In this study, researchers reviewed electronic health record data to assess whether the coronavirus disease 2019 pandemic was associated with disruptions in diabetes care processes of A1C testing, retinal screening, and nephropathy evaluation among patients receiving care with Wake Forest Baptist Health in North Carolina. Compared with the pre-pandemic period, they found an increase of 13–21 percentage points in the proportion of patients delaying diabetes care for each measure during the pandemic. Alarmingly, delays in A1C testing were greatest for individuals with the most severe disease and may portend an increase in diabetes complications.

The coronavirus disease 2019 (COVID-19) pandemic led to interruptions in clinical care in March 2020 and exposed gaps in public health infrastructure across the United States, leading to high morbidity and mortality (1,2). The burden of COVID-19 mortality is disproportionately higher among certain population subgroups in the United States, including older adults, Black and Hispanic Americans, individuals with lower education and lower income, and those from rural communities (3,4). These demographic and socioeconomic factors align with the U.S. population subgroups for which diabetes, a disease with high health care utilization and costs, has a greater burden of morbidity and mortality (58). Early evidence indicates that health care utilization for preventive, elective, and emergent care unrelated to COVID-19 remains lower than in the pre-pandemic months (914).

For individuals with diabetes, regular assessment of A1C, retinal examinations, and nephropathy evaluations are crucial for diabetes management, risk assessment, and prevention of complications (15). Early reports suggest a decrease in aggregate volumes of A1C testing after the start of the COVID-19 pandemic (16,17). What remains unclear is whether and to what extent within-individual A1C testing has been delayed during the pandemic. Furthermore, it remains to be seen whether within-individual screening for eye and kidney complications has been disrupted during the pandemic.

Evaluation of population-level associations are informative for surveillance, but these disruptions in diabetes care at the aggregate level may not hold at the individual level (i.e., they may represent an “ecological fallacy”) (18,19). Evaluation of within-individual disruptions in diabetes care directly assess patient-specific disruptions in diabetes care before and after the start of the COVID-19 pandemic that may be masked when assessed at the aggregate level. Our objective was to assess whether and for which population subgroups the COVID-19 pandemic was associated with within-individual disruptions in diabetes care, including A1C testing, retinal screening exams, and nephropathy evaluation among the population of Wake Forest Baptist Health (WFBH) patients with diabetes in 2018–2021.

Data Sources

We used electronic health record (EHR) data for individuals who received care at clinics affiliated with WFBH. This health system has a catchment area that includes a 24-county region of western North Carolina and southwestern Virginia. All clinics in the health system use the EpicCare (Verona, WI) EHR. This database includes demographic information, Current Procedural Terminology (CPT) codes, diagnosis codes, medication codes, laboratory data, insurance data, and International Classification of Diseases, 10th Revision (ICD-10), codes. We combined EHR data with census data from the American Community Survey (ACS) 2014–2018 at the Zip Code Tabulation Areas (ZCTA) level (20). The Wake Forest School of Medicine Institutional Review Board approved the study protocol and waived patient consent, as the study data were de-identified.

Study Design and Time Frame

Individuals ≥21 years of age with prevalent diabetes during 2018 were eligible for inclusion. Diabetes prevalence was identified with ICD-10 diagnosis codes E10, E11, and E13 from outpatient, emergency department, and inpatient encounters from 1 January through 31 December 2018. We obtained EHR data on all encounters from 1 January 2018 through 28 February 2021. We excluded individuals who died prior to 1 March 2021. We chose 1 March 2020 as the start date of the COVID-19 pandemic and created three contiguous 12-month periods of comparison surrounding the start of the pandemic; period 1 spanned 1 March 2018 through 29 February 2019, period 2 was 1 March 2019 through 29 February 2020, and period 3 was 1 March 2020 through 28 February 2021.

Cohort Entry and Diabetes Care

Primary care provider (PCP) encounters were identified using CPT codes for evaluation and management visits (Supplementary Table S1). To reduce the potential for including individuals in the sample who did not receive regular diabetes care from WFBH, we required at least one PCP encounter in both pre-pandemic periods 1 and 2. To ensure continuation of care within WFBH during the pandemic, we required patients to have at least one EHR encounter of any type during period 3. To enable comparison of the outcome levels during the pre-pandemic and pandemic periods, we required patients to have at least two outcome-specific encounters in the EHR during both pre-pandemic periods and at least one outcome-specific encounter during the pandemic period for that respective outcome.

We performed a sensitivity analysis that included individuals who did not have one outcome-specific encounter in the year after the start of the pandemic but who did have an EHR encounter, and we set the date of encounter censoring at 28 February 2021. Compared with the primary analysis, this sensitivity analysis provided a less conservative estimate for delayed care by including those who went for longer than the full pandemic year without an outcome-specific encounter, but it assumes that patients did not receive care outside of the WFBH system.

The American Diabetes Association recommends regular assessment of A1C and annual nephropathy evaluation and comprehensive eye examinations for people with diabetes (15). Supplementary Table S1 presents CPT codes that were used to determine retinal exam screening and nephropathy evaluation.

Delayed Care

For each diabetes care process—A1C testing, retinal screening, and nephropathy evaluation—we created a binary variable for existence of a “delayed test” since the most proximal prior respective process measure. We defined a delayed A1C test as a duration of >6 months since the most recent A1C measurement and defined as “not delayed” a duration between A1C tests of ≤6 months (Supplementary Table S1). For retinal and nephropathy screening, a delayed test was defined as a duration of >12 months since the most recent respective screening, and a test was deemed not delayed if the duration between assessments was ≤12 months. Supplementary Figure S1 illustrates the study design, using A1C testing as an example.

