Diabetes and Mortality Among 1.6 Million Adult Patients Screened for SARS-CoV-2 in Mexico

Background: Whether diabetes is associated with COVID-19-related mortality remains unclear. Methods: In this retrospective case-series study we examined the risk of death associated with self-reported diabetes in symptomatic adult patients with laboratory-confirmed COVID-19 who were identified through the System of Epidemiological Surveillance of Viral Respiratory Disease in Mexico from January 1 through November 4, 2020. Survival time was right-censored at 28 days of follow-up. Results: Among 757,210 patients with COVID-19 included in the study, 120,476 (16%) had diabetes and 80,616 died. Patients with diabetes had a 49% higher relative risk of death than those without diabetes (Cox proportional-hazard ratio; 1.49 (95% confidence interval [CI], 1.47-1.52), adjusting for age, sex, smoking habit, obesity, hypertension, immunodeficiency, and cardiovascular, pulmonary, and chronic renal disease. The relative risk of death associated with diabetes decreased with age (P=0.004). The hazard ratios were 1.66 (1.58-1.74) in outpatients and 1.14 (1.12-1.16) in hospitalized patients. The 28-day survival for inpatients with and without diabetes was, respectively, 73.5% and 85.2% for patients 20-39 years of age; 66.6% and 75.9% for patients 40-49 years of age; 59.4% and 66.5% for patients 50-59 years of age; 50.1% and 54.6% for patients 60-69 years of age; 42.7% and 44.6% for patients 70-79 years of age; and 38.4% and 39.0% for patients 80 years of age or older. In patients without COVID-19 (878,840), the adjusted hazard ratio for mortality was 1.78 (1.73-1.84). Conclusion: In symptomatic adult patients with COVID-19 in Mexico, diabetes was associated with higher mortality. This association decreased with age.

INTRODUCTION 1 0 1 present study, a COVID-19 case was defined as a patient with suspected viral respiratory disease 1 0 2 who had SARS-CoV-2 infection confirmed by reverse-transcriptase-polymerase-chain-reaction 1 0 3 test. Patients who tested negative for SARS-CoV-2 were referred to as non-COVID-19 cases,  Incidence rates of death were expressed as cases per 100,000 person-days. We used Cox 1 0 7 proportional-hazards regression to calculate the hazard ratio and 95% confidence intervals (CIs) 1 0 8 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 for mortality. There was no violation of proportional-hazards assumption. Survival time was 1 0 9 right-censored at 28 days of follow-up from the admission date (date of the patient's visit).

