Changes in Healthcare Utilization Among Medicare Beneficiaries With Diabetes 2 Years Into the COVID-19 Pandemic

HIGHLIGHTS

• •Healthcare services were reduced among persons with diabetes during the COVID-19 pandemic.
• •The decline was larger in the initial months of the pandemic but persisted through 2021.
• •Changes in utilization varied by age group, sex, race/ethnicity, and urbanicity.
• •The increase in telehealth visits did not compensate for the decrease in in-person visits.

Abstract

Introduction

The COVID-19 pandemic abruptly impacted healthcare service delivery and utilization. However, the impact on older adults with diabetes in the U.S. is unclear. This study aimed to estimate changes in healthcare utilization among older adults with diabetes during the initial 2 years of the COVID-19 pandemic compared with the changes in the 2 years before and to examine the variation in utilization changes by demographic and socioeconomic characteristics.

Methods

In this study, we analyzed changes in utilization, measured by the average use of healthcare services per 1,000 persons with diabetes, using medical claims for Medicare fee-for-service beneficiaries aged ≥67 years. Utilization changes by setting (acute inpatient, emergency room, hospital outpatient, physician office, and ambulatory surgery center) and by media (telehealth and in person) were examined for 22 months of the pandemic (March 2020–December 2021) compared with those during prepandemic period (March 2018–December 2019). We also estimated utilization changes by beneficiaries’ age group, sex, race/ethnicity, and residential urbanicity.

Results

The study sample consisted of approximately 6 million beneficiaries with diabetes each month. In the first 2 years of the pandemic, the average use of healthcare services by setting was 5%–17% lower than the prepandemic level for all types of services. Phase 1 (March 2020−May 2020) had the largest decrease in utilization: physician office visits changed by −51.2% (95% CI= −55.0%, −47.5%), ambulatory surgery center procedures changed by −45.1% (95% CI= −49.8%, −40.4%), emergency room visits changed by −36.9% (95% CI= −39.0%, −34.7%), acute inpatient stays changed by −31.5% (95% CI= −33.6%, −29.3%), and hospital outpatient visits changed by −27% (95% CI= −29.3%, −24.8%). The reduction in utilization varied by sociodemographic subgroup. During the pandemic, the use of telehealth visits increased by 511.1% (95% CI=502.2%, 520.0%) compared with that in the prepandemic period. The increase was smaller among rural residents.

Conclusions

Medicare beneficiaries with diabetes experienced a reduction in the use of healthcare services during the COVID-19 pandemic, some of which persisted through 2 years into the pandemic. Telehealth visits increased but not enough to overcome decreases in in-person visits. Understanding these patterns may help to optimize the use of healthcare resources for diabetes management in the postpandemic era and during future emergencies.

INTRODUCTION

The coronavirus disease 2019 (COVID-19) pandemic abruptly impacted healthcare service delivery and utilization worldwide.¹ On one hand, providers postponed treatment for nonurgent conditions, such as in-person office visits and elective procedures.^2,3 By contrast, patients decreased their use of healthcare services owing to fear of contagion and/or social distancing policies.^3,4 Both effects together led to a significant reduction in healthcare service utilization. For example, in the U.S., studies estimated that the decrease in the use of the emergency room (ER) and outpatient services was 42% and 70%, respectively, in April 2020 compared with that in the same weeks in 2019.^3,5

The COVID-19 pandemic had an especially adverse impact on persons with chronic conditions, such as diabetes. Epidemiologic studies show that having diabetes increased the risk of having COVID-19–related hospitalization by 3–4 times,⁶ and people with diabetes accounted for one third of in-hospital deaths with COVID-19.⁷ Conversely, if the pandemic disrupted necessary medical care for diabetes management, insufficient or delayed care may have worsened patient health outcomes.

To examine the impacts of the reduction in healthcare utilization on diabetes-related health outcomes, the first step is to measure the magnitude of the reduction. However, there have been few attempts at such estimations. Existing studies have examined ER visits,⁵ outpatient visits,^3,8 telehealth visits,⁹ and elective and nonelective care¹⁰ during the pandemic. Nonetheless, these studies restricted their analyses to specific types of healthcare services, thus leaving the comprehensive picture of pandemic-induced healthcare service reduction unclear. Moreover, these studies examined the general population, and few have focused specifically on persons with diabetes, who are among the most disproportionately affected groups during the pandemic.^6,7 In addition, almost all previous studies had a relatively short study period limited to the first few months of the pandemic. Understanding the impact of the pandemic more comprehensively and over a longer period will be important as more data become available.

This study examined the changes in healthcare utilization during the COVID-19 pandemic among Medicare beneficiaries with diabetes. First, we evaluated changes in healthcare utilization from the 2 years before to the 2 years after the outbreak of the pandemic by setting, in total, and by subgroup. Then, we compared utilization changes in telehealth visits with those in in-person visits.

