Identifying Factors Associated with Self-Reported Adult Telehealth Utilization: Evidence from Mississippi

Abstract

Background:

Mississippi faces significant health disparities and barriers to health care access, particularly in its most rural areas. Telehealth offers a promising solution to address these challenges, but its adoption remains uneven. The purpose of this study was to investigate the potential factors associated with self-reported telehealth utilization among adult Mississippi residents, focusing on individual-, household-, and area-level characteristics.

Methods:

Data were collected from a state-representative survey of adult Mississippi residents (N = 821) using both online- and phone-based platforms, supplemented with secondary internet quality and local health care access data. A two-stage hurdle regression model was used to examine factors associated with telehealth use and conditional on any use, utilization frequency. A regression estimating associations with the use of in-person medical care was also estimated for comparison purposes.

Results:

Telehealth use was significantly associated with specific health conditions and health insurance status. However, local internet quality did not significantly impact the likelihood of telehealth use aside from a marginally significant association with local upload speed. Findings suggest that other demographic- and health-related factors may play a more prominent role. We also find differential telehealth utilization rates by region, suggesting that area-level characteristics like health care infrastructure may affect telehealth use likelihood.

Conclusions:

Telehealth adoption in Mississippi is associated with individual factors like health and insurance status rather than broadband access alone. Efforts to expand telehealth use should also address noninfrastructure barriers, such as digital literacy and awareness, particularly in rural and underserved populations.

Introduction

The state of Mississippi has long faced disparities in health outcomes and health care access. Mississippi consistently ranks at or near the bottom of all U.S. states on nearly every health metric, including clinical care, overall quality of care, and health systems quality.^1,2 These limitations underscore the critical need to expand health care access throughout the largely rural and low-income state.

One potential method for expanding health care access in underserved areas is telehealth. The Health Resources Services Administration defines telehealth as “…the use of electronic information and telecommunications technologies to support long-distance clinical health care and health administration.”³ Common telehealth technologies include videoconferencing, internet services, and remote monitoring tools. Recently, telehealth utilization has increased significantly, representing a growing share of the health care environment. Telehealth has been associated with improved health care access among patients with diabetes,⁴ acute respiratory illness,⁵ mental health conditions,⁶ and those in rural areas.^7,8

Recent research underscores telehealth’s potential to address health disparities. One study examined the impact of telemedicine compared to in-person primary care on visit completion rates in a rural Appalachia.⁹ The authors found that telemedicine significantly improved health care access for individuals facing geographic barriers. These findings are relevant for understanding how telehealth might address similar disparities in other rural areas.

Despite its promise, barriers to telehealth adoption persist. Broadband internet access, a proposed “superdeterminant” of health care access, is understandably critical for telehealth participation.¹⁰ In Mississippi, broadband access is limited, particularly in the state’s rural areas where only 63.4% of the population has fixed high-speed broadband compared to 97.8% in urban areas.¹¹ These disparities may prevent individuals from accessing telehealth regardless of need. Additionally, digital literacy and awareness have been identified in previous research as potential barriers that may be more prevalent in underserved communities.¹²

Mississippi’s telehealth regulation, including parity laws enacted in 2013 and 2014, has positioned the state as a leader in telehealth policy.¹³ However, policy alone cannot ensure utilization, and factors such as health status, insurance coverage, and regional infrastructure should be considered.

This study examines the self-reported utilization of telehealth among Mississippi residents, focusing on factors like area-level characteristics, broadband access, demographics, and economic status. Using a state-representative survey of adult residents, combined with secondary data on internet quality and local health care access, we employ a two-stage hurdle regression model to assess the likelihood and frequency of telehealth use. The findings provide insights into the barriers and facilitators of telehealth adoption, offering evidence-based recommendations for improving access. Results indicate that some health conditions and insurance coverage are significantly associated with self-reported telehealth adoption, while broadband access/quality may play a lesser role. These findings highlight the complexity of telehealth adoption and the need for targeted strategies to address the needs of underserved populations in Mississippi.

Methods

SURVEY DATA

To obtain a representative dataset of adult Mississippi residents, survey data were collected using two primary modalities, online and telephone, to maximize coverage and reduce potential biases caused by geographic disparities in broadband access. Specifically, while collecting online-only surveys is more cost effective, relying on online data collection alone would most likely prevent certain populations with limited internet from participating. To address this limitation, phone survey data were also collected by the Survey Research Laboratory (SRL) at Mississippi State University in parallel with online survey data collection.

