Introduction
South Africa has the highest prevalence of diabetes mellitus (DM) in Africa, with 1 in 9 adults living with diabetes.(1) In 2021, South Africa reported a prevalence of 11.3% and 4,234,000 total cases of diabetes in adults.(2) Similarly, the burden of SARS-CoV-2 (COVID-19) in South Africa between 2020 and 2022 was significant, marked by high infection rates, mortality, and substantial strain on the healthcare system.
COVID-19 appears to be associated with diabetes in a bidirectional manner. Patients living with diabetes (PLWD) experience a more severe form of COVID-19 pneumonia,(1) and many patients presenting with COVID-19 pneumonia are diagnosed with new-onset diabetes.(3) Studies have shown a higher prevalence of diabetes among patients with severe COVID-19.(7) Other studies showed that PLWD had more severe pneumonia, were more likely to be hospitalized due to COVID-19, and had higher rates of mortality compared to those without diabetes.(5,8,9) Among PLWD, risk factors for mortality include being older, having more comorbidities, specifically chronic obstructive pulmonary disease (COPD), congestive cardiac failure (CCF), chronic kidney disease (CKD), and having a glycated haemoglobin (HbA1C) >10% compared with those with HbA1C of less than 6.5%.(4)
Diabetes is generally associated with an increased risk of infections, as hyperglycaemia is thought to cause immune system dysfunction.(5) The presence of diabetes and degree of hyperglycaemia seem to be independently associated with increased severity and mortality in COVID-19 infection.(6) Monocytes are found in high density in the lungs of patients with COVID-19, and hyperglycaemia with glycolysis in these cells promotes viral replication and cytokine production through a mitochondrial reactive oxygen species (ROS)/hypoxia-inducible factor (HIF1) pathway. The monocyte-derived cytokines then drive T cell dysfunction and epithelial cell death.(7) In those with DM or impaired glucose tolerance, glycaemic deterioration is a typical complication of COVID-19 infection, and this appears to be associated with the levels of inflammatory cytokines that affect the skeletal muscle and liver, which mediate the bulk of insulin-dependent glucose uptake.(6) Furthermore, HbA1C is an independent predictor of natural killer cell (NKC) activity in patients with Type 2 diabetes, which plays a crucial role in immunity. Elevated HbA1C levels are associated with reduced NKC activity, increasing susceptibility to respiratory viruses.(6)
While studies conducted in the Western Cape and KwaZulu-Natal concur with global findings,(8,9) there is no data on the relationship between diabetes and COVID-19 in the North-West Province of South Africa. According to Statistics SA, the North-West Province is the third poorest province after Limpopo and the Eastern Cape, and it is known that poverty and low income are associated with a higher prevalence of diabetes and diabetes-related complications.(10) We analyzed data from a public tertiary-level hospital in the North-West Province of South Africa to describe diabetes in a cohort of adults diagnosed with COVID-19 and identify factors associated with adverse health outcomes.
Methods
Study design and setting
This retrospective study was conducted at the Klerksdorp/Tshepong Hospital Complex (KTHC), a tertiary public hospital in the North-West Province. We included adults (≥18 years) who presented at KTHC between June 2020 and May 2021 (the peak of the first wave of the pandemic in South Africa) and diagnosed with COVID-19 (nasal or nasopharyngeal swab sample tested with rRT-PCR assay; National Health Laboratory Services (NHLS)).
Data collection
Patient demographic and clinical information were extracted from KTHC medical records. An Excel-based data collection tool was developed for the study. All data was captured in this tool and then exported into SAS 9.4 (SAS Institute Inc., Cary, NC, USA) for analysis.
Study variables
Demographics
Age, race, sex, and model of care were extracted from patient medical records. Age was defined according to the WHO age group classification for 2023.(11) The model of care was categorized as inpatient or outpatient based on whether the patient was managed in the hospital (inpatient) or as an outpatient.
Clinical variables
Self-reported co-morbid conditions (e.g., HIV, hypertension, heart failure etc.), and data on TB treatment were extracted from medical records. COVID-19 disease severity at presentation was classified as mild in asymptomatic patients who did not require oxygen, moderate if they required oxygen through nasal prongs or a face mask, and severe if they required high-flow nasal oxygen or ventilation (invasive or non-invasive).