METHODS

Study Population

This study used a database of all medical claims for Medicare fee-for-service (FFS) beneficiaries between 2018 and 2021. Medicare is the primary health insurance for adults aged ≥65 years in the U.S. We used Medicare enrollment Part A and Part B files, which contain demographics and medical claims for all Medicare beneficiaries covered by the FFS program.

The study sample comprised beneficiaries with diabetes who were fully enrolled in both Parts A and B for 24 months, including the index month, that is, the month when services were measured, and 23 months before it. Because a 24-month window was required to identify diabetes, the age range for the study sample was ≥67 years.

Measures

Diabetes was identified using ICD-10-CM codes (E10, E11, and E13). A beneficiary was considered to have diabetes if this person had (1) at least 1 diabetes-related inpatient claim or (2) at least 2 diabetes-related outpatient claims that occurred on different days within a 24-month window. The requirement of at least 2 outpatient claims reduces the possibility of false positives because the first claim might indicate a screening visit.

Claims were used to estimate changes in utilization for various types of healthcare services. First, we examined claims for acute inpatient stays, ER visits, hospital outpatient visits, physician office events, and ambulatory surgery center (ASC) procedures. Acute inpatient stays were unique admissions at acute care hospitals. We considered 3 types of outpatient visits: (1) hospital outpatient visits, which were healthcare services provided at hospital-based clinics, rural health clinics, and federally qualified health centers; (2) physician office events, which were services provided by healthcare professionals in a provider office setting; and (3) ASC procedures, which were procedures performed at ASCs. During the pandemic period (March 2020–December 2021), we further determined whether the visit or admission was COVID-19 related. COVID-19–related visits were identified using ICD-10-CM codes B97.29 and U07.1 before April 2020 and U07.1 afterward.^11,12

Second, we explored the use of telehealth versus in-person visits with healthcare professionals in outpatient settings. We identified telehealth visits using the Health Common Procedure Coding System codes provided by the Centers for Medicare and Medicaid Services¹³ combined with procedure code modifier (95, GT, GQ, or G0) or the place of service code (value of 2).¹⁰ Note that Centers for Medicare and Medicaid Services expanded telehealth-eligible services after the onset of the pandemic, and our study used the list of services updated in August 2022.

Statistical Analysis

To estimate the changes in healthcare utilization, we calculated the average use of healthcare services per 1,000 beneficiaries with diabetes each month. We compared utilization during the pandemic period, defined as March 2020 to December 2021, with that during the prepandemic period, defined as March 2018 to December 2019. We chose March 2020 as the start of the pandemic because that is the month that the U.S. and the WHO declared a public health emergency because of COVID-19.^14,15 January and February 2020 were excluded from the study period because the COVID-19 cases were ramping up but not yet declared a pandemic.

We first plotted the average use of healthcare services to visualize changes during the study period. Then, we quantified the changes by estimating the following linear regression:

$$\text{ln}{\left(\text{Average}\text{use}\right)}_{sm}=\alpha +\sum _p{\beta }^{p}Phas{e}_m^p+{\delta }_s+{\delta }_m+{\epsilon }_{sm}$$

The dependent variable is the natural log of the average use of each type of healthcare service per 1,000 beneficiaries with diabetes in state $s$ in month $m$, where $m$ equals 1 to 22. We divided the pandemic period into 4 phases: Phase 1, March 2020–May 2020, when most states responded to the pandemic by announcing state emergency status and issuing shelter-in-place orders; Phase 2, June 2020–December 2020, when states released the shelter-in-place orders gradually; Phase 3, January 2021–June 2021, when vaccination became available, and the vaccination rate increased; and Phase 4, July 2021–December 2021, when the increase in the vaccination rate slowed, and new variants dominated COVID-19 cases. The cutoff of phases was based on data patterns and COVID-19–related policies and news. We also tested sensitivity by moving 1 month in and out of a phase and did not find significant differences. The variable of interest, $Phas{e}_m^p$, is an indicator for each phase, where $p$ equals 1 to 4, corresponding to the 4 phases. The parameter ${\beta }^p$ represents the changes in the use of healthcare services in each phase during the pandemic compared with changes in the corresponding months during the prepandemic period. In a separate regression, we also estimated the total impact of the pandemic by replacing the phase indicators with a single indicator for all pandemic months. In the regression, we controlled for the state-fixed effects, ${\delta }_s$, and month-fixed effects, ${\delta }_m$. State-fixed effects control for time-invariant unobserved characteristics for the state, and month-fixed effects control for unobserved characteristics that evolve over time but are constant across states.¹⁶ SEs were clustered at the state and month levels.

To explore whether the changes varied by subgroup, we estimated the regression by age group (67–74, 75–84, and ≥85 years), sex, race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, and others), and urbanicity (on the basis of the 2013 urban–rural classification scheme for counties provided by the Centers for Disease Control and Prevention's National Center for Health Statistics¹⁷). This study was exempt from IRB or ethics review. SAS, version 9.4 (SAS Institute, Cary, NC) was used for statistical analyses.