All online survey data collection was contracted out to Qualtrics, who utilized a panel-group approach to target state-average quotas for age, race/ethnicity, and gender. This panel was carefully selected to represent Mississippi’s population and included population-proportional sampling within each of the state’s Cooperative Extension Regions.¹⁴ The phone survey specifically targeted respondents in counties below Mississippi’s median broadband access level based on the Federal Communications Commission’s (FCC’s) 2020 Broadband Deployment Data.¹¹ The SRL employed dual-frame random digit dialing, incorporating both landline and cellular numbers from a purchased call list. This dual-frame approach ensured that we reached a more comprehensive cross-section of the population living in areas with limited internet access.

Through this two-pronged data collection strategy, we achieved a more inclusive sample, with targeted oversampling via phone in counties with low broadband access. All online and phone survey participants were compensated with a $10 gift card provided by the respective companies (Qualtrics or SRL). Survey data collection took place between August 2022 and November 2022. A total of 1,191 surveys (partially and fully completed) were collected, including 891 online and 300 phone surveys. After data cleaning, which involved dropping 370 incomplete online surveys from Qualtrics with missing information, the final analysis sample included 821 observations. While a response rate for the online survey is not provided by Qualtrics given the nature of how they collect data from internally recruited participant panels, the cooperation rate (number of completed surveys divided by the sum of complete responses and participant refusals) for the phone survey was 19%.

The survey questions covered a range of sociodemographic and health-related topics, including self-reported internet access, physical/mental health status, and health care utilization. The survey addressed participants’ interaction with health care providers through in-person and telehealth visits, including telehealth formats such as video calls, phone calls, email, text messaging, and patient portals. Participants reporting any telehealth use were asked about specific communication platforms and their frequency of use, specifically for more intensive forms like video and phone calls. Respondents also provided data on in-person health care visits to facilitate a comparative analysis with telehealth use.

SECONDARY DATA

To supplement our survey data, we incorporated area-level internet speed measures from Ookla speed test data aggregated at the ZIP-code level.¹⁵ Ookla data provide upload and download speeds from real internet speed tests rather than information self-reported by internet service providers, offering a more accurate picture of area-level internet quality by aggregating individual test results within ∼610.8 m² tiles and compiling them at the ZIP-code level. Data for this portion of our analysis were drawn from the second quarter of 2022.

Local health care access data were sourced from the Area Health Resources File (AHRF)¹⁶ and Lightcast,¹⁷ providing county-level information on health care providers and facilities. The number of doctors per 1,000 residents includes active MDs, DOs, and specialists as of 2019, based on AHRF data. To further contextualize health care access, we calculated the number of health care employees and businesses per 1,000 residents, drawing on 2021 data from Lightcast, with health care businesses excluding hospitals and specialty facilities. Regional classifications follow the Mississippi Economic Development Council’s framework.¹⁸

QUANTITATIVE METHODS

We utilized both descriptive statistics and regression analysis. The primary model was a two-stage hurdle regression, which was used to estimate associations at both the extensive and intensive margins of telehealth utilization. The first stage used a logistic regression to estimate the likelihood of any self-reported telehealth utilization over the previous 12 months (at time of survey). In the second stage, for those who reported any telehealth, we applied a negative binomial model to estimate the number of telehealth visits. We also ran a logistic regression on any self-reported in-person medical visits to offer a comparison with telehealth utilization estimates.

This methodology allowed for a comprehensive examination of telehealth use across Mississippi, with consideration for local internet quality and demographic characteristics, which are essential in evaluating health care accessibility.

Results

Table 1 shows summary statistics for our full analysis sample. Roughly 32.5% of respondents used telehealth at some point during the previous 12 months. Among those reporting any telehealth utilization, the average number of video call or telephone visits with a health care professional was 2.71 ± 2.79. When asked about in-person medical care, 82.5% of the sample reported having any in-person medical visits during the previous 12 months, with a mean number of in-person visits of 4.57. Fig. 1 shows the number of respondents reporting any telehealth or in-person medical visits during the last 12 months. Additionally, we further separate respondents into categories of: no medical visits reported, telehealth only, in-person only, and in-person and telehealth. While a significant number of respondents reported using only in-person care during the past 12 months (52.5%), 21 respondents (roughly 3%) in our sample reported only using telehealth visits, implying that most respondents using telehealth in our sample also used in-person care.

Fig. 1.

Distribution of self-reported medical visits by visit type. Total observations = 821. Data source: Telehealth survey conducted by authors.

Table 1.