A diabetic emergency was defined as a patient with diabetic ketoacidosis (DKA), a hyperglycaemic hyperosmolar state (HHS), or hypoglycaemia. The anti-diabetic medication used before or during the hospital stay was recorded and categorized as oral, injectable, or a combination of both.
Laboratory
We defined two groups: those without diabetes and those with Type 1 or Type 2 DM (the numbers with Type 1 were too small, n=7, to analyse separately). DM status was obtained from the self-reported medical history on presentation at KTHC. For those without a history of diabetes, if hyperglycaemic on admission, an HbA1C was done, and a diagnosis of diabetes was made if the HbA1C was >6.5%. For PLWD, control was defined for those who had HbA1C recorded, either previous (more than three months) or current (within the last three months), according to a target of <7% (‘good’ control) as recommended by the American Diabetes Association. The HbA1C of 7–8% was defined as ‘fair’ control and ‘poor’ at levels >8%.(12)
A cut-off >5 mg/L for C-reactive protein (CRP) levels was used as a marker of inflammation. The estimated glomerular filtration rate (eGFR) was categorized as normal if >90 ml/min, mildly reduced between 60-89, mild to moderate between 45-59, moderate to severe between 30-44, severely reduced between 15-29, and as kidney failure if less than 15 ml/kg/min.(13) D-dimer was categorized as normal or elevated using a cut-off of <0.23 mg/L vs. ≥0.23 mg/L, respectively.(14)
Intervention and Outcomes
If supplemental oxygen was required, it was documented as nasal prongs, facemasks, non-rebreather masks, or ventilatory support (using high-flow nasal oxygen, non-invasive ventilation, or invasive ventilation). Health outcomes were extracted from the patient's medical records. Adverse health outcomes of interest included inpatient vs. outpatient management, COVID-19 complications resulting in admission to the ICU, and all-cause mortality.
Data Analysis
Continuous data are presented as mean and standard deviation if normally distributed, median and interquartile range if not normally distributed, or as proportions for categorical variables. Patient characteristics and outcomes were compared between groups (i.e., those without diabetes and those with diabetes). We compared differences in characteristics and laboratory results using the ANOVA or T-test for parametric data (normal distribution), the Kruskal-Wallis or Mann-Whitney test for non-parametric data (not normal distribution), and the chi-squared or Fischer's exact test (scarce data) for proportions. The denominator was adjusted to reflect the missing values for missing or incomplete data (e.g., where a variable could not be calculated). Statistical significance was set at p<0.05 with a 95% confidence interval.
Using logistic regression, we explored the association between demographic and clinical characteristics and outcomes of interest (all-cause mortality, inpatient care, and admission to ICU). Variables were selected for subsequent multivariate analysis if the univariate assessment demonstrated a p-value <0.2. Only the crude RR and 95% confidence interval are presented for all-cause mortality outcomes because of the limited number of outcomes and the absence of data on length of stay.
The study was approved by the University of the Witwatersrand Human Ethics Research Committee (protocol M210154).
Results
Patient characteristics and the prevalence of diabetes
Of the patients tested for COVID-19 between June 2020 and May 2021, 846 were positive, and 360 (42.5%) were managed as outpatients. The prevalence of diabetes among those who tested positive for COVID-19 was 20.2% (n=171), 29 of whom were managed as outpatients and 142 as inpatients. The prevalence of diabetes was higher among those managed as inpatients (142/486; 29.2%) versus outpatients (29/360; 8.1%) (p<0.001).
Demographics
Patients were mainly female (61%), Black African (94%), and between the ages of 18-49 years (57%) (Table 2.1A). Compared to non-diabetics, PLWD were more likely to be managed as inpatients, were older, presented with moderate or severe COVID-19, and were more likely to present with one or more co-morbidities in addition, in particular, hypertension and CKD (p<0.05) (Supplementary Table 2.1B).
