Methods

Study design and participants
We conducted a retrospective study of COVID-19 patients admitted to Wuhan Union West Hospital and Wuhan Red Cross Hospital to ascertain whether FBG was an independent predictor for 28-day mortality in patients with COVID-19 without a previous diagnosis of diabetes. COVID-19 infection was laboratory-confirmed in accordance with the interim guidance formulated by the WHO [15]. The aforementioned hospitals were two mandatorily designated hospitals for the treatment of COVID-19 patients in China. The institutional ethics committees of Wuhan Union Hospital (No. 0036) reviewed and approved this study protocol. No patients or medical staff involved in patient care took part in the study design and statistical analyses.
All consecutive patients included in this study had a definitive outcome (died, discharged or still hospitalised) within 28 days, with the time period spanning from 24 January 2020 to 10 February 2020. All patients received standard treatment, including antiviral therapy, respiratory support (nasal cannulation, mask oxygenation, high-flow nasal cannula oxygen therapy, non-invasive positive pressure ventilation or invasive mechanical ventilation), symptomatic and supportive treatment and antimicrobial therapy, as appropriate, to prevent or treat secondary infections, which was in accordance with the COVID-19 diagnosis and treatment protocols released by the National Health Commission of the People’s Republic of China [16]. This retrospective project did not interfere with the course of medical management.
A total of 1258 confirmed COVID-19 patients were admitted into the two hospitals. Of these, 653 patients were ruled out for one of the following reasons: (1) no definitive 28-day outcome since they were transferred to another hospital; (2) missing key clinical information (e.g. demographic or clinical data); (3) no FBG data available at admission (for one of the following reasons: [a] patients had a blood glucose measurement taken before admission; [b] patients were tested for blood glucose 24 h after admission; [c] patients received a random blood glucose test but were not tested for FBG; [d] patients did not receive a blood glucose test since this was not routinely conducted for every COVID-19 patient); (4) having previously diagnosed diabetes. The flow diagram of patient selection is detailed in Fig. 1.
Fig. 1 Flow diagram of patient selection. aReasons for no FBG measurement at admission: 13 patients had a blood glucose measurement taken before admission; 25 patients were tested for blood glucose 24 h after admission; 84 patients received a random blood glucose test but were not tested for FBG; and 81 patients did not receive a blood glucose test since this was not routinely conducted for every COVID-19 patient
Patients were discharged when they met the following discharge criteria: (1) body temperature returned to normal, lasting for more than 3 days; (2) respiratory symptoms significantly improved; (3) imaging examinations revealed that acute exudative lesions were significantly improved; (4) two real-time RT-PCR tests for the presence of SARS-CoV-2 virus yielded negative results (with two samples of respiratory specimens taken over 24 h apart) [16].

Data collection
We obtained data from the electronic records of the relevant departments. The following data were collected: demographics, clinical data (symptoms, past medical history, admission FBG and in-hospital complications) and the data on 28-day outcomes.
Past medical histories were obtained from hospital databases or by self-reporting, including diabetes, hypertension, chronic lung disease, chronic heart disease, chronic liver disease, chronic kidney disease, cerebrovascular disease and carcinoma, which were diagnosed according to standard criteria. The common complications that developed after hospitalisation included acute respiratory distress syndrome (ARDS), acute cardiac injury, acute kidney injury, acute liver injury, cerebrovascular accident, coagulopathy and secondary infection.

