Social vulnerability, lower broadband internet access, and rurality associated with lower telemedicine use in U.S. Counties

Abstract

Objective

Our objective was to determine how social vulnerabilities, broadband access, and rurality relate to telemedicine use across the United States through large-scale analysis of real-world telemedicine data.

Materials and Methods

We conducted a retrospective, observational study of dyadic U.S. telemedicine sessions that occurred January 1, 2022 to December 31, 2022, linked to the 2020 Centers for Disease Control and Prevention Social Vulnerability Index (SVI) and the National Center for Health Statistics Urban-Rural Classification Scheme for Counties. We examined county-level telemedicine use rates (sessions per 1000 population) in relation to SVI indexes, broadband internet access, and rurality classifications using polynomial regression and data visualization.

Results

We found a negative, nonlinear association between overall social and socioeconomic status vulnerabilities and telemedicine use. Telemedicine rates in urban counties exceeded that of rural counties. There was more variability in telemedicine use for the urban counties according to social vulnerability and broadband access.

Discussion

Rurality and broadband access demonstrated a greater effect on telemedicine use than social vulnerability, and the relationship between social vulnerability, broadband access, and telemedicine use differed for rural versus urban areas.

Conclusion

This observational study of nearly 8 million U.S. telemedicine sessions showed that rurality and broadband access are key drivers of telemedicine use and may be more important than many social vulnerabilities in determining community-level telemedicine use. We also found nuanced differences in the relationship between social vulnerability and telemedicine use between rural and urban counties, and at different levels of broadband access.

Background and significance

Telemedicine addresses critical barriers to healthcare access, including lack of local access to specialty care providers, transportation, and childcare.^1–6 However, current evidence suggests that access to telemedicine since the onset of the COVID-19 pandemic has not been equitable in the United States.^7–9 Multiple studies of telemedicine utilization have indicated patients who are Asian, Black, Hispanic or Latino/a, non-English speaking, lower income, female, older, or Medicaid recipients were less likely to use telemedicine.^10–15 This is concerning given people of color experience substantial health disparities and the enormous potential to address existing and emerging health disparities through telemedicine. Existing evidence for telemedicine use disparities is largely anecdotal or based on local or regional studies, and we currently lack a comprehensive understanding of how social vulnerability affects telemedicine use in the United States.

Studies of telemedicine in the COVID-19 pandemic era indicate that older adults, particularly those with chronic conditions and disabilities, are less likely to use telemedicine.^10–16 Contributing factors identified in the literature include a lack of social support, unfamiliarity with technology, less computer ownership, and internet access.^11–14 Additionally, many older adults have reported not wanting to use telemedicine for a first visit to a healthcare provider or new health problems.¹¹ Racial disparities have been observed for telemedicine use, specifically disadvantaging persons of Black, Hispanic or Latino/a, and Asian race.^11–13,16 Among these sub-populations, contributing factors include poverty, financial hardships, disproportionate prevalence of chronic diseases compared to non-Hispanic White individuals, and limited access to education and economic opportunities.¹⁷ The intersection of socioeconomic status and demographic factors contributes to telemedicine inequity. Although 90% of adults in the United States use the internet, only 65% of Latino and 71% of Black adults have home broadband access compared to 80% of white adults.¹³ As a result, Black and Hispanic or Latino/a adults were more likely to rely on mobile devices to complete their telemedicine visits which may be insufficient.^11,13

In healthcare, a lack of digital literacy and access means that patients cannot benefit from innovations including patient portals, mobile health applications, wearable sensors, and telehealth.¹ Digital access is often a matter of cost: whether a patient can afford to purchase devices, such as smartphones and personal computers, and whether they can afford to pay for continued and periodic broadband service. It is also a matter of location, as households in rural areas are less likely to have good connectivity and access to broadband service, and when they do, the costs are often higher than in urban areas because it is expensive to build networks in rural areas.¹⁸ A substantial proportion of U.S. households lack broadband access and there is a clear digital divide in broadband access, according to age, ethnicity, household income, and geographical location. People living in rural areas experience health disparities and are likely to benefit from telemedicine services but often lack digital infrastructure and may contend with poor broadband and cellular connectivity.^19,20 Certainly, persons living in rural areas have historically experienced barriers to telemedicine that include broadband access, technology cost, and insufficient access to technology.^11–13,16 Those living in rural areas tend to be older, of a lower socioeconomic class and less formally educated.^11,21 These factors lead to technical issues such as limited wireless internet service and an inability to use technology.^12,21

