Abstract
ABSTRACT
Background
Older adults in real‐world Chinese integrative‐care settings often present with uneven patterns of functional limitation, cognitive impairment, and clinical vulnerability. Whether routinely collected geriatric and Traditional Chinese Medicine (TCM) constitution records can be organized into clinically interpretable profiles remains insufficiently described.
Methods
We conducted a cross‐sectional analysis of 402 adults aged 60 years or older assessed during 2024–2025 in three Shanghai institutions: a hospital TCM department, a nursing home, and an elderly care center. Of 435 screened records, 33 (7.6%) were excluded due to missing data. Four routine continuous variables were analyzed: Barthel Index, education‐adjusted Montreal Cognitive Assessment (MoCA), a study‐specific tumor risk score, and a study‐specific infection risk score. Standardized variables were summarized by principal component analysis (PCA; PC1 and PC2, cumulative variance 70.8%) and K‐means clustering. TCM constitution and concurrent care disposition were examined as external descriptive variables and were not used to derive clusters.
Results
Participants had a mean age of 78.0 ± 9.6 years; 206 (51.2%) were women. Among candidate solutions (
k
= 2–6), the gap statistic and silhouette coefficient both favored
k
= 2 (gap = 0.152, SE = 0.016; silhouette = 0.287), but a three‐profile solution (silhouette = 0.252; gap = 0.118, SE = 0.019) was retained as an exploratory descriptive summary based on prespecified clinical interpretability criteria. The three profiles were:
Functionally Preserved Profile
(
n
= 149, 37.1%; Barthel 76.9 ± 14.8, MoCA 18.2 ± 3.9, tumor risk 9.6 ± 4.5, infection risk 25.0 ± 11.8),
Cognitive‐Functional Vulnerability with Low Tumor‐Risk Profile
(
n
= 101, 25.1%; Barthel 47.0 ± 18.2, MoCA 12.0 ± 4.6, tumor risk 4.9 ± 3.1, infection risk 34.3 ± 13.5), and
Multidomain High‐Risk Profile
(
n
= 152, 37.8%; Barthel 37.8 ± 19.8, MoCA 12.0 ± 5.1, tumor risk 15.2 ± 5.6, infection risk 46.8 ± 15.2). The Cognitive‐Functional Vulnerability profile had the lowest tumor risk score among the three profiles while its MoCA score was identical to that of the Multidomain High‐Risk Profile, confirming that the profiles represent distinct combinations of domain‐specific vulnerabilities rather than a single ordinal gradient. A preserved‐function/impaired‐cognition subgroup was identified in 73 participants (18.2%). Among nine TCM constitution labels, Yang deficiency differed across profiles after Bonferroni correction (
p
= 0.003).
Conclusions
Routine multidimensional assessment data can describe clinically recognizable vulnerability profiles among older adults in Chinese integrative‐care settings. The three‐profile solution represents an exploratory descriptive summary rather than a statistically optimal classification (silhouette coefficient = 0.252, indicating weak cluster separation; gap statistic favored
k
= 2). The most actionable observation was the functional‐cognitive dissociation subgroup, which may be missed if function alone is assessed. TCM constitution contributed descriptive context but did not independently define the profiles. These findings require external, prospective validation before clinical deployment.