Characteristics of COVID-19 patients at presentation (n=846).
| Variable | Category | Total | No Diabetes | Type 1 or 2 |
|---|---|---|---|---|
| N=846 | N=675 (79.8%) | N=171 (20.2%) | ||
| Current HbA1C (n=129) | <7% “good” | 19/675 (2.8%) | 13/110 (11.8%) | |
| 7-8% “fair” | - | 20/110 (18.2%) | ||
| >8% “poor” | - | 77/110 (70.0%) | ||
| Newly diagnosed | - | 11/171 (6.4%) | ||
| Model of care | Inpatient | 486 (57.5%) | 344 (51.0%) | 142 (83.0%) * |
| Outpatient | 360 (42.5%) | 331 (49.0%) | 29 (17.0%) | |
| Gender | Male | 330 (39.0%) | 266 (39.4%) | 64 (37.4%) |
| Female | 516 (61.0%) | 409 (60.6%) | 107 (62.6%) | |
| Age (n=387) | Median (IQR) | 45.0 (32.0-58.0) | 40.5 (31.0-55.0) | 58.0 (47.0-64.0) |
| Young adults (18-34) | 122/387 (31.5%) | 113/310 (36.4%) | 9/77 (11.7%) | |
| Early middle-age (35-49) | 97/387 (25.1%) | 83/310 (26.8%) | 14/77 (18.2%) | |
| Late middle-age (50-65) | 119/387 (30.8%) | 80/310 (25.8%) | 39/77 (50.6%) * | |
| Older adults (>65 years) | 49/387 (12.6%) | 34/310 (11.0%) | 15/77 (19.5%) * | |
| Race | Black African | 54 (6.4%) | 632 (93.6%) | 160 (93.6%) |
| Other | 792 (93.6%) | 43 (6.4%) | 11 (6.4%) | |
| Severity of COVID-19 | Mild | 585 (69.2%) | 514 (76.1%) | 71 (41.5%) |
| Moderate | 206 (24.3%) | 134 (19.9%) | 72 (42.1%) * | |
| Severe | 55 (6.5%) | 27 (4.0%) | 28 (16.4%) * | |
| Laboratory Results | ||||
| Serum creatinine (n=572) | Median (IQR) | 86 (70-109) | 85 (71-107) | 88 (69-114) |
| Normal | 405/572 (70.8%) | 311/423 (73.5%) | 94/149 (63.1%) | |
| Elevated | 167/572 (29.2%) | 112/423 (26.5%) | 55/149 (36.9%) * | |
| eGFR (mL/min/1.73m2) (n=226/572=40%) | G1 Normal (≥90) | 12/226 (5.3%) | 9/157 (5.7%) | 3/69 (4.4%) |
| G2 Mild (60-89) | 44/226 (19.5%) | 31/157 (19.8%) | 13/69 (18.8%) | |
| G3a Mild – Moderate (45-59) | 60/226 (26.5%) | 47/157 (30.0%) | 13/69 (18.8%) | |
| G3b Mod – Severe (30-44) | 70/226 (31.0%) | 41/157 (26.1%) | 29/69 (42.0%) | |
| G4 Severe (15-29) | 31/226 (13.7%) | 21/157 (13.3%) | 10/69 (14.5%) | |
| G5 Kidney failure (<15) | 9/226 (4.0%) | 8/157 (5.1%) | 1/69 (1.5%) | |
| CRP (n=509) | Median (IQR) | 66 (3.0 – 155) | 46 (0 – 120) | 115 (49 – 196) |
| Normal (≤5 mg/L) | 136/509 (26.7%) | 126/369 (34.2%) | 10/140 (7.1%) | |
| Elevated (>5 mg/L) | 373/509 (73.3%) | 243/369 (65.9%) | 130/140 (92.9%) * | |
| D-Dimer (n=453) | Median IQR | 0.29 (0.0 – 0.93) | 0.23 (0.0 – 0.79) | 0.49 (0.20 – 1.36) |
| Normal <0.23 mg/L) | 197/453 (43.5%) | 164/336 (48.8%) | 33/117 (28.2%) | |
| Elevated (≥0.23 mg/L) | 256/453 (56.5%) | 172/336 (51.2%) | 84/117 (71.8%) * | |
±patient self-report
Abbreviations: COPD - chronic obstructive pulmonary disease; CRP - C-reactive protein; eGFR - estimated glomerular filtration rate; HbA1C - glycated haemoglobin; HIV Human immunodeficiency virus; IQR inter-quartile range; TB tuberculosis;
significant difference between two groups; p-value <0.05.
