Abstract and Introduction
The COVID-19 pandemic has disproportionately affected people with diabetes, who are at increased risk of severe COVID-19.* Increases in the number of type 1 diabetes diagnoses[1,2] and increased frequency and severity of diabetic ketoacidosis (DKA) at the time of diabetes diagnosis have been reported in European pediatric populations during the COVID-19 pandemic. In adults, diabetes might be a long-term consequence of SARS-CoV-2 infection.[4–7] To evaluate the risk for any new diabetes diagnosis (type 1, type 2, or other diabetes) >30 days† after acute infection with SARS-CoV-2 (the virus that causes COVID-19), CDC estimated diabetes incidence among patients aged <18 years (patients) with diagnosed COVID-19 from retrospective cohorts constructed using IQVIA health care claims data from March 1, 2020, through February 26, 2021, and compared it with incidence among patients matched by age and sex 1) who did not receive a COVID-19 diagnosis during the pandemic, or 2) who received a prepandemic non–COVID-19 acute respiratory infection (ARI) diagnosis. Analyses were replicated using a second data source (HealthVerity; March 1, 2020–June 28, 2021) that included patients who had any health care encounter possibly related to COVID-19. Among these patients, diabetes incidence was significantly higher among those with COVID-19 than among those 1) without COVID-19 in both databases (IQVIA: hazard ratio [HR] = 2.66, 95% CI = 1.98–3.56; HealthVerity: HR = 1.31, 95% CI = 1.20–1.44) and 2) with non–COVID-19 ARI in the prepandemic period (IQVIA, HR = 2.16, 95% CI = 1.64–2.86). The observed increased risk for diabetes among persons aged <18 years who had COVID-19 highlights the importance of COVID-19 prevention strategies, including vaccination, for all eligible persons in this age group,§ in addition to chronic disease prevention and management. The mechanism of how SARS-CoV-2 might lead to incident diabetes is likely complex and could differ by type 1 and type 2 diabetes. Monitoring for long-term consequences, including signs of new diabetes, following SARS-CoV-2 infection is important in this age group.
Retrospective cohorts were constructed using two U.S. medical claims databases: IQVIA¶ and HealthVerity.** Patients who were aged <18 years on their index encounter date and who were continuously enrolled in a closed payor system throughout the study period†† were followed from their index date§§ until the end of the study period. Patients were excluded from the analysis if they had preexisting diabetes, defined as one or more International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis codes for diabetes (E08–E13) in the 1–13 months preceding their index date.
In the IQVIA database, patients with a COVID-19 diagnosis (ICD-10-CM codes B97.29 or U07.1)¶¶ during March 1, 2020–February 26, 2021, were defined as having COVID-19. Patients with COVID-19 were matched by age and sex to pandemic and prepandemic period comparison groups.***The pandemic period non–COVID-19 group comprised patients without COVID-19–related ICD-10-CM codes during March 1, 2020–February 26, 2021.††† The prepandemic period ARI group comprised patients with a diagnosis of ARI§§§ (Supplementary Table 1, https://stacks.cdc.gov/view/cdc/113062) during March 1, 2017–February 26, 2018. A prepandemic non-ARI group consisted of those in this age group whose records did not include ARI ICD-10-CM codes during March 1, 2017–February 26, 2018.
In HealthVerity, the COVID-19 group comprised patients aged <18 years whose record included an ICD-10-CM diagnosis code for COVID-19 or a positive SARS-CoV-2 polymerase chain reaction (PCR) test result during March 1, 2020–June 28, 2021. The pandemic period non–COVID-19 group consisted of those who had a negative SARS-CoV-2 PCR test result and no record of COVID-19 diagnosis codes or positive SARS-CoV-2 test results during the same period. Both groups were identified within a subset of CDC-licensed HealthVerity data that includes patients with a health care encounter possibly related to COVID-19 (Supplementary Table 2, https://stacks.cdc.gov/view/cdc/113062). There was no prepandemic comparison period for the HealthVerity data.