Patient Characteristics

For each individual, we quantified baseline characteristics using pre-pandemic EHR data. We used the Elixhauser comorbidity algorithm to group all ICD-10 diagnosis codes identified in all encounters during period 1 into 27 nondiabetic conditions (21). Based on the count of the number of conditions identified for each individual, we defined a three-level variable: 0–1, 2–4, and ≥5 comorbidities. We categorized insurance status identified in the first encounter of period 1 as Medicare, private, Medicaid, uninsured, or other. Use of insulin was determined by the presence of ICD-10 code Z79.4. We categorized baseline glucose control, using the latest A1C test result in period 1, into four groups: <7% (<53 mmol/mol), 7–7.9% (53–63 mmol/mol), 8–8.9% (64–74 mmol/mol), and ≥9% (≥75 mmol/mol).

Area-Level Characteristics

We characterized multiple area-level demographic and socioeconomic measures. Area of residence was determined by linking ZCTA data with 2010 Rural-Urban Commuting Area codes and defined as metropolitan, micropolitan, small town, or rural (22). We created categorical variables for ACS community-level percentage of community residents who identify as non-Hispanic Black, identify as Hispanic, report difficulty with the English language, were currently unemployed, were living at or below the poverty level, and were currently uninsured. Categorical cut points for each ACS variable differed based on the distribution in the sample and to ensure adequate numbers for estimation of a delayed exam within each subgroup.

Statistical Analysis

We identified 65,629 individuals with a medical encounter and ICD-10 code indicating diabetes in 2018. After excluding individuals who died before the start of the pandemic and those who were determined to be not receiving regular diabetes care within WFBH, analytic sample sizes were n = 25,323 for A1C, n = 8,675 for nephropathy evaluation, and n = 1,882 for retinopathy screening (exclusions shown in Supplementary Table S2).

We estimated the proportion of patients delaying diabetes care and the duration between subsequent diabetes care assessments in the two contiguous 12-month periods before the start of the pandemic as a control period. To understand whether the pandemic was associated with disruptions in diabetes care, we estimated the proportion of patients delaying diabetes care and the duration between subsequent diabetes care assessments from the date of the last assessment before the start of the pandemic period to the date of the first assessment during the pandemic period.

To formally assess differences in diabetes care between the pre-pandemic and pandemic periods, we used fixed-effects linear regression models to estimate within- person incidence of delaying diabetes care and the duration between subsequent diabetes care assessments. This identification strategy treats individuals as their own control subject and provides an effective way to control for systematic time-invariant patient characteristics observed and unobserved (23).

We obtained heteroskedasticity-consistent standard errors clustered at the patient level. As a test of validation of our identification strategy, we compared the duration between A1C tests for the three most proximal A1C tests before the start of the pandemic, occurring during periods 1 and 2, as an internal control test. The objective of this internal validation was to determine whether the duration between A1C tests differed before the start of the pandemic. No difference in the duration between pre-pandemic A1C tests would suggest that differences during the pandemic were associated with the pandemic and not some other process.

We assessed for differences in delaying diabetes care according to subgroups for demographics, clinical characteristics, and ACS socioeconomic measures. All analyses were performed with Stata, v. 16.1, statistical software (StataCorp, College Station, TX).

Characteristics of the diabetes patient population (96% with type 2 diabetes) are presented in Table 1. Patients’ mean age was 63 years (SD 13 years), 53% were female, and 71% identified as non-Hispanic White. More than half of the population had an A1C <7%, and 31% had a record of current insulin use. Patients’ health insurance coverage was predominantly Medicare or private insurance. Less than one-third of the sample had zero or one Elixhauser comorbidity. Nearly 80% of patients had a metropolitan residence.

TABLE 1

Demographic, Clinical, and Socioeconomic Characteristics Measured in 2018 of Patients With Prevalent Diabetes in the WFBH System (N = 25,323)

CharacteristicValue
Mean age, years 63.3 ± 12.8 
Median age, years 64 (55–72) 
Female sex
Missing 
13,383 (53)
1 (<1) 
Race/ethnicity
Non-Hispanic White
Non-Hispanic Black
Hispanic
Other/multiple specified
Unknown 

18,062 (71)
5,517 (22)
840 (3)
863 (3)
41 (<1) 
Type 1 diabetes 1,050 (4.2) 
A1C group, % (mmol/mol)
<7 (<53)
7–7.9 (53–63)
8–8.9 (64–74)
≥9 (≥75)
Missing 

13,285 (53)
5,946 (24)
2,801 (11)
3,247 (13)
44 (<1) 
Mean A1C, % 7.3 ± 1.6 
Mean A1C, mmol/mol 56 ± 17.5 
Median A1C, % 6.9 (6.2–7.9) 
Median A1C, mmol/mol 52 (44–63) 
Insulin medication use 7,750 (31) 
Insurance status
Medicare
Private
Medicaid
Uninsured
Other 

14,794 (58)
9,616 (38)
117 (<1)
698 (3)
98 (<1) 
Elixhauser comorbidities
0–1
2–4
≥5 

7,845 (31)
13,526 (53)
3,952 (16) 
Area of residence
Metropolitan
Micropolitan
Small town
Rural
Missing 

20,102 (79)
4,087 (16)
729 (3)
404 (2)
1 (<1) 
CharacteristicValue
Mean age, years 63.3 ± 12.8 
Median age, years 64 (55–72) 
Female sex
Missing 
13,383 (53)
1 (<1) 
Race/ethnicity
Non-Hispanic White
Non-Hispanic Black
Hispanic
Other/multiple specified
Unknown 

