We undertook a national retrospective cohort study, comparing clinical outcomes in those treated for COVID-19 in Scotland by treatment and across time-periods when different sub-lineages were prevalent.

The study cohort were all individuals treated for COVID-19 between December 21, 2021, and September 26, 2022. For this analysis, the drugs of interest were molnupiravir, nirmaltrelvir + ritonavir, remdesivir, sotrovimab, sarilumab, and tocilizumab. Casirivimab + imdevimab was also licensed for use in Scotland at the same time, but it was excluded from this analysis due to low uptake across the country following reports of lack of efficacy against Omicron sub-lineages14. Baricitinib was also excluded as it was not captured in the study data sources18.

As a person could be treated for multiple infections, records were separated into estimated infection-treatment episodes of at most 40 days in duration. Within an episode, an individual may still have been treated with multiple medications, either sequentially if they did not respond to the first treatment option or at the same time if they were at especially high risk and/or had severe symptom presentation. Cases in which the first administration or prescription of each therapy were initiated within three days of each other were classified as combination treatments; all other episodes were categorised by the first therapy given.

Patients were sub-grouped according to whether they were treated in the community or in a hospital setting. Group 1 patients were defined as those treated for acute COVID-19, during a hospital admission of at least one night’s duration, with COVID-19 as the primary cause. Group 2 patients were defined as those treated during a hospital admission, of at least one night’s duration, without COVID-19 as the primary cause. Group 3 patients were treated outside of an acute hospital admission (or during day visits, for medication administration). Finally, Group 4 patients were those treated during a currently uncoded hospital admission of at least one night’s duration, and thus with insufficient data to classify them into Group 1 or 2. As sarilumab and tocilizumab were only indicated for patients with severe acute COVID-19 requiring admission, treatment(s) for which an admission could not be identified were excluded, as this was a likely case of treatment in a specialist inpatient unit (such as cancer or maternity departments) for which no data were available.

Information regarding the prescription/administration of any of the drugs of interest was captured from multiple data sources, at the Health Board (Scottish regional authority for health care service delivery) level18. First, data were purposely collected through spreadsheets (henceforth Public Health Scotland Order, or PHSO), with weekly updates provided directly by the regional Health Boards (who manage their territories hospitals, district nursing services and healthcare planning). At least one report was received from 13/14 Health Boards (Supplementary Fig. 2) between December 21, 2021, and September 26, 2022 (with different end points per Health Board in the PHSO data, as listed in Supplementary Methods). Second, data on drug exposure were available for six Health Boards from the Hospital Electronic Prescribing and Medicines Administration System (HEPMA). Finally, data for one Health Board (NHS Lothian) were extracted from the prescribing information system (PIS)19. For all data sources, prescriptions had to be associated with a valid Community Health Index (CHI; Scotland’s unique patient identifier number) number to enable linkage to other sources; as such, 79% of all prescriptions dispensed were retained for analysis.

We used data from the Scotland-wide Early Pandemic Evaluation and Enhanced Surveillance of COVID-19 (EAVE II) platform20,21, as described in Supplementary Methods. Comorbidities were estimated from inpatient admissions and medical procedures, and primary care consultations and prescriptions, with a five-year look-back period. Four sources were used to estimate the incidence of outcomes.

Clinical outcomes estimated were inpatient admissions (and re-admissions) of at least one night’s duration, ICU admissions (of any duration), and deaths, both all-cause and specifically for COVID-19, within 28 days of treatment initiation. The incidence risk of each outcome was estimated using the denominator of those with at least 28 days of follow-up in the relevant data source, or an outcome within 28 days. As an exception, however, for COVID-19 inpatient admissions, an extended minimum follow-up period was used of 42 days (28 days x1.5). The outcome was still censored at 28 days, but only those with an event within 28 days or no event within 42 days were included. This was due to the lag in SMR01 reporting of approximately 6 weeks, based on discharge date and the time required for clinical coding to take place. This more stringent follow-up requirement reduced the risk of classifying someone as having no outcome when it was simply yet to be reported, but without excluding valid data, particularly on more recently emerging variants and sub-lineages. Additionally, those in Groups 1, 2 and 4 who were yet to be discharged from hospital within 28 days of treatment initiation were excluded from hospital admission analyses. Finally, we estimated any acute COVID-19 event as any inpatient hospital admission, ICU admission, or death within 28 days.

The high-risk conditions estimated were chemotherapy, radiotherapy, blood cancer, respiratory cancer, cirrhosis, chronic kidney disease (CKD stage 3 + ), prescription of immunosuppressants, a neurological condition (Parkinson’s disease, motor neurone disease, multiple sclerosis or cerebral palsy), rheumatoid arthritis or systemic lupus erythematosus, a solid organ transplant, a stem cell transplant, a bone marrow transplant, HIV/AIDS, sickle cell disease, Down’s syndrome, and splenectomy22.

Patient demographic data were extracted from primary care registries: age, sex, Scottish Index of Multiple Deprivation (SIMD), and the Urban-Rurality index. The number of COVID-19 vaccinations prior to treatment was extracted. Finally, the first positive reverse-transcription polymerase chain reaction (RT-PCR) test, lateral flow test (LFT), or inpatient admission with primary cause as COVID-19 in the 28 days preceding treatment initiation was selected as the first date of diagnosis to enable the time between diagnosis and treatment to be estimated.

Univarible logistic regression was used to assess risk factors (high-risk comorbidities and patient characteristics) for severe COVID-19 outcomes (hospital admissions, ICU admissions, and death) within 28 days of treatment initiation, in Group 3 patients. A stratified logistic regression was also conducted, to compare the adjusted odds ratio (aOR) of severe COVID-19 outcomes within 28 days of treatment initiation across time periods according to the most prevalent COVID-19 sub-lineage, in Group 3 patients treated with sotrovimab, nirmaltrelvir + ritonavir, or molnupiravir.

Patient and public contributors were involved in the design and interpretation of this study. Details of their involvement are presented in Supplementary Notes.

The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist for cohort studies is presented in Supplementary Table 1.

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.