The Impact of Telehealth Implementation on Underserved Populations and No-Show Rates by Medical Specialty During the COVID-19 Pandemic

Abstract

A unique and sudden need for virtual medical visits created by the coronavirus disease 2019 (COVID-19) pandemic has led to an unprecedented expansion of telemedicine across nearly all medical specialties in the United States. In addition to providing essential medical services during the pandemic, telemedicine has the potential to expand health care access to underserved populations by eliminating traditional barriers to care such as transportation needs, distance from specialty providers, and approved time off from work. However, the literature regarding telehealth accessibility for low-income, non-English-speaking, and minority patients remains limited. Through a cross-sectional analysis comparing 2019 clinic visits with 2020 telehealth visits at the UMass Memorial Medical Center, we demonstrate specialty-specific changes in patient demographics, including a younger population, fewer non-English-speaking patients, and a relative preservation of minority, Medicaid, and Medicare patients among telehealth visits in comparison to clinic visits. We also demonstrate that nonsurgical specialties had significantly lower no-show rates and the greatest number of telehealth visits. Overall, our findings highlight the potential shortcomings of telemedicine in servicing non-English-speaking patients, while maintaining that it is an important tool with the potential to improve access to health care, particularly in nonprocedural specialties.

Introduction

Telemedicine has been a rapidly growing and continuously evolving method of health care delivery for more than two decades.¹ Although telemedicine has been implemented by a variety of specialties, particularly during the coronavirus disease 2019 (COVID-19) pandemic, it is believed that specialties best suited for this platform are those that involve either chronic disease management (e.g., psychiatry, primary care) or a significant visual component (e.g., dermatology, radiology).^2–4 Until recently, a number of barriers have limited the widespread implementation of telemedicine services. One of the greatest barriers has been a restriction on insurance coverage for televisits, with per-plan restrictions on qualifying specialties and services.²

The COVID-19 pandemic brought with it sweeping regulatory reform and legislative mandates ensuring broad coverage for telehealth services in all specialties across the country over a relatively short period.⁵ The dramatic increase in telemedicine during this time has provided a unique opportunity to gain insight into the utility of televisits across all specialties, as well as its potential to increase health care access for underserved populations. Telehealth has been associated with significant reductions in no-show rates when used in dermatology, suggesting that telemedicine may address traditional barriers to and disparities in access to care while also improving efficiency.^6,7 Limited data are available evaluating the impact of transitioning care to telehealth amid the COVID-19 pandemic across specialties. Our study compares data from 49 specialties to highlight differences in completed visits (Fig. 1), no-show rates (Fig. 2), and patient demographics (Table 1) between pre-pandemic clinic visits and mid-pandemic televisits.

Fig. 1.

Line graph depicts telemedicine adoption rate (2020 Televisits as a percentage of 2019 clinic visits). Bar graph depicts the total number of completed clinic and televisits per specialty.

Fig. 2.

No-show rates for 2019 clinic and 2020 televisits by specialty type. **p < 0.05 ***p < 0.01.

Table 1.

Patient Demographics and No-Show Rates in 2019 Clinic Visits and 2020 Televisits by Specialty Category

