Study design and settings

This retrospective cohort study was conducted using data from September 2019 to October 2021. The data were provided by the Longevity Improvement & Fair Evidence (LIFE) Study, which is a research database project managed by Kyushu University (Fukuoka, Japan).15 The LIFE Study collects claims data from enrollees of Japan’s National Health Insurance System and Latter-Stage Older Persons Health Care System residing in participating municipalities. The National Health Insurance System is a public insurance programme for persons aged 0–74 years, and the Latter-Stage Older Persons Health Care System is a public insurance programme for all persons aged ≥75 years and persons aged 65–74 years with specific diseases. The claims data include recorded diagnoses based on the International Classification of Diseases, 10th Revision (ICD-10) codes, drug prescriptions and medical treatments during clinical encounters.

Study participants

The study participants comprised National Health Insurance System enrollees and Latter-Stage Older Persons Health Care System enrollees residing in Kobe City (Hyogo, Japan), which is a LIFE Study participant municipality. To comparatively evaluate the psychiatric sequelae of COVID-19 infection, we constructed a primary cohort and two control cohorts.

The primary cohort consisted of patients who had been hospitalised with a diagnosis of COVID-19 (ICD-10 codes: U071, U072, B34.2) between March 2020 and July 2021. Next, we constructed an influenza control cohort comprising patients who had been hospitalised with a diagnosis of influenza (ICD-10 codes: J09-11) between March 2020 and July 2021, as well as an RTI control cohort comprising patients who had been hospitalised with a diagnosis of any other RTI (J00-06, J12-18, J20-22) between March 2020 and July 2021. The use of patients with influenza and patients with other RTI as the control cohorts is similar to the method employed in a previous study.12 We chose this approach because patients with influenza or other RTI may experience acute symptoms similar to patients with COVID-19 and could also be susceptible to the same post-acute sequelae.

We limited the analysis to patients who did not die during the index hospitalisation. Patients with a diagnosis of influenza or other RTI were excluded from the primary cohort, patients with a diagnosis of COVID-19 or other RTI were excluded from the influenza control cohort and patients with a diagnosis of COVID-19 or influenza were excluded from the RTI control cohort. The patients were followed for 3 months and 6 months after the index infection. Using the claims data, we identified patients hospitalised between 1 March 2020 and 31 July 2021 for the 3-month follow-up (until 31 October 2021), and patients hospitalised between 1 March 2020 and 30 April 2021 for the 6-month follow-up (until 31 October 2021).

Outcome measures

The study outcome was the occurrence of psychiatric sequelae after COVID-19 infection, influenza infection or other RTI.

At present, there is no universally accepted definition for post-acute COVID-19 symptoms. WHO has proposed the following clinical case definition: ‘Post COVID-19 condition occurs in individuals with a history of probable or confirmed SARS-CoV-2 infection, usually 3 months from the onset of COVID-19 with symptoms and that last for at least 2 months and cannot be explained by an alternative diagnosis.’16 Other studies have also reported that a large proportion of patients still experience psychiatric symptoms 6 months after COVID-19 onset.9 14 Based on these reports, we examined the occurrence of psychiatric sequelae 1–3 months after the index infection (designated ‘acute psychiatric sequelae’) and 4–6 months after the index infection (designated ‘delayed psychiatric sequelae’).

The psychiatric sequelae were identified using ICD-10 codes and grouped into five categories: (1) mood/anxiety/psychotic disorder (ICD-10 codes: F20–29, F30–F39, F40–F48), (2) mood disorder (F30–F39), (3) anxiety disorder (F40–F48), (4) psychotic disorder (F20–F29), and (5) insomnia (F51.0 or G47.0). Patients who had received diagnoses of any of these psychiatric disorders 6 months before the index infection were excluded from the analysis. For the analysis of delayed psychiatric sequelae (4–6 months after the index infection), we also excluded patients who received diagnoses of any of the psychiatric disorders in the first 3 months after the index infection.

Statistical analysis

The patient characteristics were described using means with SD for continuous variables and numbers with percentages for categorical variables. To account for variations in patient characteristics, we acquired information on patient sex, age and comorbidities. Age was analysed using three categories (≤64 years, 65–74 years, and ≥75 years). These categories were selected because Japan’s health insurance system categorises persons aged 65–74 years and ≥75 years as ‘early elderly’ and ‘late elderly’, respectively.17 The following comorbidities were identified and included in the analysis: hypertension (ICD-10 codes: I10–I16), diabetes (E10–E14), chronic lower respiratory disease (J40–J47), heart disease (I20–I25, I30–I52), chronic kidney disease (N18, I12), chronic liver disease (K70, K72, K73, K74, K76.0, K76.1, K76.6, K76.8), stroke (I63), dementia (F01, F02, F03, G31.0, G31.83), cancer (C00–D49, C81–C96), and rheumatoid arthritis (M05, M06).

This analysis sought to compare the incidences of psychiatric sequelae after COVID-19 infection with the incidences after influenza infection or other RTI. To produce comparable estimates, we used propensity score matching to control for variations between the primary cohort and control cohorts.18 We first calculated the propensity scores for becoming infected with COVID-19 and influenza using multivariable logistic regression models that adjusted for patient sex, age and comorbidities. Patients with COVID-19 were then matched with patients with influenza using these propensity scores, and the incidences of psychiatric sequelae were compared. The same process was used to match the patients with COVID-19 with the patients with RTI. Matching was performed through 1:1 nearest neighbour matching without replacement using a calliper width of 0.2 SDs of the logit of the propensity score. Absolute standardised differences below 0.10 were regarded as showing good balance between the cohorts.

To evaluate the risk of psychiatric sequelae occurrence after COVID-19 infection, we constructed multivariable logistic regression models for each of the five psychiatric sequelae categories during 1–3 months (acute psychiatric sequelae) and 4–6 months (delayed psychiatric sequelae) after infection. These models included COVID-19 infection as the independent variable of interest, with influenza infection or other RTI as the reference category. The covariates included patient sex, age and comorbidities. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated.

We then conducted a sensitivity analysis to assess the robustness of our results. In Japanese claims data, patients may receive suspected diagnoses of psychiatric disorders. Therefore, such patients may not be undergoing any pharmacological treatment despite having recorded diagnoses. To provide a more stringent identification of the outcomes, we defined the psychiatric sequelae not only through ICD-10 codes but also through corresponding prescriptions of hypnotics/sedatives, anxiolytics, antipsychotics and other central nervous system agents. Furthermore, the inclusion of these prescriptions would enable us to focus on more severe psychiatric disorders that require pharmacological treatment. These drug prescriptions were identified using Japanese drug codes recorded in the claims data. Based on these newly defined outcomes, we recalculated the ORs and 95% CIs of psychiatric sequelae occurrence after COVID-19 infection compared with influenza or other RTI.

All statistical analyses were performed using Stata V.17 (Stata Corp, College Station, Texas, USA). Statistical significance (two sided) was set at 5%.