Data in this retrospective study were obtained from the electronic health records (EHRs) of a large health system providing diverse specialty care, located in a major urban center in California. Data from patients with telehealth encounters at the health system in the 1-year period from December 2021 to December 2022 were included in this study. We selected this time frame as a later phase of the COVID-19 PHE when telehealth care was well-established, but in-person restrictions had been loosened, and telehealth use had settled from its peak pandemic levels. For this study, telehealth visits were defined as video-enabled and telephone encounters between a patient and any provider type.

This study was approved by the University of California, San Francisco Institutional Review Board (review #22-35996). As a retrospective medical record review of deidentified patient data that had been previously collected as part of clinical care and quality improvement, it was deemed exempt from an informed consent process and HIPAA (Health Insurance Portability and Accountability Act) authorization. Data sharing was not included in institutional review board approval.

All adult patients (≥18 years) in the health system residing in a rural California zip code (see Assigning Rurality section) who had at least 1 telehealth encounter in the study period (December 2021 to December 2022) were included in the dataset (N=9359). The study population was drawn from a geographically diverse area of California and included residents of zip codes in a radius of hundreds of miles from the health center. According to our identified zip codes, rural patients comprised 4% of total telehealth patients at the health center during the study period.

Rural patients were identified using Rural-Urban Commuting Area (RUCA) codes [35] zip code approximations from the Washington, Wyoming, Alaska, Montana, Idaho Rural Health Research Center [36]. Along with their wide use in health services research, we chose RUCA codes because they provide a more granular breakdown of rurality than alternative systems that are based at the county level, such as Rural-Urban Continuum Codes. Given high levels of within-county variation in population density and distribution in many California counties, county-level classifications may collapse important rural-urban distinctions. RUCA codes avoid this by classifying based on the smaller scale of US Census tracts [37]. RUCA codes are assigned to US Census tracts based on population density, measures of urbanization, and daily commuting flows. The Washington, Wyoming, Alaska, Montana, Idaho Rural Health Research Center zip code database combines RUCA values from census tracts that comprise specific zip code areas [37], allowing researchers to apply RUCA codes to more easily accessed patient zip codes.

We used a 4-level urban-rural categorization of RUCA codes [38]: urban; large rural city or town (micropolitan), the most populous or least rural level; small rural town; and isolated small rural town, the most rural level. All California zip codes in the 3 rural categories were included (Figure 1). These RUCA groupings allowed us to analyze demographic and telehealth encounter characteristics of a diverse rural population with more nuance, reflective of meaningful measures of rural population density and resource proximity.

Statistical analysis was conducted from July 1 to October 17, 2023. We conducted descriptive statistical analyses of all patient demographic and telehealth encounter variables, with results presented as distributions for all categorical variables and measures of central tendency for the only continuous variable, patient age. Descriptive statistics were calculated for the total sample by rurality level and by telehealth modality.

Pearson chi-square test or Fisher exact test was used as appropriate to assess for significant associations between categorical variables. Because age was negatively skewed in this sample, the Kruskal-Wallis H test was used to test associations with continuous age.

Data were analyzed with Stata BE (version 17.0; StataCorp). For this study, statistical significance was determined at P values <.05.

There were 9359 unique patients with an address in a rural California zip code who had at least 1 telehealth encounter with the health system from December 2021 to December 2022 (Table 2). Our data were complete for all variables and did not contain missing values. The majority lived in large rural town zip codes (n=6393, 68.3%), 16.5% (n=1543) lived in small rural town zip codes, and 15.2% (n=1423) lived in isolated rural town zip codes. Of 506 rural zip codes in California, 331 (65.4%) were represented in the sample. One quarter of patients (n=2358, 25.2%) lived in just 6 zip codes, which were all large rural towns, and 50.7% (n=4745) of patients lived in 21 zip codes, of which 19 were large rural and 2 small rural zip codes.

Mean age, dichotomous age, race or ethnicity, preferred language, payer, and encounter provider were all significantly associated with patient level of rurality (Table 2). Sex, patient portal activation, and encounter specialty were not significantly associated with rurality. The mean age of the sample was 56.1 (SD 17.0; median 59.4, IQR 42.8-69.4) years and increased as rurality increased. While patients aged 65 years or older made up 37.2% (n=3485) of the sample, the proportion of those older than 65 years was 39.3% (n=606) and 41.3% (n=588) in small rural town and isolated rural town zip codes, respectively.

