Detecting Language-Related Disparities in Spiritual Care Using Artificial Intelligence.

INTRODUCTION

Language barriers perpetuate health care disparities, yet traditional data sources overlook inequities in relationship‑based care. This study uses artificial intelligence to examine language‑related disparities in spiritual care, a domain largely documented in unstructured clinical notes.

METHODS

We analyzed electronic health records of 7,360 deceased adults aged ≥65 years from a tertiary hospital (2010-2023). Natural language processing identified spiritual care documentation, and structured flowsheets captured formal chaplain visits. Logistic regression assessed associations between patient language and spiritual care access and timing.

RESULTS

Non‑English speakers had 59% lower odds of receiving formal spiritual care (odds ratio = 0.41, p = .003) and received support nearly 10 months later (median 136 vs. 432 days, p = .020). Among non‑English speakers, non-chaplain providers more often first documented spiritual needs.

DISCUSSION: Artificial intelligence-supported analysis of unstructured documentation revealed language‑related disparities missed by structured data, offering a culturally grounded approach to identifying inequities in relationship‑based care.