A Time-to-Event Comparison of Immune and Endocrine Biomarkers and Latent Profiles in Hospitalisation: An Outcome-wide Approach.

BACKGROUND

Early identification of risk for hospitalisation is crucial to reducing public health burden. Immune and endocrine-related markers are robust indicators of disease in epidemiological studies, but their value has not been consistently established with severe disorders requiring hospitalisation. Patterning of biomarker expression through latent profile analysis (LPA), may improve predictive accuracy for clinical outcomes above individual biomarkers alone.

METHOD: Four biomarkers (C-reactive protein; fibrinogen; leukocytes; insulin growth-factor-1) measured in the English Longitudinal Study of Ageing (ELSA) in 2008 were linked to administrative data on hospitalisations obtained from Hospital Episode Statistics (HES). Hospitalisation for 12 disease classes was monitored from 2008-2018 (n=4,940). Analyses were adjusted for genetic predisposition and a wide set of confounders.

FINDINGS: There were 9,419 cases of hospitalisation over the 10-year follow-up period. LPA of the four biomarkers indicated a three-profile solution offered greatest parsimony, categorised as low-risk [52.43%]; moderate-risk [35.89%]; and high-risk [11.68%] inflammatory status. Profiles offered greater specificity than individual biomarkers with a risk gradient in hospitalisation for sleep, circulatory, endocrine, and genitourinary disorders, where the magnitude of associations was notably higher in the high-risk group. Profiles also identified risk for infection-related hospitalisation not identified by individual biomarkers alone (HR: high-versus-low-risk=1.38; 95%CI=1.08-1.78, p=0.011). No associations emerged in hospitalisation for digestive, nervous, or skin disorders.

INTERPRETATION: LPA enabled more precise risk-stratification and subgroup-specific analyses, with profiles better characterising health outcomes requiring hospitalisation than individual biomarkers.

FUNDING: Biotechnology and Biological Sciences Research Council (BBSRC); Economic and Social Research Council (ESRC); National Institute on Aging.