OP66 #Income trajectories and health: a latent class growth mixture model approach in understanding society (UK household longitudinal study). (September 2018)

• Record Type: Journal Article
• Authors: Moldovan, AL; Benzeval, M; Clarke, P
• Is Part Of: Journal of epidemiology and community health. Volume 72(2018)Supplement 1
• Page Start: A32
• Page End: A32
• Publication Date: 2018-09
• Subjects: income trajectories biomarkers; Public health -- Periodicals; Epidemiology -- Periodicals; 614.4
• DOI: 10.1136/jech-2018-SSMabstracts.65
• Languages: English
• ISSNs: 0143-005X

Abstract

Abstract : Background: Most studies examining income and health have used average income at one point in time or low income over time. These are global measures and cannot really capture change in income over time. In this study, we propose a novel approach to studying long-term income trajectories and how they influence health measured by biomarkers to capture different dimensions of health, which might be influenced by different income trajectories. To this end, we use Latent Class Growth Mixture Models (LCGMM) to identify the types of household-level income trajectories that are associated with cholesterol, glycated haemoglobin (HbA1c), fibrinogen, c-reactive protein (CRP), haemoglobin (Hgb), and Gamma-Glutamyl Transferase (GGT), in the UK population from 1991–2010. Methods: We use household income (quartiles/year), Waves 1–18 of the British Household Panel Study (a stratified random sample), and biomarkers from Wave 3 of the Nurse Health Assessment component of the UK Household Longitudinal Survey into which BHPS is now incorporated. We select individuals with at least nine waves of income data and biomarker information (n=1449). Biomarkers are used as logged continuous measures. We fit LCGMMs (estimator: Maximum Likelihood with robust S.E.) using MPlus 8.0. The Bayesian Information Criterion and the Bootstrapped Likelihood Ratio Test are used to assess model fit and determine the number of classes. We run LCGMM with biomarkers as distal outcomes, using the manual BCH procedure, to test equality of means across classes. We control for age, gender, education, employment, household type, self-rated health at baseline (Wave 1 of BHPS). We conduct sensitivity analyses to check whether results change when removing respondents using medications found to affect biomarker values. Results: Eight types of income trajectories are identified representing both stability and volatility: 'stable high' (11.8%); 'high-decreasing-medium' (12.8%); 'low-increasing-medium' (15.7%); 'medium-increasing-high' (11.6%); 'stable medium' (17.9%); 'stable low' (15.7%); 'high-medium-low' (3.7%); 'medium-low' (10.8%). After adjusting for confounders, we find that the group of individuals living in low-income households throughout the twenty years ('stable low') have a higher logged Total-HDL cholesterol mean ratio (LogM=0.865; M=∼2.63), compared to both the 'medium-increasing-high' (LogM=0.862; M=∼2.38; LogM diff.=0.103, S.E.=1.993) and 'high-medium-low' (LogM=0.862; M=∼2.37; LogM diff.=0.102, S.E.=1.969) groups. These differences are not significant when removing respondents using statins. Findings for other biomarkers will be presented. Conclusion: We find differences in cholesterol between: 1) persistent poverty and upward volatility; 2) persistent poverty and downward volatility. Our results suggest that persistent poverty is more harmful to household members, compared to fluctuations and eventual poverty. Some of our other health measures respond differently to income stability and variability.