Summary Statistics for Final Analysis Sample

	MEAN	SD	MIN	MAX
Telehealth	0.325	0.469	0	1
# Videocall or telephone visits^a	2.712	2.786	1	20^b
In-person	0.825	0.381	0	1
# In-person visits^a	4.569	3.987	1	20^b
Self-report physical health^d	3.153	1.054	1 (poor)	5 (excellent)
Self-report mental health^d	3.410	1.142	1 (poor)	5 (excellent)
Self-reported diagnosis of physical/mental health conditions
Hypertension	0.491	0.5	0	1
Heart disease	0.118	0.323	0	1
Cholesterol	0.358	0.48	0	1
Cancer	0.114	0.319	0	1
Diabetes	0.178	0.383	0	1
Asthma	0.171	0.376	0	1
Depression	0.382	0.486	0	1
Anxiety	0.423	0.494	0	1
Multidiagnoses	0.614	0.487	0	1
No diagnoses	0.202	0.402	0	1
Types of self-reported health insurance coverage
Employer provided	0.392	0.489	0	1
Marketplace coverage	0.096	0.295	0	1
Medicare	0.259	0.439	0	1
Medicaid	0.166	0.372	0	1
Other	0.013	0.115	0	1
No insurance	0.153	0.361	0	1
Types of household internet access
Broadband only	0.635	0.482	0	1
Multitype including BB	0.063	0.244	0	1
Satellite only	0.079	0.27	0	1
Mobile hotspot only	0.089	0.285	0	1
Other only	0.012	0.11	0	1
Multitype excluding BB	0.011	0.104	0	1
No home internet	0.111	0.314	0	1
Public transportation use	0.055	0.228	0	1
Driving for travel	0.935	0.248	0	1
# Motor vehicles owned^a	2.113	1.256	0	9
Age^a	49.985	16.167	18	99
Male	0.390	0.488	0	1
Female	0.610	0.488	0	1
White	0.737	0.441
African American	0.210	0.407	0	1
Other race	0.054	0.225	0	1
Marital status
Married	0.482	0.500	0	1
Unmarried couple	0.093	0.290	0	1
Divorced	0.128	0.334	0	1
Widowed	0.061	0.239	0	1
Separated	0.054	0.225	0	1
Single and never married	0.183	0.387	0	1
# Adults in household^a	2.130	0.905	1	7
Educational attainment
Less than high school	0.076	0.264	0	1
High school	0.245	0.430	0	1
Some college/associates	0.419	0.494	0	1
Bachelor’s or higher	0.261	0.439	0	1
Employment status
Full-time	0.447	0.497	0	1
Part-time	0.019	0.138	0	1
Self-employed	0.082	0.274	0	1
Retired	0.196	0.397	0	1
Other^c	0.256	0.437	0	1
Household income level
Under $25 k	0.333	0.471	0	1
$25 k to under $50 k	0.259	0.439	0	1
$50 k to under $150 k	0.348	0.477	0	1
$150 k or higher	0.049	0.215	0	1
County-level healthcare access variables
# Doctors^a,e	1.994	2.131	0	8.864
# Healthcare workers^a,e	22.220	12.959	0.000	64.052
# Healthcare businesses^a,e	23.001	20.104	5.545	242.653
Zip-code-level internet quality variables
Avg. download speed (Mbps)^a	163.259	76.968	4.054	395.632
Avg. upload speed (Mbps)^a	95.164	76.694	1.384	330.453
Avg. latency (ms)^a	54.434	52.015	3.000	331.511
Broadband speed requirements (FCC, 2015)
Download speed ≥25 Mbps	0.983	0.130	0	1
Upload speed ≥3 Mbps	0.966	0.182	0	1
Latency ≥50 ms	0.343	0.475	0	1
Total observation	821

Presents continuous variables. All other variables are indicator variables.

20^bindicates that the participant reported >20 visits in the past 12 months. The total number of observations is 2 out of 267 telehealth respondents and 17 out of 660 in-person visit respondents.

Includes out-of-work, homemaker, student, unable to work, or choose not to work.

Five-point Likert scale.

indicates that the variable is measured as per 1,000 people. County- and ZIP-code-level data come from the Area Health Resources File (AHRF), U.S. Dept. of Health and Human Services, Lightcast (https://lightcast.io/), and Ookla Speedtest data (https://github.com/teamookla).

Moving to our other variables of interest, roughly 38% of the sample reported being in very good or excellent physical health, and 47% reported very good or excellent mental health. Among our set of self-reported physical and mental health condition diagnoses, the most reported conditions were hypertension, anxiety, and depression at 49.1%, 42.3%, and 38.2% of the sample, respectively. More than 61% of respondents reported having been ever diagnosed with more than one of the listed conditions. Roughly 85% of respondents reported having some type of health insurance with the most common types being employer provided and Medicare. When asked about their household’s internet access, 63.5% of the sample reported having broadband only. Alternatively, more than 11% of respondents reported having no home internet, reflecting the low level of internet access in Mississippi.

