Files

Download

Download Full Text (218 KB)

Division

East Florida

Hospital

HCA Florida Aventura Hospital

Specialty

Hospice & Palliative Care

Document Type

Poster

Publication Date

2025

Keywords

polypharmacy, LLM, AI, artificial intelligence, ChatGPT, DeepSeek, palliative care

Disciplines

Palliative Care

Abstract

Background: Polypharmacy is common in older adults and increases risks of adverse drug events, falls, and hospitalizations. Deprescription can mitigate these harms but is often hindered by time constraints, fragmented guidelines, and complex comorbidities. Artificial intelligence (AI) large language models (LLMs) may provide decision-support for deprescription in geriatric and palliative care.

Objective: To compare ChatGPT and DeepSeek in their deprescription recommendations across standardized polypharmacy vignettes.

Methods: Three validated clinical vignettes, varied by activities of daily living (ADL) status and cardiovascular disease (CVD) history, were presented to ChatGPT (GPT-4/5) and DeepSeek (2024 version). Each vignette was tested in three replicates per model. Outputs were analyzed for (1) binary deprescription decisions, (2) medications targeted, and (3) replicate consistency. Results: In patients without CVD, both models consistently recommended deprescription across all ADL scenarios. In patients with CVD, ChatGPT was more variable (e.g., deprescription rates ranged from 0–67% across replicates), whereas DeepSeek consistently recommended deprescription in 100% of scenarios. Both models most often targeted gastrointestinal and pain medications. ChatGPT demonstrated high replicate stability in the no-CVD scenarios, while DeepSeek showed complete stability across all scenarios.

Conclusion: Both ChatGPT and DeepSeek show promise as adjuncts for deprescription decision-support in geriatrics. DeepSeek demonstrated greater conservatism in CVD cases but higher internal consistency, whereas ChatGPT was more aggressive but variable. LLMs may augment deprescription reviews but require expert oversight and validation against established frameworks.

Original Publisher

HCA Healthcare Graduate Medical Education

Comparing ChatGPT and DeepSeek for Deprescription Support in Geriatric and Palliative Care: An Exploratory Study Using Clinical Vignettes

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.