Machine learning models identify predictive features of patient mortality across dementia types

• Published in: Communications medicine Vol. 4; no. 1; pp. 23 - 13
• Main Authors: Zhang, Jimmy, Song, Luo, Miller, Zachary, Chan, Kwun C. G., Huang, Kuan-lin
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 28.02.2024; Springer Nature B.V; Nature Portfolio
• Subjects: 692/617/375/132; 692/617/375/132/1283; 692/700/1750; Datasets; Dementia; Machine learning; Medicine; Medicine & Public Health; Missing data; Mortality; Neuropathology; Neuropsychology; Patients; Survival analysis; Variables; United States > US
• ISSN: 2730-664X; 2730-664X
• DOI: 10.1038/s43856-024-00437-7

Background Dementia care is challenging due to the divergent trajectories in disease progression and outcomes. Predictive models are needed to flag patients at risk of near-term mortality and identify factors contributing to mortality risk across different dementia types. Methods Here, we developed machine-learning models predicting dementia patient mortality at four different survival thresholds using a dataset of 45,275 unique participants and 163,782 visit records from the U.S. National Alzheimer’s Coordinating Center (NACC). We built multi-factorial XGBoost models using a small set of mortality predictors and conducted stratified analyses with dementiatype-specific models. Results Our models achieved an area under the receiver operating characteristic curve (AUC-ROC) of over 0.82 utilizing nine parsimonious features for all 1-, 3-, 5-, and 10-year thresholds. The trained models mainly consisted of dementia-related predictors such as specific neuropsychological tests and were minimally affected by other age-related causes of death, e.g., stroke and cardiovascular conditions. Notably, stratified analyses revealed shared and distinct predictors of mortality across eight dementia types. Unsupervised clustering of mortality predictors grouped vascular dementia with depression and Lewy body dementia with frontotemporal lobar dementia. Conclusions This study demonstrates the feasibility of flagging dementia patients at risk of mortality for personalized clinical management. Parsimonious machine-learning models can be used to predict dementia patient mortality with a limited set of clinical features, and dementiatype-specific models can be applied to heterogeneous dementia patient populations. Plain language summary Dementia has emerged as a major cause of death in societies with increasingly aging populations. However, predicting the exact timing of death in dementia cases is challenging, due to variations in the gradual process where cognitive decline interferes with the body’s normal functions. In our study, we build machine-learning models to predict whether a patient diagnosed with dementia will survive or die within 1, 3, 5, or 10 years. We found that the prediction models can work well across patients from different parts of the US and across patients with different types of dementia. The key predictive factor was the information that is already used to diagnose and stage dementia, such as the results of memory tests. Interestingly, broader risk factors related to other causes of death, such as heart conditions, were less significant for predicting death in dementia patients. The ability of these models to identify dementia patients at a heightened risk of mortality could aid clinical practices, potentially allowing for earlier interventions and tailored treatment strategies to improve patient outcomes. Zhang, Song et al. develop machine-learning models to predict dementia patient mortality at one-, three-, five-, and ten-year thresholds. The models identify dementia patients at risk of mortality using a limited set of dementia-related predictors, and the stratified analyses reveal shared and distinct predictors of mortality across dementia types.