key: cord-0806709-4bw26042
authors: Gregory, Justin M.; Slaughter, James C.; Duffus, Sara H.; Smith, T. Jordan; LeStourgeon, Lauren M.; Jaser, Sarah S.; McCoy, Allison B.; Luther, James M.; Giovannetti, Erin R.; Boeder, Schafer; Pettus, Jeremy H.; Moore, Daniel J.
title: COVID-19 Severity Is Tripled in the Diabetes Community: A Prospective Analysis of the Pandemic’s Impact in Type 1 and Type 2 Diabetes
date: 2020-12-02
journal: Diabetes Care
DOI: 10.2337/dc20-2260
sha: 550c768a204ec4a3ff63f09d051bb304491ed6ae
doc_id: 806709
cord_uid: 4bw26042

OBJECTIVE: To quantify and contextualize the risk for coronavirus disease 2019 (COVID-19)–related hospitalization and illness severity in type 1 diabetes. RESEARCH DESIGN AND METHODS: We conducted a prospective cohort study to identify case subjects with COVID-19 across a regional health care network of 137 service locations. Using an electronic health record query, chart review, and patient contact, we identified clinical factors influencing illness severity. RESULTS: We identified COVID-19 in 6,138, 40, and 273 patients without diabetes and with type 1 and type 2 diabetes, respectively. Compared with not having diabetes, people with type 1 diabetes had adjusted odds ratios of 3.90 (95% CI 1.75–8.69) for hospitalization and 3.35 (95% CI 1.53–7.33) for greater illness severity, which was similar to risk in type 2 diabetes. Among patients with type 1 diabetes, glycosylated hemoglobin (HbA(1c)), hypertension, race, recent diabetic ketoacidosis, health insurance status, and less diabetes technology use were significantly associated with illness severity. CONCLUSIONS: Diabetes status, both type 1 and type 2, independently increases the adverse impacts of COVID-19. Potentially modifiable factors (e.g., HbA(1c)) had significant but modest impact compared with comparatively static factors (e.g., race and insurance) in type 1 diabetes, indicating an urgent and continued need to mitigate severe acute respiratory syndrome coronavirus 2 infection risk in this community.

patients (6, 7) , patients referred to a registry (8) , or an analysis of mortality as the sole outcome without directly verifying patient-specific factors (9, 10) . Thus, there remains a need to ascertain a wider range of clinical outcomes across a broader spectrum of patients with type 1 diabetes infected with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2).

Although clinical outcome data for the SARS-CoV-2 virus in type 1 diabetes are lacking, some evidence suggests people with type 1 diabetes are prone to worse outcomes when infected with other respiratory viruses (11) . Both type 1 diabetesrelated factors, such as dysglycemia, type 1 diabetes duration, BMI, and vascular disease, and demographic factors, such as age, race/ethnicity, and social determinants of health, may modify these risks.

To address these critical gaps in our knowledge, we aimed to answer two questions. First, among patients with COVID-19, to what extent does a diagnosis of type 1 diabetes increase risk for hospitalization, greater severity of illness, and death compared with patients without diabetes or those with type 2 diabetes? Second, which covariates increase risk for a more severe outcome among people with type 1 diabetes? To answer these questions, we conducted a prospective cohort study to identify case subjects with COVID-19 using the electronic health record (EHR), categorizing each patient by diabetes category: no diabetes, type 1 diabetes, or type 2 diabetes.

We retrieved data by querying the Epic Clarity data warehouse at Vanderbilt University Medical Center (VUMC). This data warehouse encompasses the entire EHR at VUMC, a network of 137 primary care, urgent care, and hospital facilities that manages .2 million ambulatory and inpatient visits annually. During the prospective study period between 17 March and 7 August 2020, VUMC tested all patients admitted to the hospital for any reason and all patients prior to any surgical procedure. Additionally, tests were conducted on patients presenting to VUMC primary care, after-hours, urgent care, and minor medical clinics. Using our EHR query, we identified all individuals with a positive COVID-19 test at VUMC, as determined by a positive Centers for Disease Control and Prevention SARS-CoV-2 real-time PCR diagnostic panel, with probes and primers provided by Integrated DNA Technologies (Coralville, IA). We then categorized these case subjects by diabetes category, as shown in the flowchart in Supplementary  Fig. 1 , according to problem list diagnoses. If a patient's problem list included both type 1 diabetes and type 2 diabetes, the primary investigator (J.M.G.) adjudicated which diagnosis was more likely. Type 1 diabetes was assigned if the patient had a history of having autoantibodies associated with type 1 diabetes or required multiple daily injections. Type 2 diabetes was assigned if the patient required multiple oral antihyperglycemic medications. We retrieved prespecified clinical outcomes related to COVID-19 and covariates thought to potentially affect illness severity (listed  in Supplementary Appendix Tables 1-3) , as documented in the EHR. Beginning with the first case subject with COVID-19 in March 2020, we prospectively repeated the query every 2 weeks to identify case subjects, outcomes, and covariates. We also accessed publicly available data sets from the Tennessee Department of Health to determine the number of positive and total COVID-19 tests that were conducted in the local community around VUMC (i.e., Davidson County and its six surrounding counties, accessed on https:// www.tn.gov/content/tn/health/cedep/ ncov/data.html).