The majority of rural telehealth patients (n=6508, 69.5%) were White, while 14.5% (n=1352) were Latinx, and 10.2% (n=951) were another race or ethnicity. Race or ethnicity categories represented in the category combined other included Asian (n=142, 1.5%), Black or African American (n=105, 1.1%), Native American or Alaska Native (n=161, 1.7%), and other (n=340, 3.6%). Isolated rural town zip codes had less racial or ethnic diversity: 77.4% (n=1101) of patients in these zip codes were White. English was the preferred language for 95.4% (n=8926) of the sample, with 4.1% (n=383) of telehealth patients preferring Spanish, and 0.5% (n=50) preferring 1 of 21 other languages. Aligning with the Latinx race or ethnicity, notably fewer primary Spanish speakers were in the isolated rural town grouping (n=17, 1.2%).

At 44.8% (n=4193) of the sample, the largest payer group was Medicare, followed by other insurance at 35.3% (n=3304) and Medi-Cal at 19.9% (n=1862) of the sample. Nearly a quarter (n=975, 23.3%) of Medicare recipients were aged 18 to 64 years. There were more Medicare recipients in small rural towns and isolated rural towns (n=742, 48.1% and n=702, 49.3%, respectively), while the isolated rural town grouping had notably fewer other insurance patients (n=451, 31.7% compared to n=3304, 35.3% overall).

Mean age, dichotomous age, race or ethnicity, preferred language, payer, patient portal activation, encounter provider, and encounter specialty were all significantly associated with telehealth modality (Table 3). The level of rurality and sex were not significantly associated with modality.

Most of the telehealth encounters were video visits, at 92.9% (n=8690) of encounters. Before collapsing telephone encounter types, unscheduled telephone encounters comprised 0.7% (69/9359) of included telehealth encounters; 10.3% (69/669) of telephone encounters were unscheduled. The majority (60/69, 87%) of unscheduled telephone encounters were with “resource” providers, a broad category of providers that included counselors and chaplains.

Video users were younger than telephone patients, with a mean age of 55.8 (SD 17.00; median 59.0, IQR 42.4-69.1) years compared to 59.6 (SD 16.2; median 63.0, IQR 49.0-72.1). Patients 65 years or older had 8.7% (n=302) of their telehealth encounters as telephone compared to only 6.3% (n=367) of those younger than 65 years. Telehealth modality also differed substantially by race or ethnicity. Telephone use was highest among Latinx patients (n=123, 9.1%), 2 percentage points higher than the sample and 2.5 percentage points higher than among White patients (n=430, 6.6%). Preferred Spanish language speakers had nearly double the telephone use compared to preferred English language patients, at 12.3% (n=47) and 7% (n=621), respectively.

Medicare patients had the highest use of telephone modality, followed by Medi-Cal patients and patients with other insurance. Patient portal status was strongly associated with telehealth encounter modality (χ ²₂=219.7; P<.001). Patients with activated portals had only 6% (n=515) of their encounters as telephone, while those with portals that were pending activation had 21.1% (n=146) of their encounters as telephone.

Patient portal activation status was significantly associated with sex, mean age, dichotomous age, preferred language, and payer (Table 4). We were not able to test the associations of portal activation to race or ethnicity, provider, and specialty due to small cell counts. A large majority of the sample (n=8577, 91.6%) had activated patient portals, while 7.4% (n=693) were pending activation, and 1% (n=89) were inactivated. Patient portal activation was higher for patients who were of female sex, younger than 65 years of age, of White race or ethnicity, primary English speakers, and not on public insurance. These differences were particularly pronounced among Latinx patients (n=1183, 87.5% activated patient portals compared to n=8577, 91.6% overall) and patients who were primary Spanish speakers (n=295, 77% activated patient portals). There was some variation in patient portal activation across provider types, ranging from 89.8% (other providers: n=449) to 93.5% (both nurse practitioners: n=1116 and physician assistants: n=435).

In this study, we used 3 levels of rurality to characterize a population of rural-dwelling California adults who used telehealth services at an urban medical center from December 2021 to December 2022. Patients who lived in more rural zip codes were older, and a much higher proportion were White and primary English speakers. This aligns with other research showing that rural populations tend to be on average older and less racially and ethnically diverse [3,4]. Older age among more rural patients is of particular concern, as challenges associated with more rural status (eg, distance to services and weather disruptions) may be more impactful for older adults, compounding health care access challenges. Older adults also have lower digital access [41] and higher telehealth unreadiness [42], evidenced in our study by fewer video visits and lower patient portal use among older patients. Older patients may, in part, prefer telephone visits over video due to lower digital comfortability. Interventions to increase health care access through telehealth use among rural older adults could include digital access and comfortability assessments, digital education, and support for rural connectivity.

