Patient-Level Factors Associated with Utilization of Telemedicine Services from a Free Clinic During COVID-19

Introduction

The coronavirus (COVID-19) pandemic initially resulted in interruptions to health care access through closures or downscaled operations of numerous health care organizations.^1–3 Many organizations began adopting digital health technologies to remotely provide health care while minimizing in-person visits to moderate the risk of COVID-19 exposure.^4,5 In addition, the use of these technologies enabled clinicians to follow their patients for chronic care management.^5,6 Since continuity of care has been associated with a lower number of visits to the emergency department (ED) for ambulatory care-sensitive conditions,^7,8 televisits may allow EDs to focus their resources on emergent cases and COVID-19 patients.

Recent evidence suggests that the pandemic may disproportionately affect underserved populations.^9–12 Although telemedicine adoption has permitted some patients to receive uninterrupted care, organizations with fewer financial resources and safety-net clinics may struggle to provide this mode of care due to lack of resources to implement telemedicine technology among other competing financial priorities. Furthermore, there are disparities in the adoption and use of digital health technologies among different types of patients in the United States.^13–18 For instance, telemedicine use is negatively associated with older age,^19,20 rural residence,²¹ and being insured by public insurers.²⁰

However, many of these studies have been performed primarily with patients who have insurance. Those without insurance may have systemic differences from those with insurance, such as prior unmet medical needs and limited options for care. Thus, it is unclear what implementation factors are critical for telemedicine delivery to these underserved populations. To our knowledge, no study has been conducted with a focus of telemedicine usage among patients without insurance.

To address this gap, we aim to evaluate factors associated with telemedicine use among patients of a student-run free clinic network during the COVID-19 pandemic. We hypothesize that utilizers are more likely to be chronically ill, younger, white, more educated, and primarily speak English. This study’s findings may be useful to health system leaders and policymakers assessing the digital divide’s impact on this underserved population, especially during COVID-19. Furthermore, findings may also inform health care organizations considering the implementation of telemedicine systems for use beyond the COVID-19 pandemic.

Methods

STUDY DESIGN

A retrospective, cross-sectional design was used to determine factors related to telemedicine use. Thus, we utilized the Strengthening the Reporting of Observational Studies in Epidemiology statement to guide our reporting below.²²

SETTING AND SAMPLE

This study occurred in a student-run free clinic network associated with an academic medical center in North Central Florida that is staffed by volunteers. This clinic network provides free primary care services for patients who are uninsured or underinsured in the local community. Before COVID-19, patients could access clinic services through walk-in visits or appointments through any of the four primary care locations during the evenings and no telemedicine services were offered. In March 2020, the clinic network ceased seeing patients in-person and instead offered audio-only by phone and video telemedicine appointments via Zoom (Zoom Video Communications, San Jose, CA). In late July 2020, in-person care was offered again alongside the telemedicine option. Patients could request either method when requesting appointments.

DATA SOURCE

All patient sociodemographic variables, visit type, and comorbidities were pulled from the free clinic network’s electronic health record (EHR). To compare the demographic makeup of patients seeking telemedicine services with those who opted for only in-person visits, our study period focused on the telemedicine postimplementation period (March 2020 to September 2020). We only focused on the adult population (aged 18 and older).

Utilizing geographical information system processing techniques, the team used U.S. census data to calculate the distance between patients’ home residence and the location of their preferred clinic. Institutional review board approval was obtained from the University of Florida.

STUDY MEASURES

The outcome metric assessed if a patient used the telemedicine service at least once. Sociodemographic covariates from patient-reported check-in questionnaires in the EHR included the following: (i) age, (ii) sex, (iii) race/ethnicity, (iv) marital status, (v) highest education level attained, (vi) primary language spoken, (vii) housing status, and (viii) driving distance in miles between the patient’s home and clinic.