Annex to Table 2.1A: Comorbidities among COVID-19 patients at presentation (n=846).
| Variable | Category | Total | No Diabetes | Type 1 or 2 |
|---|---|---|---|---|
| N=846 | N=675 (79.8%) | N=171 (20.2%) | ||
| Pulmonary TB | No | 550/570 (96.5%) | 437/455 (96.0%) | 113/115 (98.3%) |
| Yes | 20/570 (3.5%) | 18/455 (4.0%) | 2/115 (1.7%) | |
| Hypertension | No | 498/846 (58.9%) | 465/675 (68.9%) | 33/171 (19.3%) |
| Yes | 348/846 (41.1%) | 210/675 (31.1%) | 139/171 (80.7%) * | |
| HIV | Negative | 665/845 (78.7%) | 527/674 (78.2%) | 138/171 (80.7%) |
| Positive | 180/845 (21.3%) | 147/674 (21.8%) | 33/171 (19.3%) | |
| Viral load (n=92/180=51%) | Undetectable (<20 copies/ml) | 26/92 (28.3%) | 25/74 (33.8%) | 1/18 (5.6%) |
| Detectable (≥20 copies/ml) | 66/92 (71.7%) | 49/74 (66.2%) | 17/18 (94.4%) | |
| CD4 cell counts (n=89/180=49%) | ≤200 cells/mm3 | 34/89 (38.2%) | 29/73 (39.7%) | 5/16 (31.2%) |
| 200 – 500 cells/mm3 | 37/89 (41.6%) | 30/79 (41.1%) | 7/16 (43.8%) | |
| ≥500 cells/mm3 | 18/89 (20.2%) | 14/73 (19.2%) | 4/16 (25.0%) | |
| Cardiac failure ± | No | 810/846 (95.7%) | 648 (96.0%) | 162/171 (94.7%) |
| Yes | 36/846 (4.3%) | 27/675 (4.0%) | 9/171 (5.3%) | |
| Hypothyroidism± | No | 837/846 (98.9%) | 669/675 (99.1%) | 168/171 (98.2%) |
| Yes | 9/846 (1.1%) | 6/675 (0.9%) | 3/171 (1.8%) | |
| COPD/Asthma± | No | 810/846 (95.7%) | 644/675 (95.4%) | 166/171 (97.1%) |
| Yes | 36/846 (4.3%) | 31/675 (4.6%) | 5/171 (2.9%) | |
| Chronic kidney disease± | No | 810/846 (95.7%) | 653/675 (96.7%) | 157/171 (91.8%) |
| Yes | 36/846 (4.3%) | 22/675 (3.3%) | 14/171 (8.2%) * |
±patient self-report
Abbreviations: COPD - chronic obstructive pulmonary disease; CRP - C-reactive protein; eGFR - estimated glomerular filtration rate; HbA1C - glycated haemoglobin; HIV Human immunodeficiency virus; IQR inter-quartile range; TB tuberculosis;
significant difference between two groups; p-value <0.05.
Laboratory findings
Among those with diabetes, the HbA1C results were available in 64% of patients (n=110/171) within the last three months, and of these, a majority (70%) had poorly controlled diabetes (defined as HbA1C >8%), and 11 patients were newly diagnosed (median HbA1C 9.5% IQR 7.5-10.9). On further analysis of those with poorly controlled diabetes, 35% had an HbA1C between 8-10%, 27% between 10-12% and 38% ≥12%.