For county-level local health care access variables collected from the AHRF and Lightcast, per 1,000 population counties in our sample had roughly 2 doctors, 22 health care workers, and 23 health care businesses. For ZIP-code-level internet speeds, the average download speed was 163 megabits per second (Mbps) with a minimum of 4.05 Mbps and a maximum of 395.6 Mbps. The average upload speed for ZIP codes in our sample was 95.16 Mbps, and the average latency was 54.4 milliseconds. Using the pre-2024 FCC broadband speed definition of 25 Mbps download/3 Mbps upload, 98.3% of respondents lived in a ZIP code with average download speeds ≥25 Mbps, and 96.6% lived in a zip code with average upload speeds ≥3 Mbps. Additionally, Fig. 2 shows the geographic distribution of Ookla Speedtest locations across the state of Mississippi, highlighting the high level of coverage in the state.

Fig. 2.

Average download (left) and upload (right) internet speed tested by Ookla at ZIP-code level. Darker colors present faster download and upload speeds; each dot represents the centroid of the tile, which is a basic geographic unit used to aggregate speed test results. Data source: Ookla Speedtest data (https://github.com/teamookla).

Table 2 shows the results of our primary regression analysis. Columns 1 and 3 show estimates for the logistic regression first stage and negative binomial second stage of the hurdle model, respectively, focusing on factors associated with any telehealth use and frequency of use. For comparison purposes, Column 2 shows odds ratio (OR) coefficients from an equivalent regression to Column 1 but estimated for in-person care rather than telehealth. The first stage logistic regression results in Column 1 of Table 2 indicate that self-reported diagnoses of asthma and depression were significantly associated with higher odds of any telehealth utilization (OR = 1.935, p < 0.01; OR = 1.968, p < 0.01, respectively). Individuals with health insurance coverage, including employer-provided insurance (OR = 1.794, p < 0.1), marketplace coverage (OR = 3.380, p < 0.01), Medicare (OR = 2.658, p < 0.05), and Medicaid (OR = 2.973, p < 0.01), were significantly more likely to use telehealth compared to those without insurance. However, home internet access, including self-reported broadband (OR = 0.860, p > 0.1), was not a significant predictor of telehealth use. Alternatively, we do find a positive and marginally significant association for respondents with broadband-quality upload speeds (OR = 10.509, p < 0.1). We also find some evidence to suggest that the likelihood of any telehealth use decreases with age, increases with income, and is smaller for females (OR = 0.72, p < 0.1). Finally, we find evidence that self-reported public transportation use is associated with a higher likelihood of any telehealth use (OR = 3.386, p < 0.01).

Table 2.