	PRIMARY CARE			ADULT NONSURGICAL			PEDIATRIC NONSURGICAL			ADULT SURGICAL			PEDIATRIC SURGICAL
PATIENT DEMOGRAPHICS	CLINIC	TELEVISIT	p	CLINIC	TELEVISIT	p	CLINIC	TELEVISIT	p	CLINIC	TELEVISIT	p	CLINIC	TELEVISIT	p
Mean age (SD)	44 (26)	48 (24)	<0.001	57 (21)	55 (21)	<0.001	13 (8)	12 (8)	<0.001	55 (20)	52 (19)	<0.001	12 (12)	8 (8)	<0.001
Gender
Female	58.1% (10,999/18,917)	63.2% (5,781/9,141)	<0.001	55.0% (19,395/35,234)	56.9% (11,231/19,740)	<0.001	48.2% (1,438/2,983)	49.1% (979/1,993)	1.00	58.9% (25,676/43,565)	59.3% (5,495/9,269)	1.00	44.0% (1,112/2,526)	43.0% (166/386)	1.00
Male	41.9% (7,918/18,917)	36.8% (3,360/9,141)		45.0% (15,839/35,234)	43.1% (8,509/19,740)		51.8% (1,545/2,983)	50.9% (1,014/1,993)		41.1% (17,889/43,565)	40.7% (3,774/9,269)		56.0% (1,414/2,526)	57.0% (220/386)
Primary language
Non-English	11.2% (2,123/18,916)	8.7% (797/8,343)	<0.001	10.8% (3,791/35,218)	9.1% (1,799/19,734)	<0.001	11.1% (332/2,983)	11.3% (225/1,992)	1.00	12.1% (5,291/43,552)	9.1% (840/9,265)	<0.001	8.6% (216/2,524)	4.2% (16/385)	0.011
English	88.8% (16,793/18,916)	91.3% (8,343/9,140)		89.2% (31,427/35,218)	90.9% (17,935/19,734)		88.9% (2,651/2,983)	88.7% (1,767/1,992)		87.9% (38,261/43,552)	90.9% (8,425/9,265)		91.4% (2,308/2,524)	95.8% (369/385)
Race/Ethnicity
Asian	5.1% (970/18,917)	3.1% (284/9,142)	<0.001	2.9% (1,022/35,233)	2.8% (553/19,740)	<0.001	3.9% (117/2,983)	3.1% (62/1,993)	0.886	2.9% (1,273/43,565)	2.5% (231/9,269)	<0.001	2.5% (64/2,526)	2.3% (9/386)	1.00
Black or African American	10.0% (1,897/18,917)	8.4% (763/9,142)		5.1% (1,791/35,233)	5.1% (1,009/19,740)		9.1% (270/2,983)	7.7% (153/1,993)		5.9% (2,565/43,565)	6.9% (640/9,269)		6.5% (163/2,526)	9.1% (35/386)
LatinX	19.3% (3,651/18,917)	16.7% (1,527/9,142)		12.1% (4,266/35,233)	12.4% (2,453/19,740)		21.2% (633/2,983)	23.4% (466/1,993)		15.1% (6,581/43,565)	15.4% (1,426/9,269)		19.2% (486/2,526)	18.7% (72/386)
Other	4.2% (801/18,917)	3.6% (330/9,142)		2.9% (1,008/35,233)	2.8% (562/19,740)		3.9% (117/2,983)	4.2% (84/1,993)		3.2% (1,399/43,565)	2.6% (237/9,269)		4.6% (116/2,526)	4.1% (16/386)
Unknown	0.9% (165/18,917)	0.7% (63/9,142)		1.4% (485/35,233)	0.9% (168/19,740)		2.0% (61/2,983)	2.2% (43/1,993)		1.2% (540/43,565)	0.6% (53/9,269)		3.6% (91/2,526)	4.9% (19/386)
White	60.4% (11,433/18,917)	67.6% (6,175/9,142)		75.7% (26,661/35,233)	76.0% (14,995/19,740)		59.8% (1,785/2,983)	59.5% (1,185/1,993)		71.6% (31,207/43,565)	72.1% (6,682/9,269)		63.6% (1,606/2,526)	60.9% (235/386)
Insurance payor
Medicaid	30.8% (5,610/18,202)	27.2% (2,360/8,671)	<0.001	18.0% (6,031/33,454)	18.5% (3,449/18,691)	<0.001	50.2% (1,454/2,895)	48.9% (943/1,930)	1.00	22.6% (9,098/40,330)	21.8% (1,875/8,617)	0.459	43.9% (1,069/2,436)	49.3% (176/380)	0.052
Medicare	20.5% (3,729/18,202)	25.5% (2,207/8,671)		32.1% (10,723/33,454)	30.8% (5,755/18,691)		1.0% (28/2,895)	1.3% (25/1,930)		26.7% (10,775/40,330)	26.7% (2,298/8,617)		1.5% (37/2,436)	0.3% (1/380)
Private	48.7% (8,863/18,202)	47.3% (4,104/8,671)		50.8% (16,700/33,454)	50.8% (9,487/18,691)		47.1% (1,413/2,895)	49.8% (962/1,930)		50.7% (20,457/40,330)	51.6% (4,444/8,617)		54.6% (1,330/2,436)	53.4% (203/380)
No-show rate	12.4% (2,675/21,592)	11.2% (1,152/10,294)	0.002	12.9% (5,230/40,464)	10.5% (2,304/22,044)	<0.001	15.9% (564/3,547)	13.7% (316/2,309)	0.102	10.4% (5,035/48,600)	13.1% (1,403/10,672)	<0.001	11.0% (312/2,838)	15.5% (71/457)	0.037

Bold values indicate statistically significant p values (<0.05).

SD, standard deviation.

