LLMs may enhance geriatric polypharmacy management in primary care
ChatGPT demonstrated potential in providing useful clinical decision support for primary care physicians managing polypharmacy and deprescription.
Source: Getty Images
By - A research team from the Mass General Brigham MESH Incubator found that the large language model (LLM) ChatGPT can support clinical decision-making around medication management and deprescription for geriatric patients, according to a recent Journal of Medical Systems study.
The researchers underscored that polypharmacy – the use of five or more medications – is common among older adults but can increase the risk of adverse drug interactions. Further, they indicated that the prevalence of polypharmacy in older patients hovers around 40 percent.
The increasing number of these individuals on Medicare seeing more specialists has also led to upticks in primary care physicians overseeing medication management. However, challenges like primary care shortages and the rapidly aging population make effective polypharmacy difficult.
The study authors asserted that tools to aid in polypharmacy management are key to tackling the increasing burden of care, leading them to evaluate the potential of LLMs for managing polypharmacy and deprescription.
Deprescribing unnecessary drugs can help reduce the risk of adverse drug interactions, but the decision-making process around deprescription is often time-consuming and complex, the research team noted.
To assess ChatGPT’s utility in this area, the team tasked the LLM with generating yes/no binary deprescribing decisions in response to standardized clinical vignettes – pulled from a study investigating general practitioners’ deprescribing decisions – and choosing which of several medications to deprescribe based on list-based prompts.
The researchers fed the clinical scenarios, which all featured an elderly patient taking a combination of medications but varied in terms of cardiovascular disease history and degree of impairment in activities of daily living (ADL), into ChatGPT 3.5.
The LLM consistently recommended deprescribing medications in patients without a history of cardiovascular disease in yes/no binary deprescribing decisions. In patients with cardiovascular disease history, ChatGPT was more likely to leave a patient’s medication regimen unchanged.
In patients both with and without a history of cardiovascular disease, ADL status did not significantly impact the model’s decisions.
The LLM also tended to disregard patient pain level, preferring to deprescribe pain medications over drugs like antihypertensives and statins.
The study further revealed that ChatGPT’s responses varied when researchers presented the same clinical vignette in new chat sessions, highlighting potential inconsistencies in the reported clinical deprescribing trends incorporated into the model’s training.
However, the research team emphasized that the tool’s deprescribing decisions varying based on ADL status, cardiovascular disease history and medication type is in many ways concordant with the logic of general practitioners, indicating that specifically trained LLMs could provide valuable clinical decision support for primary care physicians in the context of geriatric polypharmacy management.
“Our study provides the first use case of ChatGPT as a clinical support tool for medication management,” said senior corresponding author Marc Succi, MD, associate chair of Innovation and Commercialization at Mass General Brigham Radiology and executive director of the MESH Incubator, in a news release. “While caution should be taken to increase accuracy of such models, AI-assisted polypharmacy management could help alleviate the increasing burden on general practitioners. Further research with specifically trained AI tools may significantly enhance the care of aging patients.”
Artificial intelligence (AI) approaches are increasingly being researched for potential applications in medication adherence and management.
In May, leadership from the mHealth Impact Laboratory at the Colorado School of Public Health sat down with HealthITAnalytics to discuss the role of AI in medication adherence, including challenges, key considerations and potential benefits.
