sizes. The project is paid and is intended to produce a peer-reviewed publication; students will be co-authors commensurate with their contribution.

Project Tracks (join one or more)

• Code organization & reproducibility  –  Structure and document the project’s R/Python codebase on GitHub (version control, README, reproducible pipeline); light data-analysis involvement as time allows.
• Data wrangling & pre-processing  –  Clean and merge patient-level claims/utilization data, construct analysis-ready cohorts and variables (e.g., observation windows, comorbidity flags, admission counts) in R or Python.
• Monte Carlo power simulation  –  Simulate zero-inflated Poisson/negative-binomial patient data from preliminary estimates to evaluate statistical power across sample sizes and effect sizes; streamline this work into a well-organized, potentially publishable R or Python package.
• Causal inference (matching & weighting)  –  Use R packages for propensity-score matching, weighting (e.g., entropy balancing/IPTW), and covariate balance diagnostics to support a causal estimate of the intervention’s effect, and help draft the corresponding analysis and results.

Skills Needed (many can be developed on the job)

• Comfort programming in R and/or Python; prior coursework in statistics or biostatistics is a plus
• Git/GitHub 
• Interest in causal inference, count-data/GLM models, or simulation methods – track-specific, no prior expertise required
• Self-directed, detail-oriented, and comfortable working with sensitive health data
• Interest in contributing to a manuscript for publication

Interested? Contact Mike Arciero – michael.arciero@maine.edu