Study Data and Methods

We conducted a cross-sectional analysis of ambulatory clinic and telehealth appointments across three UMass Memorial Health Center campuses during the months of May and June in both 2019 and 2020. The institutional review board of the University of Massachusetts designated this study IRB-exempt as a quality improvement initiative. De-identified patient and appointment data were compiled by the UMass Data Analytics Core using the institution's data lake, which sources data from the electronic medical record (Epic) and hospital billing office.

Data from 2019 were used to analyze baseline trends in clinic visits, with a sample size of just over 1,100,000 appointments. Data from 2020 were used to analyze trends in televisits in the setting of the COVID-19 pandemic, with a sample size of nearly 46,000 appointments. All televisits were conducted as live appointments between the physician and patient by using Doximity or AmWell technology. Completed and no-show appointments were included, whereas cancelled or rescheduled appointments were excluded. To limit the analysis to appointments that could potentially be conducted virtually, all procedural appointments were excluded.

The data included appointments from 49 medical specialties. To summarize the data by common specialty practices, each specialty was grouped into one of the following five categories for data analysis: Primary Care, Adult Non-Surgical, Pediatric Non-Surgical, Adult Surgical, and Pediatric Surgical (Table 2). Specialty-specific adoption of telehealth during the COVID-19 pandemic was measured by dividing the number of completed 2020 televisits by the number of completed 2019 nonprocedural clinic visits (Fig. 1). No-show rates were calculated as the number of no-show appointments divided by the total number of appointments (Fig. 2). To identify differences in sample characteristics between clinic visits and televisits, patient demographics, including age, race and ethnicity, sex, insurance payor, and primary language, were analyzed by using data from completed visits only. Statistical analysis was performed with Fisher's exact test, and two-tailed p-values <0.05 were considered statistically significant. The Bonferroni method was applied to correct p-values.

Table 2.

Total Number of Completed 2019 Clinic Visits and 2020 Televisits in 49 Medical Specialties, Grouped into 5 Specialty Groups by Common Practice

SPECIALTY	APPOINTMENT TYPE
	2019 CLINIC	2020 TELEVISIT
Primary care	18,917	9,142
Adolescent medicine	630	488
Family medicine	5,371	3,047
Geriatrics	537	358
Internal medicine	8,103	4,202
Pediatrics	4,276	1,047
Adult nonsurgical	35,234	19,740
Anticoagulation	330	89
Cardiology	5,487	3,858
Dermatology	4,899	1,560
Diabetes services	3,306	2,661
Endocrinology	2,387	1,378
Gastroenterology	2,465	1,593
Infectious diseases	1,168	525
Muscular dystrophy	261	70
Nephrology	778	368
Neurology	2,198	1,946
Nutrition	801	454
Oncology	4,646	1,932
Psychiatry	990	658
Pulmonary	3,013	1,093
Reproductive endocrinology and infertility	185	118
Rheumatology	2,320	1,437
Pediatric non-surgical	2,983	1,993
Neonatology	124	33
Pediatric cardiology	330	197
Pediatric endocrinology	992	675
Pediatric genetics	243	145
Pediatric hematology and oncology	278	68
Pediatric nephrology	224	137
Pediatric neurology	193	254
Pediatric pulmonology	599	484
Adult surgical	43,565	9,269
Bariatrics	1,283	1,399
Cardiothoracic surgery	252	64
General surgery	1,855	365
Gynecologic oncology	1,003	173
Neurosurgery	478	236
Obstetrics and gynecology	4,280	639
Ophthalmology	3,647	641
Orthopedic surgery	15,079	2,110
Otolaryngology	3,060	595
Plastic surgery	1,672	150
Radiation oncology	3,137	222
Transplant	2,341	976
Urogynecology	813	173
Urology	2,718	1,112
Vascular surgery	1,211	294
Wound care	736	120
Pediatric surgical	2,526	386
Pediatric neurosurgery	96	50
Pediatric orthopedics	1,585	73
Pediatric surgery	365	155
Pediatric urology	480	108
Grand total	103,225	40,530

Study Results

Telehealth Implementation During the COVID-19 Pandemic

Between May and June of 2019, 117,697 outpatient nonprocedural visits were completed across the three largest UMass Memorial Medical Center campuses. During the same months in 2020, 45,933 telehealth visits were completed, accounting for 39% of the 2019 patient load (Table 2). The degree of telehealth implementation during COVID-19 varied by specialty. Telehealth visits in primary care, adult nonsurgical, and pediatric nonsurgical specialties accounted for 48.3%, 56.0%, and 66.8% of their respective 2019 clinic visits. In contrast, pediatric surgical specialties and adult surgical specialties saw only 15.3% and 21.3%, respectively (Fig. 1). Adult nonsurgical specialties saw the greatest number of televisits, with 19,740 completed visits (Fig. 1).