18,062 (71)
5,517 (22)
840 (3)
863 (3)
41 (<1) 
Type 1 diabetes 1,050 (4.2) 
A1C group, % (mmol/mol)
<7 (<53)
7–7.9 (53–63)
8–8.9 (64–74)
≥9 (≥75)
Missing 

13,285 (53)
5,946 (24)
2,801 (11)
3,247 (13)
44 (<1) 
Mean A1C, % 7.3 ± 1.6 
Mean A1C, mmol/mol 56 ± 17.5 
Median A1C, % 6.9 (6.2–7.9) 
Median A1C, mmol/mol 52 (44–63) 
Insulin medication use 7,750 (31) 
Insurance status
Medicare
Private
Medicaid
Uninsured
Other 

14,794 (58)
9,616 (38)
117 (<1)
698 (3)
98 (<1) 
Elixhauser comorbidities
0–1
2–4
≥5 

7,845 (31)
13,526 (53)
3,952 (16) 
Area of residence
Metropolitan
Micropolitan
Small town
Rural
Missing 

20,102 (79)
4,087 (16)
729 (3)
404 (2)
1 (<1) 

Data are mean ± SD, n (%), or median (interquartile range).

Overall Diabetes Care Testing and Screening Visits

Compared with the corresponding calendar months during the two contiguous 12-month periods before the start of the pandemic, we observed an initial decrease in overall A1C testing during March and April 2020 (Figure 1). Overall, A1C testing frequency did rebound to pre-pandemic levels during June 2020. From July through December 2020, overall A1C testing frequency was lower than in the pre-pandemic periods, at ∼75–85% of the rate of prior years. For overall frequency of retinal screening visits and nephropathy evaluations, we observed a decrease in overall testing during March and April 2020 compared with the same calendar months of the two contiguous 12-month periods before the start of the pandemic (Supplementary Figures S2 and S3). Overall retinal screening visits, but not nephropathy evaluations, did rebound to their pre-pandemic frequency by June 2020. From July through December 2020, overall retinal screening visit frequency was similar to pre-pandemic years, whereas overall nephropathy evaluation frequency was lower than pre-pandemic years, at ∼60–70% of the rate of prior years.

FIGURE 1

Frequency of overall A1C testing during three contiguous 12-month periods from 1 March 2018 to 28 February 2021.

FIGURE 1

Frequency of overall A1C testing during three contiguous 12-month periods from 1 March 2018 to 28 February 2021.

Close modal

Within-Individual Diabetes Care Testing and Screening Visits

In the period before the start of the pandemic, the proportions of patients experiencing delayed diabetes care were 41% for A1C testing, 36% for retinal screening, and 19% for nephropathy evaluation (Table 2). The mean duration between the penultimate and ultimate test was 173 days for A1C testing, 233 days for retinal screening, and 187 days for nephropathy evaluation. During the pandemic period, there was an increase in the proportion of patients delaying diabetes care for each measure, ranging from an additional 13 to 21 percentage-point increase in delayed care for the first visit during the pandemic. These estimates can be interpreted as the change in the proportion of individuals with a delayed diabetes care assessment during the pandemic period compared with before the pandemic, with a positive value indicating an increase in the proportion of delayed care during the pandemic.

TABLE 2

Individual Incidence of Delayed Diabetes Care and Mean Time Interval Between Diabetes Care Process Measures

Care ProcessNIncidence of Delayed Testing, % (95% CI)Time Interval Between Tests, days (95% CI)
Proportion With Delay of Last Pre-Pandemic Test (reference)Percentage-Point Change in Proportion With Delay of First Pandemic Period TestDuration Between Last Two Pre-Pandemic Tests (reference)Excess Time Interval for First Pandemic Period Test
A1C test 25,323 41.3 (40.9–41.6) 20.6 (19.9–21.4) 172.5 (171.7–173.3) 47.9 (46.3–49.5) 
Retinal exam 1,882 36.2 (35.1–37.2) 12.7 (10.6–14.8) 232.7 (229.2–236.2) 64.8 (57.7–71.8) 
Nephropathy evaluation 8,675 19.0 (18.4–19.6) 18.6 (17.4–19.9) 186.9 (184.7–189.1) 116.4 (112.0–120.1) 
Care ProcessNIncidence of Delayed Testing, % (95% CI)Time Interval Between Tests, days (95% CI)
Proportion With Delay of Last Pre-Pandemic Test (reference)Percentage-Point Change in Proportion With Delay of First Pandemic Period TestDuration Between Last Two Pre-Pandemic Tests (reference)Excess Time Interval for First Pandemic Period Test
A1C test 25,323 41.3 (40.9–41.6) 20.6 (19.9–21.4) 172.5 (171.7–173.3) 47.9 (46.3–49.5) 
Retinal exam 1,882 36.2 (35.1–37.2) 12.7 (10.6–14.8) 232.7 (229.2–236.2) 64.8 (57.7–71.8) 
Nephropathy evaluation 8,675 19.0 (18.4–19.6) 18.6 (17.4–19.9) 186.9 (184.7–189.1) 116.4 (112.0–120.1) 

For A1C testing, a delay was defined as a time interval >180 days. For retinal screening exam and nephropathy evaluation, a delay was defined as a time interval >365 days.

The corresponding mean duration between the last pre-pandemic care and first pandemic care was an additional 48 days for A1C testing, 65 days for retinal screening, and 116 days for nephropathy evaluation compared with the pre-pandemic duration between diabetes care. Supplementary Table S3 presents the results of our internal validation. The mean duration between A1C tests before the start of the pandemic differed by 2 days; this finding supports a similar duration between successive A1C tests before the pandemic.

