In this retrospective observational study, the prevalence of diabetes in patients with non-mild COVID-19 cases was 34.3%. Among the diabetic patients, 45.0% were unaware of their underlying diabetes condition before admission. Diabetes was independently associated with an increased risk of in-hospital death in patients with COVID-19. Notably, patients with undiagnosed diabetes who were newly defined by HbA1c testing at admission had an increased risk of mortality during hospitalisation similar to that of patients with diagnosed diabetes, compared with their non-diabetic counterparts.
Diabetes has been garnering attention in terms of its prevalence and impact in the COVID-19 population. A report on the largest case series of COVID-19 in China, conducted by the Chinese National Emergency Response Epidemiology Team, showed that the prevalence of diabetes among 44,672 confirmed Chinese mainland patients with COVID-19 was 5.3% . Observational studies and meta-analyses reported that the prevalence of pre-existing diabetes in Chinese patients with COVID-19 ranged from 8.2–19.0% [8, 21–23]. Here, we showed a much higher prevalence of diabetes (34.3%) in patients with COVID-19. This could be due to two reasons. First, our patients were from one of the national intensive care centres for COVID-19 that only admitted moderate to critical patients. The patients in our study were older and had more severe conditions than those in the nationwide analysis [20, 21]. Therefore, a higher prevalence of diabetes was expected in this study, similar to that reported by medical centres in Western countries [6, 11, 24, 25]. This might also suggest an association between pre-existing diabetes and an increased severity of COVID-19. Second, we included patients with newly diagnosed diabetes defined by HbA1c testing at admission. By contrast, most previous studies reported the prevalence of diabetes as a comorbidity according to patient histories of those with COVID-19, and patients who were included in non-diabetic groups had no available HbA1c data  or some of them had HbA1c levels over 6.5% . In the most recent national epidemiological survey involving 75,880 adult participants, the prevalence of overall, self-reported, and newly diagnosed diabetes based on the American Diabetes Association criteria were 12.8%, 6.0%, and 6.8%, respectively, in China . In agreement with that study, we found that approximately 50% of diabetic patients (elevated HbA1c levels) were undiagnosed before admission. HbA1c was first introduced into the American Diabetes Association diagnostic criteria of diabetes in 2010 . HbA1c testing can well represent average blood glucose levels within 2–3 months before testing and is not influenced by factors such as acute infection, stress, or recent medications that could alter glucose metabolism, like corticosteroids. Moreover, HbA1c testing does not require fasting. Therefore, HbA1c is a reasonable diagnostic parameter for the quick identification of the background glucose metabolic state in severe and critical patients with COVID-19. Because diabetes is one of the most common comorbidities in patients with COVID-19 and is associated with poor outcomes, HbA1c testing at admission can provide important information for patient assessment and help identify those who have not been diagnosed but are at great risk.
It has been shown that diabetic patients have poorer COVID-19 outcomes. The prevalence of diabetes is much higher in patients with COVID-19 treated in intensive care units than in those treated in general wards . Diabetic patients with COVID-19 had a higher risk of developing severe or critical illness  and having multiple-organ damage, and a higher mortality rate than non-diabetic patients [10, 11, 14, 15, 20]. Whats’ more, the risk of COVID-19 hospitalizaiotn was elevated in community people with poorly controlled diagnosed diabetes, and even in those with undiagnosed diabetes (A1C ≥ 6.5% at baseline) . Similar to previous studies [8, 21], our data indicated that diabetes, together with advanced age, a high qSOFA score, and coagulation disorders, was a risk factor for in-hospital death in moderate to critical patients with COVID-19. Similarly, diabetes was also previously reported as a major risk factor for mortality in severe acute respiratory syndrome in 2003 and Middle East respiratory syndrome [29, 30]. Thus far, there is no established effective therapy for reducing the mortality rate of COVID-19. However, a recent study reported that a well-controlled blood glucose level in diabetic patients during hospitalisation was associated with a markedly reduced mortality from COVID-19, in comparison with poorly controlled glycaemia . Therefore, identifying undiagnosed diabetes provides awareness of the background glycaemic disorder, thereby facilitating appropriate intervention for at-risk patients with coronavirus infections, including glucose monitoring and glycaemic control, and possibly better outcomes.