Patient Demographics

When comparing 2019 clinic visits with 2020 televisits, at least one statistically significant difference in patient demographics (e.g., gender, age, race/ethnicity, insurance payor) was observed for each specialty group except the nonsurgical pediatric specialties (Table 1). The percentage of female patients in the televisit group increased across all specialties except pediatric surgery, with statistically significant increases seen in primary care, adult nonsurgical, and adult surgical specialties. The percentage of patients whose primary language was not English decreased significantly with televisits across all specialties, except the pediatric nonsurgical group. The mean age of all patients seen in 2019 clinics was not significantly different from those seen in 2020 televisits (51 [standard deviation, SD = 24], 51 [SD = 23], p = 1.00). However, significant differences in mean age between clinic visits and televisits were noted within specialty subgroups. Mean patient age increased significantly with televisits conducted by the primary care group, but it decreased significantly with televisits conducted by the other four specialty subgroups.

All primary care and adult specialties were found to have statistically significant differences in the racial and ethnic composition of patients seen in clinic versus telehealth. Specifically, the percentage of white patients increased with televisits by 0.3–7.2%, with the largest change seen in the primary care group. Corresponding decreases were observed across all other racial and ethnic groups. In contrast, patient race or ethnicity in pediatric specialties demonstrated the opposite trend, but these results were not statistically significant. Analysis of individual specialties revealed significant decreases in white patients seen via telehealth in dermatology, muscular dystrophy, and otolaryngology (73.2% vs. 68.0%, p = 0.005, 55.6% vs. 54.3%, p = 0.005, 70.5% vs. 65.6%, p = 0.005). The racial and ethnic composition of all 2019 clinic patients was similar to that of the hospital system's catchment area, with minority patients making up a slightly greater percentage of the patient population than would be expected based on the catchment area population (Table 3). This finding is preserved in the 2020 televisit population, despite a small decrease in the overall percentage of minority patients seen via televisits as compared with clinic visits (Table 3).

Table 3.

Race and Ethnicity of the 2019 Clinic Population, 2020 Televisit Population, and Hospital System Catchment Area Population

RACE/ETHNICITY	CLINIC	TELEVISIT	CATCHMENT AREAa
Asian	3.34%	2.81%	5.40%
Black or African American	6.48%	6.42%	6.10%
LatinX	15.13%	14.67%	12.20%
White	70.42%	72.22%	75.70%

The “Other” and “Unknown” categories are excluded from this table, as there are no comparable data available for the catchment area.

^a2019 demographic data for Worcester County.¹²

Insurance payor breakdown for televisits was significantly different from clinic visits in the primary care and adult nonsurgical groups. In primary care specialties, there was a 5.0% increase in Medicare patients in the televisit group, whereas percentages of Medicaid and private insurance enrollees dropped by 3.6% and 1.4%, respectively. Opposite findings were observed in adult nonsurgical televisits, with a 1.3% decrease in Medicare patients and 0.5% increase in Medicaid enrollees. No significant differences in payor breakdown were observed among the other three specialty groups.

No-Show Rates

The no-show rate for all 2019 clinics was 11.8%, whereas the no-show rate for all 2020 televisits was 11.5% (p > 0.05). When separated by specialty type, primary care and adult nonsurgical specialties demonstrated significant reductions in no-shows with 2020 televisits as compared with 2019 clinic visits (12.4% vs. 11.2%, p < 0.001, 12.9% vs. 10.5%, p < 0.001). Similar trends were seen in pediatric nonsurgical televisits but were not statistically significant. In contrast, statistically significant increases in no-show rates were observed in adult and pediatric surgical specialty 2020 televisits as compared with 2019 clinic visits (Fig. 2).

Discussion

Fewer non-English-speaking patients were seen via telehealth compared with in-person clinic visits across all specialties, possibly reflecting variable patient comfort with virtual visits depending on language fluency. In addition, it is possible that limited availability of telephone interpreters and technological challenges with incorporating interpreters into virtual visits contributed to these observed decreases. However, there was an appreciable number of non-English-speaking patients seen via televisits, indicating that telehealth could still be a valuable resource for these patients. As telehealth continues to be used during the ongoing COVID-19 pandemic, it is vital that we recognize this platform's shortcomings for our interpreter-dependent patients. Focus should be placed on incorporating high-quality interpreter services into telemedicine platforms to ensure access for patients who require these services.

