Background Multimorbidity, the co-existence of two or more long-term health conditions, is a public health concern due to its association with adverse health outcomes. Recent research has shown that people with multimorbidity including depression have particularly high service use and premature mortality. However, since existing research is largely based on cross-sectional studies, information is lacking about whether people develop depression before or after long-term physical health conditions. Our aim was to examine the association between the number of long-term physical health conditions and subsequent risk of depression.

Methods We included data from 113,007 UK Biobank participants with linked primary care data and without depression at baseline who had complete data available. We ascertained the number of long-term physical health conditions from participant self-report and linked primary care, hospital and cancer registry data at baseline and during follow-up. We identified depression during follow-up using primary care and hospital records. We performed Cox proportional hazards models to evaluate time to incident depression by number of physical health conditions at baseline and time-varying number of physical health conditions, sequentially adjusted for baseline sociodemographic and clinical characteristics.

Results Participants had between 0 and 19 physical health conditions at baseline and gained between 0 to 13 conditions during follow-up. During an average follow-up of 6.8 years, 4376 (3.9%) participants were diagnosed with depression. In fully adjusted models, compared to those with no physical morbidity at baseline, hazard ratios (HR) for depression increased with the number of physical health conditions at baseline with HRs (95% CI) of 1.24 (1.13, 1.38), 1.46 (1.32, 1.62), 1.73 (1.55, 1.93), 1.89 (1.67, 2.15), 2.31 (2.05, 2.61) for participants with 1, 2, 3, 4 and 5 or more physical health conditions at baseline, respectively. The associations were stronger in analyses using the number of morbidities as a time-varying covariate, with hazard ratios (95% CI) for depression of 1.35 (1.20, 1.52), 1.49 (1.32, 1.68), 1.83 (1.62, 2.07), 2.25 (1.97, 2.57) and 3.09 (2.73, 3.50) among those with 1, 2, 3, 4, and 5 or more morbidities, respectively.

Conclusion In a large prospective cohort study the presence of multiple long-term physical health conditions was strongly associated with subsequent depression in a dose-response relationship. Our study suggests that existing studies using a count of physical morbidities at baseline may have underestimated the association with risk of depression. A key limitation was that only complete cases were included. However, a multiple imputation is currently in process.