Since the study site has four primary care clinics and some patients may visit more than one, we ascertained which clinic site they self-reported as their preferred clinic site to use in our driving distance calculations. To obtain an accurate estimate of the distance between the preferred clinic site and a patient’s residence, we calculated driving distance based on roads rather than straight-line distances. Earlier work has shown driving distances can be more precise than straight-line distances. However, distance could not be calculated for patients who were homeless, listed a postal office box as their mailing address, or did not provide a home address. Driving distance was treated as a categorical variable (more than 31 miles, 15–30 miles, 0–15 miles, and unknown). Distance calculations were done through ArcGIS StreetMap Premium (ESRI, Redlands, CA).

We assessed for the presence of the following chronic conditions that are known to be prevalent in our patient population: (i) hypertension, (ii) hyperlipidemia, (iii) depression, and (iv) diabetes mellitus. We included a variable that represents the average number of chronic conditions patients have from this list of conditions. Based on self-reported data on intake questionnaires, we also examined for prior hospitalization history (measured as a dichotomous variable). Lastly, we created a dichotomous variable to represent if the patient was ever seen as a new patient visit during the study period.

ANALYTIC APPROACH

To address missing data, we coded an additional level within each categorical variable for patients who had missing data for a given variable. Descriptive statistics were used to describe the sample’s characteristics across all variables. We performed bivariate analyses through the unpaired Student’s t-test with unequal variances for all continuous variables, and the Pearson’s chi-square test for all categorical variables.

Since the literature is well-established about what factors influence telemedicine use, we used a theory-based approach when determining which variables to include in the multivariate logistic regression analyses. We reported both the unadjusted and adjusted odds ratios (aORs) for all variables that were placed in the regression model. Since each chronic condition studied makes up the variable on the total number of chronic conditions, we ran one model using the total number of chronic conditions and another model that instead includes the presence or absence of each chronic condition. A multicollinearity check revealed that missing variables were correlated with one another, and so, we removed all but one of the missing variables from our models. A p-value of <0.05 was interpreted as significant. All bivariate and multivariate analyses occurred using Stata SE 16.0 (Stata-Corp LP, College Station, TX).

Results

SAMPLE CHARACTERISTICS

Our final sample was 198 patients, where 112 (56.7%) patients were telemedicine utilizers. Of patients who utilized telemedicine, the number of visits ranged from 1 to 5, with an average of 1.54 visits. A majority of patients in the sample were female (58.1%), white (31.3%), not married (49.0%), completed postsecondary education (57.1%), had primary language as English (72.7%), lived in an apartment or house (36.9%), were uninsured (71.2%), employed (39.4%), and lived within 15 miles of their clinic (61.1%) (Table 1). With the exception of one patient, patients overwhelmingly elected for audio-only telemedicine over telemedicine with video.

FACTORS ASSOCIATED WITH TELEMEDICINE USE

In unadjusted analyses, telemedicine use was associated with white patients (42.0% vs. 17.4%, p < 0.001), those who were not married (58.0% vs. 37.2%, p < 0.001), those having secondary education (72.3% vs. 37.2%, p < 0.001), those living within 15 miles of clinic (77.7% vs. 40.0%, p < 0.001), those having no insurance (86.6% vs. 51.2%, p < 0.001), those who are unemployed (40.2% vs. 25.6%, p < 0.001), those who are established patients (96.4% vs. 47.7%, p < 0.001), and those self-reporting no prior hospitalization history (85.7% vs. 43.0%, p < 0.001) (Tables 1 and 2).

After controlling for other factors, older patients ( aOR = 0.96, 95% confidence interval [CI] 0.94–0.99), male patients (aOR = 0.40, 95% CI 0.18–0.92), and those establishing care as a new patient (aOR = 0.01, 95% CI 0.00–0.07) were less likely to use telemedicine. Furthermore, patients who lived within 15 miles of clinic (aOR = 4.59, 95% CI 1.54–13.72) were more likely to use telemedicine (Table 3). Further analyses revealed that carrying any of the chronic conditions studied were unrelated to the likelihood of using telemedicine.