Subgroup Analyses for Within-Individual Diabetes Care

We did not observe differences in the proportion of delayed A1C testing in the period before the pandemic for subgroups defined by age, sex, or race/ethnicity (Table 3). Individuals with type 1 diabetes had a lower proportion of A1C testing delay in the pre-pandemic period and greater increases in the proportion of testing delay during the pandemic compared with individuals with type 2 diabetes. With respect to insurance status subgroups, the proportion of delayed A1C testing in the pre-pandemic period was highest among uninsured individuals (48%), corresponding to a mean duration between A1C tests that was 29 days longer than for those with Medicare. We observed that higher A1C level and greater comorbidity burden were both associated with higher proportion of delaying A1C evaluation in the pre-pandemic period. We also observed differences in the proportion of patients delaying A1C testing during the pandemic period across subgroups of A1C level and comorbidity burden; specifically, there was an increase in the proportion of patients delaying A1C testing during the pandemic period corresponding to higher A1C and greater burden of comorbidities. Supplementary Table S4 presents the proportion of patients delaying A1C testing by insulin use, area of residence, and ACS population-level demographic and socioeconomic subgroups.

TABLE 3

Assessment of Individual Incidence of Delayed A1C Testing and Mean Time Interval Between A1C Testing According to Demographic and Clinical Subgroups

SubgroupNIncidence of Delayed Testing, % (95% CI)Time Interval Between Tests, days (95% CI)
Proportion With Delay of Last Pre-Pandemic Test (reference)Percentage-Point Change in Proportion With Delay of First Pandemic Period TestDuration Between Last Two Pre-Pandemic Tests (reference)Excess Time Interval for First Pandemic Period Test
Age-group, years
18–64
≥65 

12,729
12,594 

41.6 (41.1–42.2)
40.9 (40.4–41.4) 

21.3 (20.3–22.4)
20.0 (19.0–21.0) 

177.2 (176.0–178.4)
167.7 (166.7–168.8) 

49.0 (46.6–51.4)
46.8 (44.8–48.9) 
Sex
Female
Male 

13,383
11,939 

40.8 (40.3–41.3)
41.7 (41.2–42.3) 

21.8 (20.8–22.9)
19.3 (18.2–20.4) 

172.3 (171.2–173.4)
172.7 (171.6–173.9) 

50.7 (48.5–52.8)
44.8 (42.5–47.2) 
Race/ethnicity
Non-Hispanic White
Non-Hispanic Black
Hispanic
Other specified 

18,062
5,517
840
863 

40.9 (40.4–41.3)
42.0 (41.2–42.8)
44.5 (42.5–46.5)
42.3 (40.3–44.3) 

20.4 (19.5–21.2)
21.4 (19.8–23.0)
19.6 (15.7–23.6)
21.8 (17.7–25.9) 

169.5 (168.6–170.4)
179.7 (177.8–181.5)
180.6 (175.8–185.5)
182.3 (177.6–187.0) 

46.6 (44.8–48.4)
51.7 (48.0–55.3)
51.7 (41.9–61.5)
47.1 (37.7–56.5) 
Diabetes type
Type 1
Type 2 

1,050
24,273 

33.0 (31.1–34.8)
41.6 (41.3–42.0) 

27.2 (23.5–30.9)
20.4 (19.6–21.1) 

163.1 (159.1–167.1)
172.9 (172.1– 173.7) 

59.2 (51.2–67.2)
47.4 (45.8–49.0) 
Insurance status
Medicare
Private
Medicaid
Uninsured 

14,794
9,616
117
698 

40.6 (40.1–41.1)
41.8 (41.2–42.4)
37.6 (31.5–43.7)
48.0 (45.7–50.3) 

20.4 (19.4–21.3)
21.0 (19.7–22.2)
23.1 (10.9–35.2)
23.5 (18.9–28.1) 

168.3 (167.3–169.2)
177.0 (175.6–178.3)
190.7 (173.5–207.9)
197.2 (191.0–203.3) 

47.4 (45.5–49.4)
47.2 (44.4–49.9)
64.4 (30.1–98.8)
67.3 (54.9–79.6) 
A1C, % (mmol/mol)
<7 (<53)
7–7.9 (53–63)
8–8.9 (64–74)
≥9% (≥75) 

13,285
5,946
2,801
3,247 

49.8 (49.3–50.3)
34.3 (33.5–35.0)
28.6 (27.5–29.7)
29.7 (28.6–30.7) 

17.3 (16.3–18.3)
22.6 (21.0–24.1)
26.1 (23.9–28.3)
26.0 (23.9–28.1) 

185.2 (184.1–186.2)
160.3 (158.8–161.8)
153.5 (151.1–155.9)
159.1 (156.6–161.5) 

44.4 (42.2–46.5)
45.7 (42.6–48.7)
56.8 (52.0–61.6)
59.0 (54.0–64.0) 
Elixhauser comorbidities
0–1
2–4
≥5 

7,845
13,526
3,952 

45.6 (45.0–46.3)
40.6 (40.1–41.1)
34.8 (33.8–35.7) 

18.8 (17.4–20.1)
20.8 (19.8–21.8)
23.9 (22.0–25.8) 

179.2 (177.7–180.6)
171.5 (170.4–172.6)
162.7 (160.6–164.9) 

43.1 (40.2–45.9)
48.4 (46.3–50.5)
55.9 (51.6–60.1) 
SubgroupNIncidence of Delayed Testing, % (95% CI)Time Interval Between Tests, days (95% CI)
Proportion With Delay of Last Pre-Pandemic Test (reference)Percentage-Point Change in Proportion With Delay of First Pandemic Period TestDuration Between Last Two Pre-Pandemic Tests (reference)Excess Time Interval for First Pandemic Period Test
Age-group, years
18–64
≥65 

