# Data-driven frailty and reserve phenotypes in older outpatients: a cluster analysis of Comprehensive Geriatric Assessment

## TL;DR

This study uses data analysis to identify four distinct health profiles in older adults based on their frailty and resilience levels, using a comprehensive geriatric assessment.

## Contribution

The novel contribution is the identification of four distinct frailty and reserve phenotypes using PCA and cluster analysis of CGA data in older outpatients.

## Key findings

- PCA identified four components explaining 73.5% of variance, including a Frailty Axis and Reserve Capacity.
- Cluster analysis revealed four distinct phenotypes: Vulnerable Low-Complexity, Resilient High-Reserve, Resilient Frailty, and Globally Frail.
- Educational attainment was linked to clinical reserve, while comorbidities and lab markers were associated with frailty.

## Abstract

The progressive aging of the population represents a critical public health challenge. Within this context, the management of frailty has emerged as a central priority in geriatric care, with Comprehensive Geriatric Assessment (CGA) widely recognized as the gold-standard tool for its evaluation. This study aimed to stratify a large cohort of older adults using a multidimensional approach based on CGA, employing Principal Component Analysis (PCA) and Cluster Analysis to identify distinct phenotypic profiles.

A cross-sectional study was conducted on 1055 outpatients aged ≥65 years, assessed at the Geriatric Outpatient Service of the University of Cagliari between 2020 and 2024. All participants underwent a CGA. PCA was performed on selected CGA variables, and the resulting components were used for a hierarchical cluster analysis. Post-hoc comparisons between clusters were conducted using ANOVA, Chi-squared or Fisher tests, as appropriate.

PCA identified four principal components explaining 73.5% of total variance. The first component represented a Frailty Axis, while the second reflected Reserve Capacity. Cluster analysis based on these two axes revealed four distinct phenotypes: (I) Vulnerable Low-Complexity (younger patients with low comorbidity but significant cognitive, functional and nutritional impairments), (II) Resilient High-Reserve (low comorbidity with preserved cognitive, functional, and nutritional status and high educational attainment), (III) Resilient Frailty (high comorbidity, functional and nutritional deficits but preserved cognitive reserve) and (IV) Globally Frail (older patients with high comorbidity with multidomain impairments).

These findings demonstrate the ability of CGA, combined with PCA-informed clustering, to identify clinically meaningful frailty and resilience patterns in older adults. The study highlights the role of educational attainment as a key factor contributing to clinical reserve; conversely, it showed that demographic characteristics, laboratory markers, and comorbidities align with frailty.