Discussion

This study identified patient-level factors associated with receiving telemedicine care from a free clinic network during the COVID-19 pandemic, offering insights on the digital divide among a predominantly uninsured population and suggesting areas to optimize telemedicine delivery. To our knowledge, this is the first study to do so among free or reduced cost safety-net clinics, which may have systemic differences in patient case mix and infrastructure compared with nonfree clinics.

Overall, we found over half (56.6%) of our patients accessed our primary care services through telemedicine during the ongoing COVID-19 pandemic. Patients who were younger, female, established patients of the clinic, and lived closer to the clinic were more likely to receive care via telemedicine after controlling for other factors. Specifically, no differences were observed in telemedicine use for any chronic condition studied after controlling for other factors. We offer implications for practice and policy below.

Some disparities in telemedicine use across age, sex, and distance to clinic persisted even among a patient population that is predominantly uninsured. Additional work is still needed to determine other factors that may be hindering further adoption among this underserved population.

Our study did not observe disparities in race/ethnicity, which is consistent with some studies examining various types of digital health utilization.^23–26 At the same time, other studies reported that black patients were less likely to use digital health tools when compared with white patients.^{17,19,27–29} Differences in findings may stem from our sample having overrepresentation of factors, such as having no insurance.

Disparities over language preference were also not found despite having overrepresentation of Hispanic patients in our sample. This finding may be due to the existing infrastructure in these free clinics to offer Spanish translation services to these patients.³⁰ Findings on language preference’s association with telemedicine use were consistent with another study that examined a pediatric specialty clinic that treats a large portion of underserved patients.³¹

In addition, our finding that older patients were less likely to use telemedicine is consistent with several other studies.^19,20,29 Based on our patients’ predominant preference for audio-only telemedicine, our results suggest that exclusive reliance on phone visits/consults with older patients to overcome age disparities may not be sufficient especially when working with medically underserved populations.

Lastly, the finding that patients living closer to clinics were more likely to use telemedicine was surprising, suggesting that the use of telemedicine to save driving time and associated costs for patients living further away may not be fully realized when working with the medically underserved. This may stem from the relatively few clinics that provide free primary care, which may partially explain why patients are willing to drive further to obtain free care (compared with clinics using a sliding-scale model). Patients living farther away from the free clinics were also more likely to live in rural areas, suggesting that patients may be experiencing issues (e.g., lack of infrastructure for provisioning telecommunications and internet) with utilizing telemedicine.

Future qualitative research may be needed to identify reasons for telemedicine nonutilization from free clinic patients as other factors may influence telemedicine use, such as cultural factors, privacy concerns, patient preferences for in-person communication, or personal thoughts on the suitability of telemedicine for their chief complaints.

We found that among patients with chronic conditions that we studied, per-condition utilization of telemedicine visits ranged from 15% (hyperlipidemia) to 33% (hypertension). We speculate that the relatively low utilization rates across the studied chronic conditions may be due to patient preferences for in-person visits to manage their conditions in comparison with virtual care,^32,33 or the inherent infeasibility of select conditions to be adequately managed through telemedicine (e.g., chest pain) due to the need for a physical examination or point-of-care testing (e.g., blood glucose readings) to guide appropriate treatment. In our study, management of depression and various metabolic diseases was comparable among patients utilizing telemedicine and those who opted for in-person visits, suggesting that telemedicine appears to be facilitating continuity of care during the ongoing pandemic.

Nonetheless, the relatively low utilization rate of telemedicine to treat chronic conditions suggests that additional research is needed to optimize the delivery of telemedicine care for chronic conditions (e.g., diabetes, hypertension) that can potentially lead to greater and more expensive health care utilization if care is foregone.^34–39 For instance, one validated method has been recently published on how to perform a neurological examination using video telemedicine.⁴⁰ Further work is also needed to assess the quality of care of telemedicine delivered in free clinic settings to determine if these patients seek subsequent health care modalities (i.e., in-person visits to free clinics, ED visits) for the same chief complaint as their index telemedicine visit.

