Study design and setting
All patients with COVID-19 in Hong Kong were identified through the Clinical Data Analysis and Reporting System (CDARS), which is a territory-wide administrative database managed by the Hospital Authority (HA), the statutory body operating 42 public hospitals and 120 clinics for the population sized 7.4 million in Hong Kong20. This database collects anonymized demographic, clinical and service data, and has been used extensively for large population studies21,22. The Hospital Authority Hong Kong West Cluster/The University of Hong Kong institutional review board (UW 22-258) approved and waived the requirement for obtaining patient informed consent, and all methods were performed in accordance with relevant guidelines and regulations.
Selection of participants
Patients who fulfilled the following inclusion criteria were included: (1) a diagnosis of COVID-19 under the International Classification of Diseases, Ninth Revision, Clinical Modified (ICD-9-CM); (2) admitted to hospitals under the Hospital Authority during February 22 to April 15, 2022. We defined the index date as the date of admission of an episode coded for COVID-19.
Our exposure of interest was Molnupiravir and Nirmatrelvir-Ritonavir among hospitalized COVID-19 patients, with an increased risk of deterioration, including old-age and chronic disease patients. We included hospitalized patients who were aged ≥ 60 years; or younger patients aged ≥ 18 with at least one chronic disease. Patients with prescription records of Molnupiravir or Nirmatrelvir-Ritonavir within 4 days of hospital admission were allocated into the Molnupiravir and Nirmatrelvir-Ritonavir treatment groups; patients who received both antivirals were excluded from this study. Considering the delayed access to antivirals in early stage of rolling out, patients received Molnupiravir or Nirmatrelvir-Ritonavir later than 4 days from the index date were excluded.
The primary outcomes were 28-days all-cause mortality and respiratory mortality, as detailed in Supplementary Table 1. Secondary outcomes were circulatory shock, respiratory failure, acute kidney injury, coagulopathy, acute liver impairment, and combined organ dysfunction, a composite outcome defined as presence of any one of these other secondary outcomes. Acute organ dysfunction is defined by vasopressor initiation, mechanical ventilation, elevation of lactate, or changes in total bilirubin, platelets or creatinine relative to specified baseline values23, and the unprecedented prescription of dexamethasone, as directed by the clinical guideline24. We defined these outcomes in reference to a sepsis surveillance toolkit, published by the US Centers for Disease Control and Prevention (CDC), which is a validated approach based on administrative data to facilitate health care facilities to monitor the incidence and outcomes of patients who developed sepsis according to the Sepsis-3 criteria, known as an Adult Sepsis Event (ASE)25. ASE is defined as (1) presumed serious infection, signified by obtained microbiological examination (for example, blood culture) and ≥ 4 consecutive days of antimicrobials (or up until 1 day before mortality, discharge to hospice, transfer to another acute care hospital, or transition to comfort measures) starting within 2 calendar days of when blood cultures were obtained, plus (2) evidence of concurrent organ dysfunction, signified by any of 6 binary indicators of cardiovascular, pulmonary, renal, hepatic, coagulation, or perfusion dysfunction. This approach for the identification of sepsis patients in Hong Kong data from CDARS has been validated, with high sensitivity (0.93), high specificity (0.86) and area under receiver operating characteristic curve (0.90)26.
The follow-up of each patient was commenced from the index date, defined as the date of hospital admission, for 28 days or until date of mortality. For secondary outcomes, follow-up time was defined as time from the index date to time of event, or loss of follow up.
Baseline data was collected for each patient including sex, age, and socioeconomic status. The socioeconomic status of each patient was determined by the Social Deprivation Index (SDI) of their area of residence27. SDI is derived from six variables to describe the conditions of social deprivation for each residential district: the proportions of the population with unemployment, monthly household income < US$250, no schooling at all, one-person household, never-married status, and subtenancy. The data comes from population census published by the government. Each of these six variables had significant factor loading for a specific principal factor, and all of them are deemed to be representative indicators of social disadvantage in the published literature and in the setting of the Hong Kong population. SDI for each district was calculated by taking the average of these six selected variables28.
Baseline data was also collected for comorbidities (Charlson Comorbidity Index (CCI), diabetes mellitus, hypertension, stroke, congestive heart failure, atrial fibrillation, schizophrenia, cirrhosis, depression, chronic kidney disease, rheumatoid arthritis, obesity and alcohol abuse) and chronic medication use (antiplatelets and anticoagulants, ACE inhibitors and angiotensin receptor blockers, beta blockers, calcium channel blockers, diuretics, statins systemic corticosteroids, bronchodilators and inhaled corticosteroids, cancer drugs, and rheumatological drugs). The differences in patient characteristics between treatment groups were expressed in terms of standardized mean differences (SMD), where covariates with SMD < 0.2 were considered balanced. The International Classification of Diseases, Ninth Revision (ICD-9) codes used for classification of the comorbidities are available in Supplementary Table 2.
Baseline characteristics were expressed as the mean (Standard deviation [SD]) for continuous variables and frequency (%) for categorical variables. Inverse probability treatment weighting (IPTW) based on propensity scores using the aforementioned variables was used to construct a weighted cohort of patients to address potential indication bias due to nonrandomized allocation of patients to the treatment group. IPTW balances baseline characteristics in comparator groups by weighting each patient by the inverse probability of receiving the treatment. Hazard ratios (HR) for the association of Nirmatrelvir-Ritonavir or Molnupiravir use and the study outcomes were estimated over the entire follow-up with weighted Cox proportional hazards regression using weights obtained by IPTW.
Alternatively, given the small event rate, we repeated all analyses utilizing bootstrapping methods followed by sampling with replacement to obtain precise estimates of HRs, incidence rate differences (IRD) and their 95% confidence intervals. Details regarding the bootstrapping method are documented in Supplementary Fig. 1.
Prespecified subgroup analyses were conducted by stratifying the study population by age groups (5-years increments from < 60 to ≥ 80), sex, SDI, and presence of diabetes. Further analyses were conducted to investigate the number of mortalities stratified by individual components of organ dysfunction. The number of organ dysfunctions and length of stay were also stratified by survival status at the end of follow-up.
All significance tests were two-tailed and considered significant when the P value was less than 0.05. All analyses were conducted using RStudio version 1.4.171729, using the following packages: tidymodels was used to obtain bootstrapped dataframes30, for the subsequent computation of hazard ratios using the survival package31, and weighted incidence rate differences via the fmsb package32,33. The package ‘WeightIt’ was used to compute the propensity scores and inverse probability weights32. The package comorbidity was used to compute the CCI34.
Ethics approval and consent to participate
This study was approved, and the requirement for obtaining patient informed consent was waived, by the Hospital Authority Hong Kong West Cluster/The University of Hong Kong institutional review board (UW 22-258).