To further characterize factors that modulate risk for COVID-19 severity, we conducted a detailed chart review on all case subjects in patients with type 1 diabetes and attempted to contact them by telephone. After we obtained written, informed consent via electronic document, patients completed a survey over the phone, including questions about their medical history and social determinants of health (as listed in Supplementary  Appendix Tables 1-3) .

The research team managed study data using secure Research Electronic Data Capture (REDCap) electronic data tools (12) . The Institutional Review Board of Vanderbilt University approved the study protocol.

To quantify the magnitude of COVID-19 severity, we defined an ordinal outcome variable for illness severity with six mutually exclusive levels that had to occur within 14 days of a positive PCR test for SARS-CoV-2:

c No hospitalization c Hospitalization for any reason without any respiratory support c Hospitalization for any reason with lower acuity respiratory support (e.g., oxygen by nasal canula, nonrebreather mask, or continuous or bilevel positive airway pressure) c Intensive care unit (ICU) admission for any reason c Endotracheal intubation and mechanical ventilation for any reason c Death for any reason Hospitalization was analyzed as a binary outcome using logistic regression. Disease severity was analyzed using the proportional odds ordinal logistic regression model. In all models, we included diabetes status and adjusted for a set of prespecified covariates that could potentially confound the association between diabetes status and illness severity. These covariates included age, sex, BMI, diabetes category, smoking history, race, and a history of hypertension. Results are presented as odds ratios (ORs) with corresponding 95% CIs. Forall analyses, multiple imputation (10 imputed data sets) for missing covariates was performed using the "aregImpute" function available in the rms package for R version 3.6.3 (13).

After detailed chart review and patient surveys, we further analyzed the risk factors listed in Supplementary Appendix Table 3 for worsening illness severity among the patients with type 1 diabetes. To quantify the effect size of each independent variable on the ordinal variable for illness severity, we determined the unadjusted OR for worse illness severity for each independent variable using univariate ordinal regression.

During the prospective study period between 17 March and 7 August 2020, VUMC obtained 83,437 SARS-CoV-2 PCR results for 69,701 unique individuals at VUMC, 6,451 of whom tested positive for COVID-19 (7.7% positive). By comparison, 45,456 of 372,779 SARS-CoV-2 PCR tests (12.2%) were positive in Davidson County, TN, and its six surrounding counties over the same interval. Of the COVID-19-positive individuals at VUMC, 6,138 had no diabetes, 40 had type 1 diabetes, and 273 had type 2 diabetes. Table 1 summarizes demographic and clinical characteristics of these patients. The median age and BMI of patients with type 1 diabetes were similar to the group without diabetes, while patients with type 2 diabetes were two decades older and had a BMI ;20% higher than the other two groups. Pediatric patients (i.e., aged ,18 years) comprised 9.4%, 20.0%, and 0% of the groups with no diabetes, type 1 diabetes, and type 2 diabetes, respectively. The type 1 diabetes cohort had a higher percentage for White race and a lower percentage for Hispanic ethnicity. The percentage of individuals taking medications to treat hypertension or who had a diagnosis of hypertension or asthma was lowest for those without diabetes, intermediate for those with type 1 diabetes, and highest for those with type 2 diabetes.

