Evaluating an Artificial Intelligence Powered Medication Adherence Program's Targeting by Patients' Socially Vulnerability.

Author(s)

Jones C1, Leung K2, Flores CA3, Schultz L3, Marano T3, Chen C3, Chan S3, Kinley J3, Sivakolunthu Vel B3, Wonders P4, Taitel M5
1AllazoHealth, Saint Joseph, MI, USA, 2AllazoHealth, San Antonio, TX, USA, 3AllazoHealth, New York City, NY, USA, 4Walgreen Co, Hawthorn Woods, IL, USA, 5Walgreen Co, Deerfield, IL, USA

Presentation Documents

BACKGROUND: The CDC’s Social Vulnerability Index (SVI) scores census tracts based on socioeconomic status, household composition & disability, minority status & language, and housing type & transportation. In 2020, a national retail pharmacy chain used an Artificial Intelligence (AI) powered program to target in-person, telephonic, SMS and email interventions to patients with the goal of improving the population’s adherence rate to Diabetes, Hypertension, and Statin medications. The AI program did not explicitly attempt to provide more or less support to patients based on their SVI score. To assure health equity, it is prudent to evaluate AI programs for unintended bias.

OBJECTIVES: The objective was to determine whether patients classified as socially vulnerable, were more or less likely to be targeted by AI for adherence improving interventions.

METHODS: This retrospective study used the SVI dataset and prescription, intervention, and AI targeting data between 1/1/2020 and 12/24/2020 from Walgreens pharmacy and AllazoHealth, a healthcare AI company. Patients who qualified for CMS specified hypertension, diabetes, or statin adherence measures were included in the study. Patients living in the 20% of Census Tracts ranked most vulnerable on the SVI were classified as Highly Vulnerable; the remaining patients were classified as Less Vulnerable. Two-proportion z-tests were performed comparing the proportion of Highly Vulnerable patients to the proportion of Less Vulnerable patients targeted for medication adherence interventions.

RESULTS: Highly Vulnerable patients were 11.1% more likely than Less Vulnerable patients to be targeted by the AI for interventions for Diabetes (P< .00001), 15.6% more likely for Hypertension (P< .00001) and 17.7% more likely for Statins (P< .00001). Sample sizes were 349,514 (Diabetes), 1,044,568 (Hypertension), and 1,174,871 (Statins).

CONCLUSIONS: Highly Vulnerable patients were more likely to be targeted by the AI for interventions than Less Vulnerable patients, thus assuring advanced pharmacy services were offered to vulnerable populations.

Conference/Value in Health Info

2021-05, ISPOR 2021, Montreal, Canada

Value in Health, Volume 24, Issue 5, S1 (May 2021)

Code

PPM6

Topic

Health Policy & Regulatory, Health Service Delivery & Process of Care, Methodological & Statistical Research, Patient-Centered Research

Topic Subcategory

Adherence, Persistence, & Compliance, Artificial Intelligence, Machine Learning, Predictive Analytics, Health Disparities & Equity, Pharmacist Interventions and Practices

Disease

Cardiovascular Disorders, Diabetes/Endocrine/Metabolic Disorders, Personalized and Precision Medicine

Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on ispor.org. Update my browser now

×