This was a multicenter retrospective cohort study. The study protocol was reviewed and approved by the Medical Ethics Committee of the First Affiliated Hospital of Air Force Medical University (Xijing Hospital) (Approval No.: KY20222118-F-1), and the requirement for informed consent was waived due to the retrospective nature of the study. The study was conducted in accordance with the STROBE guidelines.

Adult patients hospitalized with severe or critical COVID-19 at three tertiary general hospitals in Shaanxi Province between 1 December 2022 and 30 November 2023 were eligible for inclusion. The diagnostic criteria for severe or critical COVID-19 were based on the Diagnostic and Treatment Protocol for Novel Coronavirus Infection (Trial Version 10) issued by the National Health Commission of China (27). Severe COVID-19 was defined as meeting any of the following criteria: Respiratory rate of ≥30 breaths/min, oxygen saturation level of ≤93% at rest, or arterial partial pressure of oxygen to fraction of inspired oxygen ratio of ≤300 mmHg. Critical COVID-19 was defined as respiratory failure that necessitates mechanical ventilation, shock, or other organ failure requiring intensive care. Complete definitions are provided in Supplementary Table S1.

The inclusion criteria were as follows: (i) laboratory confirmation of SARS-CoV-2 infection by a positive nucleic acid or antigen test and (ii) fulfillment of the diagnostic criteria for severe or critical COVID-19. The exclusion criteria were as follows: (i) age <18 years, (ii) receipt of any antiviral drugs (including nirmatrelvir/ritonavir, azvudine, or molnupiravir) before the baseline date, and (iii) use of other antiviral drugs (including azvudine and molnupiravir) during the observation period after baseline.

Clinical data were retrospectively extracted from the electronic medical record systems of the three participating hospitals by three independent data collectors. To improve accuracy and consistency, all extracted data were reviewed and verified by two additional reviewers. The following variables were collected: (i) demographics, including sex, age, and smoking history; (ii) prior SARS-CoV-2 infection; (iii) comorbidities, including chronic lung disease, hypertension, heart disease, cerebrovascular disease, diabetes, chronic liver disease, chronic kidney disease, cancer, and immunocompromised status (immunodeficiency disorders or long-term use of glucocorticoids/immunosuppressants); (iv) disease course, including the date of COVID-19 symptom onset and the date of hospital admission; (v) treatments received before admission; (vi) disease severity, including changes in COVID-19 severity during hospitalization and the corresponding dates; (vii) inpatient antiviral therapy, including the name, dosage, and start and end dates of antiviral drugs administered during hospitalization; (viii) concomitant therapies during hospitalization, including glucocorticoids, antibiotics, tocilizumab, baricitinib, and anticoagulants; (ix) SARS-CoV-2 nucleic acid test results; and (x) clinical outcomes, including the date of discharge and the discharge status (improved, deceased, or other).

The exposure of interest was treatment with nirmatrelvir/ritonavir during the hospitalization period following a diagnosis of severe or critical COVID-19. Based on this exposure, patients were categorized into the nirmatrelvir/ritonavir group and the non-antiviral group.

The baseline date was defined separately for each group. In the nirmatrelvir/ritonavir group, the baseline date was the date of the first dose administered after patients met the criteria for severe or critical COVID-19 during hospitalization (the median time from meeting the criteria to receiving the first dose was 1.0 days, with an interquartile range [IQR] of 0.0–1.0 days). In the non-antiviral group, the baseline date was defined as the date when patients first met the criteria for severe or critical COVID-19 during hospitalization. The end of the observation period was defined as the earliest of hospital discharge, death, or 28 days after the baseline date.

The primary outcome was 28-day mortality, which refers to death from any cause within 28 days from the baseline date while the patient was still hospitalized. Secondary outcomes included in-hospital mortality, defined as death from any cause during hospitalization regardless of its timing in relation to the baseline date, and the length of hospital stay after the baseline date.