Supplementary Table 1 summarizes unadjusted COVID-19 hospitalization rates and illness severity for patients categorized by diabetes status. Whereas hospitalization was three times more likely to occur in patients with type 1 diabetes than patients without diabetes (22.5% vs. 7.1%; P , 0.001), hospitalization in the group with type 2 diabetes occurred two times more often than in the group with type 1 diabetes (44.3% vs. 22.5%; P 5 0.009), which was two decades younger. No deaths occurred in the group with type 1 diabetes, but deaths occurred in 0.5% of patients without diabetes (P 5 0.67 vs. type 1 diabetes) and 4.8% of patients with type 2 diabetes (P 5 0.368 vs. type 1 diabetes and P , 0.001 vs. no diabetes). Unadjusted for other covariates, illness severity in patients with type 1 diabetes was worse than in those without diabetes (P , 0.001), but less severe than in those with type 2 diabetes (P , 0.001).

Because of key differences in baseline risk factors between the groups (e.g., older age in the group with type 2 diabetes), we used multivariable regression to adjust for age, race, sex, hypertension, smoking, and BMI. Figure 1 summarizes adjusted ORs for hospitalization rates and illness severity. After statistical adjustment, groups with both type 1 and type 2 diabetes had similarly increased odds of hospitalization and worsening illness severity relative to individuals without diabetes ( Fig. 1A and C). In type 1 diabetes, the adjusted ORs were 3.90 (95% CI 1.75-8.69) for hospitalization and 3.35 (95% CI 1.53-7.33) for worsening illness severity. For type 2 diabetes, the adjusted ORs were 3.36 (95% CI 2.49-4.55) for hospitalization, 3.42 (95% CI 2.55-4.58) for worsening illness severity, and 3.21 (95% CI 1.54-6.70) for death. Across all ages, patients with type 1 diabetes had a higher probability of hospitalization than patients without diabetes (Fig. 1B ). Next to age, the presence of diabetes was the most important factor in the multivariable ordinal regression model for illness severity (Fig. 1D ).

Because an EHR query is limited in its ability to capture some risk factors for a diabetes. Clinical factors significantly associated with greater COVID-19 illness severity (P , 0.05) included having a previous diagnosis of hypertension, higher glycosylated hemoglobin (HbA 1c ), taking any antihypertensive medication other than an ACE inhibitor or angiotensin receptor blocker, having at least one DKA admission in the past year, and not using a continuous glucose monitor (CGM). There was insufficient evidence to determine whether patients with a past medical history of microvascular disease tended to have worsening illness severity.

Of the 11 patients previously diagnosed with hypertension, 55% were hospitalized. By comparison, only 12% of the 25 patients not diagnosed with hypertension were hospitalized. The unadjusted OR for greater illness severity was 7.06 in the hypertensive patients versus patients without hypertension (P 5 0.020).

Social determinants of health were also significantly related to severity outcome. Eleven percent of patients who identified race as White were hospitalized, and the unadjusted OR for greater illness severity was 0.091 for White race versus not White race (P 5 0.007). Seventy-one percent of patients who identified as Black or African American were hospitalized, and the unadjusted OR for worse illness severity was 10.4 for Black race versus not Black race (P 5 0.009). While 8% of patients with private insurance were hospitalized, 60% of patients with public insurance only were hospitalized and 67% of patients with no insurance were hospitalized. The unadjusted OR for increased illness severity was 30.7 for public or no insurance versus private insurance (P 5 0.001). Although the survey included questions about income category and highest level of education, we found patients who were hospitalized were less likely to take phone calls or report this information.

Similarly, patients using higher levels of technology to manage their diabetes were less likely to have worse outcomes. Whereas only 9% of patients using a CGM as the primary means of glucose monitoring were hospitalized, 47% of patients using a blood glucose monitor as the primary means of glucose monitoring were hospitalized. The unadjusted OR for worse illness severity was 8.70 for a blood glucose monitor versus CGM (P 5 0.016). While only 6% of patients using an insulin pump were hospitalized, 33% of patients using multiple daily injections were hospitalized. The unadjusted OR for greater illness severity was 7.07 for multiple daily injections versus insulin pump use (P 5 0.085).

Two principal themes emerge from these data regarding the severity of COVID-19 in type 1 diabetes. First, after adjustment for age, race, and other risk factors, the odds of a COVID-19-related hospitalization and greater illness severity for patients with type 1 diabetes are three-to fourfold higher than for patients without diabetes. This increased risk is approximately the same for patients with type 2 diabetes ( Fig. 1A and C) . Second, COVID-19 outcome severity in type 1 diabetes is associated with glycemic, vascular, and socioeconomic risk factors. Before adjustment for baseline characteristics that differed between groups, patients with type 1 diabetes appeared tohave a risk for hospitalization and greater illness severity that was intermediate between the group with no diabetes and the group with type 2 diabetes. Importantly, however, once our analysis adjusted the odds of hospitalization and greater illness severity for other known COVID-19 risk factors, especially age, both groups with diabetes had similar odds of worsening morbidity compared with the group without diabetes. The three-to fourfold adjusted OR for hospitalization and greater illness severity seen in this investigation is comparable to the adjusted OR for mortality recently reported by the National Health Service (NHS) in England. In this large population study, the adjusted ORs for in-hospital death with COVID-19 were 3.5 for people with type 1 diabetes and 2.0 for people with type 2 diabetes (both relative to people without diabetes) (9) .

