Study design and data sources

A cohort design was applied to longitudinal, nationwide, register-based data on all people who were living in Denmark between 27 February 2020 and 15 October 2021. The start date marked the first confirmed SARS-CoV-2 case in Denmark^{Reference Hansen, Michlmayr, Gubbels, Mølbak and Ethelberg19} and the end of follow-up was defined by the recency of data. Information on all residents living in Denmark was obtained from the Danish Civil Registration System.^{Reference Pedersen20} Using the unique personal identification number assigned to all individuals, a linkage to data from the following registers was facilitated: the Danish Microbiology Database,^{Reference Voldstedlund, Haarh and Mølbak21} the Population Education Register,^{Reference Jensen and Rasmussen22} the National Patient Register,^{Reference Lynge, Sandegaard and Rebolj23} the Psychiatric Central Research Register,^{Reference Mors, Perto and Mortensen24} the National Prescription Registry,^{Reference Kildemoes, Sorensen and Hallas25} the Homeless Register^{Reference Nielsen, Hjorthøj, Erlangsen and Nordentoft11} and the Central Criminal Register.²⁶ The National Patient Register and the Psychiatric Central Research Register provided dates and diagnoses from for all medical in-patient contacts since 1994 and all psychiatric in-patient contacts since 1969 respectively. Out-patient and emergency department contacts have been recorded since 1995 in both registers. Sociodemographic information, including socioeconomic status, was obtained from Statistics Denmark.

Study population

All individuals who lived in Denmark and were 15 years of age or older, i.e. were born on 15 October 2006 or earlier, were included. The following subgroups, which have been suggested as being vulnerable to SARS-CoV-2 infection, were identified: people who had (a) primary school or less education, (b) chronic medical conditions, (c) disability pension, (d) a history of mental disorder, including severe mental illness, (e) a history of a substance use disorder, (f) history of homelessness and (g) history of imprisonment. Information on highest obtained education was collected from the Population Education Register and based on the status in September 2019 (Supplementary Table 1, available at https://dx.doi.org/10.1192/bjp.2022.194). Persons who had been approved for disability pension prior to 2020 were identified through their socioeconomic status. People with chronic medical conditions were identified using diagnoses for 31 medical conditions recorded according to ICD-10 in the National Patient Register^{Reference Momen, Plana-Ripoll, Agerbo, Benros, Børglum and Christensen27} and which had been diagnosed prior to 27 February 2020. In addition, records of medication prescribed for chronic medical conditions were obtained from the National Prescription Registry (Supplementary Table 2).^{Reference Mors, Perto and Mortensen24,Reference Schmidt, Schmidt, Sandegaard, Ehrenstein, Pedersen and Sørensen28} Severe mental disorder was defined as having been diagnosed with schizophrenia, bipolar disorder or depressive disorder at some point after 1969 and recorded in the Psychiatric Central Research Register. Persons with substance use disorders were identified using the same data source as for mental disorders, as well as data on treatment from the Registry of Drug Abusers Undergoing Treatment for the period 1996–2018 and supplemented with information from the National Registry of Alcohol Treatment during 2006–2018. In addition, we used information on drugs used to treat addictive disorders from the National Prescription Registry since 2015. Homeless people were identified using a previously developed algorithm, which primarily was based on the Homeless Register, with nationwide information on homeless shelter contacts in Denmark during 1999–2020.^{Reference Nilsson, Laursen, Osler, Hjorthøj, Benros and Ethelberg13} Persons who at some point during 1991–2020 had been imprisoned were identified from the Central Criminal Register.