RESULTS

The number of Medicare FFS beneficiaries with diabetes declined from 6.3 million to 5.9 million between 2018 and 2021. The demographic composition of the study population also changed slightly, with fewer people aged ≥85 years, fewer non-Hispanic Black persons, fewer Hispanic persons, and more males (Table 1).

Table 1.

Descriptive Statistics of Medicare Fee-for-Service Beneficiaries With Diabetes, Estimated in June of Each Year

	2018		2019		2020		2021
	Number	Percentage	Number	Percentage	Number	Percentage	Number	Percentage
Total	6,345,561	100.0	6,302,830	100.0	6,140,884	100.0	5,894,898	100.0
By age group
67–74	2,789,069	44.0	2,777,891	44.1	2,723,101	44.3	2,580,921	43.8
75–84	2,517,601	39.7	2,502,239	39.7	2,432,507	39.6	2,385,131	40.5
≥85	1,038,891	16.4	1,022,700	16.2	985,276	16.0	928,846	15.8
By race and ethnicity
Non-Hispanic White	4,820,761	76.0	4,779,208	75.8	4,660,005	75.9	4,496,673	76.3
Non-Hispanic Black	659,196	10.4	644,832	10.2	610,925	10.0	554,957	9.4
Hispanic	444,048	7.0	437,198	6.9	420,947	6.9	395,263	6.7
Others	421,556	6.6	441,592	7.0	449,007	7.3	448,005	7.6
By sex
Male	2,999,987	47.3	3,004,451	47.7	2,952,161	48.1	2,850,743	48.4
Female	3,345,574	52.7	3,298,379	52.3	3,188,723	51.9	3,044,155	51.6
By urbanicity
Rural	1,478,508	23.3	1,466,630	23.3	1,427,888	23.3	1,360,220	23.1
Urban	4,867,053	76.7	4,836,200	76.7	4,712,996	76.8	4,534,678	76.9

Note: This study used medical claims for Medicare fee-for-service beneficiaries aged ≥67 years.

In Figure 1, we plotted the unadjusted average use of healthcare services by setting for the prepandemic and pandemic periods. Healthcare utilization for all services decreased sharply during the initial phase of the pandemic and then began to rebound after May 2020. However, almost 2 years after the pandemic outbreak, acute inpatient stays and ER visits were still lower than prepandemic levels. The proportion of utilization that was COVID-19 related peaked in the winter of 2020 for all service types (Appendix Figure 1, available online).

Figure 1.

Average use of healthcare services before (March 2018–December 2019) and during (March 2020–December 2021) the COVID-19 pandemic among Medicare fee-for-service beneficiaries with diabetes.^a

^aAcute inpatient stays were unique admissions at acute care hospitals. Hospital outpatient visits were healthcare services provided at hospital-based clinics, rural health clinics, and federally qualified health centers. Physician office events were services provided by healthcare professionals in a provider office setting. ASC procedures were procedures performed at ASCs.

Apr., April; ASC, ambulatory surgery center; Aug., August; Dec., December; ER, emergency room; Feb., February; Jan., January; Jun., June; Jul., July; Mar., March; Nov., November; Oct., October; Sep., September.

Overall, utilization during the pandemic was lower than the prepandemic level for all types of services (Table 2). Total changes were the largest for ER visits and acute inpatient stays, which were decreased by 16.8% (95% CI= −17.5%, −16.1%) and 14.3% (95% CI= −14.9%, −13.7%), respectively. However, in the initial phase of the pandemic, that is, from March to May 2020, physician office events and ASC procedures had the largest reductions: −51.2% (95% CI= −55%, −47.5%) and −45.1% (95% CI= −49.8%, −40.4%), respectively. Unlike other service types, ASC procedures had a higher level of utilization in the last phase of the study period than in the prepandemic period (3.9% higher).

Table 2.

Percentage Changes in Utilization by Setting During the COVID-19 Pandemic Among Persons With Diabetes