Based on our moderate rate of telemedicine utilization (56.6%), one key finding in our study was that it may be feasible to deliver telemedicine services in a free care setting. Safety-net organizations, such as free clinics, may have organizational-level barriers (e.g., limited information technology [IT] support, lower operating margins) to implement and sustain a video-based telemedicine model. However, offering audio-only models as one type of telemedicine service may improve the reach of televisits’ benefits to underserved populations. Another unique finding from our study was the disproportionate number of patients in the free clinics who elected to have audio-only telemedicine services rather than those with video components, suggesting that there may be barriers to utilizing telemedicine with video. In contrast, other care settings may see an even mix of audio-only and video telemedicine encounters.⁴¹

Although underserved populations tend to have less access to internet and smartphones,^42,43 there is a gradual increase in smartphone ownership among these populations.⁴⁴ Our study team found that over 90% of patients seen at our free clinics self-reported smartphone ownership,⁴⁵ suggesting many patients seen should have had video capabilities on their smartphones. Notably, internet availability has been documented to influence telemedicine use in rural areas,⁴³ and other studies have also found that these underserved populations often utilize only their smartphone to access the internet due to cost concerns with home internet alternatives.⁴⁶

Consequently, organizations delivering telemedicine to underserved populations may need to consider systemic factors these populations face when accessing video telemedicine through their mobile phones. For instance, data usage may be greater when utilizing telemedicine with video compared with audio-only telemedicine. Patients with limited data usage allowances may elect to forego telemedicine with video. Thus, additional research is needed to identify and evaluate interventions that address this specific digital divide in access to telemedicine with video components. For instance, the U.S. Veterans Affairs health system has partnered with one mobile carrier to allow its data plans to better support telemedicine services by not counting data used against veterans’ data use limits.⁴⁷

Future qualitative studies should also explore barriers and facilitators of audio-only telemedicine implementation in safety-net settings. Further policy work is also needed in developing the telemedicine infrastructure for safety-net organizations, such as developing grants to hire more IT staff and implementing device loan programs, to improve and sustain gains in telemedicine adoption among safety-net organizations beyond the COVID-19 pandemic.

The results should be interpreted with several limitations. There was a relatively low sample size used, which may have limited our abilities to detect differences in some of the clinical and sociodemographic variables. However, this free clinic network traditionally sees a lower patient volume compared with nonfree clinics due to operating hours on four weeknights. We also did not collect information on whether visits were for a full evaluation and management service, urgent visits, or for brief reasons (e.g., medication refills). We also could not account for whether free clinic staff encouraged or discouraged the use of audio-only telemedicine. Lastly, study data came from one free clinic network, limiting generalizability to other organizations due to possible systemic organizational factors.

Future research should confirm this study’s findings utilizing data from multiple free clinics and assess other factors that may influence telemedicine adoption among patients (e.g., patient preferences for in-person communication, familiarity with technology use).

Conclusions

In this cross-sectional study in a free clinic network, more than half of the patients utilized telemedicine. Of those who did, many elected to have audio-only encounters rather than video-facilitated telemedicine. Patients who were younger, female, established patients of the clinic, and lived closer to the clinic were more likely to use telemedicine. Additional research is needed to identify optimal ways to deliver chronic care management via telemedicine to safety-net patients who are accessing the internet via smartphones, such as developing more affordable and robust models to support video telemedicine.

Authors’ Contributions

This work represents the original research of the authors. This work has not been previously published. O.T.N., K.T., A.K.W., K.M., and C.L. conceptualized the study. O.T.N., K.M., and C.W. analyzed the data. O.T.N., K.T., and R.W.N. interpreted the data. O.T.N., K.T., and N.K. drafted the article. K.T., A.K.W., K.M., C.W., and RWN provided critical revisions to the article. All authors approved the submission.

Disclosure Statement

No competing financial interests exist.

Funding Information

No funding was received for this article.