Objective

While there is growing evidence that access to telemedicine and digital health may not be equitable, most studies that examine these inequities are limited to specific programs or sites in limited geographic areas. To date, we lack a clear understanding of how varied dimensions of social vulnerability affect telemedicine use, and how these effects may vary across different communities and geographical areas of the United States. Here, to gain insight into national patterns in the United States, we analyzed telemedicine use rates in relation to social vulnerability for a large observational sample of telemedicine session data provided by a commercial telemedicine company, Doxy.me, Inc. These data are national in scope, agnostic to setting, payment system, or healthcare provider type and represent approximately 8%-30% of U.S. telemedicine sessions, enabling insight into national telemedicine use patterns.²² Our objective was to determine if U.S. telemedicine use decreases as community social vulnerability increases and to assess the role of rurality in the relationship between social vulnerability and telemedicine use.

Materials and methods

We conducted a retrospective, observational study of dyadic telemedicine sessions that occurred in the United States between January 1, 2022 to December 31, 2022, linked to the 2020 Centers for Disease Control & Prevention (CDC) Social Vulnerability Index (SVI) and the National Center for Health Statistics (NCHS) Urban-Rural Classification Scheme for Counties.^23,24 We assigned geoidentifiers for client location in each session and examined the density of county-level telemedicine use (sessions per 1000 population) in relation to percentile rankings for social vulnerability, classifications of rurality, and broadband internet access using methods of polynomial regression and data visualization.

Ethical review and approval

The dataset was determined as HIPAA de-identified using the expert determination method. The University of Utah Institutional Review Board reviewed this study and determined it to be non-human subjects research.

Data

Doxy.me, Inc. provisioned a dataset for a random sample of 8 000 000 telemedicine sessions that occurred on its commercial platform between January 1, 2022 and December 31, 2022, during the COVID-19 public health emergency. Doxy.me is a HIPAA-compliant, commercial telemedicine platform in common use across the United States by diverse healthcare organizations, clinics, and independent healthcare providers to conduct synchronous, web-based virtual visits. Estimates of the doxy.me platform’s use in U.S. telemedicine vary from 8% to 30% of all U.S. telemedicine sessions and reflect diverse healthcare settings, locations, provider types, and specialties. The regional distribution of doxy.me telemedicine sessions approximates the observed geographic distribution of national Medicare claims for telemedicine services.²² However, HIPAA-compliant and privacy-preserving features of the platform preclude robust comparisons of patient demographics, services provided, or provider characteristics at this time.

Client location

Doxy.me, Inc. provided geospatial coordinates indicating the approximate location of client and provider participants in each session. This geolocation was based upon each participant’s I.P. address, using a free I.P. geolocation service, ipstack.²⁵ We assigned geoidentifiers to client location using ArcGIS Pro 3.1.0.²⁶ We derived the U.S. county in which clients were located from the geoidentifer.

Social vulnerability

We assigned community social vulnerability indicators based upon the county in which clients were located, using the 2020 CDC SVI.²³ Here, with the objective of examining overall patterns and relationships, we analyzed the percentile rankings of counties according to overall social vulnerability and 4 themes: Socioeconomic Status, Household Characteristics, Racial and Ethnic Minority Status, and Housing Type and Transportation. The measurement of these indicators is detailed in documentation published by the CDC and summarized in Table 1.

Table 1.

Overview of CDC Social Vulnerability Index.²³

Index a	Variable name	Constituent measurements
Overall Social Vulnerability	RPL_THEMES	All indicators listed below.
Socioeconomic Status	RPL_THEME1	Below 150% poverty Unemployed Housing cost burden No high school diploma No health insurance
Household Characteristics	RPL_THEME2	Aged 65 and older Aged 17 and younger Civilian with a disability Single-parent households English language proficiency
Racial and Ethnic Minority Status	RPL_THEME3	Hispanic or Latino (of any race) Black or African Americanb Asianb American Indian or Alaska Nativeb Native Hawaiian or Pacific Islanderb 2 or more racesb Other racesb
Housing Type and Transportation	RPL_THEME4	Multi-unit structures Mobile homes Crowding No vehicle Group quarters

^aPercentile ranking of vulnerability, count-level.

^bNot Hispanic or Latino.