		By age group, years			By race				By sex		By urbanicity
	Overall	67–74	75–84	≥85	Non-Hispanic White	Non-Hispanic Black	Hispanic	Others	Male	Female	Rural	Urban
Acute inpatient
Total	−14.3	−14.3	−14.1	−13.8	−15.0	−11.3	−9.4	−13.3	−13.4	−15.2	−15.4	−13.8
	(−14.9, −13.7)	(−15.0, −13.6)	(−14.8, −13.4)	(−14.6, −13.0)	(−15.6, −14.3)	(−13.4, −9.1)	(−11.4, −7.3)	(−14.6, −11.9)	(−14.1, −12.8)	(−15.8, −14.5)	(−16.4, −14.3)	(−14.4, −13.1)
Phase 1	−31.5	−31.8	−31.0	−31.1	−33.3	−19.0	−28.3	−32.5	−30.1	−32.9	−31.8	−30.8
	(−33.6, −29.3)	(−34.2, −29.4)	(−33.1, −28.8)	(−33.8, −28.3)	(−35.5, −31.1)	(−25.8, −12.2)	(−34.3, −22.3)	(−36.3, −28.7)	(−32.2, −27.9)	(−35.1, −30.6)	(−35.5, −28.2)	(−33.1, −28.6)
Phase 2	−10.1	−9.7	−9.7	−10.4	−11.2	−8.8	−0.3	−7.1	−9.0	−11.2	−10.3	−9.9
	(−10.8, −9.3)	(−10.6, −8.8)	(−10.6, −8.8)	(−11.5, −9.2)	(−12.1, −10.4)	(−12.1, −5.5)	(−3.9, 3.2)	(−9.5, −4.7)	(−9.8, −8.1)	(−12.0, −10.4)	(−11.6, −8.9)	(−10.7, −9.1)
Phase 3	−14.3	−14.1	−14.1	−14.1	−14.5	−8.0	−11.5	−13.5	−13.8	−14.8	−15.9	−13.6
	(−15.0, −13.6)	(−15.0, −13.2)	(−15.0, −13.2)	(−15.3, −12.9)	(−15.3, −13.8)	(−12.1, −3.9)	(−15.4, −7.6)	(−15.7, −11.2)	(−14.6, −12.9)	(−15.6, −14.0)	(−17.4, −14.5)	(−14.4, −12.8)
Phase 4	−10.7	−11.0	−10.8	−9.0	−10.5	−13.5	−8.3	−10.7	−10.0	−11.4	−12.4	−9.9
	(−11.3, −10.0)	(−11.9, −10.2)	(−11.7, −10.0)	(−10.1, −7.9)	(−11.3, −9.8)	(−17.8, −9.1)	(−11.7, −4.9)	(−13.0, −8.3)	(−10.8, −9.1)	(−12.1, −10.6)	(−14.2, −10.6)	(−10.6, −9.1)
ER
Total	−16.8	−16.8	−16.4	−16.7	−16.1	−20.1	−18.4	−17.3	−15.2	−18.1	−15.1	−17.2
	(−17.5, −16.1)	(−17.5, −16.0)	(−17.1, −15.6)	(−17.5, −15.9)	(−16.9, −15.4)	(−21.9, −18.3)	(−20.0, −16.9)	(−18.5, −16.1)	(−15.9, −14.5)	(−18.8, −17.3)	(−16.2, −14.0)	(−17.9, −16.5)
Phase 1	−36.9	−35.9	−36.7	−37.9	−37.5	−33.6	−41.2	−38.5	−33.3	−39.9	−33.9	−37.2
	(−39.0, −34.7)	(−38.1, −33.7)	(−38.9, −34.5)	(−40.4, −35.5)	(−39.7, −35.2)	(−39.4, −27.9)	(−46.1, −36.3)	(−42.2, −34.9)	(−35.4, −31.3)	(−42.2, −37.5)	(−37.9, −29.9)	(−39.4, −35.0)
Phase 2	−16.2	−16.2	−15.4	−16.6	−15.7	−21.7	−17.0	−15.7	−14.0	−17.9	−13.5	−16.9
	(−17.0, −15.4)	(−17.1, −15.2)	(−16.4, −14.5)	(−17.6, −15.6)	(−16.6, −14.8)	(−24.8, −18.6)	(−19.4, −14.6)	(−17.8, −13.7)	(−14.8, −13.2)	(−18.8, −17.0)	(−15.3, −11.6)	(−17.7, −16.1)
Phase 3	−15.8	−16.2	−15.4	−15.2	−14.9	−16.8	−16.7	−17.0	−15.3	−16.2	−15.9	−15.8
	(−16.5, −15.1)	(−17.1, −15.3)	(−16.2, −14.6)	(−16.2, −14.3)	(−15.6, −14.1)	(−20.2, −13.4)	(−19.5, −13.9)	(−18.9, −15.1)	(−16.0, −14.5)	(−17.0, −15.4)	(−17.2, −14.7)	(−16.6, −15.1)
Phase 4	−8.4	−8.5	−8.3	−7.8	−7.2	−14.7	−10.4	−8.7	−7.4	−9.2	−6.8	−9.0
	(−9.1, −7.7)	(−9.5, −7.5)	(−9.2, −7.5)	(−8.7, −6.9)	(−7.9, −6.4)	(−17.9, −11.5)	(−13.0, −7.8)	(−10.4, −6.9)	(−8.2, −6.6)	(−10.0, −8.4)	(−8.3, −5.3)	(−9.8, −8.3)
HOP
Total	−5.7	−5.2	−6.3	−5.4	−5.6	−6.9	−5.5	−6.1	−5.6	−5.7	−5.9	−6.1
	(−6.4, −5.0)	(−5.9, −4.4)	(−7.1, −5.6)	(−6.1, −4.8)	(−6.3, −4.8)	(−8.0, −5.9)	(−6.4, −4.6)	(−7.0, −5.2)	(−6.3, −4.9)	(−6.5, −5.0)	(−7.5, −4.2)	(−6.9, −5.4)
Phase 1	−27.0	−28.0	−28.8	−21.2	−30.0	−19.5	−20.3	−29.6	−26.2	−27.8	−26.2	−27.8