12,729
12,594 

41.6 (41.1–42.2)
40.9 (40.4–41.4) 

21.3 (20.3–22.4)
20.0 (19.0–21.0) 

177.2 (176.0–178.4)
167.7 (166.7–168.8) 

49.0 (46.6–51.4)
46.8 (44.8–48.9) 
Sex
Female
Male 

13,383
11,939 

40.8 (40.3–41.3)
41.7 (41.2–42.3) 

21.8 (20.8–22.9)
19.3 (18.2–20.4) 

172.3 (171.2–173.4)
172.7 (171.6–173.9) 

50.7 (48.5–52.8)
44.8 (42.5–47.2) 
Race/ethnicity
Non-Hispanic White
Non-Hispanic Black
Hispanic
Other specified 

18,062
5,517
840
863 

40.9 (40.4–41.3)
42.0 (41.2–42.8)
44.5 (42.5–46.5)
42.3 (40.3–44.3) 

20.4 (19.5–21.2)
21.4 (19.8–23.0)
19.6 (15.7–23.6)
21.8 (17.7–25.9) 

169.5 (168.6–170.4)
179.7 (177.8–181.5)
180.6 (175.8–185.5)
182.3 (177.6–187.0) 

46.6 (44.8–48.4)
51.7 (48.0–55.3)
51.7 (41.9–61.5)
47.1 (37.7–56.5) 
Diabetes type
Type 1
Type 2 

1,050
24,273 

33.0 (31.1–34.8)
41.6 (41.3–42.0) 

27.2 (23.5–30.9)
20.4 (19.6–21.1) 

163.1 (159.1–167.1)
172.9 (172.1– 173.7) 

59.2 (51.2–67.2)
47.4 (45.8–49.0) 
Insurance status
Medicare
Private
Medicaid
Uninsured 

14,794
9,616
117
698 

40.6 (40.1–41.1)
41.8 (41.2–42.4)
37.6 (31.5–43.7)
48.0 (45.7–50.3) 

20.4 (19.4–21.3)
21.0 (19.7–22.2)
23.1 (10.9–35.2)
23.5 (18.9–28.1) 

168.3 (167.3–169.2)
177.0 (175.6–178.3)
190.7 (173.5–207.9)
197.2 (191.0–203.3) 

47.4 (45.5–49.4)
47.2 (44.4–49.9)
64.4 (30.1–98.8)
67.3 (54.9–79.6) 
A1C, % (mmol/mol)
<7 (<53)
7–7.9 (53–63)
8–8.9 (64–74)
≥9% (≥75) 

13,285
5,946
2,801
3,247 

49.8 (49.3–50.3)
34.3 (33.5–35.0)
28.6 (27.5–29.7)
29.7 (28.6–30.7) 

17.3 (16.3–18.3)
22.6 (21.0–24.1)
26.1 (23.9–28.3)
26.0 (23.9–28.1) 

185.2 (184.1–186.2)
160.3 (158.8–161.8)
153.5 (151.1–155.9)
159.1 (156.6–161.5) 

44.4 (42.2–46.5)
45.7 (42.6–48.7)
56.8 (52.0–61.6)
59.0 (54.0–64.0) 
Elixhauser comorbidities
0–1
2–4
≥5 

7,845
13,526
3,952 

45.6 (45.0–46.3)
40.6 (40.1–41.1)
34.8 (33.8–35.7) 

18.8 (17.4–20.1)
20.8 (19.8–21.8)
23.9 (22.0–25.8) 

179.2 (177.7–180.6)
171.5 (170.4–172.6)
162.7 (160.6–164.9) 

43.1 (40.2–45.9)
48.4 (46.3–50.5)
55.9 (51.6–60.1) 

For A1C testing, a delay was defined as a time interval >180 days.

Supplementary Tables S5 and S6 present the proportions of patients delaying and the duration of delay for retinal screening and nephropathy evaluation before and during the pandemic according to subgroups defined by age, sex, race/ethnicity, diabetes type, insurance status, A1C level, and comorbidity burden. The proportion of patients with a delay in retinal screening in the pre-pandemic period was lower for individuals with type 1 diabetes than for those with type 2 diabetes. During the pandemic, the proportion of patients with a delay in retinal screening was greater for patients with type 1 diabetes than for those with type 2 diabetes. The proportion of patients with a delay for nephropathy evaluation was higher for individuals with type 1 diabetes than for those with type 2 diabetes in the pre-pandemic period, but not during the pandemic. We observed that higher A1C level and greater comorbidity burden were both associated with higher proportions of patients delaying both retinal screening and nephropathy evaluation in the pre-pandemic period. We did not observe subgroup differences in delaying either of these diabetes care measures during the pandemic period, but confidence limits were wide because of smaller subgroup sample sizes.

Supplementary Tables S7 and S8 present the proportion of patients with a delay in retinal screening and nephropathy evaluation before and during the pandemic by insulin use, area of residence, and ACS population-level demographic and socioeconomic subgroups. We observed differences in the proportion of patients delaying retinal screening during the pre pandemic period across subgroups of insulin use, area of residence, percentage of community unemployed, and percentage identifying as Hispanic, although precision around these estimates was low. We did not observe differences in the proportion of patients delaying retinal screening during the pandemic period across these same subgroups. We did not observe material differences in the proportion of patients delaying nephropathy evaluation before or during the pandemic across insulin use, area of residence, and ACS population-level demographic and socioeconomic subgroups.