Whereas our data agree with an increasing body of evidence suggesting hospitalization rates are low in individuals without diabetes and rise considerably after 40 years of age (14), our analysis additionally suggests the probability of hospitalization is substantially higher for patients with type 1 diabetes than patients without diabetes until nearly the eighth decade of age (Fig. 1B) . Although the NHS investigators reported low absolute risk for mortality for patients with type 1 diabetes ,40 years of age, when we examine the more common outcome of hospitalization, we find the probability ranges between ;15% and ;22% for patients ,40 years of age, after adjustment for BMI. This probability is considerably greater than the ;5% risk seen over that age range in individuals without diabetes. Thus, the increased risk for COVID-19-related hospitalization remains markedly higher for people with type 1 diabetes until the seventh decade of life.

An EHR query is limited in its ability to accurately capture many risk factors for COVID-19 morbidity, particularly among people with type 1 diabetes. For this reason, our detailed chart review of 37 out of the 40 COVID-19-positive patients with type 1 diabetes enhanced our ability to analyze associations between potential clinical covariates and outcome severity. Our analysis suggests chronic hyperglycemia and vascular disease, social determinants of health, and decreased use of diabetes technology correlate significantly with outcome severity; these factors represent common characteristics between type 1 and type 2 diabetes and suggest that addressing modifiable factors will reduce risk in all patients with diabetes. These relationships are not surprising given the linkage between these risk factors and worse illness with COVID-19 reported in previous studies that did not focus on type 1 diabetes. For example, the heightened risk of poor outcomes for patients with type 1 diabetes with hypertension reported in this study is consistent with findings from previous meta-analyses of the broader Chinese population (15, 16) . Whereas previous reports have indicated proportionally higher rates of hospitalizations from COVID-19 among Black patients and those with public insurance (17), this study is the first to show a similar finding in the population with type 1 diabetes. Although our investigation identified an inverse correlation between diabetes technology use and illness severity, we found that patients with worse illness severity answered questions about socioeconomic status at a lower rate. Thus, our analysis cannot exclude the possibility that greater amounts of diabetes technology use are a surrogate for higher socioeconomic status. Somewhat surprisingly, we did not observe a significant correlation between type 1 diabetes duration and COVID-19 severity. Moreover, we did not find statistically significant associations between the illness severity and age and sex within the type 1 diabetes cohort. We suspect these relationships would have been stronger had the sample size been larger, since these correlations were robust in the larger EHR cohort.

Interestingly, only 1 out of 40 patients with type 1 diabetes diagnosed with COVID-19 was also hospitalized for DKA. This finding contrasts with a preliminary report from the T1D Exchange Collaboration, which found 46% of 33 COVID-19positive patients had DKA (8). One possible reason for this significant discrepancy is that clinicians and patients were more likely to refer more severe case subjects to the T1D Exchange multicenter registry. Similarly, physicians participating in the registry may have been unaware of less severe case subjects in the community. This tendency for referral bias may have been particularly true during the early weeks of the unfolding pandemic when the preliminary report was published.

Strengths of our study include the ability of the EHR query to broadly and prospectively identify case subjects with COVID-19 across our entire EHR in an unbiased fashion. Further, testing at Vanderbilt captures a wide cross-section of illness severity. In contrast with two recent studies characterizing clinical characteristics of patients with type 1 diabetes hospitalized with COVID-19 (6,7), our analysis included not only every hospitalized patient, but also numerous patients in outpatient primary care, minor medical, and urgent care clinics (i.e., many patients withmilder symptoms). Additionally, our hospital tests all patients immediately prior to an elective surgical procedure (i.e., many entirely asymptomatic patients). Because this approach does not rely on clinics to report case subjects with COVID-19 in type 1 diabetes, it is less prone to referral bias than other current approaches (8) . Furthermore, our approach determined the presence of COVID-19 in a straightforward, standardized fashion across the EHR: to have COVID-19 for our study, the patient simply had to have a positive SARS-CoV-2 PCR test. In addition, our phone survey allowed us to collect more detailed information about social determinants of health. Thus, our approach complements that of our colleagues who are also working to characterize COVID-19 in type 1 diabetes.