	(−29.3, −24.8)	(−30.4, −25.6)	(−31.1, −26.4)	(−23.3, −19.0)	(−32.4, −27.6)	(−23.3, −15.7)	(−22.8, −17.9)	(−32.5, −26.7)	(−28.4, −23.9)	(−30.1, −25.5)	(−30.0, −22.5)	(−30.1, −25.4)
Phase 2	−2.5	−1.5	−3.3	−3.1	−2.7	−4.1	−1.3	−4.2	−2.3	−2.6	−1.4	−2.8
	(−3.2, −1.8)	(−2.2, −0.8)	(−4.0, −2.6)	(−3.9, −2.3)	(−3.5, −2.0)	(−5.9, −2.3)	(−2.7, 0.0)	(−5.5, −2.9)	(−3.0, −1.6)	(−3.3, −1.9)	(−4.6, 1.8)	(−3.6, −2.1)
Phase 3	−1.0	−0.6	−1.3	−1.3	−0.1	−4.6	−2.4	−0.5	−1.2	−0.8	−1.4	−1.5
	(−2.0, −0.1)	(−1.6, 0.4)	(−2.3, −0.3)	(−2.3, −0.2)	(−1.2, 0.9)	(−6.3, −3.0)	(−4.0, −0.9)	(−1.7, 0.7)	(−2.1, −0.2)	(−1.8, 0.2)	(−3.6, 0.9)	(−2.4, −0.5)
Phase 4	−3.4	−2.6	−3.7	−4.4	−2.1	−6.2	−6.0	−2.0	−3.4	−3.3	−5.5	−3.9
	(−4.0, −2.8)	(−3.3, −2.0)	(−4.3, −3.0)	(−5.2, −3.5)	(−2.7, −1.4)	(−7.6, −4.9)	(−7.4, −4.7)	(−3.0, −1.0)	(−4.0, −2.8)	(−4.0, −2.7)	(−8.8, −2.2)	(−4.5, −3.3)
Physician office
Total	−12.6	−11.7	−13.2	−14.7	−12.6	−11.4	−12.9	−12.6	−11.6	−13.6	−14.5	−12.1
	(−13.8, −11.5)	(−12.8, −10.6)	(−14.3, −12.0)	(−16.1, −13.4)	(−13.8, −11.5)	(−12.8, −10.1)	(−14.2, −11.6)	(−13.9, −11.4)	(−12.7, −10.5)	(−14.9, −12.4)	(−16.0, −13.1)	(−13.3, −11.0)
Phase 1	−51.2	−47.3	−52.6	−59.8	−51.0	−49.7	−52.4	−53.3	−47.8	−54.6	−51.9	−51.0
	(−55.0, −47.5)	(−50.8, −43.8)	(−56.4, −48.8)	(−63.9, −55.7)	(−54.7, −47.2)	(−53.9, −45.5)	(−56.7, −48.1)	(−57.1, −49.5)	(−51.4, −44.3)	(−58.5, −50.7)	(−56.4, −47.4)	(−54.7, −47.3)
Phase 2	−9.7	−8.8	−9.9	−12.4	−9.3	−10.9	−10.5	−10.5	−8.4	−10.9	−11.9	−9.3
	(−10.6, −8.7)	(−9.7, −7.8)	(−10.9, −8.9)	(−13.4, −11.4)	(−10.3, −8.4)	(−12.6, −9.3)	(−11.9, −9.1)	(−11.7, −9.3)	(−9.4, −7.5)	(−11.9, −9.9)	(−13.5, −10.2)	(−10.2, −8.3)
Phase 3	−6.7	−6.4	−7.3	−6.8	−6.9	−4.3	−6.0	−6.2	−6.3	−7.1	−8.9	−6.1
	(−7.9, −5.5)	(−7.6, −5.2)	(−8.5, −6.0)	(−8.1, −5.5)	(−8.1, −5.7)	(−6.1, −2.6)	(−7.6, −4.4)	(−7.5, −4.8)	(−7.5, −5.2)	(−8.4, −5.9)	(−10.8, −7.0)	(−7.3, −4.9)
Phase 4	−2.7	−2.5	−3.1	−2.8	−3.0	0.0	−2.9	−1.4	−2.5	−2.9	−4.5	−2.0
	(−3.5, −2.0)	(−3.3, −1.8)	(−3.9, −2.4)	(−3.7, −2.0)	(−3.7, −2.2)	(−1.2, 1.3)	(−4.0, −1.7)	(−2.3, −0.4)	(−3.3, −1.8)	(−3.7, −2.1)	(−6.6, −2.5)	(−2.8, −1.3)
ASC
Total	−5.2	−5.4	−5.6	−4.2	−5.2	−7.1	−5.1	−5.2	−3.9	−6.6	−6.6	−4.9
	(−6.5, −4.0)	(−6.6, −4.1)	(−6.9, −4.3)	(−5.5, −2.8)	(−6.4, −3.9)	(−9.2, −5.0)	(−6.9, −3.3)	(−7.0, −3.5)	(−5.1, −2.8)	(−8.0, −5.2)	(−8.4, −4.8)	(−6.1, −3.6)
Phase 1	−45.1	−44.6	−46.0	−45.2	−44.8	−46.9	−49.4	−52.5	−40.5	−50.0	−44.7	−45.4
	(−49.8, −40.4)	(−49.2, −39.9)	(−50.8, −41.1)	(−50.0, −40.5)	(−49.4, −40.1)	(−53.3, −40.6)	(−55.4, −43.3)	(−59.3, −45.7)	(−44.9, −36.1)	(−55.1, −44.9)	(−50.1, −39.3)	(−50.3, −40.6)
Phase 2	−1.1	−0.9	−1.1	−2.0	−0.8	−7.0	−0.6	−2.0	0.6	−2.9	−4.1	−0.5
	(−2.1, −0.2)	(−1.9, 0.1)	(−2.1, −0.0)	(−3.2, −0.7)	(−1.7, 0.2)	(−9.8, −4.2)	(−3.0, 1.8)	(−4.1, 0.1)	(−0.4, 1.5)	(−4.0, −1.8)	(−6.2, −2.0)	(−1.5, 0.5)
Phase 3	0.8	0.2	0.2	3.4	0.8	−0.8	1.6	2.0	1.5	0.0	2.4	1.2
	(−0.6, 2.1)	(−1.2, 1.5)	(−1.2, 1.6)	(1.9, 5.0)	(−0.6, 2.1)	(−4.4, 2.8)	(−1.1, 4.2)	(−0.1, 4.1)	(0.2, 2.8)	(−1.5, 1.4)	(−1.2, 6.1)	(−0.2, 2.6)
Phase 4	3.9	3.4	3.7	6.2	3.5	6.3	5.1	7.4	3.6	4.2	0.7	4.3
	(3.1, 4.7)	(2.6, 4.3)	(2.8, 4.5)	(5.0, 7.4)	(2.6, 4.3)	(2.9, 9.7)	(2.8, 7.4)	(5.8, 9.0)	(2.8, 4.5)	(3.3, 5.0)	(−1.7, 3.2)	(3.5, 5.1)