A quarter of our sample was comprised of patients from race or ethnicity groups other than White, in line with the rural United States as a whole [43]. However, at the time of the 2020 US Census [44], rural California counties [45] had a higher percentage of residents who were American Indian or Alaska Native (5.6%), Asian (2.1%), Black or African American (1.7%), and Hispanic or Latino (22.8%) than in our sample. While these data do not support a direct comparison because of different rurality measures (counties are the smallest scale for which US Census data are consistently available, as the US Census Bureau QuickFacts data tool provides statistics only for counties and for cities and towns with a population of 5000 or more), this may indicate that fewer rural individuals from these race or ethnicity groups are using telehealth at this urban health center. This is significant, given evidence that rural American Indian or Alaska Native and populations of color experience worse health outcomes than rural White populations [6,7,46]. Rural American Indian or Alaska Native and populations of color contend with complex barriers to realizing health as a result of legacies of colonization and slavery [6,47]. For these populations, patient-centeredness and cultural tailoring [48] will be of central importance for successful implementation and equitable use of telehealth services.

Our findings align with existing research showing higher video visit use by White patients compared to patients of other races or ethnicities [29-32]. In our rural sample, patients who were Latinx had the lowest video visit use despite being younger and living less rurally. These findings agree with a majority of studies showing lower video use among Hispanic or Latino patients [30,31,39], although Drake et al [29] found higher video use among rural and urban Hispanic patients in North Carolina. Research has also found that while Hispanic or Latino individuals used less video visits, they had higher overall telehealth use compared to non-Hispanic White individuals [24,32,49]. We did not include a comparison to in-person patients at the health center, and more research is needed to explore how rural Latino patients use in-person versus telehealth specialty services at distant health centers.

Multiple other studies have shown that patients with limited English proficiency (LEP) have fewer video and more telephone visits than English-proficient patients [28,30,31,33]. While these studies contained mixed rural-urban samples, we confirmed this finding in our entirely rural sample, where video use disparity was greatest among Spanish-speaking patients. Patients with LEP experience multiple barriers to health care access overall and, consequently, worse health outcomes [50]. These barriers may be confounded in rural areas with fewer services and lower diversity. Video visit disparities may exacerbate this issue. While video access is limited by patient-level LEP barriers, such as mistrust and perceived discrimination [51,52], clear provider- and system-level barriers also exist. Patients with LEP may not be offered video visits [25,33], a lack of language-concordant front office staff poses challenges to patients with LEP in obtaining appointments [51] and coordinating care [52], and difficulties bringing an interpreter on video platforms may also deter providers from offering video visits to patients with LEP [25,53]. Integrated video translation services, LEP community outreach, and digital access assessment, as well as the availability of language-concordant outreach materials, front office or call center staff, and patient portals have all been identified as important areas for intervention [50,53].

The patterns we found of lower video visit use among older patients, Latinx patients, and Spanish-speaking patients are similar to those reported in studies early in the COVID-19 pandemic [28-31]. Our video use findings also concur with more recent national data on older patients and Latinx patients [32]. The persistence of video visit disparities after the initial phases of the COVID-19 PHE, when systems- and patient-level telehealth barriers were likely highest, related to implementation and scale-up challenges, underscores the need for ongoing research and policy attention to understand this issue. As others have noted [25,27,28], telephone visits likely support overall access for vulnerable populations; therefore, while efforts should be made to address video barriers, policy should continue to support telephone visit availability and reimbursement.

As a proxy measure of digital engagement, an inactive patient portal may indicate patients at risk of digital access disparities [54,55], and our findings appear to substantiate this. Video visits were less common among patients whose portals were inactive or pending activation than among those with active portals, a finding we anticipated based on other studies [31,34,39]. On the other hand, our finding that neither telehealth modality nor patient portal status was significantly associated with rurality level was novel. Previous research has found that rural patients were significantly less likely to have video visits [28,34] and significantly less likely to have an activated patient portal [39]. However, another study found that while rurality was not associated with 3 measures of technology access, video and portal use were both positively associated with living in isolated rural census tracts [40].

In this context, our findings contribute to a complex picture of digital access and telehealth use patterns among rural populations. Our paper contributes to understanding telehealth modality use between patients living at different levels of rurality. One potential explanation for our finding of no association is that these other studies used nonrural comparison groups, while our sample was entirely rural. Another possible explanation is the use of different methodologies to define rurality, for example, RUCA codes versus Rural-Urban Continuum Codes, as well as different geographic units, such as census tract, zip code, or county [56,57]. Finally, rural populations in the United States are heterogeneous [8,43,58], and these findings may represent meaningful variation between these rural populations.