Rurality

We determined rurality of U.S. counties using the National Center for Health Statistics (NCHS) Urban-Rural Classification Scheme for Counties, 2013, classifying counties belonging to metropolitan areas as urban, and those belonging to non-metropolitan areas as rural.²⁴

Broadband internet access

The 2020 CDC SVI report also provided county level estimates of percentage of households without a computer with broadband internet access. The distribution of percentages was normal so we categorized into evenly distributed terciles: High, Medium, and Low.

Telemedicine use

We limited this analysis to dyadic telemedicine sessions conducted by a single provider with a single client. We defined a dyadic telemedicine session as any session between one patient and one provider (both within the United States with a geoidentifiable state and county), with a duration between 5 and 120 minutes, as previously described in Cummins et al.²² Telemedicine sessions were aggregated to the county level. County level SVI, and RUCA information were merged. We determined county population using the 2020 CDC SVI. We normalized telemedicine use by county population, analyzing the number of sessions per 1000 population.

Analysis

We performed descriptive statistics, data visualization, and polynomial regression analysis using R version 4.3.1, packages: stats (lm function) and ggplot2. Initial plots showed a nonlinear relationship between telemedicine sessions and social vulnerability indexes. To capture the nonlinear relationship and to keep model parsimony without overfitting we utilized a 2nd-degree polynomial function as our model. To mitigate multicollinearity, we centered the data on the 50th percentile of each SVI index. We applied Lind and Mehlum’s procedure to ensure quadratic relationship using Stata version 17.0 and utest testing.²⁷ We added the 3-level factor of Broadband Access and then the Urban/Rural factor with interaction term. Significance testing was done with alpha < 0.05, B-coefficients, and confidence intervals are reported.

Results

We found that 7 334 130 dyadic telemedicine sessions met the inclusion and exclusion criteria; these sessions occurred in the United States between January 1 and December 31, 2022. 93% of the sessions (6 841 221) occurred in metropolitan counties (Urban) and 7% of the sessions (492 909) occurred in non-metropolitan counties (Rural), by client location. Of the 3143 counties defined by NCHS in 2013, our dataset represents 3110 counties (99%), with 1163 being Urban and 1947 being Rural counties. High broadband internet counties numbered 1041, Medium broadband counties = 1038, and Low broadband counties = 1031.

Social vulnerability, broadband access, and telemedicine use

The results of polynomial regression analyses of telemedicine use per 1000 residents across the 5 social vulnerability indexes (described in Table 1) while controlling for broadband internet access, are presented in Table 2. Model goodness of fit, represented by R², was statistically significant for all SVI indexes. The SVIs of Overall Vulnerability and Socioeconomic Status showed significant B-coefficients for curvature indicating nonlinear effects. Lind and Mehlum procedures were done for these 2 indexes to verify the nonlinear relationship. Both indexes passed Lind and Mehlum criteria, having significant steep slopes at both ends of the data range and the vertex exist within the range. B-coefficients for Overall Vulnerability indicated a nonlinear association (Curvature = −13.70, P < .001) of telemedicine use with Overall Vulnerability Index. The negative curvature indicates an inverted or upside-down U-shape. Counties with a low Overall Vulnerability Index score tend not to use as much telemedicine. The same goes for counties with a high Overall Vulnerability Index score. The counties in the middle of the Overall Vulnerability Index tend to use telemedicine more. Counties classified as Medium broadband access made 7.18, P < .001 less telemedicine sessions per 1000 residents than High broadband counties; counties classified as Low made 9.17, P < .001 less telemedicine per 1000 residents than High broadband counties. This finding for broadband access is similar for all other SVI indexes; where counties High in broadband access used more telemedicine sessions. Figure 1 shows graphs of each SVI Index plotted against telemedicine sessions per 1000 residents with the levels of Broadband Access classification (High, Medium, and Low) as a stratification factor. Results for the SVI of Socioeconomic Status were similar those of the Overall Vulnerability Index. The SVI of Household Characteristics showed negative slope (Slope = −5.25, P < .001) but no curvature (Curvature = 4.08, P = .16). The negative slope indicates that counties with a low Household Characteristic index score utilized telemedicine more than counties with a high Household Characteristic index score. The SVI of Racial and Ethnic Minority showed a positive slope (Slope = 3.67, P < .001) with no curvature (Curvature = −3.51, P < .244). The positive slope indicates that as the percentage of racial and ethnic minorities increases among the counties, the use of telemedicine also increases. The SVI of Housing Type and Transportation had similar results; a positive slope with no curvature (Slope = 4.14, P > .001 and Curvature = −4.94, P = .09). As the SVI of Housing Type and Transportation index increases telemedicine use increases.