Among PLWD with mild COVID-19, the median HbA1C was 10.4% (IQR 8.0-12.2%; n=37), 9.7% in those with moderate disease (IQR 7.8-12.4; n=45), and 8.9% in those with severe COVID-19 (IQR 7.1-11.0; n=21) (p=0.610). More than a third of patients with moderate or severe COVID-19 had diabetes (100/261). Among those with a reported HbA1C, over two-thirds had an HbA1C above 8% (48/71; 68%). PLWD presented with higher serum creatinine, CRP, and D-dimer levels than those without diabetes (p<0.05). CRP also differed significantly by the severity of COVID-19, ranging from 8 mg/L (IQR 0-72) in those with mild disease to 112 mg/L (IQR 59-196) in those with moderate and 185 mg/L (IQR 139-293) in those with severe disease (p<0.001). The median D-dimer in patients was 0.10 mg/L (IQR 0-0.45), 0.47 mg/L (IQR 0.24-1.27), and 0.95 ng/ml (IQR 0.3-2.27) in those with mild, moderate, and severe diabetes and COVID-19 respectively (p<0.001).
Clinical findings
The renal function in patients differed significantly in relation to the severity of COVID-19, with a median value for serum creatinine of 80 umol/L (IQR 68-101), 87 umol/L (IQR 72-110) and 105 umol/L (IQR 80-150) in those with mild, moderate and severe COVID-19, respectively (p<0.05). Those with CKD were more likely to be managed as inpatients, admitted to ICU (RR 2.91 95% CI 1.43-5.94), and more likely to die (aRR 14.31 95% CI 5.90-34.76) compared to those without CKD.
A third of patients with Type 1 or 2 DM (51/171) presented with two or more co-morbidities, whereas only 14% of those without diabetes (92/675) had two or more co-morbidities. Therefore, PLWD were two times more likely to present with two or more co-morbidities compared to those without diabetes (RR 2.19 95% CI 1.62-2.95), specifically hypertension and CKD.
Patient pathway analysis
Among PLWD, seven patients had Type 1 diabetes and were all inpatients, while 164 had Type 2 diabetes, with a ratio of 1 outpatient for every 4.9 inpatients. This ratio was 1:1.04 for COVID-19 patients without diabetes, whereby 83% of PLWD compared to 51% of those without diabetes were managed as inpatients (Figure 2.1).
Summary of patient flow for COVID-19 patients with or without diabetes. Mod = moderate; Interv = intervention; O2 = oxygen (nasal prongs, facemask, non-rebreather mask), other = ventilator support using high flow nasal oxygen, non-invasive ventilation or intubation; ICU = Intensive Care Unit
Among the inpatients with COVID-19 and diabetes, 23% (33/142) had complications and required admission to ICU, whereas only 8% of the COVID-19 non-diabetic inpatients were admitted to ICU (Figure 2.1). Most PLWD were on insulin (99/155; 63.9%); with some receiving both oral agents and insulin (39/155; 25.2%) and a minority on oral medication alone (16/155; 10.3%). There were no recorded cases of hyperosmolar hyperglycemic state (HHS) or hypoglycaemic coma over the study period in the 275 patients for whom this data was available. A single case of DKA was recorded in a patient with Type 2 diabetes and moderate COVID-19.
In those without diabetes, COVID-19 was classified as mild (n=514) in a majority (76%), moderate in 20%, and severe in 4% of patients. On the contrary, among those with COVID-19 and diabetes, the disease was classified as severe in 16% (n=28), moderate in 42% of patients, and mild in the rest. No patients with Type 1 diabetes had severe COVID-19; it was mild (n=5) or moderate Covid-19 (n=2). Just over half (51%) of the patients being treated for severe COVID-19 had diabetes (n=28/55).
All patients with moderate or severe COVID-19 were managed as inpatients, whereas 38% of those with mild COVID-19 were admitted (225/585). Over a third of the PLWD (n=62) and COVID-19 did not require any intervention, whereas 44% (n=76) required oxygen via nasal prongs or facemasks (compared to 22% of patients without diabetes). An additional 19% of PLWD (n=33) required high-flow nasal oxygen or ventilation (versus 5% of those without diabetes).
In terms of mortality, seven patients who required oxygen by facemask died, whereas 11 patients who required ventilation died of COVID-19 complications. Of the 18 deaths, eight had diabetes, and ten did not. Among the eight PLWD that died, only one received oxygen by facemask, whereas the rest (n=7) required ventilation. For those without diabetes, six received oxygen by the facemask, and four required ventilation.
