Discussion
In our sample of patients diagnosed with COVID-19 and type 2 diabetes or at risk of type 2 diabetes, 22% were hospitalized, 5% were admitted to ICUs, 2% were intubated, and 3% died within 30 days of receiving a COVID-19 diagnosis from March 2020 to February 2021.
Vulnerable populations experienced worse outcomes with COVID-19 infections throughout the pandemic in 2020, including the elderly, people of color, and those with pre-existing or comorbid conditions, such as diabetes.1 15–18 Our findings were consistent with prior studies. As our analysis included data until early 2021, we were also able to show that the prevalence of severe outcomes decreased over time. Compared with the initial period (March to June 2020), patients diagnosed with COVID-19 in later periods were ~40%–65% less likely to experience the most severe COVID-19 outcomes. Although this shift in outcomes may reflect changing demographics in that healthier patients were infected later in the pandemic, the shift also suggests advances in the science and understanding of COVID-19 and the implementation of medical protocols to better respond to and manage COVID-19.
In the analyses of racial differences, we found non-Hispanic Black and Hispanic patients with type 2 diabetes were younger and had higher HbA1c levels (online supplemental table S1). These observations were consistent with population-level data reported in prior studies. In an analysis of the National Health and Nutrition Examination Survey (NHANES) data, Wang et al reported that Mexican American and non-Hispanic Black adults had a significantly younger mean age at diabetes diagnosis (mean age 47.2 and 44.9 years, respectively) relative to non-Hispanic White adults (mean age 51.8 years).19 Another NHANES study found non-Hispanic Black and Mexican American adults were significantly less likely to achieve HbA1c targets (60.4% and 55.7%, respectively) as compared with non-Hispanic White adults (68.3%).20 After controlling for covariates, our data indicated that patients who were non-Hispanic Black or Hispanic had significantly higher odds of hospitalization than White patients, and the associations were consistent across age, sex, and COVID-19 diagnosis periods. Although disparities were not apparent in the most severe outcome category (ie, ICU admission, intubation, or death), individuals in minority groups may have higher COVID-19 incidence or be diagnosed later, leading to increased odds of hospitalization. Once admitted to a hospital, the disease prognosis of these patients was comparable to that of their White counterparts. These findings were consistent with studies conducted in large health systems in Houston and Milwaukee,21 22 in which non-Hispanic Black and Hispanic patients had a higher likelihood of hospitalization, but there were no differences in ICU utilization, in-hospital mortality, ventilator use, or treatment parameters.22
Patients with type 2 diabetes using insulin had significantly higher odds of adverse events, which was expected, as insulin use is an indicator of an advanced stage of diabetes. This finding was consistent with other large studies.23–25 For instance, among 64 892 veterans with diabetes and COVID-19, insulin use was associated with higher odds of hospitalization and risk of death.25 A meta-analysis of 33 studies reported an OR of 1.70 (95% CI 1.33 to 2.19) as comparing patients on insulin to non-users.26 Because the coronavirus replicates faster in a high glucose environment and glucose fluctuations make it more challenging to treat COVID-19,2 patients with uncontrolled diabetes may experience more severe COVID-19 outcomes. This suggests that providers treating patients with COVID-19 with diabetes should remain vigilant, and patients with diabetes on insulin may require further attention regarding COVID-19 prevention and management.
Our study also suggested that metformin use was associated with a protective effect; patients with diabetes on metformin had ~40% lower odds of severe COVID-19 outcomes than those not taking metformin, after adjusting for HbA1c levels. Some previous studies noted similar reduced risks in patients on metformin.24 25 In a nationwide observational cohort study in England, metformin use was associated with 23% reduced mortality in people with type 2 diabetes.25 A meta-analysis reported a pooled OR of 0.54 (95% CI 0.47 to 0.62) for metformin use and mortality.26 Several potential mechanisms might explain the protective effect of metformin, either by reducing the likelihood of SARS-CoV-2 infection or by decreasing COVID-19 severity. Metformin reduces blood glucose levels; worse glucose control has been associated with higher mortality and end-organ complications in patients with COVID-19.18 Metformin could decrease endothelial injury, an important factor and therapeutic target in mitigating COVID-19 complications.27 28 Metformin also inhibits neutrophil extracellular trap release, alleviating the development of downstream lung injury.29–31 Metformin could also decrease the viral cycle, with efficacy against Middle East respiratory syndrome and COVID-19.32 33 While we do not have data regarding medication dosage or combination therapies to support the causal relationship between metformin and less severe COVID-19 outcomes, further research on these mechanisms may contribute to the prevention of COVID-19 and other viral infections.