Table 2.

Polynomial regression results of telemedicine use per 1000 with broadband internet access across social vulnerability indexes.

	R2	B	95% CI for B
			Lower bound	Upper bound
Overall Vulnerability	0.100 c
Constant		17.67 c	16.77	18.57
Slope		0.25	−1.37	1.88
Curvature		−13.70 c	−19.44	−7.96
BA: Medium		−7.18 c	−8.24	−6.13
Low		−9.17 c	−10.32	−8.03
Socioeconomic Status	0.0974 c
Constant		17.35 c	16.44	18.27
Slope		−0.30	−2.00	1.40
Curvature		−10.77 c	−16.54	−5.00
BA: Medium		−7.16 c	−8.23	−6.09
Low		−8.98 c	−10.18	−7.78
Household Characteristics	0.107 c
Constant		15.53 c	14.64	16.43
Slope		−5.25 c	−6.81	−3.69
Curvature		4.08	−1.63	9.78
BA: Medium		−6.41 c	−7.47	−5.36
Low		−7.96 c	−9.06	−6.85
Racial and Ethnic Minority	0.101 c
Constant		16.68 c	15.86	17.50
Slope		3.67 c	2.18	5.16
Curvature		−3.51	−9.43	2.40
BA: Medium		−6.75 c	−7.80	−5.69
Low		−9.18 c	−10.26	−8.09
Housing Type and Transportation	0.103 c
Constant		17.16 c	16.27	18.05
Slope		4.14 c	2.63	5.66
Curvature		−4.94	−10.67	0.80
BA: Medium		−7.25 c	−8.30	−6.21
Low		−9.79 c	−10.85	−8.72

^a P < .05,

^b P < .01,

^c P < .001.

Figure 1.

Telemedicine use per 1000 with broadband internet access across social vulnerability indexes. (A) Overall Vulnerability, (B) Socioeconomic Status, (C) Household Characteristics, (D) Racial and Ethnic Minority Status, and (E) Housing Type & Transportation.

Rurality, social vulnerability, broadband internet access, and telemedicine use

We added the factor of Urban/Rural and the interaction between Urban/Rural with Broadband Internet Access to the model. Goodness of fit, R², increased for all SVI indexes indicating improved modeling with the. The results of the polynomial regressions across social vulnerability indexes are detailed in Table 3. For the SVI of Overall Vulnerability Index, there was no slope and a significant negative curvature (Slope = −0.95, P = .247; Curvature = −12.40, P < .001). Counties with Medium and Low broadband access engaged in fewer telemedicine sessions than counties with High broadband access (B-coefficients: −2.94, P < .001; −3.85, P < .001, respectively). Urban use was higher than rural use (B-coefficient = 8.47, P < .001). The interaction between Urban/Rural and High/Medium/Low broadband access was statistically significant. Compared to the reference category (Rural counties with High broadband access) being in Urban counties with Medium broadband access was associated with a decrease in telemedicine use by 4.13 calls per 1000 residents (P < .001). The reduction in telemedicine sessions was 5.94 sessions per 1000 residents in Urban counties with Low broadband access (P < .001). Table 3 presents results for all other SVIs. Modeling showed similar results; Counties High in broadband access used telemedicine more and Urban counties used telemedicine more. The interaction between Rurality and Broadband access showed that Rural counties with High broadband access used telemedicine more than Urban counties with Medium and Low broadband access. Figure 2A plots telemedicine use per 1000 residents with the SVI Overall Vulnerability Index for Rural counties as well as Urban counties. The red lines indicate High, green indicates Medium, and blue indicates Low broadband internet access. There is less variability of telemedicine use within rural counties than urban counties. The 3 lines (red, green, and blue) are much closer together in rural counties. For urban counties, High broadband access has greater variation, peaking at approximately 22 telemedicine sessions per 1000 residents.

Table 3.

Polynomial regression results of telemedicine use per 1000 with broadband access and urban/rural across social vulnerability indexes.

