When an AI system flags a patient as a potential opioid misuser, the consequences can be immediate and severe — denied refills, untreated pain, and eroded trust in care. At Chapman University’s School of Pharmacy, researchers are documenting exactly how that happens and teaching students to interrogate the tools they will inherit.
The bias hiding in prescription monitoring
NarxCare, an AI platform embedded in many state Prescription Drug Monitoring Programs, uses predictors like multiple pharmacy visits, overlapping prescriptions, and frequent physician changes to score misuse risk. In practice, those signals often reflect life circumstances rather than abuse. Patients who move frequently due to high housing costs, lack reliable transportation, or live in medically underserved areas get flagged and denied medication.[3]
California’s Department of Justice never allowed NarxCare in the state, a decision Chapman researcher Sherry Yun Wang calls fortunate. In other states, she has heard patients describe crying from untreated pain after being cut off by algorithmic decisions.[3] The pattern reveals a core tension: tools designed for population-level risk assessment get applied to individual patients without context.
Research that follows the harm
Wang, an assistant professor of pharmacy, arrived at Chapman in 2021 as AI was transforming healthcare but remained absent from clinical training. She secured a $250,000 National Institutes of Health grant in 2025 to study bias in AI prescription tools and was named Women in AI’s North America Researcher of the Year.[3] Her work spans epidemiology, health analytics, pharmacoeconomics, and clinical pharmacy — all focused on how algorithmic decisions affect real patients.
To address the problem systematically, Wang founded the Healthcare AI Initiative (HAI) at Chapman. The initiative examines ethical questions around data privacy and develops equitable machine learning approaches for clinical decision-making.[3] The goal is not to reject AI but to ensure it serves all populations fairly.
Teaching students to use — and question — AI
Wang developed an AI elective within the School of Pharmacy where students with no coding experience use AI tools to run geospatial analyses and answer clinical research questions in real time.[3] The course moves beyond theory: learners upload data, run models, and interpret results under guidance that emphasizes critical evaluation over blind adoption.
Jennifer Totonchy, associate dean of research and graduate programs, frames the stakes plainly: “Part of embracing this technology is recognizing that we can’t send graduates into the world without knowing how to use these tools.”[3] The elective treats AI literacy as a clinical competency, not a technical specialty.
An institutional backbone for responsible use
Chapman’s AI Hub provides the infrastructure that makes this work possible across campus. The hub maintains curated, IS&T-reviewed lists of approved AI tools for faculty, students, and staff, along with syllabus statement templates, on-demand training, and plain-language glossaries.[2][2] Separate guidelines govern research publication, sensitive data protection, and administrative workflow automation.[2][2]
This structure means the pharmacy school’s efforts sit within a university-wide framework. When Wang’s students need tool access or policy clarity, the hub’s review process and training modules are already in place. The hub also publishes open-access guidelines and papers for the broader community.[2][2]
Why this model matters beyond one campus
Chapman is also a lead partner in a National Science Foundation–funded initiative helping Emerging Research Institutions build the infrastructure, staffing, and oversight needed to adopt generative AI responsibly.[5] That work extends the same principles — equity-centered design, critical literacy, institutional guardrails — to universities with fewer resources.
Meanwhile, Chapman engineers have demonstrated the positive side of healthcare AI: a smartphone app that estimates hemoglobin from a fingernail selfie, delivering lab-comparable results for over 1.4 million screenings and enabling at-home monitoring for kidney disease and cancer patients.[5] The contrast is instructive. The same institution producing accessible diagnostic tools is also exposing how prescription algorithms can harm vulnerable patients.
The lesson for health education is clear. AI will be part of clinical practice whether curricula address it or not. Chapman’s approach — pairing bias research with hands-on training, all supported by campus-wide governance — offers a replicable path. Students learn to use the tools. They also learn where the tools fail.
Want to see how Chapman structures its AI policies and training? Explore the AI Hub’s public resources and approved tool lists to compare with your own institution’s approach.[2][2]
