Clinical Characteristics and Risk Factors for Mortality of COVID-19 Patients With Diabetes in Wuhan, China: A Two-Center, Retrospective Study

In December 2019, pneumonia of an unknown cause was detected in Wuhan, China, which was later named coronavirus disease 2019 (COVID-19) by the World Health Organization. The virus that caused this epidemic was named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The outbreak of COVID-19 swept across China and other countries, arousing global concern. As of 24 April 2020, a total of 2,626,321 COVID-19 cases were confirmed worldwide, and 181,938 patients had died (1). Compared with severe acute respiratory syndrome (SARS) and Middle East respiratory syndrome (MERS), COVID-19 has a lower mortality among confirmed cases. However, elderly patients with underlying comorbidities, including diabetes, hypertension, and coronary heart disease, are at greater risk of poor outcomes (2,3).

Diabetes is one of the leading causes of morbidity, and it causes enormous health and financial burdens worldwide (4). Connections between diabetes and increased susceptibility to infections, including respiratory tract, urinary tract, and soft-tissue infections, have long been accepted (5). The available evidence demonstrates that diabetes predisposes people to developing infectious diseases, and patients with diabetes are at greater risk of infection-related mortality (6,7). Furthermore, diabetes has been associated with a poor prognosis and increased pneumonia-associated mortality (8,9). Previous studies demonstrated that diabetes is one of the major comorbidities in COVID-19 patients. Wang et al. (10) and Guan et al. (11) reported that patients with diabetes accounted for 10.1% and 7.4% of COVID-19 patients, respectively. Recent publications showed that 20–30% of nonsurviving COVID-19 patients had underlying diabetes (3,12). This evidence indicates that COVID-19 patients with diabetes might be at a higher risk of death. Thus, the clinical characteristics and risk factors for in-hospital mortality of COVID-19 patients with diabetes need to be explored.

In this study, we aimed to describe the demographic features, clinical data, treatments, and outcomes of COVID-19 patients with diabetes. We also compared the characteristics and risk factors for in-hospital death of the patients who had diabetes with those of age- and sex-matched patients without diabetes.

This two-center, retrospective study was conducted at Renmin Hospital of Wuhan University and Zhongnan Hospital of Wuhan University, which are two major tertiary hospitals in Wuhan that serve as government-designated hospitals for the treatment of COVID-19. All the confirmed COVID-19 patients with diabetes who were discharged or died from 1 January 2020 to 8 March 2020 were identified. The patients with diabetes included in our study had a clear diagnosis of diabetes by their physicians on the electronic medical records. The age of those with diabetes (64.0 [interquartile range [IQR], 56.0–72.0] years) in our study was significantly different from the overall COVID-19 population (47.0 [IQR, 35.0–58.0] years) in China (11). Older age and male sex have also been demonstrated to be associated with in-hospital death of COVID-19 patients (3,12). Thus, to adjust age and sex, an age- (±2 years) and sex-matched COVID-19 patient without diabetes was randomly selected for each patient with diabetes according to previously reported similar methods (13,14). Whenever more than one patient was available for each patient with diabetes, the match was randomly selected from those available. This study was approved by the Institutional Ethics Boards of Renmin Hospital of Wuhan University (No. WDRY2020-K060) and Zhongnan Hospital of Wuhan University (No. 2020020), Wuhan, China. The ethic committees in these two hospitals waived informed consent. Oral consent was obtained when we contacted patients or their families for information about patients’ diabetes history.

