Prevalence of pre-diabetes, newly diagnosed diabetes, and associated COVID-19 disease severity in patients hospitalized at Sir Ketumile Masire Teaching Hospital, Botswana

1 Introduction

The coronavirus disease 2019 (COVID-19), caused by the novel SARS-CoV-2 virus, was declared a global pandemic by the WHO on March 11, 2020 (1–3). The pandemic resulted in over 765 million confirmed cases and more than 6.9 million deaths worldwide (4). The disease predominantly affects the respiratory tract, with severity ranging from asymptomatic or mild to severe or critical illness (5, 6). A key observation throughout the pandemic has been the increased morbidity and mortality in vulnerable populations, particularly among the elderly and individuals with pre-existing comorbid conditions, including diabetes (7, 8). Various studies show an increased severity of the COVID-19 disease in the diabetic population (9–11).

Diabetes mellitus is a disorder of glucose metabolism that is a leading cause of morbidity and mortality, often attributed to cardiovascular diseases (12). It is also the leading cause of end-stage renal disease, adult-onset blindness, and non-traumatic lower extremity amputations (13–15). Diabetes is one of the fastest-growing global health emergencies of the 21st century. The International Diabetes Federation estimates that approximately 473 million people were living with diabetes in 2019, with figures projected to rise to 700 million by 2045 (12). Three-quarters (75%) of those with diabetes were living in low- and middle-income countries (16). The diabetic population had a disproportionate burden of the COVID-19 disease, as studies have shown a strong association between diabetes and worse outcomes (17–19).

The precise mechanisms linking COVID-19 and diabetes are still under investigation, but several hypotheses have been proposed (20–22). The SARS-CoV-2 virus, like its predecessor Middle East Respiratory Syndrome (MERS), uses its spike glycoprotein for binding to host angiotensin-converting enzyme 2 (ACE2) for cell entry (23–25). These receptors are highly expressed not only in the lungs but also in various other organs, including the pancreas, liver, and adipose tissue. Viral binding to pancreatic beta-cells may directly impair insulin production. Furthermore, the intense inflammatory response triggered by the infection, known as a “cytokine storm”, can induce significant insulin resistance and cellular damage, leading to severe hyperglycemia (13, 26). These acute effects may exacerbate pre-existing low-grade inflammation and metabolic dysregulation common in diabetes, leading to worse outcomes. However, our study did not measure pancreatic function and cannot distinguish these mechanisms.

While the association between established diabetes and severe COVID-19 is well documented (17, 27). Although reports from other sub-Saharan African countries, including South Africa (28, 29), Nigeria (30), and Kenya (31), had begun to describe the relationship between diabetes and COVID-19 outcomes, fewer studies had examined the prevalence of previously undiagnosed pre-diabetes and diabetes among hospitalized patients in this region (32), and on how these glucose abnormalities relate to disease severity.

We therefore undertook this study to address this gap in Botswana and to contribute to the emerging sub-Saharan African evidence base. We determined the prevalence of pre-diabetes and newly diagnosed diabetes and factors associated with COVID-19 disease severity and in-hospital mortality among this cohort.

2 Methods

2.1 Study design and setting

We conducted a single-center, retrospective cross-sectional study at Sir Ketumile Masire Teaching Hospital (SKMTH) in Gaborone, Botswana. SKMTH was designated as the primary COVID-19 isolation and management hospital at the time of the study. The study period was from June 2020 to June 2021, during which 1,832 patients were admitted with confirmed SARS-CoV-2 infection.

2.2 Study population and sampling

The study population included all adult patients (age ≥18 years) admitted to SKMTH with a laboratory-confirmed SARS-CoV-2 positive test. Patients with a known history of diabetes mellitus, pregnant patients, and those with incomplete medical records were excluded. A total of 992 files met the inclusion criteria. From this eligible population, a simple random sampling technique was used to select 100 patient files for review. At the time of study planning, no Botswana-specific prevalence was available, and our sample size was based on the available regional prevalence estimate of pre-diabetes from Uganda. This calculation was intended primarily to support the estimation of the prevalence of dysglycemia in the study population. Analyses of associations with clinical outcomes were therefore considered exploratory (32).

2.3 Data collection

A standardized data abstraction sheet was used to collect epidemiological, clinical, laboratory, treatment, and outcome data from patient records. Patient information such as age, sex, marital status, employment, location of residence, and comorbidities (e.g., hypertension, HIV, sepsis, asthma) was collected. Laboratory data included standard blood counts, renal and liver function tests, creatinine kinase, and, most importantly, HbA1c, random blood sugar (RBS), and fasting blood sugars (FBS). Oxygenation requirements were used to classify disease severity as mild (low-flow oxygen), moderate (CPAP/high-flow oxygen), or severe (invasive mechanical ventilation).

