Laboratory and clinical frailty indices identify distinct patterns of vulnerability across the adult age range in the Novara cohort study

Background Frailty may involve biological dysregulation that precedes, accompanies, or occurs partly independently of clinically expressed deficits. Laboratory and clinical frailty indices quantify vulnerability through different sets of deficits, but whether their divergence reflects measurement characteristics or distinct manifestations of ageing-related vulnerability remains unclear. We investigated laboratory and clinical frailty within the same population, focusing on individual-level discordance. Methods Within the Novara Cohort Study, 73 laboratory analytes were evaluated in 1,186 community-dwelling adults aged 18–89 years using repeated elastic-net modelling across multiple imputed datasets and cross-validation partitions. Twenty-three analytes selected consistently across all model fits constituted the final laboratory signature associated with the clinical Frailty Index (FI). The resulting laboratory Frailty Index (FI-Lab) was then calculated independently using prespecified age- and sex-specific reference intervals. Clinical FI and FI-Lab were compared in 1,160 participants at population and individual levels; discordance, reduced versus full laboratory panels, and sensitivity to coefficient weighting were examined. Results Clinical FI and FI-Lab were weakly correlated (Spearman ρ = 0.11) and showed low individual-level agreement (quadratic-weighted κ = 0.11), despite partly overlapping population-level distributions. Discordance increased with age: older age was associated with relatively higher FI-Lab, whereas female sex, depressive symptoms, concern about falling, and multimorbidity were associated with relatively higher clinical FI. Clinical FI showed stronger associations with psychosocial and functional characteristics, whereas FI-Lab was more closely associated with cardiometabolic factors. The 23-marker FI-Lab preserved the health-burden associations of the full 76-analyte index and showed similarly modest discrimination of physical and cognitive impairment. Coefficient weighting increased correlation with clinical FI but did not eliminate individual-level discordance. Conclusion A parsimonious frailty-related laboratory signature revealed substantial individual-level discordance from clinically expressed frailty, with distinct demographic and phenotypic correlates. Persistence of this discordance across alternative scoring approaches, together with the similar performance of the reduced and full laboratory panels, suggests that it may reflect more than scoring differences alone. Clinical-laboratory discordance may therefore represent an informative phenotype for longitudinal investigation to determine whether divergent profiles reflect different stages of frailty or distinct trajectories of ageing-related vulnerability.