Regression Results for Self-Reported Telehealth and In-Person Medical Visits

	TELEHEALTH (ANY)(FIRST HURDLE)	IN-PERSON (ANY)	TELEHEALTH COUNT(SECOND HURDLE)
	(1)LOGISTIC REGRESSION(ODDS RATIOS)	(2)LOGISTIC REGRESSION(ODDS RATIOS)	(3)NEGATIVE BINOMIAL
Self-report physical health status (ref = poor)
Fair	0.823 (0.423, 1.601)	0.470 (0.123, 1.803)	−0.005 (0.342)
Good	1.297 (0.658, 2.554)	0.271 (0.072, 1.022)	−0.152 (0.374)
Very good	0.875 (0.422, 1.814)	0.290 (0.074, 1.142)	−0.640 (0.439)
Excellent	0.959 (0.390, 2.359)	0.370 (0.086, 1.590)	−0.483 (0.546)
Self-report mental health status (ref = poor)
Fair	1.630 (0.810, 3.280)	0.658 (0.218, 1.986)	−0.253 (0.339)
Good	1.006 (0.503, 2.010)	1.020 (0.338, 3.081)	−0.550 (0.358)
Very good	0.752 (0.354, 1.595)	1.196 (0.369, 3.878)	−0.240 (0.410)
Excellent	0.932 (0.421, 2.065)	1.173 (0.350, 3.929)	−0.252 (0.452)
Diagnosed physical/mental health conditions
Hypertension	0.847 (0.611, 1.173)	3.519*** (2.179, 5.681)	−0.350* (0.209)
Heart disease	1.083 (0.668, 1.757)	2.061 (0.729, 5.823)	0.399 (0.270)
Cholesterol	1.330 (0.926, 1.910)	2.665*** (1.467, 4.842)	−0.130 (0.218)
Cancer	0.943 (0.592, 1.505)	1.057 (0.453, 2.470)	−0.055 (0.262)
Diabetes	0.997 (0.669, 1.486)	1.477 (0.708, 3.081)	0.290 (0.239)
Asthma	1.935*** (1.316, 2.844)	0.746 (0.414, 1.344)	0.049 (0.216)
Depression	1.968*** (1.327, 2.921)	0.518* (0.289, 0.928)	−0.053 (0.265)
Anxiety	1.413 (0.950, 2.102)	3.602*** (1.998, 6.496)	0.144 (0.256)
# In-person visit	1.126*** (1.083, 1.170)		0.079*** (0.019)
# Telehealth visit		1.259** (1.069, 1.482)
Types of health insurance (ref = no insurance)
Employer-provided only	1.794* (1.026, 3.136)	4.080*** (2.215, 7.517)	0.112 (0.337)
Marketplace coverage only	3.380*** (1.657, 6.892)	5.858*** (2.270, 15.117)	0.297 (0.438)
Medicare only	2.658** (1.411, 5.004)	2.601** (1.237, 5.468)	0.535 (0.364)
Medicaid only	2.973*** (1.609, 5.492)	4.646*** (2.286, 9.443)	0.414 (0.374)
Other only	3.493*** (1.718, 7.104)	5.285*** (1.897, 14.725)	0.150 (0.412)
Multi-insured	3.071*** (1.516, 6.218)	2.531 (0.976, 6.562)	0.510 (0.388)
Types of home internet access (ref = no internet)
Broadband only (wired/wireless)	0.860 (0.524, 1.411)	0.952 (0.474, 1.913)	−0.415 (0.294)
Multitypes including Broadband	2.032 (0.974, 4.239)	0.542 (0.203, 1.444)	0.044 (0.389)
Satellite only	0.835 (0.415, 1.677)	0.302** (0.119, 0.771)	0.173 (0.416)
Mobile hotspot only	0.766 (0.388, 1.510)	0.680 (0.303, 1.528)	−0.899** (0.445)
Other only	0.864 (0.194, 3.845)	0.183** (0.045, 0.744)	0.464 (0.798)
Multitypes excluding Broadband	2.532 (0.662, 9.678)	3,942,675.676 (0, inf.)	−0.430 (0.846)
Public transportation use	3.386*** (1.723, 6.652)	0.742 (0.297, 1.858)	0.148 (0.287)
Age group (ref = Ages 18–39)
Ages 40–49	0.894 (0.575, 1.390)	0.583* (0.342, 0.994)	−0.140 (0.265)
Ages 50–59	0.618* (0.391, 0.976)	1.039*** (0.559, 1.931)	0.065 (0.272)
Ages 60–69	0.723 (0.432, 1.208)	1.209*** (0.605, 2.417)	−0.467 (0.308)
Age 70 older	0.356** (0.179, 0.71)	3.353* (1.101, 10.213)	−0.593 (0.432)
Female	0.720* (0.522, 0.992)	2.391*** (1.588, 3.599)	−0.018 (0.200)
African American Other races	1.066 (0.739, 1.538)	0.961 (0.600, 1.539)	0.099 (0.221)
Married	1.335 (0.945, 1.886)	1.631* (1.055, 2.520)	−0.231 (0.214)
# Adults in household	0.935 (0.791, 1.106)	0.838 (0.680, 1.032)	−0.036 (0.106)
Bachelor’s degree or higher	1.133 (0.785, 1.636)	2.800*** (1.573, 4.983)	0.019 (0.222)
Full-time employee	1.311 (0.909, 1.891)	0.634 (0.396, 1.015)	0.270 (0.246)
Income group (ref = under $25k)
$25 k to under $50k	1.743** (1.156, 2.630)	1.129 (0.681, 1.870)	0.106 (0.258)
$50 k to under $150k	1.729** (1.109, 2.696)	1.333 (0.772, 2.302)	0.367 (0.262)
$150 k or higher	2.712** (1.246, 5.903)	2.499 (0.715, 8.735)	0.271 (0.475)
# Doctors per 1,000 people	1.058 (0.943, 1.186)	0.863 (0.731, 1.018)	0.104 (0.069)
# Health care employment per 1,000 people	0.991 (0.975, 1.007)	1.019 (0.997, 1.043)	−0.010 (0.011)
# Health businesses per 1,000 people	0.995 (0.984, 1.008)	0.991 (0.979, 1.003)	−0.003 (0.006)
Broadband (25 Mbp download speed)	0.914 (0.253, 3.301)	2.329 (0.458, 11.855)	0.646 (0.844)
Broadband (3 Mbp upload speed)	10.509* (0.991, 111.395)	0.108 (0.009, 1.327)	17.546 (4,224.248)
Broadband latency (Longer than 50 ms)	0.806 (0.560, 1.158)	1.824** (1.123, 2.962)	0.176 (0.227)
Regions in MS (ref = Eco Dev Reg 1, Northwest of MS)
Eco Dev Reg 2	2.578** (1.240, 5.362)	1.942 (0.808, 4.664)	0.088 (0.467)
Eco Dev Reg 3	1.117 (0.459, 2.718)	0.999 (0.379, 2.633)	−0.460 (0.589)
Eco Dev Reg 4	2.413** (1.170, 4.976)	1.254 (0.557, 2.819)	−0.073 (0.468)
Eco Dev Reg 5	3.246** (1.387, 7.599)	4.163** (1.421, 12.198)	−0.754 (0.559)
Eco Dev Reg 6	3.547** (1.369, 9.187)	2.900 (0.785, 10.717)	0.247 (0.569)
Eco Dev Reg 7	2.253* (1.038, 4.890)	2.189 (0.878, 5.457)	−0.050 (0.489)
Eco Dev Reg 8	2.260* (1.077, 4.741)	1.192 (0.529, 2.684)	0.118 (0.469)
Observations	821	821	821
Log likelihood	−413.892	−254.822	−844.361
AIC	941.783	623.645	1,918.722