Three significant limitations of this study warrant consideration. First, data were accrued from a single academic health system largely serving urban and suburban populations. Although the VUMC system includes many outpatient testing locations throughout the local community (e.g., primary care, minor medical, and urgent care clinics), we were unable to access detailed information about case subjects with COVID-19 identified outside the VUMC system. Second, although all hospitalized and presurgical patients at VUMC received COVID-19 testing during the prospective study period, our study cannot exclude the possibility that clinicians in the outpatient setting were more likely to test patients with diabetes than without diabetes because of the perception that patients with diabetes were at greater risk. Third, although we extensively characterized risk factors and outcomes for patients with COVID-19 in this study, the sample size remains relatively small compared with whole-population studies. This smaller sample size prevented us from conducting some multivariate regression analyses within the group with type 1 diabetes and from better characterizing COVID-19 outcomes in Hispanic patients with type 1 diabetes. Additionally, more rare outcomes, such as death, are difficult to adequately characterize. This is likely the reason for the striking difference seen in the death rate in our study (0 out of 40 COVID-19-positive patients with type 1 diabetes) and that of the whole-population study done by NHS England (9, 10) , which focused on mortality as the outcome of interest. Because the present investigation involved a comparatively smaller sample size, it can neither support nor refute the NHS study's finding of high mortality in COVID-19-positive patients with type 1 diabetes. Instead, our analysis complements the NHS study by showingtheadjustedoddsofhospitalization and the severity of illness with COVID-19 are three-to fourfold higher in type 1 diabetes versus a comparable population of individuals who do not have diabetes.

Our data have important implications for the community with diabetes as colder temperatures emerge in the Northern hemisphere. As social activities move inside, humidity decreases, and the will to maintain social distancing practices wanes, experts have expressed considerable concern that SARS-CoV-2 transmission will increase sizably (18) . Our investigation suggests that as COVID-19 hospitalizations rise, patients with both type 1 and 2 diabetes will comprise a disproportionally higher number of those admissions and, once hospitalized, demonstrate a greater degree of illness severity. Thus, unless the type 1 diabetes community reamplifies its efforts to mitigate the spread of SARS-CoV-2 in the coming months, physicians should anticipate an imminent escalation in the number of these patients with severe illness, including both adults and children. In light of these data, we call on our colleagues to emphasize the importance of social distancing measures and hand hygiene, with particular emphasis on patients with diabetes, including those in the most vulnerable communities who our study affirms will face the most severe impact.

Clinical course and outcomes of critically ill patients with SARS-CoV-2 pneumonia in Wuhan, China: a single-centered, retrospective, observational study

China Medical Treatment Expert Group for Covid-19. Clinical characteristics of coronavirus disease 2019 in China

Baseline characteristics and outcomes of 1591 patients infected with SARS-CoV-2 admitted to ICUs of the Lombardy region

Covid-19 in critically ill patients in the seattle regioncase series

the Northwell COVID-19 Research Consortium. Presenting characteristics, comorbidities, and outcomes among 5700 patients hospitalized with COVID-19 in the New York City area [published correction appears

CO-RONADO investigators. Type 1 diabetes in people hospitalized for COVID-19: new insights from the CORONADO study

Characteristics and diabetes control in adults with type 1 diabetes admitted with COVID-19 infection

Type 1 diabetes and COVID-19: preliminary findings from a multicenter surveillance study in the U

Associations of type 1 and type 2 diabetes with COVID-19-related mortality in England: a whole-population study

Risk factors for COVID-19-related mortality in people with type 1 and type 2 diabetes in England: a population-based cohort study

Risk of infection in type 1 and type 2 diabetes compared with the general population: a matched cohort study

Research electronic data capture (REDCap)-a metadata-driven methodology and workflow process for providing translational research informatics support

Prevalence of comorbidities and its effects in patients infected with SARS-CoV-2: a systematic review and metaanalysis

How the pandemic might play out in 2021 and beyond

Acknowledgments. The authors thank Alexandria DuBois, Vanderbilt University Medical Center (Nashville, TN), for administrative support throughout the study. helped design the study, analyze the data, and write the manuscript. J.C.S. assisted with the study design, conducted statistical analysis, and wrote the manuscript. All authors critically reviewed the manuscript. J.M.G. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.