Sensor-Based Digital Biomarkers for Early Identification of Cognitive Frailty: A Systematic Review.

Area of Science:

• Gerontology
• Biomedical Engineering
• Digital Health

Background:

• Cognitive Frailty (CF), the combination of Physical Frailty (PF) and Mild Cognitive Impairment (MCI), is a significant health risk in aging populations.
• Early CF detection is challenging due to varied criteria and complex assessments.
• Sensor-based digital biomarkers offer a potential solution for objective and accessible CF assessment.

Purpose of the Study:

• To systematically review and synthesize evidence on sensor-based digital biomarkers for Cognitive Frailty (CF) detection.
• To identify the most effective sensors and digital biomarkers associated with CF.
• To evaluate the performance of machine learning models developed for CF prediction.

Main Methods:

• Systematic literature review across eight databases following PRISMA 2020 guidelines.
• Qualitative and quantitative synthesis of data from 20 eligible studies.
• Analysis of sensor types (e.g., IMUs), digital biomarkers (e.g., gait, physical activity, body composition, cardiovascular indices), and predictive models.

Main Results:

• Inertial Measurement Units (IMUs) were the most common sensors, primarily measuring gait and motor function.
• Gait features (Stride Length, Velocity) and physical activity (MVPA) showed significant associations with CF.
• Models for CF prediction reported AUCs up to 0.911 and pooled sensitivity/specificity of 0.81/0.80, indicating moderate-to-good discriminative ability.

Conclusions:

• Sensor-based digital biomarkers, particularly gait and physical activity metrics, show considerable potential for early Cognitive Frailty (CF) detection.
• Heterogeneity in study designs and definitions necessitates standardized protocols and rigorous validation for clinical translation.
• Further research is needed to refine sensor technology and modeling approaches for reliable CF assessment.