A generalizable biopsychosocial risk score for stratifying disease vulnerability in healthy populations

Abstract

Prospective risk stratification of future adverse health outcomes before clinical onset among ostensibly healthy individuals is an important goal of preventive and precision medicine, yet current tools remain largely disease-specific and do not provide an integrated measure of health risk for early prevention and risk stratification. We developed the Risk Score for Disease Vulnerability (RS4DV) based on 85 accessible biopsychosocial measures, which was constructed using a Light Gradient Boosting Machine developed in a UK Biobank discovery cohort ( n = 391,193) and evaluated in a held-out UK Biobank test cohort ( n = 90,864). Its prospective risk-stratification value was subsequently evaluated among held-out participants free of baseline diagnoses ( n = 35,193). Over more than a decade of follow-up, baseline RS4DV stratified long-term health outcomes; each one-unit increase in RS4DV was associated with disease progression (HR = 2.60, 95% CI: 2.51–2.70) and all-cause mortality (HR = 4.03, 95% CI: 3.68–4.41). Multi-omics analyses identified genomic, proteomic, and neuroimaging correlates, supporting biological relevance across organizational levels. For practical feasibility, we developed a six-item version of RS4DV using routinely accessible measures. The simplified model showed moderate predictive performance, high concordance with the full model, and cross-cohort transferability in the 1970 British Birth Cohort under zero-shot evaluation. RS4DV provides a practical framework for prospective disease-risk stratification from routine biopsychosocial measures.