The underlying mechanism of the impact of diabetes on the prognosis of COVID-19 is still under investigation. The dysregulated immune response caused by diabetes may contribute to increased disease severity. Diabetic patients with COVID-19 have more neutrophils and a higher rate of lymphopenia , which is in agreement with our findings of higher white blood cell counts and lower lymphocyte counts in diabetic patients than in non-diabetic patients. In addition, diabetes may cause a chronic inflammatory state, elevating the levels of pro-inflammatory cytokines, such as interleukin-1 (IL-1) and IL-6, and further aggravate cytokine storms in some patients with COVID-19 [31, 32]. However, our study did not show a significant difference in serum IL-6 levels between groups. Angiotensin-converting enzyme 2 (ACE2) may be another underlying mechanism for the detrimental effects of diabetes on the prognosis of COVID-19. SARS-CoV-2 gains entry into host pneumocytes by binding to ACE2 . Diabetic patients were reported to have a higher expression of ACE2, thereby facilitating viral uptake and increasing the risk of severe infection . Moreover, glucose can also directly increase the viral load and upregulate the expression of ACE2 and IL-1β in SARS-CoV-2-infected monocytes in a dose-dependent manner, suggesting that individuals with elevated circulating glucose levels may be more susceptible to SARS-CoV-2 infection and more likely to develop severe illness . Therefore, the cause for worse prognosis in diabetic patients with COVID-19 is multifactorial.
Our study has several advantages. Our study focused on the clinical outcomes of both undiagnosed and diagnosed diabetes in patients with COVID-19. The HbA1c determination method used in our center is comparable to the National Glycohemoglobin Standardisation Programme standard. By testing the HbA1c level at admission, we reduced the omission diagnostic rate of diabetes and prevented the overdiagnosis of diabetes because of stress-induced hyperglycaemia. The high percentage of undiagnosed diabetes, together with the similarly worse clinical outcome of undiagnosed and diagnosed diabetes compared with non-diabetes, highlighted the importance of screening for undiagnosed diabetes by HbA1c detection in patients with COVID-19. Moreover, the patients included in this study were admitted at a single medical centre and underwent treatments following uniform clinical guidelines, thereby reducing bias resulting from different treatment methods. Finally, we presented survival curves of diabetic and non-diabetic patients with COVID-19, while most previous studies only showed final outcomes without time-kinetic changes. Shi et al.  reported survival curves of patients with COVID-19, in which the survival probability of patients with diabetes was lower than that of sex- and age-matched patients without diabetes. However, in their study, patients in the non-diabetic control group had no available HbA1c data, and the fasting glucose levels in some cases were over 11.1 mmol/L, indicating a high possibility of patients with undiagnosed diabetes in the control group.
Our study has some limitations. First, it was a single-centre study with a limited number of patients. We enrolled as many patients as we could and excluded patients who were transferred from Fangcang hospitals to reduce bias from pre-admission treatment. Second, not all patients had their HbA1c level tested during the hospitalization, particularly those in the non-diabetic group, as not all medical teams in our COVID-19 wards had members specialising in endocrinology. At the very beginning of the pandemic in Wuhan, some medical staff had not realised the potential benefit of evaluating and managing glucose metabolism in patients with COVID-19. Therefore, some patients did not undergo HbA1c testing; thus, the prevalence of diabetes in our COVID-19 population may even be higher than what we reported in this study. Nevertheless, in the subgroup analysis of 140 patients with available HbA1c data, the association between lower survival probability and diabetes (overall, diagnosed, or undiagnosed diabetes) was consistent with the results of the primary analysis of all 233 patients. Third, we only included IL-6 in the risk factor analysis and did not analyse other inflammatory biomarkers, such as serum C-reactive protein (CRP) or ferritin, in the present study. Most patients were tested for high-sensitivity CRP (hsCRP), rather than CRP, because hsCRP was incorporated into the biochemical analysis at our medical centre. hsCRP levels are more associated with systemic low-grade inflammation than with acute inflammatory conditions, such as COVID-19. In addition, IL-6, which is upstream of CRP as a sensitive marker for acute infection, was tested in most of our patients at admission. Therefore, we used IL-6 as the inflammatory biomarker in our Cox regression analysis. Although ferritin data were available in 223 patients, many were tested several days after admission, indicating that the levels could be confounded by other in-hospital factors. Thus, ferritin was excluded in the final analysis.