2.4 Classification of glucose abnormality

Participants were classified into three categories: normal glucose (HbA1c <5.7% or FBS <5.6 mmol/L), pre-diabetes (HbA1c between 5.7% and 6.4%, or FBS between 5.6-6.9 mmol/L), and newly diagnosed diabetes (HbA1c ≥6.5% or FBS ≥7.0 mmol/L, or RBS ≥11.1 mmol/L with symptoms of diabetes). The primary criterion for classification was the first available HbA1c measurement upon admission. In cases where HbA1c was missing or not ordered, classification was based on random or fasting blood glucose levels. Missing data were not imputed. Analyses were conducted using available-case denominators for each variable and complete-case data for each model. Denominators are shown in the tables.

2.5 Statistical analysis

Data were captured in REDCap and exported to STATA and R for analysis. Descriptive statistics were used to summarize participant characteristics, with proportions estimated with 95% confidence intervals using binomial exact methods. Univariate and multivariate Poisson regression with robust standard errors were performed to determine factors associated with severe disease and mortality. Variables with a p-value of <0.2 in the univariate analysis were included in the multivariate model. Where simultaneous inclusion of screened categorical variables resulted in sparse-cell instability or quasi-complete separation, the more stable specification was retained. We also conducted an HbA1c-restricted sensitivity analysis. A p-value of <0.05 was considered statistically significant.

2.6 Ethical considerations

The study was conducted in accordance with the Helsinki Declaration. This study was approved by the University of Botswana Institutional Review Board (UBR/RES/IRB/GRAD/299), the Ministry of Health (HPRD: 6/14/1), and the SKMTH ethics committee (SKMTH 01/14). Individual informed consent was waived due to the retrospective nature of the study, which used anonymous clinical data. Patient confidentiality was protected by assigning subjects case study numbers, and no personally identifiable information was entered into the database.

3 Results

3.1 Participant characteristics and comorbidities

A total of 100 patient files were reviewed, all of whom were adults aged 18 years or older (Table 1). The mean age was 49.5 years (SD 13.7), with 63% of participants falling in the 36–60-year age category. The majority of the cohort was male (73%). Most participants were of African descent (98%) and resided in urban areas (79%). The most prevalent comorbid conditions were hypertension (34%) and HIV (24%). Most participants (57%) received steroids during their hospital stay (Figure 1).

Table 1. Socio-demographic and clinical characteristics of the study participants (N = 100).

3.2 Prevalence of pre-diabetes and diabetes

Pre-diabetes was detected in 35% (95% CI: 25.7–45.2) and newly diagnosed diabetes in 43% (95% CI: 33.1–53.6) of participants. Only 22% of participants had normal glucose levels and normal HbA1c. A total of 84 (84%) participants had HbA1c values available, while 16 were classified based on fasting or random glucose alone. Pre-diabetes and newly diagnosed diabetes were common in the cohort. Because glucose values obtained during acute illness may reflect transient stress hyperglycemia, a sensitivity analysis was performed, restricting the definition of glycemic status to participants with available HbA1c measurements.

3.3 Risk factors for COVID-19 disease severity and mortality

COVID-19 severity was categorized as mild (50%), moderate (32%), or severe (18%). Univariate and multivariate Poisson regression analyses were performed to identify risk factors for severe disease (Table 2a). In the adjusted analysis, newly diagnosed diabetes was associated with severe COVID-19 (aRR=5.74; 95% CI: 1.31-25.2, p=0.02); however, the confidence interval remained wide, indicating limited precision.

Table 2a. Poisson regression table with factors associated with COVID-19 disease severity using a combined glycemic definition.

Multivariable severity model retained predictors with univariable p < 0.20 and forced inclusion of diabetes. Marital status met the screening criterion in the severity analyses but was omitted from the adjusted model because inclusion with employment status created sparse-cell instability and quasi-complete separation.

Males had a reduced risk for severe COVID-19 compared to females (aRR=0.40; 95% CI: 0.16-1.0, P = 0.05). Having sepsis was also strongly associated with a 5.9 times increased risk for severe disease (aRR=5.9; 95% CI: 1.83-18.7, P = 0.003). HIV, hypertension, and alcohol use were not significantly associated with severe disease. In the HbA1c-restricted sensitivity analysis (Table 2b), the association between newly diagnosed diabetes and severe COVID-19 was attenuated compared with the main analysis, and the confidence intervals were wider, indicating reduced precision, but still statistically significant (aRR= 3.93; 95% CI:1.14-13.57. p=0.03). These findings suggest that part of the dysglycemia observed in the full cohort may have been influenced by classification based solely on admission glucose values.