Incident diabetes was defined as one or more health care claims with a diabetes diagnosis (ICD-10-CM codes E08–E13) occurring >30 days after the index date (excluding cases of transient, resolved hyperglycemia). Frequencies of incident diabetes codes on, and DKA codes on or before, the date of the incident diabetes encounter were calculated.¶¶¶ Cox regression models were used to estimate HRs for diabetes risk. HRs were also estimated by age group and sex. Age and sex effect modifications were assessed using interaction terms. SAS (version 9.4; SAS Institute) and PANDAS (version 1.3.0; PANDAS Community) software were used to conduct all analyses. This activity was reviewed by CDC and was conducted consistent with applicable federal law and CDC policy.****
Among 80,893 patients with COVID-19 in the IQVIA database, the mean age was 12.3 years, 50.1% were female, and 0.7% were hospitalized at their index COVID-19 encounter (Table 1). Among 439,439 patients with COVID-19 in HealthVerity, the mean age was 12.7 years, 50.1% were female, and 0.9% were hospitalized at their index encounter. Diabetes was coded in 0.08% (IQVIA) and 0.25% (HealthVerity) of claims for patients with COVID-19, with the majority of diabetes diagnoses for type 1 or type 2 (IQVIA, 94.1%; HealthVerity, 94.0%). In comparison, 0.03% (IQVIA) and 0.19% (HealthVerity) diabetes cases were coded among those without COVID-19. DKA was reported in 48.5% (IQVIA) and 40.2% (HealthVerity) of patients with COVID-19 and diabetes; these proportions were higher than DKA reported in patients with diabetes without COVID (IQVIA: non-COVID 13.6%; ARI 22.0%; non-ARI 27.5%; HealthVerity: 29.7%).
In the IQVIA database, diabetes incidence was 316 per 100,000 person-years in the COVID-19 group, 118 per 100,000 person-years in the pandemic period non–COVID-19 group, 126 per 100,000 person-years in the prepandemic ARI group, and 125 per 100,000 person-years in the prepandemic non-ARI group (Table 2). Diabetes risk was 166% higher in the COVID-19 group than in the non-COVID-19 group (HR = 2.66, 95% CI = 1.98–3.56) and 116% higher than in the prepandemic ARI group (HR = 2.16, 95% CI = 1.64–2.86) (Figure). Diabetes incidence did not significantly differ between the prepandemic ARI and non-ARI groups (HR = 0.99, 95% CI = 0.84–1.15). In the HealthVerity database, diabetes incidence was 31% higher among patients aged <18 years with COVID-19 (399 per 100,000 person-years) than among those without COVID-19 (304 per 100,000 person-years; HR = 1.31, 95% CI = 1.20–1.44).
Hazard ratio for the association between COVID-19 or acute respiratory infection and new diabetes diagnosis among patients aged <18 years, by age group and sex — IQVIA PharMetrics Plus and HealthVerity claims databases,* United States, March 1, 2020–June 28, 2021†,§,¶
Abbreviations: ARI = acute respiratory infection; HR = hazard ratio, ICD-10-CM = International Classification of Diseases, Tenth Revision, Clinical Modification; Ref = referent.
†95% CIs indicated by error bars.
§Groups in IQVIA included patients aged <18 years with or without COVID-19 (COVID-19; non–COVID-19, respectively) and patients aged <18 years with or without ARI (ARI; non-ARI, respectively), during March 1, 2020–February 26, 2021, determined using presence or absence of ICD-10-CM codes for COVID-19 and ARI. The non–COVID-19 group was matched 5:1 to the COVID-19 group by age, sex, and month of encounter. The ARI group was matched 5:1 to the COVID-19 group by age and sex, and a random encounter date was selected. The non-ARI group was matched 2:1 to the ARI group by age and sex, and a random encounter date was selected. In HealthVerity, among patients aged <18 years, those with COVID-19 (COVID), determined by ICD-10-CM code or by a positive SARS-CoV-2 test result during March 1, 2020–June 28, 2021, were matched 1:1 to those with a negative SARS-CoV-2 test result (non–COVID-19) during the same period by age, sex, and month of encounter.
¶Hazard ratios are plotted on a logarithmic scale.
In the IQVIA database, risk for diabetes was similar across age groups and by sex. In the HealthVerity database, there was no association with diabetes in children aged <12 years, although a significantly increased risk was observed among all other age and sex groups. However, no age group or sex interaction terms were statistically significant.