Supplementary Table S9 shows the results of our sensitivity analysis. This analysis included individuals who had an EHR encounter but not an outcome-specific encounter during the pandemic year, and we censored an outcome-specific encounter date at 28 February 2021. The proportion of individuals who missed an evaluation during the pandemic year ranged from 12.5% for A1C testing to 44.9% for nephropathy evaluation. Compared with the primary analysis excluding these individuals, the increase in proportion of patients with a delay of an evaluation during the pandemic period was slightly higher for A1C testing (23.8 vs. 20.6 percentage points) and double for retinal screening (29.7 vs. 12.7 percentage points) and for nephropathy evaluation (39.7 vs. 18.6 percentage points).

In this EHR-based cohort of adult patients with diabetes who receive regular care from the WFBH system of western North Carolina, we observed a drop in frequency of diabetes care in the weeks after the start of the COVID-19 pandemic. Frequency of A1C testing, retinal screening exams, and nephropathy evaluations in the first 2 months of the pandemic each dropped to <50% of prior years’ total frequency. Frequency of A1C testing and retinal screening exams, but not nephropathy evaluations, were able to rebound to prior years’ assessment frequency during June 2020. Largely, retinal screening frequency during the remaining first year of the pandemic was comparable to prior years, whereas A1C testing frequency decreased in July 2020 and remained at <85% of prior years’ frequency, and the frequency of nephropathy evaluations were <70% of that in prior years.

These decreases in diabetes care procedures at the system-level resulted in within-individual delays of the first assessment during the pandemic for each of the procedures. The proportion of patients with a delayed first assessment during the pandemic increased by 35%, to 98%, among the three diabetes care process measures. Of concern, there were differences in the delaying of the first A1C test during the pandemic by important clinical subgroups. Individuals with the highest A1C and greatest burden of comorbidities had among the lowest proportion of patients delaying an A1C test before the pandemic, as would be expected given their need for close clinical monitoring to prevent complications. However, during the pandemic, both of these subgroups had the greatest respective increase in the proportion of patients with a delayed A1C test. On a positive note, there did not appear to be clinically meaningful differences in the delay of any of the diabetes care process measures during the pandemic according to age, sex, and race/ethnicity subgroups.

The mechanism for an increase in delay of diabetes care during the pandemic is not clear. At the start of the pandemic and during periods of resurgence when hospital workforce and capacity were inundated with or in preparation for COVID-19 patients, WFBH and other health systems across the nation limited elective or nonurgent procedures (24). A1C testing, retinal screening exams, and nephropathy evaluations are standards of care in diabetes management and treatment and vital to preventing complications. These procedures are not considered elective, although they could be interpreted as nonurgent relative to treating COVID-19 patients and those with other acute health conditions.

These diabetes care procedures require in-person clinical examination and laboratory evaluations and cannot be done via telemedicine. Thus, health care providers and patients may have chosen to delay a diabetes care procedure to forgo the risk of potential exposure to the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) that causes COVID-19. Fear of SARS-CoV-2 infection may have been greater for people with diabetes as a result of early evidence indicating higher risk of hospitalization and adverse outcomes for people with specific chronic conditions, including diabetes (25). Logistical and financial barriers for patients, such as lack of access to safe transportation, may also have contributed to a delay in care. The COVID-19 pandemic altered health care delivery; frequency of outpatient visits dropped, while use of telemedicine increased, and early indications suggest that overall health care utilization remains lower than in pre-pandemic years (12,14,26).

The current understanding of the impact of COVID-19 on diabetes care is limited, but our findings align with one other study that has assessed the occurrence of A1C testing (16). During the early weeks of the pandemic, population-level outpatient visits and A1C testing fell markedly among a repeated cross-sectional sample of >2 million individuals with type 2 diabetes and commercial or Medicare Advantage insurance coverage (16). Similarly, in a sample of 800,000 adults with type 2 diabetes in the United Kingdom, rates of A1C testing dropped from 20,000 tests per 100,000 person-months before the pandemic to 5,000 tests per 100,000 person-months in the months after the start of the pandemic (17). Our study extends these findings in multiple respects. First, we were able to characterize race/ethnicity of our sample participants and assess for subgroup differences in outcomes by this variable to identify potential racial/ethnic disparities in diabetes care during the pandemic. Second, we assessed for within-individual delays in diabetes care. By using individuals as their own control subject, we were able to assess delayed diabetes care during the pandemic compared with the most proximal receipt of diabetes care before the pandemic. Third, we included assessment of the additional standard-of-care procedures for people with diabetes of retinal screening exams and nephropathy evaluations (15).

Although Patel et al. (16) observed a drop in A1C testing, this decrease did not appear to be associated with worsening population-level A1C control. Other studies among individuals with type 2 diabetes suggest there is potential for worsening glycemic control during the pandemic (2730). A meta-analysis of observational studies including individuals with type 1 or type 2 diabetes suggested a differential impact of the pandemic on glucose control by diabetes type, with improved control among individuals with type 1 diabetes and worsening control among those with type 2 diabetes in the early stages of the pandemic (31). We observed a greater increase in the proportion of patients with delayed evaluation of A1C and retinal screening during the pandemic for individuals with type 1 diabetes than for those with type 2 diabetes. Data from the U.K. Clinical Research Practice Datalink Aurum database showed a dramatic population-level reduction in contacts for diabetic emergencies at the start of the pandemic; investigators suggest this reduction was from missing or forgoing contact during a diabetic emergency and rather than from the absence of a diabetic emergency (32).