Outcomes
For the outcomes of interest (i.e., admitted to ICU and all-cause mortality) and whether patients required any supplemental oxygen and/or any ventilatory support (intervention), those with diabetes had significantly poorer outcomes than those without diabetes (p<0.05) (Table 2.2).
Outcomes and interventions for patients presenting with COVID-19 (n=846).
| Variable | Category | Total | No Diabetes | Type 1 or 2 | P-value |
|---|---|---|---|---|---|
| N=846 | N=675 (79.8%) | N=171 (20.2%) | |||
| Progression | Admitted to ICU | 61/845 (7.2%) | 28/674 (4.2%) | 33/171 (19.3%) | <0.001 |
| Final outcome | Mortality | 18/846 (2.1%) | 10/675 (1.5%) | 8/171 (4.7%) | 0.010 |
| Discharge | 828/846 (97.9%) | 665/675 (98.5%) | 163/171 (95.3%) | ||
| Intervention | No intervention | 554/846 (65.5%) | 492/675 (72.9%) | 62/171 (36.3%) | <0.001 |
| O2 – oxygen * | 224/846 (26.5%) | 148/675 (21.9%) | 76/171 (44.4%) | ||
| Other intervention | 68/846 (8.0%) | 35/675 (5.2%) | 33/171 (19.3%) |
Mod - moderate; O2 – oxygen (nasal prongs, facemask, non-rebreather mask), other – ventilator support using high flow nasal oxygen, non-invasive ventilation, or intubation. Abbreviation: ICU Intensive Care Unit
Factors associated with adverse health outcomes
Model of care: inpatient versus outpatient management
Close to two-thirds of the sample (58%) were managed as inpatients. Factors associated with receiving inpatient care include Type 1 or 2 DM (RR 1.63 95% CI 1.47-1.80), being older (≥50 years; RR 1.51 95% CI 1.35-1.68), presenting with moderate to severe COVID-19 (RR 3.16 95% CI 2.42-4.12). Laboratory features associated with inpatient care included eGFR <60 mls/min (RR >2.0), high CRP (RR 2.43 95% 1.97-3.00), and high D-dimer (RR 1.70 95% CI 1.51-1.92). Comorbid conditions such as CKD, CCF, HIV and pulmonary TB were also associated with inpatient management. After adjusting for covariates, moderate to severe COVID-19 (aRR 1.16 95% CI 1.04-1.31), but not diabetes (aRR 1.02 95% CI 0.84-1.25), was associated with inpatient management (Table 2.3).
Factors associated with adverse health outcomes - including inpatient versus outpatient management, admission to ICU, and all-cause mortality (n=486).
| Inpatient care n=486 | ICU n=60 | Mortality n=18 | ||||
|---|---|---|---|---|---|---|
| Variable | Category | RR (95% CI) | aRR (95% CI) | RR (95% CI) | aRR (95% CI) | RR (95% CI) |
| Patient category | No diabetes | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 |
| Diabetes (Type 1 or 2) | 1.63 (1.47-1.80) | 1.02 (0.84-1.25) | 4.65 (2.89-7.47) | 1.99 (1.21-3.27) | 3.16 (1.27-7.88) | |
| HbA1C ≤8% | 1.42 (1.22-1.64) | 1.00 (0.90-1.11) | 4.01 (2.56-7.14) | 1.61 (0.92-2.80) | 2.81 (0.90-8.79) | |
| HbA1C >8% | 1.90 (1.74-2.06) | 1.05 (0.81-1.36) | 5.46 (3.14-9.49) | 1.62 (0.93-2.84) | 3.60 (1.16-11.20) | |
| Gender | Male | 1.01 (0.90-1.14) | 1.17 (0.72-1.90) | 1.25 (0.50-3.14) | ||
| Female | 1.0 | 1.0 | 1.0 | |||
| Age (n=387) | Young adults (18-34) | 1.0 | 1.0 | 1.0 | ^^ | |
| Early middle-age (35–49) | 1.02 (0.83-1.25) | 1.02 (0.59-1.79) | 0.60 (0.22-1.64) | ^^ | ||