0
Multivariate analyses included adjustment for age, sex, smoking habit, obesity, hypertension, 1 1 1 cardiovascular disease, chronic obstructive pulmonary disease, asthma, chronic kidney disease, 1 1 2 and immunodeficiency. These variables were chosen based on our judgment as they have been 1 1 3 associated with the severity of COVID-19 or mortality. 13,26,27 Cox regression models were 1 1 4 adjusted for age using a five-knot restricted cubic spline fitting for age. 28 Analyses within each as a continuous variable. We tested for interactions between diabetes and age, diabetes and sex, 1 1 7 and diabetes and type of patient care (outpatient vs. inpatient). The trends for the hazard ratios 1 1 8 across age groups were tested using weighted linear regression. The probability weights were 1 1 9 obtained from the inverse of the variance of the risk estimates. Since missing data among 1 2 0 predictors included in our regression models represented less than 0.75%, missing data were not 1 2 1 imputed. A complete-case analysis was performed. There were no missing data on age, sex, date 1 2 2 of hospital admission, date of symptoms onset, or date of death. We conducted three sensitivity analyses to assess the robustness of our findings: 1) full models 1 2 4 with further adjustment for pneumonia, admission to intensive care unit, intubation, and time 1 2 5 from symptoms onset to admission; 2) multilevel mixed-effect survival regression models to 1 2 6 assess the possible effect of geographical differences on our risk estimates; 29 and 3) comparison 1 2 7 of hazard ratios from analysis restricted to cases admitted before and after August 1 to address 1 2 8 the possible influence of changes to the definition of suspected viral respiratory disease. 21,25 We 1 2 9 conducted stratified analysis according to age and sex, in outpatients and inpatients. We used the 1 3 0 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 7 log-rank test to compare survival curves. All p values were two-sided. All analyses were 1 3 1 performed using Stata 14 (StataCorp LP, TX). . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint   (Table S3). In patients with COVID-19, the incidence rate of death was 1,153.1 cases per 100,000 person-1 5 5 days in those with diabetes and 292.2 cases per 100,000 person-days in those without diabetes. In 1 5 6 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2020. ; https://doi.org/10.1101/2020.11.25.20238345 doi: medRxiv preprint outpatients with COVID-19, the incidence rate of death in those with and without diabetes was, 1 5 7 respectively, 194.1 and 39.2 cases per 100,000 person-days. In hospitalized patients with 1 5 8 COVID-19, the incidence rate of death in those with and without diabetes was, respectively, 1 5 9 2,552.8 and 1,735.3 cases per 100,000 person-days. In patients without COVID-19, the incidence 1 6 0 rate of death was 344.9 cases per 100,000 person-days in those with diabetes and 55.8 cases per 1 6 1 100,000 person-days in those without diabetes. Among COVID-19 cases, our adjusted Cox proportional-hazards regression analysis showed that 1 6 3 patients with diabetes had a 49% higher relative risk of death than those without diabetes (hazard 1 6 4 ratio: 1.49 (95% confidence interval [CI], 1.47-1.52) ( Table 2). The association of diabetes with 1 6 5 mortality was mediated by age, sex, and the type of patient care (outpatient vs. inpatient) 1 6 6 (P<0.001 for all interactions). Men were at higher risk of death than women (hazard ratio: 1.65; 1 6 7 95% CI, 1.63-1.68). Compared with subjects 50 to 59 years of age, those 70 to 79 years of age 1 6 8 and those 80 years of age or older had 3-fold and 4-fold higher risk of death, respectively ( Table   1 6 9 S4). A slightly stronger association between diabetes and mortality was noted in women (hazard 1 7 0 ratio: 1.64; 95% CI, 1.59-1.68) than in men (hazard ratio: 1.41; 95% CI, 1.38-1.44). We observed 1 7 1 a stronger association between diabetes and mortality in outpatients (hazard ratio: 1.66; 95% CI, 1 7 2 1.58-1.74) compared with that in hospitalized patients (hazard ratio: 1.14; 95% CI, 1.12-1.16) 1 7 3 (Table 2). Diabetes was associated with lower survival probability in outpatients and inpatients, In non-COVID-19 cases, the adjusted hazard ratio for mortality was 1.78 (95% CI, 1.73-1.84). This association was also stronger in outpatients (hazard ratio: 1.91; 95% CI, 1.68-2.18) 1 7 7 compared with that in hospitalized patients (hazard ratio: 1.11; 95% CI, 1.07-1.14). The 28-day 1 7 8 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2020. ; https://doi.org/10.1101/2020.11.25.20238345 doi: medRxiv preprint 1 0 survival for inpatients with diabetes who had COVID-19 was lower (53.4%) compared with that 1 7 9 for those without COVID-19 (73.4%) (Figure 2). In stratified analysis among COVID-19 cases according to sex and age, diabetes was associated 1 9 0 with higher mortality in all age groups, among women and men. We observed that the relative 1 9 1 risk of death associated with diabetes decreased with age (Table 2) (trend test: P=0.004). These 1 9 2 trends were observed in women (P=0.006) and men (P=0.007) (Table S6). Although the relative 1 9 3 risk of death decreased with age, the incidence rates of death were higher in older subjects ( Table   1 9 4 S7). The relative risk of death associated with diabetes decreased with age in outpatients (trend 1 9 5 test: P=0.001) and in hospitalized patients (trend test: P=0.006) (Table S8). In hospitalized patients with COVID-19, the probability of survival at 28 days of follow-up for 1 9 7 those with diabetes compared with that among those without diabetes decreased as age increased. We did not observe substantial differences in survival between patients 70 years of age or older with diabetes and those without diabetes (Figure 3). The 28-day survival for inpatients with and 2 0 0 without diabetes was, respectively, 73.5% and 85.2% for patients 20-39 years of age; 66.6% and 2 0 1 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint older without COVID-19, we did not observe substantial differences in survival between patients 2 0 5 with diabetes and those without diabetes ( Figure S2). (>750,000) shows that those with diabetes have increased risk of death during a follow-up of 28 2 2 3 days. The relative risk of death associated with diabetes was stronger in outpatients than in 2 2 4 hospitalized patients, and decreased with age. However, the incidence rate of death was much 2 2 5 higher in hospitalized patients than in outpatients, and increased with age. Although the 28-day survival for hospitalized patients with diabetes in the youngest group (20 to 2 2 7 39 years) was about 12 percentage points lower compared with that for those without diabetes, 2 2 8 the survival difference between patients with diabetes and those without diabetes was less than 2 2 2 9 percentage points in those 70 years of age or older (Figure 3). Lower hazard ratios for death 2 3 0 associated with diabetes have been reported among older patients with COVID-19 in England 30 .