These reductions in diabetes care and worsening of glucose control are concerning given their relationship with long-term clinical complications. Given the short period of follow-up, we did not assess whether delays in diabetes care were associated with clinical complications in our study. However, other studies have assessed the relationship between process measures of diabetes care and risk for clinical complications. Individuals who maintain adherence to annual guidelines for diabetes care have dramatically lower risk (by 20–70%) for major cardiovascular and kidney complications, amputation, and death during the ensuing 2–6 years (33,34).

A couple of limitations of our work should be considered. First, our study sample selection required individuals to receive their regular diabetes care from the WFBH system. Although we included individuals who were uninsured or receiving coverage from Medicaid, we may have excluded individuals who predominately receive their care from WFBH but do so infrequently. The proportion of our cohort that included individuals of non-Hispanic Black race/ethnicity (22%) was not commensurate with the demographics of Winston-Salem, NC, and the surrounding communities (35%) (35). We used fixed-effects models to control for confounding from stable within-individual characteristics. However, some characteristics may have changed over time as a result of the pandemic and could affect the process measures of diabetes care assessed here. For example, we did not control for changes in employment and insurance coverage or for a COVID-19 diagnosis. However, our subgroup estimates by health care status did not suggest differences in diabetes care during the pandemic.

Our findings have important clinical and public health implications. First, they demonstrate a gap in health care delivery and the need to develop strategies that allow health care systems to maintain crucial components of diabetes care during periods of population-level health care interruption. Second, these results portend the potential for serious complications for the most clinically vulnerable individuals with diabetes in this health care system—those with high A1C and multiple comorbidities—and suggest the need to prioritize health care delivery to and communication with these individuals during periods of health care interruption. Our findings should be a forewarning regarding future national crises, for our own health care system and for other health care systems with similar patient populations and composition, geographic location and spread, and health care administrative policies.

In conclusion, in this EHR-based cohort of adult patients with diabetes, there was a significant decrease in population-level frequency of A1C testing, retinal screening exams, and nephropathy evaluations during the COVID-19 pandemic compared with the same calendar periods of prior years. These declines in diabetes care processes at the health care system level corresponded to within-individual delays of the first assessment during the pandemic for each of the procedures assessed.

Acknowledgments

The authors thank the patients and technical staff of the WFBH system. They also acknowledge support from the Clinical and Translational Science Institute in providing access to the study data. The data used for analysis are housed at WFBH and are not available for public distribution. The views expressed in this article are those of the authors and do not necessarily represent the views of Atrium Health or WFBH.

Duality of Interest

No potential conflicts of interest relevant to this article were reported.

Author Contributions

M.P.B., M.-Y.L., and A.H. designed this study. M.P.B. analyzed the data and wrote the manuscript. M.-Y.L. and A.H. analyzed the data, contributed to the interpretation of results, and edited and reviewed the manuscript. A.B. and B.J.W. contributed to the clinical methodology and edited and reviewed the manuscript. W.M.F. and Z.L. helped assemble the data and reviewed the manuscript. B.O. helped to assemble the data and edited and reviewed the manuscript. M.P.B. is the guarantor of this work and, as such, takes responsibility for the integrity of the data and the accuracy of the data analysis.

This article contains supplementary material online at https://doi.org/10.2337/figshare.19379516.