| Late middle-age (50–65) and older adults (>65 years)# | 1.51 (1.35-1.68) | 1.01 (0.75-1.36) | 1.48 (0.86-2.54) | ^^ | ||
| Race | Black African | 1.15 (0.88-1.51) | 0.53 (0.25-1.10) | 1.16 (0.16-8.55) | ||
| Other | 1.0 | 1.0 | 1.0 | |||
| Severity of COVID-19 | Mild | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 |
| Moderate or Severe# | 3.16 (2.42-4.12) | 1.16 (1.04-1.31) | 25.20 (10.21-62.17) | 10.44 (3.26-33.44) | 38.10 (5.10-284.81) | |
| Serum creatinine (n=572) | Normal | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 |
| Elevated | 1.26 (1.19-1.34) | 1.04 (0.40-2.72) | 2.44 (1.52-3.91) | 1.36 (0.86-2.15) | 8.49 (2.84-25.41) | |
| eGFR (mL/min/1.73m2) (n=226/572=40%) | G1 and G2 (≥60) | 1.0 | ^ | 1.0 | ^^ | |
| G3a Mild – Moderate (45-59) | 2.09 (1.92-2.29) | ^ | 1.69 (0.75-3.82) | ^^ | ||
| G3b Mod – Severe (30-44) | 2.10 (1.93-2.29) | ^ | 2.20 (1.12-4.35) | ^^ | ||
| G4 and G5 (<30) | 2.03 (1.82-2.26) | ^ | 2.54 (1.14-5.62) | ^^ | ||
| CRP (n=509) | Normal (≤5 mg/L) | 1.0 | 1.0 | 1.0 | 1.0 | ^ |
| Elevated (>5 mg/L) | 2.43 (1.97-3.00) | 1.39 (0.90-2.14) | 10.42 (2.58-42.10) | 0.98 (0.24-3.95) | ^ | |
| D-dimer (n=453) | Normal <0.23 mg/L) | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 |
| Elevated (≥0.23 mg/L) | 1.70 (1.51-1.92) | 1.02 (0.64-1.63) | 5.92 (2.58-13.58) | 1.88 (0.97-3.67) | 8.46 (1.10-65.01) | |
| Yes | 1.55 (no CI) | ^ | 2.91 (1.43-5.94) | 1.34 (0.74-2.43) | 14.31 (5.90-34.76) | |
| Newly diagnosed DM | No | 1.0 | 1.0 | 1.0 | ^ | |
| Yes | 0.87 (0.60-1.26) | 2.01 (0.86-4.69) | 0.98 (0.24-3.95) | ^ | ||
| Admitted to ICU | No | N/A | N/A | 1.0 | ||
| Yes | N/A | N/A | 25.71 (9.99-66.12) | |||
# numbers too small so categories were collapsed;
^the standard maximum likelihood estimation method failed to converge, due to zero cells or a single predictor with more than two level. Adjusted RR estimates are for the complete model (i.e., regardless of HbAIC levels).
Grey italics is the same model except that diabetes is further stratified by HbAIC in the sub-analysis (controlled versus poorly controlled).
^^age was missing and these estimates could not be calculated;
Bold = p-value <0.05; N/A not applicable.
Abbreviations: CRP - C-reactive protein; eGFR - estimated glomerular filtration rate; HbA1C - glycated haemoglobin; ICU Intensive Care Unit; TB tuberculosis; DM diabetes mellitus; RR Relative Risk; aRR adjusted Relative Risk;
Admission to ICU
From the univariate analysis, DM (RR 4.65 95% CI 2.89-7.47), moderate or severe COVID-19 (RR 25.20 95% CI 10.21-62.17), high serum creatinine (RR 2.44 95% CI 1.52-3.91) or eGFR <45 mls/min (RR >2.0), high CRP (RR 10.42 95% CI 2.58-42.10), and elevated D-dimer (RR 5.92 95% CI 2.58-13.58) were associated with admission to ICU. Patients with hypertension (RR 2.36 95% CI 1.43-3.89) or CKD (RR 2.91 95% CI 1.43-5.94) were also more likely to be admitted to ICU. Multivariate analysis showed that having DM (aRR 4.65 95% 2.89-7.47) and moderate or severe COVID-19 (aRR 10.44 95% CI 3.26-33.44) were more likely to be admitted to ICU – and patients admitted to ICU were more likely to die (RR 25.71 95% CI 9.99-66.12).
