โก Quick Summary
This study utilized large language models (LLMs), specifically GPT-4o, to predict 6-month mortality risk among patients with dementia. The model demonstrated a strong association with mortality, achieving an area under the curve (AUC) of 0.79, which could enhance hospice referral decisions.
๐ Key Details
- ๐ Dataset: 9,872 patients diagnosed with Alzheimer’s disease and related dementias (ADRD)
- ๐งฉ Features used: Discharge summaries from electronic health records (EHRs)
- โ๏ธ Technology: GPT-4o, a large language model
- ๐ Performance: AUC of 0.79, adjusted hazard ratio (aHR) of 31.02
๐ Key Takeaways
- ๐ LLMs can effectively estimate mortality risk in dementia patients.
- ๐ก GPT-4o provided highly discriminative predictions for 6-month mortality.
- ๐ฉโ๐ฌ Study population included 9,872 individuals from two academic medical centers.
- ๐ Significant findings showed a 36% mortality rate within 6 months.
- ๐ค Cox regression analysis confirmed strong associations between predictions and mortality.
- ๐ Potential for improving hospice referral decisions through automated risk assessment.
- ๐ Further research is needed to validate these findings prospectively.
๐ Background
The prevalence of Alzheimer’s disease and related dementias (ADRD) is on the rise, making it increasingly important to identify patients who may benefit from hospice care. However, access to hospice services remains a challenge, often due to difficulties in accurately predicting patient mortality. The integration of advanced technologies, such as large language models, into clinical practice could provide valuable insights and support decision-making processes.
๐๏ธ Study
This study was conducted across two academic medical centers, focusing on patients diagnosed with ADRD. Researchers employed GPT-4o to analyze discharge summaries from electronic health records (EHRs) to estimate the 6-month mortality risk without any need for retraining or preprocessing. The study aimed to assess the model’s effectiveness in stratifying mortality risk and its potential implications for hospice referrals.
๐ Results
Out of the 9,872 individuals analyzed, 3,563 (36%) died within 6 months. The predictions made by GPT-4o were statistically significant, with a log-rank p-value of less than 0.001 and an AUC of 0.79. The Cox regression models indicated a strong association between the model’s predictions and actual mortality, with an adjusted hazard ratio of 31.02 (95% CI: 27.44-35.08, p < 0.001).
๐ Impact and Implications
The findings from this study highlight the potential of GPT-4o in stratifying mortality risk among dementia patients, which could significantly enhance the decision-making process for hospice referrals. By utilizing routinely generated documentation, healthcare providers may be able to identify patients who are at higher risk of mortality more effectively, ultimately improving access to hospice care and ensuring that patients receive the support they need during critical times.
๐ฎ Conclusion
This study underscores the remarkable capabilities of large language models in predicting mortality risk among dementia patients. The ability of GPT-4o to provide accurate estimates from discharge summaries could pave the way for more informed hospice referral decisions. As we look to the future, further prospective research is essential to validate these findings and explore the broader applications of AI in healthcare.
๐ฌ Your comments
What are your thoughts on the use of AI in predicting mortality risk for dementia patients? We would love to hear your insights! ๐ฌ Leave your comments below or connect with us on social media:
Predicting hospice eligibility among dementia patients using language models.
Abstract
INTRODUCTION: Alzheimer’s disease and related dementias (ADRD) are increasing in prevalence, and access to potential benefits of hospice care remains challenging. Large language models (LLMs), like GPT-4o, applied to electronic health records (EHRs) could support decisions by estimating mortality risk.
METHODS: We analyzed patients with ADRD diagnosis from two academic medical centers. GPT-4o was used to estimate 6-month mortality risk from discharge summaries without any retraining or preprocessing. We used Cox regression to assess associations between predictions and time to death.
RESULTS: Of 9872 individuals, 3563 (36%) died within 6 months. GPT-4o predictions stratified risk of death within 6 months (log-rank pย <ย 0.001, area under the curve [AUC]ย =ย 0.79); predictions were strongly associated with mortality in Cox regression models (adjusted hazard ratio [aHR]ย =ย 31.02 95% confidence interval [CI] 27.44-35.08, pย <ย 0.001) with similar results between sites.
DISCUSSION: GPT-4o can stratify mortality risk using routinely generated documentation, potentially facilitating hospice referral decisions, but more prospective work is needed.
HIGHLIGHTS: Large language models (LLMs) can estimate 6-month mortality in patients with dementia. GPT-4o estimates of mortality risk from discharge summaries were highly discriminative (area under the curve [AUC]ย =ย 0.79). Predictions may support hospice referral decisions.
Author: [‘McCoy TH’, ‘Perlis RH’]
Journal: Alzheimers Dement
Citation: McCoy TH and Perlis RH. Predicting hospice eligibility among dementia patients using language models. Predicting hospice eligibility among dementia patients using language models. 2025; 21:e70878. doi: 10.1002/alz.70878