JMIR Publications

ABSTRACT

Background:

Traditional Chinese Medicine (TCM) constitution is a structured health-state taxonomy used in preventive care, but its relationship with routinely collected health examination data, disease-related markers, and longitudinal change remains difficult to interpret in clinical informatics settings.

Objective:

This study aimed to develop and evaluate a longitudinal clinical informatics framework for translating TCM constitution into computable, explainable, and temporally validated digital phenotypes using routine health examination records.

Objective:

This study aimed to develop and evaluate a longitudinal clinical informatics framework for translating TCM constitution into computable, explainable, and temporally validated digital phenotypes using routine health examination records.

Methods:

We conducted a retrospective longitudinal analysis of 47,417 examination-constitution records from 11,355 older adults examined between 2017 and 2026. Baseline analyses used the first available record per participant, and longitudinal analyses used 32,648 adjacent annual visit pairs. We distinguished recorded disease-history fields from examination-derived disease-related markers. The analytic framework included bidirectional disease-constitution mapping, multimarker burden modeling, lagged next-visit risk modeling, constitution-state transition analysis, and temporal validation of routine-examination-based prediction models trained on 2017-2023 records, evaluated in 2024, and tested in 2025-2026. Models were interpreted using SHAP and evaluated using discrimination, calibration, decision-curve analysis, quantitative sensitivity analysis, and exploratory uncertainty summaries.

Results:

Phlegm-dampness constitution showed the clearest digital signature. At baseline, it was characterized by higher BMI, larger waist circumference, more frequent obesity, higher blood pressure, higher fasting glucose, and higher triglycerides than balanced constitution. It was associated with higher cardiometabolic marker burden than balanced constitution (incidence rate ratio 1.65, 95% CI 1.60-1.70). In lagged annual models adjusted for current marker status and covariates, current phlegm-dampness predicted next-visit abnormal abdominal ultrasound, cardiometabolic risk clustering, diabetes-related markers, proteinuria, kidney impairment or proteinuria, dyslipidemia, and glucose abnormalities. Constitution states were dynamic, with a crude adjacent-visit group change rate of 35.6%. In temporal testing, an XGBoost model identified phlegm-dampness label presence with an area under the receiver operating characteristic curve of 0.935 and a precision-recall area under the curve of 0.898. SHAP analysis identified waist circumference, BMI, weight, age, triglycerides, and other hematologic and biochemical markers as leading predictors. Calibration and sensitivity analyses indicated temporal shift and substantial dependence on anthropometric predictors.

Conclusions:

Routine health examination data can represent TCM constitution as a measurable, longitudinal, and explainable digital phenotype. In this older adult cohort, phlegm-dampness constitution was best characterized as an adiposity-centered cardiometabolic phenotype associated with multimarker burden, next-visit disease-related markers, and predictable state transitions. The framework may help clinical informatics research incorporate culturally embedded health taxonomies, but external validation, local recalibration, workflow evaluation, and patient-facing explanation are needed before clinical deployment. Clinical Trial: not applicable