Machine learning identifies risk factors associated with long-term sick leave following COVID-19 in Danish population.

ABSTRACT

BACKGROUND: Post COVID-19 condition (PCC) can lead to considerable morbidity, including prolonged sick-leave. Identifying risk groups is important for informing interventions. We investigated heterogeneity in the effect of SARS-CoV-2 infection on long-term sick-leave and identified subgroups at higher risk. METHODS: We conducted a hybrid survey and register-based retrospective cohort study of Danish residents who tested positive for SARS-CoV-2 between November 2020 and February 2021 and a control group who tested negative, with no known history of SARS-CoV-2. We estimated the causal risk difference (RD) of long-term sick-leave due to PCC and used the causal forest method to identify individual-level heterogeneity in the effect of infection on sick-leave. Sick-leave was defined as >4 weeks of full-time sick-leave from 4 weeks to 9 months after the test. RESULTS: Here, in a cohort of 88,818 individuals, including 37,482 with a confirmed SARS-CoV-2 infection, the RD of long-term sick-leave is 3.3% (95% CI 3.1% to 3.6%). We observe a high degree of effect heterogeneity, with conditional RDs ranging from -3.4% to 13.7%. Age, high BMI, depression, and sex are the most important variables explaining heterogeneity. Among three-way interactions considered, females with high BMI and depression and persons aged 36-45 years with high BMI and depression have an absolute increase in risk of long-term sick-leave above 10%. CONCLUSIONS: Our study supports significant individual-level heterogeneity in the effect of SARS-CoV-2 infection on long-term sick-leave, with age, sex, high BMI, and depression identified as key factors. Efforts to curb the PCC burden should consider multimorbidity and individual-level risk.

The burden of post COVID-19 condition varies from one person to another due to individual characteristics such as age, sex, and having single- or multiple pre-existing conditions. Sick leave following initial SARS-CoV-2 infection is one way to quantify this burden. However, to what extent the combinations of these characteristics impact the risk of post-acute sick leave is not well understood. Here, using a machine learning method, we observe that persons infected with SARS-CoV-2 have an increased risk of taking long-term sick leave compared to persons with no history of infection. Age, high BMI, sex, and depression explained substantial effect variation on the risk of long-term sick leave after infection. This knowledge may be used to help inform patient-targeted interventions.