Note: This study used medical claims for Medicare fee-for-service beneficiaries aged ≥67 years. We compared utilization during the pandemic period, defined as March 2020 to December 2021, with that in the prepandemic period, defined as March 2018 to December 2019. Phase 1 is March–May 2020, Phase 2 is June–December 2020, Phase 3 is January–June 2021, and Phase 4 is July–December 2021. The dependent variables were the natural log of the average use of healthcare services per 1,000 beneficiaries with diabetes.

ASC, ambulatory surgery center; ER, emergency room; HOP, hospital outpatient.

Table 2 also explores utilization changes by subgroup. Persons aged ≥85 years had a lower reduction in hospital outpatient visits in Phase 1 than other age groups (−21.2% vs −28% and −28.8%). However, they had the largest reduction in physician office visits among all age groups in Phases 1 and 2. In total, non-Hispanic White persons had a larger reduction in acute inpatient stays and a lower reduction in ER visits than non-Hispanic Black and Hispanic persons.

In Table 3, we quantified the changes in telehealth and in-person visits. In total, telehealth visits increased by 511.1% (95% CI=502.2%, 520.0%) during the first 2 years of the pandemic, with the largest increase observed in 2020. In the same period, in-person visits decreased by 11% (95% CI= −11.9%, −10.1%), with the largest decrease in March–May 2020. Changes in telehealth visits were not equally observed in all subgroups. In total, persons aged ≥85 years and rural residents had a lower level of increase in telehealth usage, whereas non-Hispanic Blacks had a higher level of increase in telehealth visits. The increase in telehealth visits only partially compensated for the decrease in in-person visits, especially during the initial phase of the pandemic (Figure 2). The telehealth versus in-person visits ratio peaked at 0.19 in April 2020, then gradually decreased to 0.03–0.09 afterward.

Table 3.

