Population and data sources

We used a database recording all people tested for COVID-19 in NSW (the Notifiable Conditions Information Management System (NCIMS)) probabilistically linked to all hospitalisations (NSW Admitted Patient Data Collection; APDC) and death registry records in NSW. These data have been described previously [8]. Briefly the NCIMS is a statutory infectious diseases register that includes records all people tested for COVID-19 in NSW. Data recorded include sociodemographic details as well as the test type, date and result. For cases confirmed as positive according to nationally agreed case definitions [9], additional information is collected via interview regarding symptoms, hospitalisation and comorbidities. The NSW APDC records all inpatient care in NSW and includes coded information on the diagnosis, admission and discharge dates, and admission to ICU. The Death Registry includes information on fact and date of death. All deaths in people with a diagnosis of COVID-19 are assessed to determine if the death was caused by COVID-19 and this outcome is recorded in the NCIMS. For this study we had accessed records from NCIMS, NSW APDC, and the Death Registry to 5 October 2020.

We defined a retrospective cohort consisting of all people diagnosed with COVID-19 in NSW identified from the NCIMS between 1 January and 5 October 2020. “Severe disease” was a composite outcome of either hospitalisation or death from COVID-19 and “very severe disease” was either an ICU admission or death from COVID-19; therefore the “very severe disease” group was a subset of the “severe disease” group. As described previously [8], COVID-19 related hospitalisations and ICU admissions were identified based on linking a case with COVID-19 from NCIMS to a hospital or ICU admission in the 6-week period following the COVID-19 diagnosis. Only in-patient stays of at least one night in hospital were included. We also supplemented the linked records with information on hospitalisations in private hospitals from what was recorded in NCIMS as these may not have been captured in the linked data. COVID-19 related deaths were obtained from the NCIMS record.

Information on sociodemographic characteristics, smoking, pregnancy and selected co-morbidities was obtained from the NCIMS based on the initial interview and subsequent case interviews. Additional comorbidity information was obtained via linkage to hospitalisation records prior to the COVID-19 diagnosis, using the International Classification of Diseases Australian Modification 10 (ICD-10-AM) diagnosis code for specific illnesses. This study was approved by the UNSW Human Research Ethics Committee ({"type":"entrez-nucleotide","attrs":{"text":"HC200483","term_id":"283803172","term_text":"HC200483"}}HC200483). All methods were performed in accordance with relevant guidelines and regulations.