The mean age of telehealth users was less than that of 2019 clinic patients in every specialty group, except primary care. A significant criticism of telemedicine is that it may not be easily accessible to elderly, less technologically savvy, patients.^8–10 Interestingly, the mean age of primary care patients actually increased with televisits, indicating that difficulty with technology may not be driving the decreases observed in other specialties. Further investigations into patient comfort with and access to technology needed for virtual visits is an important next step in ensuring equal telehealth access.

Our analysis of televisits demonstrated small but significant increases in the proportion of white patients across all adult specialties, with corresponding decreases in minority patients, suggesting a potential inequity in telehealth access. Of note, pediatric specialties did not follow this trend and pediatric primary care, along with several individual specialties (dermatology, otolaryngology, and muscular dystrophy), actually demonstrated significant increases in minority patients with televisits. Moreover, we observed that the UMass Memorial Medical Center served a greater proportion of minority patients in 2019 than would be expected based on the population demographics of the hospital's catchment area. These findings were preserved in the 2020 televisit population, providing further evidence that virtual visits are an accessible and important resource for our minority patients. There are many confounding factors likely impacting the observed differences in patient demographics between clinic visits and televisits. Most notably, minority patients have made up a substantial portion of essential workers during the COVID-19 pandemic, meaning that they likely faced disproportionately greater time constraints and barriers to telehealth care during May and June of 2020.¹¹ It is important to note that the observed changes in patient race and ethnicity with televisits were small and variable across specialties. As telemedicine inevitably becomes a permanent fixture in health care, it is vital that we continue to collect and analyze patient demographic data to identify potential inequities and build a virtual health care system that expands care equitably to all populations.

Our results suggest that televisits are an important patient resource and means for improving access to care for nonprocedural specialties, given the high adoption rates and overall reduction in no-show rates. In contrast, surgical specialties had lower adoption rates and increased no-show rates with televisits, which possibly reflects patient and provider perceptions regarding the limited utility of virtual visits for surgical problems. Interestingly, the televisit adoption rate was greater in nonsurgical specialties compared with primary care. Although certain routine primary care visit types (e.g., medication follow-ups) are well suited for a virtual clinic, annual health maintenance visits often require physical exams and cannot be completed virtually. In addition, we must consider the unique and unprecedented time during which recent televisits have taken place. Many primary care providers were deployed to other hospital services during May and June, impacting their clinic schedules. External stressors due to the pandemic, undoubtedly, impacted patient behaviors regarding their health care. It is possible that the higher telehealth adoption rates observed in nonsurgical specialties is due to the fact that specialty visits may be viewed as more essential to patients than routine health maintenance visits with primary care providers.

Limitations to this study include its single-institution design with inevitable variation in telehealth implementation between specialties. As previously mentioned, the unique climate created by the COVID-19 pandemic, undoubtedly, influenced provider availability and patient behavior regarding seeking medical care, making it difficult to draw definitive conclusions from the data comparing 2019 clinic visits with 2020 televisits.

Conclusions

The implementation of telehealth services during the COVID-19 pandemic resulted in specialty-specific changes to patient populations, with many specialties seeing slightly younger patients and a greater number of patients whose primary language was English. Adult specialties saw small increases in the proportion of white patients with televisits, whereas primary care providers saw an increase in Medicaid enrollees and adult nonsurgical providers saw an increase in Medicare patients. In light of the extraordinary circumstances surrounding telehealth implementation, we view the relative preservation of minority, Medicaid, and Medicare patients in telehealth visits as a promising sign that telehealth has the potential to improve access to care for all patients. Moreover, analysis of no-show rates demonstrated that nonsurgical specialties largely benefited from televisits, whereas surgical specialties did not. Additional work must be done to ensure that ongoing telehealth expansion is done in a way that works to preserve equity and access for patients from all backgrounds.

Authors' Contributions

Every author has contributed significantly to the work, including the conception, design, analysis, drafting, revising, and providing final approval.

Disclosure Statement

No competing financial interests exist.

Funding Information

This work was conducted with support from Harvard Catalyst | The Harvard Clinical and Translational Science Center (National Center for Advancing Translational Sciences, National Institutes of Health Award UL 1TR002541) and financial contributions from Harvard University and its affiliated academic health care centers. The content is solely the responsibility of the authors and does not necessarily represent the official views of Harvard Catalyst, Harvard University and its affiliated academic health care centers, or the National Institutes of Health.