Epidemiological data indicate that the mortality rate of patients with severe or critical COVID-19 is approximately 30% (11). According to a study by Yang et al. (23), the 28-day mortality rates were 41.9% in the nirmatrelvir/ritonavir group and 77.6% in the control group, resulting in a mortality rate ratio of 0.54. Based on these estimates, we performed a sample size calculation with β set at 0.2 (power = 0.8) and a two-sided α level of 0.05. The required sample size was estimated to be 143 patients per group, resulting in a total of 286 participants. As the data were obtained from hospitalized patients, no loss to follow-up was observed.

Continuous variables were assessed for normality using the Shapiro–Wilk test. As all continuous variables were non-normally distributed, they are presented as medians (IQRs), and between-group comparisons were performed using the Mann–Whitney U test. Categorical variables are presented as numbers (percentages) and were compared using the chi-squared test or Fisher’s exact test, as appropriate. Survival probabilities were visualized using Kaplan–Meier curves, and differences between the groups were evaluated using the log-rank test. Hazard ratios (HRs) and adjusted hazard ratios (aHRs), with 95% confidence intervals (CIs), were estimated using Cox proportional hazards models. To control for potential confounders, multivariable Cox regression models were constructed by including covariates that showed a significant association (p < 0.05) in univariate analyses. Candidate covariates included age, sex, smoking history, prior SARS-CoV-2 infection, comorbidities, baseline COVID-19 severity, onset-to-baseline time (defined as the interval from symptom onset to baseline), and concomitant treatments (glucocorticoids, antibiotics, tocilizumab, baricitinib, and anticoagulants) (28). Subgroup analyses were conducted for 28-day mortality and in-hospital mortality according to age (<65 or ≥65 years), sex, presence of chronic lung disease, hypertension, heart disease, diabetes, baseline COVID-19 severity (severe or critical), and onset-to-baseline time (≤5 or >5 days). Interaction effects within the subgroups were formally tested. In addition, to further explain the finding of prolonged hospitalization, we performed a competing risks analysis by estimating cumulative incidence functions for death and discharge, with between-group comparisons conducted using Gray’s test. We also conducted a severity-stratified analysis of hospitalization duration, comparing the groups using the Mann–Whitney U test within the severe and critical subgroups. All statistical analyses were performed using SPSS version 27.0, and figures were generated using R version 4.3.2. A two-sided p-value of <0.05 was considered statistically significant.

A total of 789 patients with severe or critical COVID-19 were screened for eligibility. Of these, 403 were excluded, including 4 patients aged <18 years, 205 who received antiviral therapy before baseline, and 194 who received other antiviral agents (azvudine or molnupiravir) after baseline. Ultimately, 386 patients were included in the analysis, with 173 in the nirmatrelvir/ritonavir group and 213 in the non-antiviral group (Figure 1). Baseline characteristics are shown in Table 1. Overall, the cohort had a median age of 73.0 years (IQR: 61.0–83.0 years) and consisted predominantly of men (69.7%). Comorbidities were common (94.0%), with heart disease (64.2%), hypertension (56.0%), and diabetes (36.3%) being the most frequent. At baseline, 129 patients (33.4%) had critical COVID-19. The two groups were generally well balanced across measured baseline characteristics, including demographics, prior SARS-CoV-2 infection, comorbidities, baseline COVID-19 severity, onset-to-baseline time, and concomitant treatments (Table 1).

Within 28 days, 45 patients (11.7%) died. The 28-day mortality rate was significantly lower in the nirmatrelvir/ritonavir group than in the non-antiviral group (6.9% vs. 15.5%; p = 0.002; Table 2). The Kaplan–Meier curve for 28-day mortality is shown in Figure 2A.

Cox regression analyses evaluating the association between nirmatrelvir/ritonavir and clinical outcomes are summarized in Table 3. In univariate analysis, nirmatrelvir/ritonavir treatment was associated with a significantly lower risk of 28-day mortality compared to no antiviral therapy (HR = 0.372, 95% CI: 0.192–0.721; p = 0.003), and the full univariate results are provided in Supplementary Table S2. After adjustment for age, heart disease, diabetes, baseline COVID-19 severity, and concomitant tocilizumab use, multivariable Cox regression confirmed that nirmatrelvir/ritonavir remained associated with a significantly reduced risk of 28-day mortality (aHR = 0.346, 95% CI: 0.175–0.687; p = 0.002).

