CoChat

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

Open access

Published version of a preprint

The DOI matches its Crossref record; 61 of the 66 references it cites were found in a registry or an index; 18 of the 42 sources its claims cite were checked for support.

Abstract

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.

Overview

CoChat Summary

## TL;DR

The paper builds XGBoost models on NACC data to predict dementia patient mortality at 1-, 3-, 5-, and 10-year thresholds and finds that a parsimonious set of nine clinical features can achieve strong performance. It also shows that mortality predictors differ by dementia type, with clustering patterns such as vascular dementia with depression and Lewy body dementia with frontotemporal lobar degeneration.

## Key findings

  • Using 45,275 unique participants and 163,782 visit records from the U.S. National Alzheimer’s Coordinating Center, the authors developed mortality models that achieved AUC-ROC scores of over 0.82 for all 1-, 3-, 5-, and 10-year thresholds when using nine features.
  • The simpler two-feature models using only age and standard global CDR achieved AUC-ROC scores of over 0.76 at all four survival thresholds.
  • The nine parsimonious features were: age, independence, personal care, Trail Making Test Part B total seconds, difficulty with stove/heating water, sex, years smoked, Trail Making Test Part B commission errors, and years of education.
  • In dementia-type-specific 5-year models, all eight dementia-type models achieved AUC-ROC of over 0.79; the no-dementia model had the highest AUC-ROC at 0.873 (95% CI: 0.859–0.879).
  • Hierarchical clustering grouped dementia types into four clusters: vascular dementia with depression; FTLD with LBD; AD with other dementia; and no dementia with unknown.

## Methods

  • The study used longitudinal NACC data from 2005–2021, with the raw dataset containing 163,792 visits and 1,061 variables; after preprocessing and exclusions, the cohort analyzed included 45,275 unique participants.
  • The authors ran Kaplan–Meier survival analyses stratified by global clinical dementia rating and dementia type, then trained XGBoost models for mortality prediction at 1-, 3-, 5-, and 10-year thresholds.
  • They handled missingness by removing variables with over 40% missing values and imputing the remaining features with MICE; model performance was evaluated with training/internal-test splits plus separate date-based external test sets, and feature importance was interpreted with SHAP/TreeSHAP.

## Limitations

  • The paper states that high missingness led to preliminary elimination of many features, and that some variables were only available at certain Alzheimer’s Disease Centers; this may limit feature coverage and generalizability across sites.
  • The paper states that excluding patients lost to follow-up biases the deceased group toward shorter survival and may overestimate AUC values for longer survival thresholds.
  • Evident from the text: the models do not account for dementia progression dynamics or treatment effects, and they rely mainly on clinical/demographic variables rather than genetic or multi-omics data.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s43856-024-00437-7