All-cause mortality
Factors associated with all-cause mortality included DM (RR 3.16 95% CI 1.27-7.88), moderate or severe COVID-19 (RR 38.10 95% CI 5.10-284.81), high serum creatinine (RR 8.49 95% CI 2.84-25.41), elevated D-dimer (RR 8.46 95% 1.10-65.01), hypertension (RR 3.72 95% CI 1.34-10.34) and chronic kidney disease (RR 14.31 95% CI 5.90-34.76). From unadjusted analyses, diabetes was associated with all adverse health outcomes. After adjusting for covariates, PLWD were twofold more likely to be admitted to ICU than those without diabetes. From the sub-analysis, it should be noted that the magnitude of the association was greater among PLWD with poorly controlled diabetes (HbA1C >8%) than those with controlled diabetes (HbA1C ≤8%).
Discussion
The HbA1C did not differ significantly in PLWD with differing severities of COVID-19, which may be due to a more significant impact of current hyperglycaemia in disease severity rather than chronic control. This is in keeping with another South African study, which found no association between HbA1C and the need for ventilation or death.(15) However, PLWD were more likely to have more severe disease, require intervention or admission to ICU, and have a higher risk of mortality.
The prevalence of diabetes varied by category; from 30% among those managed as inpatients, 8% among those managed as outpatients, 51% among those with severe COVID-19, 55% among those admitted to ICU, and 44% among those that died. This is in concert with findings by Yan Yang et al. who found a higher prevalence of diabetes with increasing severity of COVID-19 infection.(8)
Most of those who were admitted for management of their COVID-19 had poorly controlled diabetes (HbA1C >8.0%). On average, South Africa has recorded poor glycaemic control, which is attributed to low awareness of the disease, sub-optimal treatment, and fragmented health systems.(16) Unlike other studies, there were fewer diabetic emergencies; fewer patients with diabetes and pulmonary TB or COPD were reported in the studied cohort, although this may be due to missing data.(17)
Consistent with results from a systematic review where data on comorbidities was extracted from 34 studies, we noted that mortality among hospitalized patients with COVID-19 in PLWD was higher compared to those without diabetes. However, our estimate (OR 3.16 95% CI 1.27-7.88) is higher than the pooled estimate from a recent systematic review and meta-analysis (aOR 1.52 95% CI 1.36–1.69).(17) We did not evaluate acute kidney injury but found an association between chronic kidney disease and mortality among hospitalized COVID-19 patients (RR 14.31 95% 5.90-34.8).
Limitations
Because of the retrospective nature of the data collection, some results were incomplete, and data analysis was limited. Information on the intensification of therapy for glycaemic management, while admitted to the hospital was also unavailable, and the duration of diabetes in PLWD was not known, all of which could impact analyses.
Due to the overwhelming nature of the pandemic and the strain caused on healthcare resources, patients deemed stable were down referred to another facility where care was continued until discharge. While the outcome (discharge, transfer, or death) was consistently recorded, the outcome date was not, which impacted the ability to calculate the length of hospital stay. We could not present the survival rate for the outcome of interest (i.e., all-cause mortality) because we could not calculate person time without the date that the outcome was assigned. For critical care studies, which have endpoints close to the therapy and disease (e.g., 28-day mortality), survival analysis is likely to detect significant differences that are not clinically significant.(18)
Conclusion
Diabetes can weaken the immune system, which could contribute to the more severe outcomes observed in COVID-19 patients with diabetes. For clinical management, this information suggests that healthcare providers must be particularly vigilant when caring for COVID-19 patients with diabetes, including closer monitoring, more intensive medical intervention, and specific strategies to manage their diabetes, including treating acute hyperglycaemia. It also serves as a call to optimise public health initiatives to prevent diabetes, increase screening programs and improve management of diabetes to minimise the burden of future pandemics on both patients and healthcare services.