Our findings point to several areas for future research and intervention. More study is needed to fully characterize rural telehealth users, specifically rural populations of color. Future studies should apply sampling methods that account for the relatively fewer people of color living in rural areas to support statistical analysis of these groups. Research is also needed to elucidate the nature of relationships between patient factors and telehealth modality use among rural patients using telehealth at distant urban medical centers. For example, research should explore how patient preference for telehealth modality relates to age, race or ethnicity, preferred language, and educational attainment, with attention to how factors such as perceived lack of access or discrimination impact modality preferences among these groups.

Policy interventions are needed to support equitable access to telehealth overall [59] as well as the appropriate application of telephone and video visit modalities [27,28]. To support access for marginalized patient populations, such interventions should include the development of culturally tailored and language-concordant telehealth patient outreach and education. Centering patient perspectives in the development of these resources, through research participation and patient advisory boards, can help ensure that the materials are accurately targeted to address the perceived needs and preferences of their intended recipients [48,59]. Patients from marginalized groups may also prefer telephone visits over video due to concerns of discrimination or lack of access [47,51,52]; culturally tailored outreach and patient advisory boards may particularly help address these concerns. For older adults, as noted, a preference for the telephone over video may arise from uncertainty around digital devices. Policy to address equitable use of video visits should therefore support assessments of not just patient digital access but also digital comfort and related patient digital education. Providers can advocate for policies such as these to be implemented within their clinics or health systems. However, to ensure appropriate video access, policies from state and federal governments must also address structural access barriers. Policy support is needed for the ongoing development of broadband access in rural and other underserved areas [60,61], free or low-cost smart device and data plans for low-income patients, and the development of telehealth resources in public spaces, such as rural libraries [62]. Providers can support structural policy changes by advocating with professional organizations or with state and federal representatives.

Finally, policymakers must stay abreast of evidence regarding the effectiveness and accessibility of telephone and video telehealth modalities to inform reimbursement decisions in support of equitable health care access. Reimbursement policy can address important provider-side telehealth barriers, for example, by incentivizing video visits or by providing payment parity for telehealth and in-person services, thereby encouraging providers to direct resources to telehealth infrastructure, workflows, and training. The recent decision to end most CMS Medicare telehealth waivers deployed under COVID-19 by October 2025, if not reversed, will likely undermine health care access overall for older patients, particularly lower-income and rural populations. The exception for rural areas is not yet well-defined. Requiring rural Medicare patients to return to clinics to access telehealth services will entail a substantial travel burden for many older rural residents [63], and the loss of audio-only reimbursement may eliminate one of the most accessible access options for patients. Further, given the role Medicare reimbursement plays in informing reimbursement decisions across payer types, the loss of CMS waivers may signal the loss of telehealth coverage for privately insured individuals as well.

Low representation of patients from several race or ethnicity groups in our sample and the choice to collapse several categories of race or ethnicity to enable tests of association were limitations of our study. The categories of race or ethnicity we combined represent distinct populations of rural residents, who experience particular structural barriers to realizing health, and focused research with these patient populations is needed. Compared to the population of rural California counties overall [44,45], our sample was comprised of substantially more female patients (n=5158, 55.1% vs 48.5% overall) and more patients aged 65 years or older (n=3485, 37.2% vs 25.8% overall). This may limit the generalizability of our findings to diverse rural patient populations, rural male patients, and rural patients younger than 65 years of age. In our data, services provided by nurse practitioners, physician assistants, or other provider types may have been billed under the physician billing code, potentially inflating the number of physician encounters. Further, while patient portal activation status and race or ethnicity showed a strong trend of association, we were not able to test this due to limitations in our statistical analysis. As discussed, this proxy measure could not fully capture broadband access or digital literacy data, and related findings should be generalized with caution. Future research could use Federal Communications Commission county-level broadband availability data and measures of patient personal device use to more accurately capture digital access. Our data also did not support comparison to rural in-person patients. Self-selection bias may have been present, as there may be systematic differences between patients who opted for telehealth and those who chose in-person care. Data were not available on the health system’s total rural population. Further research is needed to explore how these groups differ in the use of specialty care at an urban medical center. Finally, our data are from a single health system; while we recognize the limitations of a dataset from a single health system, we believe findings may be applicable to similar large urban health systems nationally, insofar as they often provide specialty care to patients over a broader geographic region.

In this sample of rural patients who used telehealth at an urban medical center, video visit use and patient portal activation were lower among patients who were older, Latino race or ethnicity, primary Spanish speakers, and publicly insured. Targeted policies are needed to support appropriate video visit use in populations at risk of access barriers, including patient digital access assessment and education, culturally tailored and language-concordant telehealth outreach and education, low-cost smart device and data plans for low-income patients, rural broadband development, and evidence-based telehealth reimbursement policy.