Data abstracted included age, sex, exposure history, history of diabetes, other underlying comorbidities (hypertension, cardiovascular disease, cerebrovascular disease, chronic pulmonary disease, chronic kidney disease, chronic liver disease, and malignancy), onset of symptoms (fever, cough, dyspnea, myalgia, etc.), vital signs at admission (heart rate, respiratory rate, and mean artery pressure), laboratory parameters on admission (blood glucose level, white blood cell count, lymphocyte count, procalcitonin, triglyceride, etc.), random blood glucose (RBG), chest computed tomographic (CT) scans, complications (acute respiratory distress syndrome [ARDS], acute cardiac injury, acute kidney injury [AKI], shock, and secondary infections), medications for treatment (antiviral, antibacterial agents, corticosteroids, Ig), treatment strategies (supplemental oxygen, noninvasive mechanical ventilation, invasive mechanical ventilation, continuous renal replacement therapy [CRRT], and extracorporeal membrane oxygenation [ECMO]), and date of discharge or death. Duration from the onset of symptoms to admission, diagnosis of COVID-19, and total hospital length of stay were also recorded. Data were abstracted using the electronic medical record systems in Renmin Hospital of Wuhan University and Zhongnan Hospital of Wuhan University. All data were reviewed by a team of experienced physicians. Any uncertain or missing records were addressed through communication with the involved health care providers, the patients, or their families. Also included in our study were 14 patients with diabetes and 9 patients without diabetes in a previous publication (10).

The diagnosis of COVID-19 was performed according to the World Health Organization interim guidance (15). The method of detection of SARS-CoV-2 using throat swabs and RT-PCR was reported previously (10). Diabetes was defined according to the guidelines of American Diabetes Association (16). Pregnant women with gestational diabetes or patients with glucocorticoid-induced hyperglycemia were identified and excluded from this study. ARDS was defined according to the Berlin definition (17). Cardiac injury was reported if serum levels of myocardial injury biomarkers (e.g., ultrasensitive troponin I) were >99th percentile of the upper reference. AKI was diagnosed according to the Kidney Disease: Improving Global Outcomes definition (18). Shock was defined according to the 2016 Third International Consensus Definition for Sepsis and Septic Shock (19). Secondary infections were diagnosed when positive cultures of pathogens were obtained from lower respiratory tract specimens or blood samples after admission (3). The calculation of average RBG was based on all the available RBG test results of the patients during this hospitalization. Average RBG = (RBG₁ + RBG₂ + RBG₃… + RBGn)/n.

The primary outcome was the in-hospital mortality of COVID-19 patients with diabetes and risk factors for the death of patients with diabetes. The secondary outcomes were the clinical characteristics, laboratory findings, incidence of complications, and the differences of risk factors for in-hospital death in COVID-19 patients with and without diabetes.

We made no assumptions regarding missing data. Categorical variables are described as frequencies and percentages based on the available data. Continuous variables are described as the medians and IQRs. We used the Pearson χ² test, the Mann-Whitney test, and the Fisher exact test for comparisons between patients with and without diabetes and between survivors and nonsurvivors, as appropriate.

To explore the risk factors associated with in-hospital death for the 306 COVID-19 patients and to assess whether diabetes was an independent risk factor for death, a multivariable Cox proportional hazards regression model was performed. Four variables, including diabetes, hypertension, cardiovascular disease, and chronic pulmonary disease, were included in the model. Age and sex were not included because the patients with and without diabetes were matched on age and sex. We also conducted Cox regression analyses to identify risk factors for in-hospital death of patients with diabetes and matched patients without diabetes. Considering the death toll was not large in our study, and to avoid overfitting in the model, five variables, including age, sex, underlying hypertension, cardiovascular disease, and chronic pulmonary disease, were chosen for the final regression models. All the variables included in the final models were based on clinical and scientific understanding, previous findings, and the results of univariable analyses. Variables were excluded from the Cox regression models if the number of events was too small (i.e., chronic kidney disease, cerebrovascular disease).

Survival curves for patients with diabetes and matched patients without diabetes were developed using the Kaplan-Meier method with the log-rank test. The statistical analyses were conducted with SPSS (version 25.0), GraphPad Prism (version 5.0), and R (version 3.6.1) software. A two-sided P value <0.05 was considered statistically significant.