	R2	B	95% CI for B
			Lower bound	Upper bound
Overall Vulnerability	0.142 c
Constant		11.97 c	10.64	13.30
Slope		−0.95	−2.55	0.66
Curvature		−12.40 c	−18.01	−6.80
BA—Medium		−2.94 c	−4.43	−1.45
Low		−3.85 c	−5.39	−2.30
Urban		8.47 c	6.97	9.98
BA—Med: Urban		−4.13 c	−6.30	−1.96
BA—Low: Urban		−5.94 c	−8.36	−3.51
Socioeconomic Status	0.140 c
Constant		11.64 c	10.30	12.98
Slope		−1.25	−2.92	0.41
Curvature		−9.33 b	−14.97	−3.70
BA—Medium		−2.89 c	−4.39	−1.39
Low		−3.68 c	−5.26	−2.11
Urban		8.49 c	6.99	9.99
BA—Med: Urban		−4.16 c	−6.33	−1.99
BA—Low: Urban		−5.90 c	−8.32	−3.47
Household Characteristics	0.152 c
Constant		9.76 c	8.45	11.08
Slope		−5.41 c	−6.93	−3.88
Curvature		6.58 a	1.01	12.16
BA—Medium		−2.23 b	−3.71	−0.75
Low		−2.80 c	−4.29	−1.32
Urban		8.65 c	7.16	10.14
BA—Med: Urban		−4.08 c	−6.24	−1.92
BA—Low: Urban		−6.28 c	−8.69	−3.87
Racial and Ethnic Minority	0.139 c
Constant		11.41 c	10.15	12.68
Slope		1.67 a	0.17	3.17
Curvature		−2.93	−8.72	2.86
BA—Medium		−2.86 c	−4.34	−1.37
Low		−4.32 c	−5.81	−2.83
Urban		8.20 c	6.67	9.72
BA—Med: Urban		−4.20 c	−6.37	−2.03
BA—Low: Urban		−5.61 c	−8.05	−3.18
Housing Type and Transportation	0.143 c
Constant		11.73 c	10.42	13.03
Slope		3.42 c	1.93	4.90
Curvature		−4.20	−9.80	1.41
BA—Medium		−3.09 c	−4.56	−1.61
Low		−4.84 c	−6.30	−3.38
Urban		8.31 c	6.81	9.81
BA—Med: Urban		−4.40 c	−6.56	−2.23
BA—Low: Urban		−5.56 c	−7.98	−3.14

^a P < .05,

^b P < .01,

^c P < .001.

Figure 2.

Telemedicine use per 1000 with broadband internet access and urban/rural across social vulnerability indexes. (A) Overall Vulnerability, (B) Socioeconomic Status, (C) Household Characteristics, (D) Racial and Ethnic Minority Status, and (E) Housing Type & Transportation.

Discussion

We conducted a retrospective, observational study of U.S. dyadic telemedicine sessions that occurred during 2022. Overall, we found a nonlinear association between social vulnerability and telemedicine use rates in U.S. counties during 2022. Telemedicine use was lowest in counties with the highest and lowest overall and socioeconomic status vulnerability and highest in counties with mid-range overall and socioeconomic status vulnerability. Telemedicine use was higher with higher levels of broadband access. However, counties with high Household Characteristics vulnerability (counties with more elderly or pediatric residents, people with a disability, single-parent households, and lower English language proficiency) had persistently lower rates of telemedicine use, despite high broadband access. Conversely, counties with high Housing Type and Transportation as well as Racial and Ethnic Minority status vulnerability showed higher telemedicine use with high broadband access. Overall, urban counties had much higher telemedicine use than rural counties. However, telemedicine use in rural counties showed little variation in relation to broadband access; use was not markedly higher in rural counties with high broadband access. We found multiple nuanced differences in the relationship between social vulnerability and telemedicine use between rural and urban settings, and by broadband access.

These findings generally validate the existing evidence that suggests age, race, ethnicity, and socioeconomic status influence telemedicine utilization in the United States.^10,14–16 However, our results indicate that the relationship of these factors to telemedicine use is intricate and context-dependent. Moreover, broadband access and rurality are the most important drivers of telemedicine use, along with specific social determinants of health. It is possible that factors unrelated to digital access, including poor local availability of telehealth programs and services or beliefs about COVID-19 public health precautions, negatively influenced telemedicine use in rural counties during the study time period, despite high broadband access. Certainly, the social and structural factors that influence telemedicine use may fundamentally differ for rural versus urban communities. Previous scholarship identified varied intrapersonal, interpersonal, and structural barriers to digital access in rural areas that include age (older), educational attainment (lower), socioeconomic status, poor digital health literacy, connectivity issues related to limited or unstable broadband access, and poor-quality infrastructure.^28,29 While barriers such as socioeconomic status and poor digital health literacy may be common to both urban and rural areas, connectivity and infrastructure are particularly problematic in rural areas. Terrain, labor cost, acquisition of equipment, and dissemination of existing resources are persistent concerns in infrastructural growth. Additionally, many rural areas lack robust healthcare delivery systems with well-developed digital health services designed for the community’s population and infrastructure. As digital communications and technology are increasingly used to deliver healthcare, further marginalization of rural populations is a concern.³⁰ It is plausible that health disparities among rural populations could increase as new technologies arise, without advancements in rural infrastructure or attention to the digital health needs of rural communities.^28,29