Note: ^*** p < 0.01, ^** p < 0.05, and ^* p < 0.1. 90% Confidence intervals shown for odds ratio coefficients in Columns 1 and 2. Standard errors shown in parentheses for Column 3. Columns 1 and 2 provide results from a logistic regression model with the coefficients in odds ratios. Column 3 shows results from the negative binomial second stage of the telehealth hurdle model.

Moving to the estimates of Column 2, unlike telehealth, in-person care was significantly associated with diagnoses of hypertension (OR = 3.519, p < 0.01) and anxiety (OR = 3.602, p < 0.01). Health insurance coverage is also strongly associated with any in-person care use as expected, with marketplace plans (OR = 5.858, p < 0.01), Medicaid (OR = 4.646, p < 0.01), and Medicare (OR = 2.601, p < 0.05) showing statistically significant associations.

The second stage of our hurdle model’s negative binomial regression results (Column 3) highlights factors influencing the frequency of telehealth visits via video call or telephone call among users reporting any telehealth. Individuals with a diagnosis of hypertension had significantly fewer telehealth visits (β = −0.350, p < 0.1), while those with more in-person visits reported higher telehealth utilization (β = 0.079, p < 0.01). Broadband internet access and other internet modalities were not significant predictors of telehealth use frequency except for a marginal negative association for respondents reporting only having mobile hotspot internet access at home (β = −0.899, p < 0.1).

Geographic regions within Mississippi showed varying impacts on telehealth use and frequency. For example, compared to the reference region (Northwest Mississippi), residents in Region 5 (East-Central Mississippi) had significantly higher odds of both telehealth (OR = 3.246, p < 0.05) and in-person care use (OR = 4.163, p < 0.05). Only Region 3, primarily including Delta counties, does not show a statistically significant difference in telehealth use.

The findings suggest that telehealth utilization is primarily associated with individual-level factors like health conditions and insurance coverage rather than local internet access/quality, although some internet-related variables show significant associations. The results emphasize the potentially complementary nature of telehealth and in-person care, as evidenced by the positive association between in-person visit frequency and telehealth use.

Discussion

The findings of this study underscore the complexity of telehealth utilization in Mississippi, where health care access and health disparities remain significant challenges. Telehealth adoption was significantly associated with some individual health characteristics, particularly the presence of certain chronic conditions such as asthma and depression. These results align with prior research indicating that individuals managing chronic health conditions are more likely to engage in telehealth services as a means of maintaining consistent health care access.^5,6 However, while previous studies have found that managing chronic conditions like diabetes through telehealth is both low cost and effective;⁴ in our study, individuals with hypertension, high cholesterol, and anxiety were more likely to use in-person care, but not telehealth. This divergence may reflect regional health care preferences, insurance coverage differences, or variations in how telehealth services are integrated into chronic disease management programs across regions.

Despite the logical importance of broadband access in facilitating telehealth services, our results suggest that neither self-reported home internet access nor measures of internet quality from Ookla speed test data were strong predictors of telehealth use, save for a marginally significant association between telehealth utilization and average internet upload speed. This finding is somewhat unexpected given prior research that has identified broadband as a critical determinant of telehealth use.¹⁰ However, our results are consistent with the growing body of evidence suggesting that broadband availability alone does not ensure telehealth uptake.¹² Instead, digital literacy, patient awareness of telehealth services, and health care provider engagement may play more significant roles in driving adoption, particularly in rural areas with historically low telehealth uptake and in-person care access.

