Incidence rates of death were expressed as cases per 100,000 person-days. We used Cox proportional-hazards regression to calculate the hazard ratio for death and its 95% confidence interval (CI). We found no violation of proportional-hazards assumption. Survival was calculated using the Kaplan–Meier method. Survival time was right-censored at 28 days of follow-up. Cox regression models were adjusted for age using a five-knot restricted cubic spline fitting for age29. Analyses within each age group (20 to 39, 40 to 49, 50 to 59, 60 to 69, 70 to 79, and ≥ 80 years) were adjusted for age as a continuous variable. We tested for interactions between diabetes and age, diabetes and sex, and diabetes and type of patient care (outpatient or inpatient). The trends for the hazard ratios across age groups were tested using weighted linear regression. The probability weights were obtained from the inverse of the variance of the risk estimates. In the eligible population of patients, no predictor included in our regression models had more than 0.31% missing data. Thus, data were not imputed30,31. A complete-case analysis was performed. There were no missing data on age, sex, date of patient evaluation, date of symptoms onset, date of death, or type of patient (outpatient or inpatient).

We conducted three sensitivity analyses to assess the robustness of our findings: (1) full models with further adjustment for pneumonia, admission to intensive care unit, intubation, and time from symptoms onset to the date of patient evaluation; (2) multilevel mixed-effect survival regression models to assess the possible effect of geographical differences on our risk estimates32; and (3) comparison of hazard ratios from analysis restricted to patients who were evaluated and notified before and after August 1 to address the possible influence of changes to the definition of suspected viral respiratory disease23,27. We conducted stratified analysis according to age and sex, in outpatients and inpatients. Given the very large population, our study had statistical power to perform a robust stratified analysis in six age groups, among outpatients and inpatients. We used the log-rank test to compare unadjusted survival curves. All p values were two-sided. All analyses were performed using Stata 14 (StataCorp LP, TX).

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.