A systematic review of the current prognostic tools used to predict the survival length of people living with advanced dementia

Background:

Dementia is a progressive neurodegenerative condition, yet it is frequently under-recognised as terminal, contributing to delayed or inadequate end-of-life care. Accurate prognostication in advanced dementia is essential to support timely palliative interventions. However, prognostic uncertainty remains high, and existing tools are inconsistently validated.

Aim:

To identify and review current prognostic models designed for survival estimation of people with advanced dementia.
Design:
Systematic review prospectively registered on PROSPERO, supported by a narrative synthesis of key findings.

Data sources:

MEDLINE, Embase, CINAHL, and PsycINFO were searched from inception to April 2025, alongside grey literature sources. Eligible studies reported development or validation of prognostic models estimating survival in advanced dementia. Risk of bias was appraised using PROBAST, and certainty of evidence using GRADE guidelines.

Results:

Six prognostic tools were identified across seven studies, including 225,553 individuals with advanced dementia across community, hospital, and nursing home settings. Key prognostic indicators included demographic factors, comorbidities, and functional dependency. Reported AUC values ranged from 0.55 to 0.76. The PRO-MADE model demonstrated the highest discrimination (AUC 0.76) and was the only tool to report detailed calibration metrics, though external validation showed reduced performance. Five studies demonstrated high risk of bias due to methodological limitations and lack of external validation.

Conclusions:

Current prognostic tools for advanced dementia show limited and inconsistent accuracy and should not be used in isolation. Further research should focus on robust external validation across various countries and healthcare settings before one model can be recommended for widespread application.