Of 1,561 COVID-19 patients, 153 patients (42 from Zhongnan Hospital and 111 from Renmin Hospital) with diabetes were included, and the prevalence of diabetes was 9.8%. The median age of the patients with diabetes was 64.0 (IQR, 56.0–72.0) years. The patients with and without diabetes were well matched for age and sex, with men and women represented approximately equally (Table 1). More patients with diabetes reported no exposure history but had a higher prevalence of hypertension (56.9% vs. 28.8%), cardiovascular disease (20.9% vs. 11.1%), and cerebrovascular disease (7.8% vs. 1.3%) (all P values <0.05). Fever and cough were the most common onset of symptoms in both groups. Patients with diabetes were more likely to require intensive care unit (ICU) admission (17.6% vs. 7.8%, P < 0.05). The characteristic ground-glass opacity on CT images is shown in Supplementary Fig. 1. Among laboratory findings, patients with diabetes had lower levels of cholesterol (3.8 vs. 4.1 mmol/L), Pao₂ (68.5 vs. 88.0 mmHg) and CD8⁺ cell count (164.0 vs. 242.0/μL), and higher levels of blood glucose (9.4 vs. 5.7 mmol/L) and procalcitonin (0.06 vs. 0.05 ng/mL) (all P values <0.05). Patients with diabetes were more likely to have ARDS (24.8% vs. 11.1%), acute cardiac injury (30.7% vs. 17.0%), secondary infections (24.2% vs. 11.1%), shock (20.9% vs. 10.5%), and AKI (12.4% vs. 3.3%) (all P values <0.05). Antiviral therapy, antibacterial therapy, and supplemental oxygen were the most common treatments in both groups. Among all of the treatment strategies, only noninvasive mechanical ventilation (13.7% vs. 5.2%, P < 0.05) was applied more commonly in patients with diabetes. Death was more common in patients with diabetes (20.3% vs. 10.5%, P < 0.05).

Diabetes history regarding duration, previous glycemic condition, previous glucose control methods, and complications are reported in Supplementary Table 1. Among patients with diabetes who died compared with survivors (Table 2), they were older (76.0 vs. 63.0 years), most were male (71.0% vs. 29.0%), and were more likely to have hypertension (83.9% vs. 50.0%), cardiovascular disease (45.2% vs. 14.8%), and present with dyspnea (54.8% vs. 28.7%) (all P values <0.05). CT analysis of the lungs in nonsurvivors revealed a higher proportion of patients with >75% involvement and fewer with <25% involvement compared with survivors. Compared with survivors, nonsurvivors had higher respiratory rate (21 vs. 20 rpm) and heart rate (88 vs. 82 bpm) (all P values <0.05). Nonsurvivors were more likely to be admitted to the ICU (64.5% vs. 5.7%) but had a significantly shorter hospital length of stay (6.0 vs. 17.0 days) (all P values <0.05). Average RBG (13.6 vs. 7.6 mmol/L, P < 0.05) and numerous laboratory parameters, including lymphocyte count, D-dimer concentration, triglyceride level, and oxygen saturation (Spo₂) among others, distinguished nonsurvivors from survivors (Table 2). The daily average RBG during the first consecutive 5 days of hospitalization tracked in 13 nonsurvivors and 32 survivors showed that nonsurvivors had a significantly higher average RBG (except day 2) (Supplementary Fig. 2A). Compared with survivors, nonsurvivors were more likely to have ARDS (100.0% vs. 5.7%), acute cardiac injury (96.8% vs. 13.9%), shock (96.8% vs. 1.6%), secondary infections (64.5% vs. 13.9%), and AKI (51.6% vs. 2.5%) (all P values <0.05). With regard to treatments, nonsurvivors were more often treated with antibiotics (100.0% vs. 74.6%), glucocorticoids (54.8% vs. 30.3%), supplemental oxygen (83.9% vs. 53.3%), noninvasive mechanical ventilation (51.6% vs. 4.1%), and invasive mechanical ventilation (29.0% vs. 1.6%) (all P values <0.05).

Among patients with diabetes, 30 patients with the highest average RBG and 30 patients with the lowest average RBG during hospitalization were analyzed (Supplementary Fig. 2B and C). The median average RBG levels were 16.4 (IQR, 14.9–17.6) mmol/L and 5.9 (IQR, 5.3–6.1) mmol/L, respectively. Half of the patients with the highest RBG had sustained detectable SARS-CoV-2 virus until death, while the RT-PCR tests turned negative in most patients with the lowest average RBG. In addition, we found that 16 of 30 patients with the highest average RBG died, whereas only 1 of 30 patients with the lowest average RBG died.