It is noteworthy that for counties with medium or high broadband access, telemedicine use increased as Ethnic and Minority status vulnerability increased. However, among rural counties, telemedicine use decreased with increased Ethnic and Minority status, regardless of higher broadband access. This indicates that factors other than broadband access are negatively influencing telemedicine use in rural counties with high Ethnic and Minority status vulnerability. Minority populations in rural areas suffer from poverty, lack of health insurance, lack of access to medical care, and language/literacy barriers that may result in low telemedicine use, regardless of broadband access. Additionally, urban areas may have greater capacity to address telemedicine use barriers for those of racial and ethnic minority status due to the presence of larger health systems, including academic health sciences centers and their resources (eg, diverse medical staff, increased clinician education among vulnerable populations). Previous studies have not explicitly considered these factors when analyzing racial and ethnic minority status in relation to telemedicine use.

Limitations

Limitations of this study relate to sampling and measurement. The data represent a substantial proportion of U.S. telemedicine sessions, encompass diverse provider types and healthcare settings, and appear geographically representative of U.S. telemedicine activity. However, they were sourced from a single telemedicine platform provider and could be biased in terms of the type of telemedicine services or providers. In terms of measurement, we aggregated the data at the county level, an approach that enabled this national-scale analysis but may not capture variation in social vulnerability among communities within counties. Finally, in this initial study, we analyzed composite indexes of social vulnerability that represent social vulnerability broadly, and in relation to geographic areas and not individuals. This broad representation of social vulnerability had the benefit of capturing multiple possible barriers to telemedicine access, informing globally effective planning and intervention. However, there could be underlying variation in the relationship between telemedicine use and the constituent indicators that comprise the SVI indexes that we analyzed. We are currently exploring these relationships in ongoing analyses.

Conclusion

This observational study of nearly 8 million U.S. telemedicine sessions showed that rurality and broadband access are key drivers of telemedicine use and may be more important than many social vulnerabilities in determining community-level telemedicine use. We also found nuanced differences in the relationship between social vulnerability and telemedicine use between rural and urban counties, and at different levels of broadband access. The results indicate that broadband access is powerful in ameliorating digital health inequities, but it is not a panacea, especially for rural communities and those with certain social vulnerabilities. Future studies should examine the differential factors that influence lower telemedicine use in rural and urban areas, as the basis for planning and prioritizing measures to address telemedicine disparities. Given the rapid growth in digital health innovation, there is an urgent need to more precisely understand the relationship between rurality, digital infrastructure, social vulnerability, and telemedicine use.

Author contributions

Mollie R. Cummins (Conceptualization, Formal analysis, Methodology, Project administration, Resources, Writing—original draft, Writing—review & editing), Bob Wong (Conceptualization, Data curation, Formal analysis, Visualization, Writing—original draft, Writing—review & editing), Neng Wan (Conceptualization, Data curation, Formal analysis, Methodology, Validation, Writing—review & editing), Jiuying Han (Formal analysis, Writing—review & editing), Sukrut D. Shishupal (Formal analysis, Writing—review & editing), Julia Ivanova (Conceptualization, Writing—original draft, Writing—review & editing), Asiyah Franklin (Writing—original draft, Writing—review & editing), Jace Johnny (Writing—original draft, Writing—review & editing), Triton Ong (Writing—review & editing), Brian E. Bunnell (Resources, Writing—review & editing), Ramkiran Gouripeddi (Conceptualization, Data curation, Methodology, Supervision, Validation, Writing—review & editing), and Brandon M. Welch (Resources, Writing—review & editing)

Funding

There was no funding source for this study; Doxy.me Inc. provided in-kind access to data and data consulting.

Conflicts of interest

B.M.W. is the founder, CEO, and shareholder of Doxy.me Inc., a commercial telemedicine company. M.C., J.I., T.O., H.S., J.F.B., B.S., and B.F.B. are employees of the same company. The authors declare no other conflicts of interest.

Data availability

The primary data that support these findings are not publicly available due to commercial restrictions. We used additional sources of data in our analysis that are publicly available from third parties and cited in the manuscript.