Ascertainment of infection with SARS-CoV-2

Individuals with a SARS-CoV-2 infection were identified as having had a PCR-confirmed SARS-CoV-2 infection. PCR tests were conducted as throat swabs in any of the free-of-charge test stations in Denmark and results were retrieved from the national Danish Microbiology Database, with the last update on 15 October 2021.^{Reference Hansen, Michlmayr, Gubbels, Mølbak and Ethelberg19,Reference Voldstedlund, Haarh and Mølbak21} Individuals were considered as infected from the date of a first record of being SARS-CoV-2 positive until the end of the follow-up. In addition, we considered people who had been admitted to a general hospital for more than 12 h and within 14 days of the date of a positive PCR test as having a SARS-CoV-2 infection-related hospital admission.^{Reference Nilsson, Laursen, Osler, Hjorthøj, Benros and Ethelberg14}

Outcome

Owing to the fact that self-harm events are under-recorded in Danish registries,^{Reference Morthorst, Soegaard, Nordentoft and Erlangsen29} we opted to examine a broader algorithm of hospital contacts that might be considered probable self-harm. This category covers self-harm regardless of intent, and includes poisoning by drugs and biological and non-medical substances, as well as lesions to the hand and forearm. Self-harm episodes were identified as presentations to either psychiatric or general hospitals, including emergency departments, and recorded in the National Patient Registry with one of the following ICD diagnoses: ICD-10 X60–X84, or where the reason for contact was coded as being suicide attempt (ALCC04) or non-suicidal self-harm (ALCC05). In addition, the following combinations of ICD diagnoses were included: a main diagnosis of a mental disorder (ICD-10 F00–F99) together with one of the following subdiagnoses: S51, S55, S59, S61, S65, S69 (cutting by sharp objects), T36–T50 (poisoning by pharmaceuticals), T52–T60 (poisoning by non-pharmaceuticals) as well as all admissions with a main diagnosis of T39, T40 (poisoning by mild analgesics; except T40.1), T42, T43 and T58 (poisoning by opioids, psychotropics or carbon monoxide). Individuals who died on the same or subsequent day as the record of the self-harm were considered as having died by suicide and excluded. The same measure has been used previously.^{Reference Morthorst, Soegaard, Nordentoft and Erlangsen29}

As a sensitivity analysis, we restricted the outcome to consist only of self-harm episodes that were exclusively identified as diagnoses (ICD-10 X60–X84) or where the reason for contact was coded as being a suicide attempt (ALCC04).

Follow-up

Participants were followed from 27 February 2020 to date of first subsequent self-harm episode, migration out of the country, death or 15 October 2021, whichever occurred first.

Statistical analyses

Poisson regression analysis was used to calculate adjusted incidence rate ratios (aIRRs) with 95% confidence intervals (CIs) per 100 000 person-years, where rates of self-harm for individuals with a positive PCR test for SARS-CoV-2 infection were calculated relative to individuals not recorded with a positive test. The outcome was defined as a first incident of a self-harm episode since onset of the pandemic. When assessing rates in relation to SARS-CoV-2 infection-related hospital admission, models were further stratified by history of mental disorder and other hospital admissions (general and psychiatric combined). In addition, self-harm rates of high-risk groups were analysed in models with high-risk status and PCR-confirmed SARS-CoV-2 infection. Analyses were adjusted for calendar time (months), age (5-year groups), gender (female, male) and country of origin (Denmark, other high-income country or low- and middle-income country). Homelessness and imprisonment were handled as time-dependent variables, whereas all other covariates were fixed as of 27 February 2020. The association between SARS-CoV-2 and self-harm was tested in a sensitivity analysis where the outcome was restricted to a more conservative definition of self-harm, as described above.

Cumulative incidences of self-harm were estimated for individuals followed from the date of a first PCR-confirmed SARS-CoV-2 infection and for the subsequent 9 months in a subsample where each case was matched on age and gender to two individuals from the general population who had not been recorded with a positive test on the date of matching. Using the Aalen–Johansen estimator, we calculated the cumulative incidence of self-harm, while accounting for competing risks from death and emigration.

The statistical analyses were performed using SAS software for Windows (version 9.4).

Ethical considerations

The project was approved by the Danish Data Protection Agency (Capital Region of Denmark: P-2020-439), Statistics Denmark and the Danish Health Data Authority. Data access was agreed by Statistics Denmark and the Danish Health Data Authority. Approval by an ethics committee and written informed consent were not required for this register-based project, according to Danish regulations. All data were de-identified and not recognisable at an individual level.