Among matched patients without diabetes who died compared with survivors (Table 2), they were older (72.0 vs. 63.0 years), more likely to have cerebrovascular disease (12.5% vs. 0), chronic kidney disease (18.8% vs. 2.2%), and present with dyspnea (68.8% vs. 35.8%) (all P values <0.05). Nonsurvivors were more likely to be admitted to the ICU (62.5% vs. 1.5%, P < 0.05). Laboratory findings, treatments, and complications are reported in Table 2.

Among the included 306 patients, multivariable analyses (Supplementary Table 2) showed that hypertension (HR 2.50, 95% CI 1.30–4.78), cardiovascular disease (HR 2.24, 95% CI 1.19–4.23), and chronic pulmonary disease (HR 2.51, 95% CI 1.07–5.90) were independent risk factors for in-hospital death. Diabetes (HR 1.58, 95% CI 0.84–2.99) was not independently associated with death after adjusting for covariables. In univariable analyses for patients with diabetes (Table 3), 23 variables, including age, male sex, hypertension, chronic pulmonary disease, cardiovascular disease, higher average RBG, and decreased lymphocyte count, among others, were related to death. In multivariable analyses, age ≥70 years (HR 2.39, 95% CI 1.03–5.56) and hypertension (HR 3.10, 95% CI 1.14–8.44) were independent risk factors for in-hospital death of patients with diabetes. For matched patients without diabetes (Table 3), Cox regression analyses indicated that age ≥70 years (HR 5.87, 95% CI 1.88–18.33) was independently associated with death. The survival curves of COVID-19 patients with diabetes and matched patients without diabetes are shown in Fig. 1.

This study analyzed the characteristics of COVID-19 patients with diabetes and sex- and age-matched patients without diabetes and identified the risk factors associated with in-hospital death of these patients. Among patients hospitalized with COVID-19, the prevalence of diabetes was 9.8%. Patients with diabetes had more underlying comorbidities, were more likely to suffer complications, had a higher proportion of ICU admissions, and more deaths compared with sex- and age-matched patients without diabetes. However, underlying hypertension, cardiovascular disease, and chronic pulmonary disease, rather than diabetes, were independently associated with in-hospital death of COVID-19 patients. Among patients with diabetes, age ≥70 years and underlying hypertension were independent risk factors for in-hospital death.

The most common symptoms of SARS-CoV-2 infection in patients with diabetes included fever, cough, and dyspnea, which were consistent with recent publications (20,21). A previous study showed that diabetes tripled the risk of hospitalization after pandemic influenza A H1N1 and significantly increased the risks of ICU admission and mortality (14,22). In our study, patients with diabetes had a significantly higher mortality and more severe disease, as verified by the higher proportion of ICU cases and the higher incidences of ARDS and multiple organ dysfunction syndrome as well as secondary infections even after sex and age adjustment.

The worse outcomes of COVID-19 patients with diabetes could be associated with underlying comorbidities. A nationwide analysis in China showed that the most prevalent comorbidity in COVID-19 patients was hypertension (16.9%), followed by diabetes (8.2%), and that the underlying diseases were associated with adverse outcomes of COVID-19 patients (23). In our study, hypertension and cardiovascular disease were more common in patients with diabetes. Although diabetes itself was not independently associated with death of COVID-19 patients in our multivariable analyses, diabetes and other comorbidities, which included cardiovascular disease and hypertension, were often closely related, and the effect of these factors could not be considered separately. Diabetes and hypertension often coexist and may act synergistically to promote adverse clinical events (24,25). The persistent hyperglycemic condition and metabolism changes in diabetes together with coexisting hypertension lead to microvascular and macrovascular changes and form a vicious cycle that further contributes to cardiovascular events (26). Recent study also suggested that hypertension was associated with increased risk of severe and fatal COVID-19, and ACE inhibitors reduced the mortality in COVID-19 patients with hypertension (27,28). This evidence further supported our finding that hypertension was independently associated with death among patients with diabetes. Thus, patients with diabetes with underlying comorbidities, especially hypertension, should attract more attention. In sex- and age-matched patients without diabetes, Cox regression analyses did not reveal underlying comorbidities as risk factors associated with death. This might be explained by the absence of diabetes and the relatively lower prevalence of other comorbidities in the matched population.