GLP-1R agonists and DPP-4 inhibitors are glucose-regulating medications known to have anti-inflammatory effects that may improve outcomes in patients with SARS-CoV-2 infection. In a multination study of TriNetX COVID-19 Research Network, the use of GLP-1R agonists was associated with significant reductions in hospital admission, respiratory complications, and mortality; the use of DPP-4 inhibitors was associated with a reduction in respiratory complications and subsequent hospitalizations.34 Moreover, in a study among veterans with diabetes and COVID-19, SGLT2 inhibitors and GLP-1R agonist were associated with lower odds of hospitalization; SGLT2 inhibitor use was also associated with lower odds of death.35 In our study, we found patients with type 2 diabetes on DPP-4 inhibitor or GLP-1R agonist had significantly lower odds of being hospitalized, whereas there were no associations with ICU admission, intubation, or death. There were no significant associations between SGLT2 use and COVID-19 adverse outcomes. The inconsistency across observational studies may be attributed to differences in study populations, strategies of adjusting for confounding, or study power because of smaller proportions of patients with diabetes on those newer medications.
Unlike some previous studies,36–38 our study did not find higher HbA1c levels were significantly associated with more severe COVID-19 outcomes. As shown in table 1, HbA1c data were missing in 26.4% of the patients with type 2 diabetes, with slightly higher percentages in those who had hospitalization (28.3%) or ICU/intubation/death (28.5%) than those without hospitalizations (25.5%). We cannot rule out the possibility that those with hospitalization or ICU/intubation/death had higher HbA1c. Nonetheless, in the multivariable regressions, we adjusted for the HbA1c levels as a categorical variable, including the ‘unknown’ category.
In patients at risk of type 2 diabetes, we found that being underweight was associated with higher odds of hospitalization, while having overweight and obesity was associated with lower odds, compared with patients of normal weight. These findings were not consistent with some prior studies.6 39 However, the at-risk patients in our study entered the cohort largely based on elevated BMI, a small proportion of the study cohort had normal weight or underweight. This group may represent a more susceptible population, that is, pre-diabetes, with different phenotypes. Likewise, the lack of significant associations between higher BMIs and adverse COVID-19 outcomes in patients with type 2 diabetes requires further study. It is possible that more aggressive care was provided to patients who were perceived as high risk, including those with both type 2 diabetes and obesity. Additional care may mitigate the negative impact of obesity.
Our study had several strengths. First, the study population consisted of patients from two states, including both hotspot and non-hotspot areas, suggesting broader generalizability. Second, our data included patients with COVID-19 diagnosed over a 12-month period, reflecting the different stages of prevention and treatment protocols before vaccines became available. Third, our data included patients from five health systems, with a substantial number of people in under-represented minority groups, from different clinical practice settings, and urban and rural areas.
There are limitations to our study. First, our patients were seen in tertiary academic health systems, and therefore, our results may not be applicable to patients seen in community settings. Second, missing information is not uncommon in EHR data, and the missingness on smoking status and HbA1c levels cannot be ignored. We cannot rule out the possibility of misclassification due to missing data. Third, EHR data may not capture COVID-19 outcomes outside the five health systems we examined, potentially underestimating the proportion of adverse events. Fourth, the association between covariates and outcomes could be overestimated due to residual confounding factors, such as insurance type and stage of COVID-19 illness at the time of diagnosis. Moreover, additional socioeconomic factors, such as education, income, and census track-level deprivation data, were not consistently coded across institutions, which limited the feasibility to study additional factors of COVID-19 hospitalizations. Finally, the original PaTH to Health: Diabetes study was a natural experiment to evaluate the effects of intensive behavioral therapy (IBT), implemented by the Centers for Medicare and Medicaid Services and subsequently by most private plans, on diabetes-related outcomes. As the BMI cut-off for the IBT services does not account for the ethnicity of the individual, Asians who reach overweight at the BMI of 23 kg/m2 and are at a greater risk for developing diabetes at a lower BMI were not included in the study population.
In conclusion, our study further emphasized that patients with type 2 diabetes or at risk of diabetes who were older, in racial minority groups, who had multiple chronic conditions, and were on insulin treatment had higher risks for severe COVID-19 outcomes. This reinforces the urgency to prevent COVID-19 and its complications, which subsequently can overburden medical resources. Given that racial disparities for COVID-19 vaccinations remain in Pennsylvania and Maryland, as well as across the USA, increased community outreach is needed to prevent COVID-19 infections and increase the public’s knowledge of and confidence in COVID-19 vaccines.