Table 2b. Poisson regression table with factors associated with COVID-19 disease severity using an HbA1c-restricted sensitivity analysis.

Multivariable severity model retained predictors with univariable p < 0.20 and forced inclusion of diabetes. Marital status met the screening criterion in the severity analyses but was omitted from the adjusted model because inclusion with employment status created sparse-cell instability and quasi-complete separation.

Of the 98 participants with available outcome data, 44 (44.9%) died during hospitalization and 54 (55.1%) were discharged alive. In the full cohort (Table 3a), newly diagnosed diabetes was associated with increased in-hospital mortality in univariable analysis (RR 2.36, 95% CI: 1.15–4.84; p=0.019), but this association did not persist after adjustment (aRR 1.97, 95% CI: 0.91–4.25; p=0.085). Male sex remained associated with lower mortality, while HIV, severe COVID-19, and sepsis were not independently associated with mortality in the adjusted model. Remdesivir use remained strongly associated with mortality.

Table 3a. Univariate multivariate analysis of risk factors of in-hospital mortality using a combined glycemic definition.

The multivariable mortality model retained predictors with univariable p < 0.20 and forced inclusion of diabetes. Estimates for remdesivir should be interpreted cautiously because they are likely confounded by indication.

In the HbA1c-restricted sensitivity analysis (Table 3b), newly diagnosed diabetes was associated with increased in-hospital mortality in both univariable (RR = 3.45, 95% CI: 1.17–10.21; p=0.025) and multivariable analyses (aRR=3.71, 95% CI: 1.49–9.26; p=0.005). Male sex remained associated with lower mortality, and HIV infection was independently associated with mortality. Severe COVID-19 and sepsis were not independently associated with mortality after adjustment, while remdesivir use remained strongly associated with mortality.

Table 3b. Univariate multivariate analysis of risk factors of in-hospital mortality using a HbA1c-restricted sensitivity analysis.

Multivariable mortality model retained predictors with univariable p < 0.20 and forced inclusion of diabetes. Estimates for remdesivir should be interpreted cautiously because they likely reflect confounding by indication.

4 Discussion

This retrospective study provides crucial insights into the clinical characteristics and outcomes of COVID-19 in patients with previously undiagnosed glucose abnormalities in Botswana. We found a significantly high prevalence of both newly diagnosed diabetes (43%) and pre-diabetes (35%) among our cohort. These combined figures (78%) are higher than those found in other countries, such as China, where a 2017 national survey found a diabetes prevalence of 12.8% (18). The high prevalence in our cohort could be attributed to several factors, including the mean age of 50 years (a high-risk group for developing type 2 diabetes) and the significant oxidative stress and tissue inflammation induced by the SARS-CoV-2 infection.

Our study provides strong evidence that newly diagnosed diabetes is a significant risk factor for severe COVID-19 disease. The adjusted relative risk of 5.74 for severe disease in diabetic patients is consistent with a study by Yi et al. (17), which also found that severe cases were more common in diabetic patients. The strong association between sepsis and severe disease (aRR=5.9) may reflect the overlap between the two conditions, as the SARS-CoV-2-induced cytokine storm can mimic sepsis and lead to similar outcomes.

An unexpected finding was the reduced risk for severe COVID-19 in males compared to females (aRR=0.40). This contradicts some international studies, such as one by Grasselli et al. in Italy, who found lower survival rates for older men (33). The difference could be due to the disproportionate number of males in our study (73%) or differences in comorbidities and age distribution between the cohorts. For mortality, the association with newly diagnosed diabetes was sensitive to the classification method. It was attenuated and no longer statistically significant in the full cohort after adjustment, but remained significant in the HbA1c-restricted sensitivity analysis. This suggests that the relationship between dysglycemia and mortality should be interpreted cautiously and may be influenced by both classification approach and limited sample size.

The significant association between Remdesivir use and mortality (aRR=9.48) should be interpreted with caution as Remdesivir was part of the standard treatment protocol for severe COVID-19 cases, this finding likely reflects confounding by indication, where sicker patients were more likely to receive the drug and, consequently, have a higher mortality rate, in contrast to other studies that showed a beneficial effects for Remdesivir used reducing the time to recovery of hospitalized patients who require supplemental oxygen and resulting in reduction in mortality and favorable safety profiles (34, 35). Similarly, corticosteroid use in 57% of participants could have affected blood sugar levels, although their short-term use and the preference for HbA1c in our study likely minimized this effect.