A recent publication demonstrated that old age was an independent predictor of mortality in COVID-19 patients (3). The age-dependent decreases in cellular and humoral immune function in elderly patients have been reported before, especially with regard to adaptive immune function (29). In our study, nonsurvivors among patients with diabetes were older compared with survivors and had obvious lymphopenia as demonstrated by significantly lower numbers of T and B cells. Furthermore, the high risks of elderly patients with diabetes could be due to their poor overall health condition and greater number of comorbidities.

Besides the above-mentioned independent risk factors, other potential risk factors related to death were also identified by our univariable analyses. Recent studies revealed lymphopenia as an important characteristic of SARS-CoV-2 infection, especially in critically ill and deceased patients (2,12). Besides, infection and destruction of lymphocytes by the SARS-CoV have been proven (30). Thus, the destruction of lymphocytes by the SARS-CoV-2 virus might be speculated but needs to be further investigated. Lymphopenia was also noticed in our study, especially in patients with diabetes. This might be explained by the previous findings that diabetes and hyperglycemia could impair the innate and adaptive immunity (31,32).

During hospitalization, underlying diabetes, illness severity, and medical treatments could contribute to the hyperglycemic condition. Hyperglycemia has been proven to be associated with increased risks of in-hospital complications and in-hospital death (33,34). In our study, we also noticed higher blood glucose levels during hospitalization in patients with diabetes who did not survive than in survivors. Thus, frequent monitoring of blood glucose and the use of oral glucose-lowering medication or insulin would be important routine procedures for patients with diabetes.

Besides the findings in our study, some other factors might also contribute to the mortality of COVID-19 patients, among which obesity would be a potential candidate. A recent study in the New York area showed that 41.7% of the patients were obese, and an increasing number of reports have linked obesity to more severe COVID-19 illness (35–37).

Despite the importance of the aforementioned findings, the current study, however, has some limitations. Firstly, data collection relied on electronic medical records. CT images of some patients transferred from other hospitals were not available in the electronic medical record systems of these two hospitals. Some important indicators were not tested in all patients. Thus, the missing data might lead to bias. Secondly, these two hospitals are designated for treatment of patients with relatively severe infection, which to some extent might lead to a higher mortality. Thirdly, the sample size and relatively small number of deaths might influence the interpretation of our findings. Fourthly, the possibility of obesity as a contributor to death in COVID-19 patients was not investigated due to the lack of information on BMI. Fifthly, patients in this study were selected on the basis of diabetes status and death or discharge. Some patients who remained alive in these two hospitals at the time of the analysis were not included in this study. Finally, because a matched design was used in our study, the roles of sex and age in the death of patients with diabetes versus those without could not be examined and should be addressed further.

In summary, the findings of our study suggested that COVID-19 patients with diabetes had worse outcomes compared with the sex- and age-matched patients without diabetes. Diabetes was not independently associated with in-hospital death, while hypertension, cardiovascular disease, and chronic pulmonary disease played more important roles in contributing to the mortality of COVID-19 patients. In-hospital death among COVID-19 patients with diabetes was associated with hypertension and advanced age, whereas only older age was independently associated with death among matched patients without diabetes. The need for early monitoring and supportive care should be addressed in these patients at high risks.

Acknowledgments. The authors thank all patients and health providers involved in this study and thank Chuanhua Yu and Lu Ma (Wuhan University School of Health Sciences) for help in statistical analysis.

Funding. This study was funded by the National Natural Science Foundation of China (No. 81800574, No. 81870442).

The funding sources for this study had no role in the study design, data collection, analyses, or interpretation, or writing of the manuscript.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Q.S., Xi.Z., and F.J. drafted the manuscript. Q.S., Xi.Z., Xu.Z., and N.H. researched the literature. Xi.Z., F.J., and G.H. contributed to the statistical analysis. Xu.Z., N.H., C.B., J.F., S.Y., Y.G., and D.X. collected the epidemiological and clinical data. C.C., X.X., L.L., H.L., and J.T. contributed to editing the manuscript, radiological analysis, and figure construction. Z.P. and W.W. conceived and supervised the study. All authors reviewed and approved the final version of the manuscript. Z.P. and W.W. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.