1.
Woolf
SH
,
Chapman
DA
,
Sabo
RT
,
Zimmerman
EB
.
Excess deaths from COVID-19 and other causes in the US, March 1, 2020, to January 2, 2021
.
JAMA
2021
;
325
:
1786
1789
2.
Woolf
SH
,
Chapman
DA
,
Lee
JH
.
COVID-19 as the leading cause of death in the United States
.
JAMA
2021
;
325
:
123
124
3.
Mackey
K
,
Ayers
CK
,
Kondo
KK
, et al
.
Racial and ethnic disparities in COVID-19-related infections, hospitalizations, and deaths: a systematic review
.
Ann Intern Med
2021
;
174
:
362
373
4.
Figueroa
JF
,
Wadhera
RK
,
Mehtsun
WT
,
Riley
K
,
Phelan
J
,
Jha
AK
.
Association of race, ethnicity, and community-level factors with COVID-19 cases and deaths across U.S. counties
.
Healthc (Amst)
2021
;
9
:
100495
5.
Menke
A
,
Casagrande
S
,
Geiss
L
,
Cowie
CC
.
Prevalence of and trends in diabetes among adults in the United States, 1988–2012
.
JAMA
2015
;
314
:
1021
1029
6.
Myers
CA
,
Slack
T
,
Broyles
ST
,
Heymsfield
SB
,
Church
TS
,
Martin
CK
.
Diabetes prevalence is associated with different community factors in the diabetes belt versus the rest of the United States
.
Obesity (Silver Spring)
2017
;
25
:
452
459
7.
Hill-Briggs
F
,
Adler
NE
,
Berkowitz
SA
, et al
.
Social determinants of health and diabetes: a scientific review
.
Diabetes Care
2020
;
44
:
258
279
8.
Walker
RJ
,
Garacci
E
,
Palatnik
A
,
Ozieh
MN
,
Egede
LE
.
The longitudinal influence of social determinants of health on glycemic control in elderly adults with diabetes
.
Diabetes Care
2020
;
43
:
759
766
9.
Mehrotra
A
,
Chernew
M
,
Linetsky
D
,
Hatch
H
,
Cutler
DM
,
Schneider
EC
.
The impact of the COVID-19 pandemic on outpatient care: visits return to prepandemic levels, but not for all providers and patients
.
10.
Jiang
J
,
Bai
G
,
Gustafsson
L
,
Anderson
GF
.
Canary in a coal mine? A look at initial data on COVID-19’s impact on U.S. hospitals
.
11.
Solomon
MD
,
McNulty
EJ
,
Rana
JS
, et al
.
The Covid-19 pandemic and the incidence of acute myocardial infarction
.
N Engl J Med
2020
;
383
:
691
693
12.
Patel
SY
,
Mehrotra
A
,
Huskamp
HA
,
Uscher-Pines
L
,
Ganguli
I
,
Barnett
ML
.
Trends in outpatient care delivery and telemedicine during the COVID-19 pandemic in the US
.
JAMA Intern Med
2021
;
181
:
388
391
13.
Patel
SY
,
Mehrotra
A
,
Huskamp
HA
,
Uscher-Pines
L
,
Ganguli
I
,
Barnett
ML
.
Variation in telemedicine use and outpatient care during the COVID-19 pandemic in the United States
.
Health Aff (Millwood)
2021
;
40
:
349
358
14.
Peterson-KFF Health System Tracker
.
Early 2021 data show no rebound in health care utilization
.
15.
American Diabetes Association
.
Introduction: Standards of Medical Care in Diabetes—2021
.
Diabetes Care
2021
;
44
(
Suppl. 1
):
S1
S2
16.
Patel
SY
,
McCoy
RG
,
Barnett
ML
,
Shah
ND
,
Mehrotra
A
.
Diabetes care and glycemic control during the COVID-19 pandemic in the United States
.
JAMA Intern Med
2021
;
181
:
1412
1414
17.
Carr
MJ
,
Wright
AK
,
Leelarathna
L
, et al
.
Impact of COVID-19 on diagnoses, monitoring, and mortality in people with type 2 diabetes in the UK
.
Lancet Diabetes Endocrinol
2021
;
9
:
413
415
18.
Greenland
S
,
Morgenstern
H
.
Ecological bias, confounding, and effect modification
.
Int J Epidemiol
1989
;
18
:
269
274
19.
Greenland
S
,
Robins
J
.
Invited commentary: ecologic studies: biases, misconceptions, and counterexamples
.
Am J Epidemiol
1994
;
139
:
747
760
20.
Agency for Healthcare Research and Quality
.
Social determinants of health database (beta version)
.
Available from https://www.ahrq.gov/sdoh/data-analytics/sdoh-data.html. Accessed 8 September 2021
21.
Elixhauser
A
,
Steiner
C
,
Harris
DR
,
Coffey
RM
.
Comorbidity measures for use with administrative data
.
Med Care
1998
;
36
:
8
27
22.
U.S. Department of Agriculture Economic Research Service
.
Rural-urban commuting area codes
.
23.
Allison
P
.
Fixed Effects Regression Models
.
Thousand Oaks, CA
,
Sage Publications
,
2009
24.
Paavola
A
.
106 Hospitals postponing elective procedures amid the COVID-19 resurgence
.
25.
Williamson
EJ
,
Walker
AJ
,
Bhaskaran
K
, et al
.
Factors associated with COVID-19-related death using OpenSAFELY
.
Nature
2020
;
584
:
430
436
26.
Bonora
BM
,
Morieri
ML
,
Avogaro
A
,
Fadini
GP
.
The toll of lockdown against COVID-19 on diabetes outpatient care: analysis from an outbreak area in northeast Italy
.
Diabetes Care
2021
;
44
:
e18
e21
27.
Önmez
A
,
Gamsızkan
Z
,
Özdemir
Ş
, et al
.
The effect of COVID-19 lockdown on glycemic control in patients with type 2 diabetes mellitus in Turkey
.
Diabetes Metab Syndr
2020
;
14
:
1963
1966
28.
Karatas
S
,
Yesim
T
,
Beysel
S
.
Impact of lockdown COVID-19 on metabolic control in type 2 diabetes mellitus and healthy people
.
Prim Care Diabetes
2021
;
15
:
424
427
29.
Tanji
Y
,
Sawada
S
,
Watanabe
T
, et al
.
Impact of COVID-19 pandemic on glycemic control among outpatients with type 2 diabetes in Japan: a hospital-based survey from a country without lockdown
.
Diabetes Res Clin Pract
2021
;
176
:
108840
30.
D’Onofrio
L
,
Pieralice
S
,
Maddaloni
E
, et al
.
Effects of the COVID-19 lockdown on glycaemic control in subjects with type 2 diabetes: the glycalock study
.
Diabetes Obes Metab
2021
;
23
:
1624
1630
31.
Eberle
C
,
Stichling
S
.
Impact of COVID-19 lockdown on glycemic control in patients with type 1 and type 2 diabetes mellitus: a systematic review
.
Diabetol Metab Syndr
2021
;
13
:
95
32.
Mansfield
KE
,
Mathur
R
,
Tazare
J
, et al
.
Indirect acute effects of the COVID-19 pandemic on physical and mental health in the UK: a population-based study
.
Lancet Digit Health
2021
;
3
:
e217
e230
33.
Bayer
FJ
,
Galusha
D
,
Slade
M
,
Chu
IM
,
Taiwo
O
,
Cullen
MR
.
Process of care compliance is associated with fewer diabetes complications
.
Am J Manag Care
2014
;
20
:
41
52
34.
Chen
Y
,
Sloan
FA
,
Yashkin
AP
.
Adherence to diabetes guidelines for screening, physical activity and medication and onset of complications and death
.
J Diabetes Complications
2015
;
29
:
1228
1233
35.
U.S. Census Bureau
.
QuickFacts: Winston-Salem, North Carolina
.
Population estimates, July 1, 2019 (V2019). Available from https://www.census.gov/quickfacts/winstonsalemcitynorthcarolina. Accessed 28 August 2021
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://diabetesjournals.org/journals/pages/license.