As shown in Table 2, 48 patients (12.4%) died during hospitalization. In-hospital mortality was significantly lower in the nirmatrelvir/ritonavir group compared to the non-antiviral group (8.1% vs. 16.0%; p = 0.003). The corresponding Kaplan–Meier curve is presented in Figure 2B.

In univariate Cox regression, nirmatrelvir/ritonavir was associated with a significantly lower risk of in-hospital mortality (HR = 0.402, 95% CI: 0.215–0.751; p = 0.004), with the full results shown in Supplementary Table S3. After adjusting for age, heart disease, diabetes, baseline COVID-19 severity, and concomitant tocilizumab use, multivariable analysis remained consistent and showed a significant reduction in mortality risk associated with nirmatrelvir/ritonavir (aHR = 0.374, 95% CI: 0.196–0.716; p = 0.003).

The median post-baseline hospitalization duration was significantly longer in the nirmatrelvir/ritonavir group than in the non-antiviral group [11.0 (IQR: 6.0–19.0) days vs. 9.0 (IQR: 5.0–14.0) days; p = 0.019] (Table 2). To assess whether longer hospitalization reflected a survivor effect or delayed recovery, we performed two supplemental analyses. First, a competing risks analysis (Supplementary Figure S1) showed a significantly lower cumulative incidence of death in the nirmatrelvir/ritonavir group (Gray’s test, p = 0.010), whereas the cumulative incidence of discharge did not differ between the groups (Gray’s test, p = 0.680), suggesting that the treatment reduced mortality without slowing discharge among survivors. Second, in severity-stratified analyses, the longer hospitalization duration was observed only in patients with critical illness [14.5 (IQR: 6.0–29.8) days vs. 9.0 (IQR: 4.0–15.0) days; p = 0.014] and not in those with severe illness [10.0 (IQR: 6.0–16.0) days vs. 9.0 (IQR: 6.0–13.0) days; p = 0.276; Supplementary Table S4]. Taken together, these findings suggest that the longer length of stay is most likely related to the mortality benefit of nirmatrelvir/ritonavir, as it reduced early deaths without delaying recovery among survivors.

Subgroup analyses for 28-day mortality are shown in Figure 3. Overall, the beneficial association of nirmatrelvir/ritonavir with mortality was consistent across most predefined subgroups. Notably, the point estimates suggested a substantial reduction in mortality risk among patients aged ≥65 years (HR = 0.418, 95% CI: 0.211–0.827), non-smokers (HR = 0.344, 95% CI: 0.167–0.707), those without chronic lung disease (HR = 0.238, 95% CI: 0.104–0.545), those without hypertension (HR = 0.261, 95% CI: 0.084–0.813), those with critical COVID-19 at baseline (HR = 0.330, 95% CI: 0.133–0.818), and those with an onset-to-baseline time of ≤5 days (HR = 0.180, 95% CI: 0.052–0.618). A significant interaction was observed between the treatment effect and chronic lung disease status (P for interaction < 0.05), suggesting that the treatment effect may differ according to this factor, with the benefit appearing more pronounced in patients without chronic lung disease. No significant interactions were identified for the other subgroup factors.

Similarly, for in-hospital mortality (Figure 4), nirmatrelvir/ritonavir was associated with a reduced risk in the same subgroups. However, for this endpoint, no significant treatment-by-subgroup interactions were detected.

Taken together, these subgroup findings suggest that the survival benefit of nirmatrelvir/ritonavir was particularly evident in older patients (≥65 years) and in those without certain comorbidities, particularly chronic lung disease. In addition, the association was maintained in critically ill patients, and earlier initiation of treatment (within 5 days of symptom onset) was associated with a more pronounced benefit.