Loading component...
Current Projects - Connected Communities Initiative
Project from Spring 2026
LLM-Informed Feature Priors for High-Cardinality EMR
- Jun Zhang
- Meenal Chaudhari
This project aims to improve clinical prediction and decision support by addressing a key limitation in how high‑cardinality electronic medical record (EMR) data are represented in predictive models. The primary goal is to evaluate whether large language model (LLM)–informed feature priors can better preserve clinically meaningful signals within sparse, high‑dimensional EMR data compared with traditional expert‑curated or purely statistical approaches. Using lung cancer risk prediction in younger non‑smokers as a proof‑of‑concept use case, the project will systematically compare three strategies for encoding high‑cardinality clinical features: expert aggregation, statistical reduction, and LLM‑informed semantic weighting - alone and in combination. The objective is to identify approaches that improve model stability, interpretability, and generalizability while retaining rare but important clinical signals. Deliverables will include comparative performance evidence and practical guidance for feature encoding strategies, supporting more reliable early disease detection and advancing data‑driven innovation aligned with OSF HealthCare’s cancer care and community health priorities.
PharmaShield: A Risk-Aware System for Resilient & Equitable Pharmacy Supply
- Jacob Gendron
- Ashish Chandra
This project aims to strengthen medication access across OSF HealthCare by transforming how pharmacy supply risk is managed under ongoing uncertainty. The primary goal is to develop a risk‑aware, decision‑support system that helps pharmacy leaders proactively protect access to critical medications (especially those supporting 340B‑eligible services) despite shortages, allocation limits and demand volatility. The project will deliver a deployable framework that converts supply and demand uncertainty into clear operational strategies, including safety stock targets and internal redistribution rules aligned with defined service‑level goals. Rather than relying on static inventory thresholds or alerts alone, PharmaShield will provide actionable guidance that balances reliability, cost and equity. During the one‑year phase, the system will be tested on selected high‑risk medications using simulation and pilot evaluation. The outcome will be an implementation‑ready tool that reduces stockout risk, supports equitable medication access, and strengthens pharmacy resilience while preserving fiscal stewardship and compliance with existing 340B requirements.
Dynamic Fall Risk Surveillance and Prevention Through AI-Enhanced Statistic
- Chris Franciskovich
- Maochao Xu
This project aims to reduce preventable inpatient falls by shifting fall prevention from static screening to continuous, data‑driven risk surveillance. The goal is to develop and validate an AI‑enhanced analytic framework that produces timely, short‑horizon fall risk estimates and clear explanations of key risk drivers, enabling more targeted and efficient prevention efforts. The one‑year seed project will deliver a retrospectively validated prototype that predicts fall risk over the next nursing shift and the next few hours, using routinely available electronic health record data. Risk estimates will update dynamically and be paired with interpretable drivers to support clinical decision‑making without increasing alert burden. A core objective is to translate these outputs into Epic‑integrable functional specifications that align with existing workflows. By providing calibrated, near‑term risk signals and actionable insights, the project establishes a foundation for proactive fall prevention, supports safer inpatient care, and prepares the system for future prospective deployment and expansion.
Advancing Zero Waste T0 Landfill at OSF through plastic waste to energy conversion
- Ryan Wolf
- Hafizur Rahman, MD
This project aims to provide OSF HealthCare with clear, defensible evidence to determine whether non‑hazardous healthcare plastics can be diverted from landfill and incineration through a viable conversion pathway that supports Zero Waste to Landfill (ZWTL) goals. In partnership with Illinois State University, the primary objective is to evaluate which plastic waste streams at an OSF HealthCare facility are technically suitable for conversion, what recovery yields are achievable and whether the resulting products have practical engineering value. The one‑year pilot phase is intentionally limited to assessment - not deployment. It will generate the data needed to support or rule out future diversion claims under UL 2799 and to inform strategic decisions about scaling. By establishing a healthcare‑specific technical baseline, the project reduces institutional risk while positioning OSF to move from waste reduction toward verified material circularity. Successful completion will enable informed consideration of a Phase II expansion aligned with long‑term sustainability and operational goals.
Become an IAI Member
Follow the link below and complete the form to become an IAI member. There is no cost to join, and all members receive communications of upcoming grant and development opportunities.
