Difficulties with prognostication prevent more patients with advanced dementia from receiving timely palliative support. The aim of this study is to develop and validate a prognostic model for 6-month and 1-year mortality in home-dwelling patients with advanced dementia.
Prospective cohort study.
The data set of 555 home-dwelling patients with dementia at Functional Assessment Staging Test stage 7 was split into derivation (n = 275) and validation (n = 280) cohorts.
Cox proportional hazards regression modeled survival in the derivation cohort using prognostic variables identified in univariate analysis. The model was validated internally and using 10-fold cross-validation. Area under the receiver operating characteristic curve measured the accuracy of the final model.
Four hundred nineteen (75.5%) patients died with a median follow-up of 47 days [interquartile range (IQR) 161]. Prognostic variables in the multivariate model included serum albumin level, dementia etiology, number of homecare admission criteria fulfilled, presence of moderate to severe chronic kidney disease, peripheral vascular disease, quality of life in late-stage dementia scores, housing type, and the Australian National Sub-Acute and Non-Acute Patient palliative care phase. The model was refined into a parsimonious 6-variable model (Palliative Support Dementia Model) consisting of age, dementia etiology, Functional Assessment Staging Test stage, Charlson Comorbidity Index scores, Australian National Sub-Acute and Non-Acute Patient palliative care phase, and 30-day readmission frequency for the prediction of 1-year mortality. The area under the receiver operating characteristic curve was 0.65 (95% confidence interval 0.59-0.70). Risk scores categorized patients into 3 prognostic groups, with a median survival of 175 days (IQR 365), 104 days (IQR 246), and 19 days (IQR 88) for the low-risk (0‒1 points), moderate-risk (2‒4), and high-risk (≥5) groups, respectively.
Palliative Support Dementia Model identifies patients at high risk of death in the next 1 year. The model produced consistent survival results across the derivation, validation, and cross-validation cohorts and will help healthcare providers identify patients with advanced dementia earlier for palliative care.

Copyright © 2020. Published by Elsevier Inc.