Impact of non-participation bias due to psychiatric illness on mortality and cardiovascular event estimates: a Danish longitudinal population study

Abstract

Objectives The impact of non-participation due to psychiatric illness on study outcomes in general population studies is insufficiently investigated. Here, we investigate the mental health bias in a population study and the potential impact on estimates of cardiovascular morbidity and overall survival.Design Data were retrieved from nationwide registries.Setting The Danish General Suburban Population Study (GESUS), a cross-sectional community study conducted in Naestved Municipality, Denmark, from 2010 to 2013.Participants 49 707 subjects invited to participate in GESUS.Main outcome measures Factors related to non-participation were examined using multivariable logistic regression and time-to-event data using Cox proportional hazards models.Results Of 21 203 (43%) participants, 823 (3.9%) had a psychiatric diagnosis. Of 28 504 non-participants, 2453 (8.6%) had a psychiatric diagnosis (OR for non-participation 1.84 (95% CI 1.69 to 2.00)). The most under-represented psychiatric disorders in participants were organic mental disorders (5.76 (3.90 to 8.48)), substance abuse (3.12 (2.14 to 4.54)) and schizophrenia (3.12 (2.33 to 4.18)). Overall, more non-participants used psychotropic drugs than participants (1.26 (1.21 to 1.31)), and psychiatric non-participants had higher psychiatric hospital service utilisation than psychiatric participants. Compared with non-psychiatric participants in a 5-year follow-up, psychiatric non-participants had higher rates of cardiovascular events (HR 2.30 (2.07 to 2.56)) and all-cause mortality (3.37 (3.01 to 3.78)) than non-psychiatric non-participants (1.65 (1.48 to 1.83) and 2.26 (2.02 to 2.54), respectively) and psychiatric participants (1.39 (1.21 to 1.59) and 1.23 (1.05 to 1.44), respectively), pinteraction<0.0001 for both outcomes.Conclusions This study demonstrates a considerable non-participation bias due to psychiatric illness in a general population health study, potentially leading to distorted estimates of somatic morbidity and mortality. Strategies for better-representing individuals with psychiatric illnesses in population health studies are needed.