Definition and measurement of FBG levels at admission
Complications were defined as the occurrence of one or more condition(s) (Table 1) that developed after hospitalisation. ARDS was defined according to WHO clinical management interim guidance [17]. Acute cardiac injury was defined as new electrocardiographic and echocardiographic abnormalities detected, or serum level of cardiac biomarkers (cardiac troponin I, cardiac troponin T, or hype-sensitive troponin I) above the upper limit of normal (ULN) [18]. Acute kidney injury was defined as an elevation of serum creatinine by 26.5 μmol/l or higher within 48 h, or serum creatinine increased to 1.5 times baseline or higher within the previous 7 days [19]. Acute liver injury was defined as alanine aminotransferase or aspartate aminotransferase levels two times above the ULN. Cerebrovascular accident was defined as the occurrence of cerebral haemorrhage or cerebral infarction during hospitalisation [20]. Coagulopathy was defined as prothrombin time prolonged by 3 s or activated partial thromboplastin time prolonged by 5 s [21]. Secondary infection was defined as the occurrence of symptoms or signs of nosocomial infection and new pathogens detected in patients’ specimens (e.g. sputum, blood) taken more than 48 h after admission [18].
Table 1 Baseline characteristics of COVID-19 patients without previous diagnosis of diabetes within 28 days after admission
Variables Total(n = 605) Non-survivor(n = 114) Survivor(n = 491) p value
Hospital
 Wuhan Red Cross Hospital 157 (26.0) 33 (21.0) 124 (79.0) 0.4178
 Wuhan Union West Hospital 448 (74.0) 81 (18.1) 367 (81.9)
Age, years
 Median (IQR) 59.0 (47.0, 68.0) 66.0 (61.0, 72.0) 56.0 (43.0, 65.0) <0.0001
 <65, n (%) 408 (67.4) 49 (43.0) 359 (73.1) <0.0001
 ≥65, n (%) 197 (32.6) 65 (57.0) 132 (26.9)
Sex
 Female, n (%) 283 (46.8) 36 (31.6) 247 (50.3) 0.0003
 Male, n (%) 322 (53.2) 78 (68.4) 244 (49.7)
Onset symptoms
 Fever, n (%) 463/530 (87.4) 88 (85.4) 375 (87.8) 0.5132
 Cough, n (%) 404/555 (72.8) 66 (66.7) 338 (74.1) 0.1308
 Expectoration, n (%) 217/521 (41.7) 43 (43.4) 174 (41.2) 0.6892
 Muscular soreness, n (%) 129/504 (25.6) 23 (24.5) 106 (25.9) 0.7813
 Fatigue, n (%) 300/528 (56.8) 58 (58.6) 242 (56.4) 0.6936
 Diarrhoea, n (%) 91/512 (17.8) 15 (15.3) 76 (18.4) 0.4774
Past history of disease 208 (34.4) 55 (48.3) 153 (31.2) 0.0005
 Hypertension, n (%) 139/543 (25.6) 34 (29.8) 105 (24.5) 0.2447
 Chronic lung disease, n (%) 18 (3.0) 4 (3.5) 14 (2.9) 0.7587
 Chronic heart disease, n (%) 55 (9.1) 13 (11.4) 42 (8.6) 0.3404
 Chronic liver disease, n (%) 16 (2.6) 3 (2.6) 13 (2.7) >0.9999
 Chronic kidney disease, n (%) 16 (2.6) 6 (5.3) 10 (2.0) 0.0531
 Cerebrovascular disease, n (%) 16 (2.6) 7 (6.1) 9 (1.8) 0.0098
 Carcinoma, n (%) 29 (4.8) 9 (7.9) 20 (4.1) 0.0853
CRB-65 score <0.0001
 0, n (%) 334 (55.2) 27 (23.7) 307 (62.5)
 1–2, n (%) 261 (43.1) 80 (70.2) 181 (36.9)
 3–4, n (%) 10 (1.7) 7 (6.1) 3 (0.6)
Admission FBG <0.0001
 <6.1 mmol/l, n (%) 329 (54.4) 35 (30.7) 294 (59.9)
 6.1–6.9 mmol/l, n (%) 100 (16.5) 21 (18.4) 79 (16.1)
 ≥7.0 mmol/l, n (%) 176 (29.1) 58 (50.9) 118 (24.0)
Complications 237 (39.2) 114 (100.0) 123 (25.1) <0.0001
 ARDS, n (%) 142 (23.5) 107 (93.9) 35 (7.1) <0.0001
 Acute cardiac injury, n (%) 80 (13.2) 63 (55.3) 17 (3.5) <0.0001
 Acute kidney injury, n (%) 76 (12.6) 65 (57.0) 11 (2.2) <0.0001
 Acute liver injury, n (%) 167 (27.6) 95 (83.3) 72 (14.7) <0.0001
 Cerebrovascular accident, n (%) 3 (0.5) 3 (2.6) 0 0.0065
 Coagulopathy, n (%) 96 (15.9) 85 (74.6) 11 (2.2) <0.0001
 Secondary infection, n (%) 79 (13.1) 69 (60.5) 10 (2.0) <0.0001
With complications
 <6.1 mmol/l, n (%) 86 (14.2) 35 (30.7) 51 (10.4)
 6.1–6.9 mmol/l, n (%) 48 (7.9) 21 (18.4) 27 (5.5)
 ≥7.0 mmol/l, n (%) 103 (17.0) 58 (50.9) 45 (9.2)
Without complications
 <6.1 mmol/l, n (%) 243 (40.2) 0 243 (49.5)
 6.1–6.9 mmol/l, n (%) 52 (8.6) 0 52 (10.6)
 ≥7.0 mmol/l, n (%) 73 (12.1) 0 73 (14.9)
Data are median (IQR) or n (%)
p values were calculated by using χ2 test, Cochran–Mantel–Haenszel χ2 test, Fisher’s exact test or Wilcoxon rank-sum test, as appropriate
Seventy-five patients (12.4%) had missing information on onset symptoms of fever; 50 (8.3%) on cough; 84 (13.9%) on expectoration; 101 (16.7%) on muscular soreness; 77 (12.7%) on fatigue; 93 (15.4%) on diarrhoea; and 62 (10.2%) on hypertension
For the test of FBG levels at admission, blood samples were collected after an overnight fast lasting at least 8 h within 24 h after admission, according to the WHO guidelines. Serum concentrations of FBG were measured by using an automatic biochemical analyser (Beckman Coulter AU5800 Analyzer, USA).