Percentage Changes in Telehealth and In-Person Visits During the COVID-19 Pandemic Among Persons With Diabetes

		By age group, years			By race				By sex		By urbanicity
	Overall	67–74	75–84	≥85	Non-Hispanic White	Non-Hispanic Black	Hispanic	Others	Male	Female	Rural	Urban
Telehealth
Total	511.1	507.9	512.7	480.6	511.9	533.4	465.8	482.9	514.3	503.7	395.6	589.8
	(502.2, 520.0)	(499.6, 516.2)	(503.4, 522.1)	(471.5, 489.8)	(503.2, 520.6)	(521.4, 545.4)	(454.1, 477.4)	(471.1, 494.7)	(505.4, 523.2)	(495.0, 512.4)	(388.0, 403.2)	(582.2, 597.3)
Phase 1	540.0	527.9	548.2	502.7	542.4	588.1	476.3	475.7	547.1	526.5	411.0	628.8
	(510.9, 569.1)	(500.6, 555.1)	(517.8, 578.6)	(470.3, 535.1)	(513.1, 571.6)	(554.1, 622.1)	(444.4, 508.2)	(440.9, 510.6)	(518.2, 576.0)	(497.6, 555.4)	(385.5, 436.5)	(602.8, 654.8)
Phase 2	557.4	547.1	558.9	530.0	557.2	587.0	486.5	522.7	565.6	545.3	431.8	640.9
	(542.1, 572.8)	(532.9, 561.3)	(542.5, 575.4)	(514.8, 545.2)	(542.1, 572.3)	(569.2, 604.7)	(467.2, 505.7)	(503.0, 542.4)	(549.4, 581.7)	(530.7, 559.9)	(419.2, 444.3)	(628.5, 653.2)
Phase 3	499.3	501.6	498.7	472.1	499.1	515.2	462.2	489.9	499.1	494.3	390.0	575.9
	(484.1, 514.5)	(487.4, 515.8)	(483.1, 514.4)	(457.5, 486.7)	(484.6, 513.5)	(495.2, 535.3)	(443.6, 480.9)	(469.7, 510.1)	(484.8, 513.4)	(479.3, 509.2)	(376.9, 403.1)	(563.6, 588.1)
Phase 4	455.0	459.3	457.3	426.7	457.8	468.3	440.9	438.2	454.4	454.5	351.4	526.6
	(440.5, 469.4)	(445.1, 473.6)	(442.3, 472.4)	(411.8, 441.7)	(443.7, 471.9)	(449.5, 487.1)	(420.4, 461.4)	(418.5, 457.8)	(440.7, 468.1)	(439.6, 469.4)	(339.1, 363.6)	(515.3, 537.8)
In-person
Total	−11.0	−11.4	−11.2	−9.8	−11.1	−10.1	−10.5	−12.4	−10.1	−11.8	−12.2	−11.0
	(−11.9, −10.1)	(−12.3, −10.4)	(−12.1, −10.2)	(−10.6, −8.97)	(−12.1, −10.2)	(−11.3, −8.85)	(−11.6, −9.35)	(−13.6, −11.3)	(−11.0, −9.17)	(−12.7, −10.9)	(−13.8, −10.7)	(−12.0, −10.1)
Phase 1	−40.4	−42.4	−41.6	−33.8	−41.6	−33.9	−39.7	−47.6	−39.0	−41.6	−37.9	−40.9
	(−43.6, −37.2)	(−45.8, −39.0)	(−44.9, −38.3)	(−36.6, −31.1)	(−44.9, −38.4)	(−37.5, −30.3)	(−43.4, −36.0)	(−51.7, −43.5)	(−42.1, −35.9)	(−44.9, −38.4)	(−42.2, −33.7)	(−44.1, −37.7)
Phase 2	−8.4	−8.7	−8.1	−7.7	−8.5	−9.3	−6.9	−9.5	−6.9	−9.6	−10.8	−8.5
	(−9.10, −7.63)	(−9.50, −7.93)	(−8.82, −7.28)	(−8.52, −6.97)	(−9.21, −7.73)	(−11.3, −7.28)	(−8.50, −5.30)	(−10.8, −8.20)	(−7.67, −6.18)	(−10.3, −8.78)	(−13.5, −8.11)	(−9.32, −7.72)
Phase 3	−7.2	−7.3	−7.2	−6.5	−7.0	−5.9	−6.0	−8.0	−6.5	−7.7	−8.3	−7.2
	(−8.13, −6.21)	(−8.33, −6.36)	(−8.19, −6.22)	(−7.51, −5.44)	(−8.03, −6.04)	(−7.58, −4.19)	(−7.51, −4.57)	(−9.36, −6.72)	(−7.45, −5.59)	(−8.69, −6.64)	(−11.1, −5.51)	(−8.22, −6.22)
Phase 4	−3.3	−2.9	−3.5	−3.4	−3.1	−3.3	−4.4	−2.7	−2.8	−3.7	−4.9	−2.9
	(−3.90, −2.73)	(−3.58, −2.28)	(−4.08, −2.88)	(−4.13, −2.74)	(−3.66, −2.46)	(−5.33, −1.25)	(−5.62, −3.26)	(−3.69, −1.75)	(−3.41, −2.21)	(−4.30, −3.05)	(−7.26, −2.60)	(−3.48, −2.25)

Note: This study used medical claims for Medicare fee-for-service beneficiaries aged ≥67 years. We compared utilization during the pandemic period, defined as March 2020 to December 2021, with that in the prepandemic period, defined as March 2018 to December 2019. Phase 1 is March–May 2020, Phase 2 is June–December 2020, Phase 3 is January–June 2021, and Phase 4 is July–December 2021. The dependent variables were the natural log of the average use of healthcare services per 1,000 beneficiaries with diabetes.