Cites

61 of 61 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Schematic diagram of the study workflow and survival analyses using the NACC dataset. A Depiction of the study workflow from data pre-processing, model training, and testing, to analyses of the machine-learning models. The survival analysis results in the NACC datasets are visualized by the Kaplan–Meier estimator curves separated by B dementia type and C standard global clinical dementia rating (CDR), where the shaded area around each line represents a 95% pointwise confidence interval.
Fig. 1 Schematic diagram of the study workflow and survival analyses using the NACC dataset. A Depiction of the study workflow from data pre-processing, model training, and testing, to analyses of the machine-learning models. The survival analysis results in the NACC datasets are visualized by the Kaplan–Meier estimator curves separated by B dementia type and C standard global clinical dementia rating (CDR), where the shaded area around each line represents a 95% pointwise confidence interval.CC BY · PubMed Central
Fig. 2 Receiver operating characteristic (ROC) curves and feature importance plots of the multi-factorial models predicting survival of NACC individuals at different time points. A ROC curves of the 1-year, 3-year, 5-year, and 10-year survival models. B Feature importance as illustrated by the SHapley Additive exPlanations (SHAP) beeswarm plots of the 1-year, 3-year, 5-year, and 10-year survival models. The SHAP values on the x-axis have log-odds units and represent the impact of the feature on the model output. For each feature, each individual patient is represented by a single point, and the x-position of a point represents the impact of that feature on that patient. The color of a point corresponds to the patient’s value for that feature along a continuous scale from low to high.
Fig. 2 Receiver operating characteristic (ROC) curves and feature importance plots of the multi-factorial models predicting survival of NACC individuals at different time points. A ROC curves of the 1-year, 3-year, 5-year, and 10-year survival models. B Feature importance as illustrated by the SHapley Additive exPlanations (SHAP) beeswarm plots of the 1-year, 3-year, 5-year, and 10-year survival models. The SHAP values on the x-axis have log-odds units and represent the impact of the feature on the model output. For each feature, each individual patient is represented by a single point, and the x-position of a point represents the impact of that feature on that patient. The color of a point corresponds to the patient’s value for that feature along a continuous scale from low to high.CC BY · PubMed Central
Fig. 3 Comparison of features predictive of patient mortality across individual dementia type-specific models. A Clustered heatmap of the top features across dementia types. Only features with a normalized SHapley Additive exPlanations (SHAP) value greater than 2.5 (explaining at least 2.5% of the prediction) in any given dementia type-specific model were included. The maximum normalized SHAP value for the clustered heatmap was set to 10 so that color contrasts were more discernable. The “No Dementia” category corresponds to patients receiving a primary etiologic diagnosis of “Not applicable, not cognitively impaired.” The “Unknown” category corresponds to patients receiving a primary etiologic diagnosis of “Missing/unknown.” “The “Other” category corresponds to patients receiving a primary etiologic diagnosis of “Cognitive impairment for other specified reasons (i.e., written-in values).” B SHAP beeswarm plots of six of the eight dementia type-specific models, excluding “Unknown” and “Other”. Features in the beeswarm plots are ranked by mean absolute SHAP value. The SHAP value distribution of the top ten most important features in each dementia type-specific model is displayed.
Fig. 3 Comparison of features predictive of patient mortality across individual dementia type-specific models. A Clustered heatmap of the top features across dementia types. Only features with a normalized SHapley Additive exPlanations (SHAP) value greater than 2.5 (explaining at least 2.5% of the prediction) in any given dementia type-specific model were included. The maximum normalized SHAP value for the clustered heatmap was set to 10 so that color contrasts were more discernable. The “No Dementia” category corresponds to patients receiving a primary etiologic diagnosis of “Not applicable, not cognitively impaired.” The “Unknown” category corresponds to patients receiving a primary etiologic diagnosis of “Missing/unknown.” “The “Other” category corresponds to patients receiving a primary etiologic diagnosis of “Cognitive impairment for other specified reasons (i.e., written-in values).” B SHAP beeswarm plots of six of the eight dementia type-specific models, excluding “Unknown” and “Other”. Features in the beeswarm plots are ranked by mean absolute SHAP value. The SHAP value distribution of the top ten most important features in each dementia type-specific model is displayed.CC BY · PubMed Central

Verification

Claims the paper makes

11 Valid · 2 Not valid · 10 Not checked · 89 Own results and interpretations

51 more claims · 1 with their cited sources checked

Create a free account to see all 112 claims and to verify the 3 references still open.

See all claims — free account
  • It may state two separate claims as one. 44
  • Probably not a substantial claim. 15
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

1 of 1

Citations

Cites

Show all 61 references

Research works printed but not linked: 3

  1. Lundberg, S. M. & Lee, S.-I. A Unified Approach to Interpreting Model Predictions. In: Advances in Neural Information Processing Systems 30 (Curran Associates, Inc., 2017).
  2. Snoek, J., Larochelle, H. & Adams, R. P. Practical Bayesian Optimization of Machine Learning Algorithms. in Advances in Neural Information Processing Systems 25 (Curran Associates, Inc., 2012).
  3. Wilson, J., Hutter, F. & Deisenroth, M. Maximizing acquisition functions for Bayesian optimization. in Advances in Neural Information Processing Systems 31 (Curran Associates, Inc., 2018).

Metadata

Authors
Zhang, J., Song, L., Miller, Z., Chan, K. C. G., & Huang, K.-L.
Affiliations
  • Tisch Cancer Institute (US)
  • The University of Queensland (AU)
  • University of Washington (US)
Corresponding author
Huang, Kuan-lin
Venue
Communications Medicine
Published
2024-02-28
Volume
4
Issue
1
Pages
23-23
ISSN
2730-664X (ISSN-L)
OpenAlex
W4392233569
Cited by
20
References
66
Citations by year
2026: 8 · 2025: 11 · 2024: 1
FWCI
5.85
Type
journal article
Language
en
Open access
Licence
Locations
  • Communications Medicine (publishedVersion, cc-by)
  • Research Square (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • PubMed Central (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Funders
U.S. Department of Health & Human Services | NIH | National Institute of General Medical Sciences (R35GM138113)
Topics
  • Health Sciences › Medicine › Psychiatry and Mental health › Dementia and Cognitive Impairment Research
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
  • Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
Keywords
dementia, mortality prediction, dementia types, vascular dementia, dementia with Lewy bodies, frontotemporal lobar degeneration, depression, neuropsychological assessment, risk stratification, XGBoost, National Alzheimer's Coordinating Center

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Machine learning models identify predictive features of patient mortality across dementia types. https://app.cochat.ai/papers/10.1038/s43856-024-00437-7
The paper
Zhang, J., Song, L., Miller, Z., Chan, K. C. G., & Huang, K.-L. (2024). Machine learning models identify predictive features of patient mortality across dementia types. Communications Medicine. https://doi.org/10.1038/s43856-024-00437-7

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.