Assessing mental health from registry data: what is the best proxy?

Abstract

Abstract Background Medical registries frequently underestimate the prevalence of health problems compared with surveys. This study aimed to determine the registry variables that can serve as a proxy for variables studied in a mental health survey. Methods Prevalences of depressive symptoms and antidepressant use from the 2018 Belgian Health Interview Survey (HIS), stratified by sex and age, were compared with same-year prevalences from INTEGO, a Belgian primary care registry. Participants aged 15 and above were included. We assessed correlation and agreement using Spearman's rho (SR), the intraclass correlation coefficient (ICC) and the limits of agreement (LOAs). Results HIS questions about depressive symptoms were compared with the following variables from INTEGO: symptom codes (SR 0.763, ICC 0.026), diagnosis codes (SR 0.653, ICC 0.195), free text (SR 0.653, ICC 0.322), antidepressant prescriptions (SR 0.793, ICC 0.391) and the combinations symptom + diagnosis codes (SR 0.640, ICC 0.270) and symptom + diagnosis codes + free text (SR 0.653, ICC 0.386). LOAs varied from [-0.120, 0.021] for antidepressants to [-0.015, 0.061] for the full combination. HIS questions about antidepressant use were compared with prescription frequencies in INTEGO of at least once (SR 0.899, ICC 0.586) and at least three times yearly (SR 0.895, ICC 0.835). Corresponding LOAs were [-0.093, -0.008] and [-0.053, 0.007]. Conclusions Correlation between the HIS and INTEGO was high, agreement fair to poor. Agreement increased by combining certain variables, by including free text, or by increasing the prescription frequency to resemble chronic use. Prevalences from INTEGO were mostly underestimates. A considerate choice of variables and prescription chronicity is needed to accurately use a registry as a surveillance tool for mental health. Key messages • The external validity of medical registries can be poor, especially compared with survey data. • Researchers should be aware of the choice of variables when using registry data as a proxy for population measures.