Figure 2.

Uses of telehealth versus in-person visits each month among Medicare fee-for-service beneficiaries with diabetes.

Note: The ratio is the number of telehealth visits to the number of in-person visits.

DISCUSSION

Using medical claims data, we examined the changes in healthcare utilization during the first 22 months of the COVID-19 pandemic among Medicare FFS beneficiaries living with diabetes. We found a reduction in the use of various types of healthcare services, especially during the initial phase of the pandemic (March–May 2020), and most of the decline persisted through 2021. Changes in utilization varied by age group, sex, race/ethnicity, and urbanicity. In addition, we observed an increase in telehealth visits during the pandemic, although the increase did not come close to compensating for the decrease in in-person visits.

Our findings are in line with those of existing studies that have also documented the sudden and large reduction in healthcare utilization during the COVID-19 pandemic in the U.S. Whaley et al. (2020) and Cantor et al. (2022) found that the use of preventive and elective care dropped after the declaration of emergency.^4,10 Chatterji and Li (2021) and Ziedan et al. (2020) estimated a 67% decrease in outpatient service use among the general population and a 40% decrease among persons with diabetes during the early stage of the pandemic (March and April of 2020).^3,18 Other studies also found a reduction in ER and various types of outpatient visits in the first few months of the pandemic among all Medicare FFS beneficiaries and among persons with diabetes.19, 20, 21, 22 We found that the level of utilization was still lower than that in prepandemic years for most types of services, even 2 years after the outbreak.

The reduction in healthcare utilization could have profound implications for healthcare systems. Furlough, quarantine, and even the adoption of telehealth have reduced clinical and nonclinical personnel in the healthcare industry.²³ Deferred or canceled care during the pandemic reduced a significant share of the revenue for many hospitals and outpatient facilities,²⁴ thus further limiting the ability to employ and retain healthcare personnel. As a result, it could be challenging to cope with the needs of patients as the pandemic continues.

A survey conducted by the WHO in May 2020 reported partial or complete disruption of services for the diagnosis and treatment of diabetes and diabetes-related complications for half of the 163 countries surveyed.²⁵ The survey also found that one third of countries did not include diabetes in their emergency preparedness plans. The fact that people living with diabetes have been disproportionately affected by the COVID-19 pandemic highlights the opportunity for including diabetes in the pandemic and other emergency preparedness and response.26, 27, 28 In developed countries, telehealth for people living with diabetes is a promising strategy. People who are older, are Hispanic, and live in rural areas may need assistance to overcome barriers to telehealth utilization, including unawareness, low healthcare literacy, insufficient digital resources, and language barriers.29, 30, 31

Inadequate healthcare utilization can adversely impact persons with diabetes because managing their condition and related complications depends on an array of services, including health professionals, medicines and technology devices, and psychosocial support. Limited or no access to any of these components may hinder the management of blood glucose, blood pressure, and lipid levels, which in turn could increase the risk of diabetes complications over time. In addition, psychological stress due to a change in care routine or financial uncertainty, along with reduced opportunities for physical activity, weight gain, and insufficient vitamin D, can further deteriorate health status.³²

Some studies found mixed results on diabetes management and complications during the initial phase of the pandemic. One study reported a substantial decline in insulin prescription fills in the U.S. during the pandemic.³³ Others found that glucose control was worse during the initial stage of the pandemic,³⁴ whereas still others found no significant changes in glucose level.³⁵ A study using hospitalized patients with diabetes in England found a significant reduction in major and minor amputations,³⁶ and another study in Canada found no association between the pandemic and limb loss for persons with diabetes.³⁷ However, these findings were based on data within the first few months of the pandemic, and many of them used small-sized and area-specific samples. As we transition from the pandemic phase to the postpandemic phase of COVID-19, we can continue to study the longer-term impact of the pandemic on health outcomes using large-scale population denominators.

Limitations

This study has several limitations. First, to monitor utilization changes over a longer period, we used medical claims records through December 2021. However, by the time of our analysis (August 2022) <5% of 2021 claims were not yet finalized. We expect a reasonably small discrepancy from analysis using the fully matured data set. Second, our study is a descriptive analysis documenting utilization changes; thus, a causal impact of the pandemic and the impact of state-level shelter-in-place policies on utilization cannot be claimed. Third, our study sample only included Medicare FFS beneficiaries aged ≥67 years with diabetes; the extent to which the findings may apply to other age groups, those with other health insurance options, and people with other chronic conditions is not known.

CONCLUSIONS

This study quantifies changes in healthcare utilization among Medicare FFS beneficiaries with diabetes in the first 2 years of the COVID-19 pandemic. Understanding these patterns may help to optimize the use of healthcare resources for diabetes management in the postpandemic era and during future emergencies.