Our findings should be interpreted in the context of emerging African literature showing that diabetes is an important comorbidity in COVID-19. In South Africa, Dave et al. (28) reported that people living with diabetes had substantially higher risks of hospitalization and death from COVID-19 in a large virtual cohort from the Western Cape. Similarly, multicenter hospital data from Kenya showed that comorbid conditions, including diabetes, were common among admitted patients and contributed to poor outcomes (31). Broader African data from critical care settings have also shown high mortality among severely ill patients with COVID-19 and have highlighted the contribution of non-communicable comorbidity to poor prognosis (36). However, most of these studies focused on patients with established comorbid disease, whereas our study specifically excluded patients with known diabetes and therefore examined the burden of previously unrecognized dysglycemia detected during admission. In this regard, our study adds Botswana-specific evidence and suggests that a substantial proportion of hospitalized patients with COVID-19 may have unrecognized metabolic dysfunction that only becomes apparent during acute illness.

The high prevalence of undiagnosed glucose abnormalities highlights a critical public health issue in Botswana. Many individuals may be living with pre-diabetes or diabetes without their knowledge, and the COVID-19 pandemic may have served as a wake-up call for these patients and the healthcare system. These findings underscore the importance of integrating routine glucose screening (e.g., HbA1c) into the care of all patients presenting with COVID-19. Early detection and management can mitigate long-term complications and improve quality of life. An important methodological consideration is that not all participants had HbA1c available at admission. Although HbA1c was available for most participants (84%), 16% were classified using fasting or random blood glucose values only. In the context of acute COVID-19 illness, corticosteroid exposure, and physiological stress, these glucose-based measures may reflect transient stress hyperglycemia rather than underlying chronic dysglycemia. To address this, we performed an HbA1c-restricted sensitivity analysis. In that analysis, the association between newly diagnosed diabetes and severe COVID-19 was attenuated but remained statistically significant, albeit with wider confidence intervals, suggesting that some of the dysglycemia identified in the primary analysis may have reflected acute illness-related hyperglycemia.

We observed a relatively high in-hospital mortality compared with the proportion classified as having severe COVID-19 suggests that the oxygenation-based severity definition may not have fully captured the overall burden of critical illness in this cohort. Severity was retrospectively classified based on oxygenation requirements documented in the clinical records, which may have underestimated true disease severity in some patients. For example, patients managed with CPAP or in high-care settings may have remained classified as moderate despite substantial clinical instability. In addition, mortality may have been influenced by complications such as pulmonary embolism, sepsis, and other forms of organ dysfunction, including renal impairment, which were not fully captured by the study variables. Because these complications were not systematically recorded, their contribution to mortality could not be formally evaluated in this analysis.

This study has several limitations. The participants with previously known diabetes were excluded to allow focused assessment of pre-diabetes and newly diagnosed diabetes. Consequently, the study cohort is not representative of all hospitalized COVID-19 patients, and the findings should be interpreted as specific to individuals without established diabetes at admission. There is a possibility of selection bias due to the study’s retrospective design and the availability of patient files with complete data. The American Diabetes Association (ADA) recommends a repeat test to confirm a diabetes diagnosis in asymptomatic individuals, which was not feasible in this study. Because participants with previously known diabetes were excluded by design, our findings apply specifically to patients with newly identified dysglycemia and should not be extrapolated to the entire hospitalized COVID-19 population. The study did not also account for the differences in specific COVID-19 variants circulating at the time, which may have unique clinical and biochemical manifestations. Lastly, although this study provides baseline data, the modest sample size and single-center study may limit the generalizability of these findings. However, as noted, SKMTH was the main isolation center at the peak of the pandemic in Botswana.

5 Conclusion

In conclusion, our study provides a primary and comprehensive description of clinical features of COVID-19 in patients with normoglycemia, pre-diabetes, and newly diagnosed diabetes. The prevalence of both pre-diabetes and diabetes is very high in our community, and diabetes is associated with increased COVID-19 disease severity and mortality. The new onset of diabetes in relation to COVID-19 provides an opportunity to observe these patients long-term. We recommend that all healthcare practitioners closely monitor glucose values for patients presenting with COVID-19 and initiate early treatment. Long-term follow-up for patients with new-onset hyperglycemia is also urgently needed, in line with regional findings that up to 73% of diabetes cases in Africa remain undiagnosed. Future studies need to follow up with these patients to determine whether the glucose derangements are temporary or permanent.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Burtram Clinton Fielding, Stellenbosch University, South Africa

Reviewed by: Mohammed Rohaim, Cairo University, Egypt

Danzil Eugene Joseph, Stellenbosch University, South Africa