Our results also provide evidence supporting the complementary nature of telehealth and in-person care. We find a strong positive association between the frequency of in-person visits and telehealth utilization, aligning with research suggesting that telehealth is not necessarily a substitute for traditional care but rather an integral component of a hybrid health care model.¹⁹ The finding that public transportation use is positively associated with telehealth use further supports the idea that telehealth serves as an accessibility-enhancing tool for populations facing mobility constraints.

Regional disparities in telehealth utilization highlight the importance of local health care contexts, which is a promising avenue for future research. As an example, residents of East-Central Mississippi (Region 5) were more likely to use both in-person care and telehealth, suggesting that regional health care infrastructure and policy implementation may significantly influence adoption. A key factor that may contribute to the higher expected likelihood of telehealth adoption observed in Region 5 is the presence of the Mississippi Band of Choctaw Indians (MBCI), the only Federally Recognized American Indian/Alaska Native Tribe in the state, and their associated services/infrastructure. Region 5 counties contain six of the MBCI’s eight tribal communities, including Pearl River which houses the MBCI’s Tribal Government.²⁰ Pearl River is also home to the Choctaw Health Center (CHC), a comprehensive health care facility that serves as the primary health services hub for both MBCI Tribal Communities and residents in surrounding communities/counties.²¹ During the COVID-19 pandemic, the CHC opened the Choctaw Health Center Telephone Clinic, allowing community members (both MBCI members and nonmembers) to access care via telehealth.²² Telehealth services may have also been even more impactful for MBCI communities during the pandemic as they were disproportionately affected by COVID-19 compared to all other groups in the state.²³ This led the CHC to integrate telehealth services during our study period as a means of expanding health care access, particularly for chronic disease management and mental health care. This finding is consistent with research indicating that health care-system-level interventions, such as telehealth clinics operated by trusted community health care providers, can enhance telehealth engagement among rural populations.⁸ Examining the role of area-level characteristics in telehealth use decision making is a promising avenue for future research.

This study has several limitations. First, it relies on survey data collected from adult Mississippians at a single point in time, which may introduce bias from unobserved confounders that could otherwise be mitigated with longitudinal data. Second, self-reported data carry the risk of misreporting, potentially biasing our estimates. Third, data collection occurred during the COVID-19 pandemic, a period of heightened telehealth usage, which may limit the generalizability of our findings to the post pandemic period. Finally, as the study focuses exclusively on Mississippi, its findings may not be directly applicable to other states.

Conclusions

This study contributes to our understanding of telehealth adoption by examining associations between health conditions, insurance coverage, and local infrastructure in Mississippi. The results suggest that while internet access may be necessary for telehealth, it is not sufficient to ensure adoption. Addressing other barriers, such as digital literacy and health care awareness, especially among older adults, low-income adults, and the uninsured, could be crucial for expanding telehealth utilization in underserved areas.

Policymakers should prioritize comprehensive strategies that include not only investments in broadband and physical infrastructure but also targeted outreach and education for populations with limited health care access. Future research should explore the role of technological literacy, provider availability, area-level characteristics, and cultural factors in shaping telehealth adoption to design more effective interventions.

Footnotes

Acknowledgments

The authors thank the participants of the 2024 Southern Economic Association and 2023 American Agricultural Economics Association conferences for their valuable feedback on earlier versions of this research. The authors also acknowledge the Mississippi Center for Clinical and Translational Research for their assistance in the early stages of this project. ChatGPT version 4.0 was used for grammatical revisions.

Authors’ Contributions

W.D. conceptualized the study, designed the survey, supervised the project, secured funding, and wrote and revised the article. A.K. cleaned the data, performed statistical analyses, created tables and figures, collected secondary area-level data, and contributed to writing and revising the article.

Ethical Approval and Informed Consent

This study involving human participants was approved by the Mississippi State University Institutional Review Board (IRB# 21-298). The research was conducted in accordance with applicable legislation and institutional requirements. Written informed consent was obtained from online survey participants, and verbal informed consent was obtained from phone survey participants.

Data Availability

Due to concerns regarding confidentiality and the risk of participant identification, the respondent-level survey data generated and analyzed in this study cannot be shared publicly. Secondary county- and ZIP-code-level data analyzed in this study are available upon reasonable request.

Disclosure Statement

The authors declare no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding Information

The collection of survey data used in this publication was supported by the National Institute of General Medical Sciences of the National Institutes of Health under Award Number 5U54GM115428. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

References

United Health Foundation. America’s health rankings annual report 2023. Available from: www.americashealthrankings.org [Last accessed: January 7, 2025 ].