3 1
Our study shows a detailed comparison of the association of diabetes with mortality across age 2 3 2 groups in a Hispanic-Latino population in Mexico, a country that has one of the highest numbers Previous studies have shown diabetes is very common in patients with 32 In our 2 3 5 study, the proportion of patients with diabetes among hospitalized patients with COVID-19 was 2 3 6 similarly high in those with and without COVID-19 (~30%). Among deceased patients, the 2 3 7 proportion of patients with diabetes was also similar in both groups (~40%). Although some 2 3 8 studies have detected an association between diabetes and mortality in subjects with COVID- 19, 4,6,7,9-13 others have not found a significant association. 2,3,8,[15][16][17][18] In our study, diabetes was 2 4 0 associated with mortality in patients with COVID-19. However, this association was not stronger 2 4 1 than that observed in patients without COVID-19. Our findings also raise concern that the 2 4 2 association of diabetes with COVID-19-related mortality varies with age (Table 2). We observed 2 4 3 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review) preprint
The copyright holder for this this version posted November 26, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 1 3 a modest association in patients 70 to 79 years of age. No association was observed in patients 80 2 4 4 years of age or older. The latter age group represented 11% of all COVID-19-related deaths in 2 4 5 our study (Table S3). Data from preliminary reports 13,26,33 suggest that 30-50% of the total 2 4 6 number of COVID-19-related deaths occur in patients 80 years of age or older. Thus, our  The present study has many strengths. The high number of fatal cases among patients with 2 5 0 COVID-19 (>80,000 deaths) and the high number of patients who had diabetes (>120,000) in the 2 5 1 population studied enabled us to conduct a stratified analysis according to age groups to obtain 2 5 2 precise estimates of the association of diabetes with mortality. The association of diabetes with 2 5 3 mortality across age groups in Latino populations has remained unknown. Our stratified analysis 2 5 4 by age was performed in a population with a number of deaths that was 15 times higher than the 2 5 5 number analyzed in a previous study conducted in a predominantly Caucasian population. 30 We 2 5 6 also performed stratified analysis among outpatients and inpatients. We observed a substantial 2 5 7 difference in the magnitude of the association of diabetes with mortality in outpatients (66% 2 5 8 higher risk) and inpatients (14% higher risk) with COVID-19, suggesting this association is 2 5 9 weaker in patients with severe COVID-19. variables since data were not available. Proper blood glucose control has been associated with 2 6 5 lower COVID-19-related mortality. 35 Another limitation of our study is that we cannot exclude 2 6 6 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2020. ; https://doi.org/10.1101/2020.11.25.20238345 doi: medRxiv preprint 1 4 the possibility that the number of deaths in patients who had COVID-19 could be underreported.

6 7
Finally, since our analysis was restricted to patients who presented symptoms for suspected viral 2 6 8 respiratory disease and only 10% of patients with mild symptoms of viral respiratory disease 2 6 9 were reported to the surveillance system, our findings may not be generalizable to populations 2 7 0 with asymptomatic or mild COVID-19. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020  In symptomatic adult patients with COVID-19 in Mexico, diabetes was associated with higher 2 8 7 mortality. The relative risk of death associated with diabetes decreased with age. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 1 6 ACKNOWLEDGMENTS 2 9 3 We thank the Secretary of Health of the Government of Mexico for providing free access to data 2 9 4 on suspected cases of viral respiratory disease in Mexico. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 1 7 FUNDING 2 9 7 No funding was received for this study. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2020. ; https://doi.org/10.1101/2020.11.25.20238345 doi: medRxiv preprint 3 0  inpatients (B) who had diabetes with and without COVID-19. Subjects were admitted from January 1 through October 7, 2020, and followed up for 28 days unless the event (death) occurred first. The solid lines represent survival probabilities and the shaded area represent the 95% confidence intervals (CIs). . Panels A to F show the probability of survival stratified according to age groups among adult inpatients with and without diabetes who had COVID-19. Subjects were admitted from January 1 through October 7, 2020, and followed up for 28 days unless the event (death) occurred first. The solid lines represent survival probabilities and the shaded area represent the 95% confidence intervals (CIs).
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted November 26, 2020. ; https://doi.org/10. 1101/2020