Disease prevalence, health-related and socio-demographic factors in the GCAT cohort. A comparison with the general population of Catalonia

Abstract

Background: Population-based cohorts play a key role in epidemiological studies. However, it is known that volunteer cohorts include a healthy volunteer bias. Assessment and characterization of this bias is needed to extrapolate results to the general population. Here, we assess the bias of the population-based cohort GCAT, encompassing 20 000 adult participants from Catalonia with electronic health record data. The aim of this study is to compare the GCAT cohort with its age-matching Catalan population, to assess their representativeness, as well as determining the weights to make results generalisable. Methods: Statistical comparisons until 2019 in multiple variables across sociodemographic, lifestyle, diseases and medication domains were performed by stratified analysis with Fisher's exact test and t-test. Electronic health records of Catalonia (SIDIAP), and registers from the statistics institute of Catalonia (IDESCAT) and Spain (INE) were used to make the comparisons. We generated weights accounting for sociodemographic, lifestyle and multimorbidity factors. Results: GCAT cohort is enriched in women and younger individuals, with higher socioeconomic status, more health conscious and healthier in terms of mortality and chronic disease prevalence. We have shown that this bias can be corrected with weighting techniques, providing a more representative sample of the general population. Conclusions: The application of multidomain weights, encompassing not only sociodemographic aspects, but also lifestyle and health-related variables, has effectively diminished the observed bias in disease prevalence estimates within the GCAT cohort. This correction has led to an enhancement of the cohort's representativeness, rendering it more akin to the general population of Catalonia.

Decoding depression by exploring the exposome-genome edge amidst COVID-19 lockdown

“…While the assumption of potential causality is reasonable since exposures precede the outcome, caution is warranted due to the possibility of bidirectional relationships between variables. The presence of healthy bias 54 and limited representation of ancestry 55 in the study sample may affect the generalizability of the findings to broader populations, however, this does not invalidate the relationships found between exposures and depression outcomes. Finally, the limited sample size may affect the statistical power, particularly in interaction analyses.…”