Radley

, Baumgartner

, Clark

, et al. 2023 scorecard on state health system performance: Appendices. New York, NY: The Commonwealth Fund; 2023. Available from: www.commonwealthfund.org/sites/default/files/2023-06/Radley_2023_State_Scorecard_APPENDICES.pdf [Last accessed: January 7, 2025 ].

HealthIT.gov. What is telehealth? How is telehealth different from telemedicine? 2019. Oct 17. Available from: www.healthit.gov/faq/what-telehealth-how-telehealth-different-telemedicine [Last accessed: January 7, 2025 ].

Appuswamy

, Desimone

. Managing diabetes in hard-to-reach populations: A review of telehealth interventions. Curr Diab Rep, 2020; 20(7):28–10.

Ashwood

, Mehrotra

, Cowling

, et al. Direct-to-consumer telehealth may increase access to care but does not decrease spending. Health Aff(Millwood), 2017; 36(3):485–491.

Douglas

, Xu

, Heggs

, et al. Assessing telemedicine utilization by using Medicaid claims data. Psychiatric Serv, 2017; 68(2):173–178; doi: 10.1176/appi.ps.201500518

Mehrotra

, Jena

, Busch

, et al. Utilization of telemedicine among rural Medicare beneficiaries. JAMA, 2016; 315(18):2015–2016.

Palozzi

, Schettini

, Chirico

. Enhancing the sustainable goal of access to healthcare: Findings from a literature review on telemedicine employment in rural areas. Sustainability, 2020; 12(8):3318.

Haggerty

, Stephens

, Peckens

, et al. Telemedicine versus in-person primary care: Impact on visit completion rate in a rural Appalachian population. J Am Board Fam Med, 2022; 35(3):475–484.

10.

Turcios

. Digital access: A super determinant of health. SAMHSA. 2023. Available from: www.samhsa.gov [Last accessed: January 7, 2025 ].

11.

Federal Communications Commission. Broadband deployment data from FCC Form 477. Washington, DC: Federal Communications Commission; 2023. Available from: www.fcc.gov/general/broadband-deployment-data-fcc-form-477 [Last accessed: January 7, 2025 ].

12.

Blandford

, Wesson

, Amalberti

, et al. Opportunities and challenges for telehealth within, and beyond, a pandemic. Lancet Glob Health, 2020; 8(11): e1364–e1365.

13.

Thomas

, Capistrant

. 50 state telemedicine gaps analysis: Coverage and reimbursement. American Telemedicine Association; 2017. Available from: www.globalmed.com/wp-content/uploads/2019/07/2017-NEW_50-State-Telehealth-Gaps-Analysis-Reimbursement_FINAL.pdf [Last accessed: January 7, 2025 ].

14.

Mississippi State University Extension. County extension offices. Mississippi State University Extension. Available from: www.extension.msstate.edu/about-extension/county-extension-offices [Last accessed: January 7, 2025 ].

15.

Speedtest by Ookla. Speedtest global performance dataset. AWS Public Dataset Registry. Available from: www.registry.opendata.aws/speedtest-global-performance / [Last accessed: January 7, 2025 ].

16.

Health Resources and Services Administration. Area health resources files (AHRF). 2025. Available from: www.data.hrsa.gov/topics/health-workforce/ahrf [Last accessed: January 7, 2025 ].

17.

Lightcast: Data and insights for the new economy. Available from: www.lightcast.io [Last accessed: January 7, 2025 ].

18.

Mississippi Economic Development Council. Regional map. Available from: www.medc.ms/regionalmap [Last accessed: January 7, 2025 ].

19.

Adepoju

, Dang

, Valdez

. Comparing in-person only, telemedicine only, and hybrid health care utilization among older adults in the United States. Telemed Rep, 2023; 4(1):93–99.

20.

Global South Studies Center. Mississippi Band of Choctaw Indians. 2016. Available from: https://gssei.wordpress.com/2016/09/06/__trashed-3/ [Last accessed: March 4, 2025 ].

21.

Choctaw Health Center. About us. Available from: https://www.choctawhealthcenter.org [Last accessed March 4, 2025 ].

22.

Choctaw Health Center. COVID-19 Information. Available from: https://www.choctawhealthcenter.org/covid19info.php?chc_sources=1&date=full01-10-2024-12-40-15 [Last accessed: March 4, 2025 ].

23.

Pulitzer Center. More Choctaws have died of COVID than those who died of the disease in Hawaii, Alaska, or Wyoming. 2020. Available from: https://pulitzercenter.org/stories/more-choctaws-have-died-covid-those-who-died-disease-hawaii-or-alaska-or-wyoming [Last accessed: March 4, 2025 ].