Assessment of pneumonia severity
CRB-65 is a generally accepted tool used for assessing the severity of pneumonia because of its simplicity and effectiveness. It is based on confusion, respiratory rate (>30/min), systolic blood pressure (≤90 mmHg) or diastolic blood pressure (≤60 mmHg), and age (≥65 years) [22, 23]. Given that pneumonia is the major clinical feature of hospitalised COVID-19 patients [24], and all participants enrolled had pneumonia, as confirmed by chest computed tomography (CT) scans, CRB-65 was used in this study to assess the severity of COVID-19. CRB-65 measures the severity of pneumonia on a 0 to 4 scale, and we grouped scores into three risk levels (CRB-65 score of 0; CRB-65 score of 1–2; CRB-65 score of 3–4) according to Ewig et al [23] and Lepper et al [9]. CRB-65 score 0, 1–2 and 3–4 are, respectively, representative of mild, moderate and severe pneumonia.

Outcome measures
All patients were categorised into three groups according to WHO guidelines in terms of admission FBG (<6.1, 6.1–6.9, and ≥7.0 mmol/l). Two outcome measures were examined: the independent risk factors for 28-day mortality and percentage differences in in-hospital complications between different FBG groups.

Statistical analysis
Descriptive statistics were used to describe patient baseline data. Categorical variables were presented as numbers with percentage proportions, and continuous variables were expressed as mean ± SD if they were normally distributed or as median (IQR) if they were not. Proportions for categorical variables were compared using the χ2 test, Cochran–Mantel–Haenszel χ2 test or Fisher’s exact test. Means of continuous variables were compared using independent group t test when the data were normally distributed. Otherwise, the Wilcoxon rank-sum test was used for medians.
For the analysis of mortality, we conducted a univariable Cox regression analysis to assess the effects of age, sex, onset symptoms, past medical history, CRB-65 score, and admission FBG on the 28-day mortality. Variables with p < 0.05 were regarded as potential risk factors and were included in multivariable Cox regression analysis by using the stepwise bidirectional selection (significance level for entry = 0.05, significance level to stay = 0.1). We conducted subgroup analysis by using Kaplan–Meier curves to assess associations between FBG or severity of pneumonia and mortality within 28 days, and tested linear trends across different groups of different FBG levels. Then we carried out a test for interaction of FBG levels and severity of pneumonia and stratified analyses according to severity of pneumonia.
Finally, univariable logistic analysis was used to assess the association between different FBG levels and in-hospital complications.
A two-sided p value <0.05 was considered to be statistically significant. All statistical analyses were performed using SAS software (version 9.4; USA).