Sociodemographic factors affecting telemedicine access: A population-based analysis.

BACKGROUND: The coronavirus disease 2019 pandemic prompted a surge in telemedicine, with the presumption that patients had computer and internet access. We sought to determine, in a population-based sample, how many Americans were using computers and the internet before the pandemic, and whether disparities existed in this.
METHODS: The National Health Interview Survey is conducted annually by the Centers for Disease Control and Prevention and is representative of the entire United States civilian non-institutionalized population. In 2018, questions were fielded regarding computer and web utilization. We evaluated sociodemographic factors associated with this.
RESULTS: Twenty-five thousand and forty-nine people, representing 245,842,992 in the population, responded to these questions. Of the responses, 19% stated they used a computer "never or almost never," 18% stated they did not use the internet, and 25% did not use email. Over the previous 12 months, 55% of responders stated they had looked up health information on the internet, 11% had filled a prescription online, 16% had scheduled a medical appointment on the internet, and 17% had communicated with a healthcare provider by email. Internet usage varied by region, age, race, education, family income, and insurance status, but not by gender.
CONCLUSION: As telemedicine becomes more prevalent, sociodemographic factors limiting computer and internet use may disadvantage certain segments of the population.

Introduction

The concept of telemedicine is not new—indeed, as early as 1879, there were reports of doctors making diagnoses via telephone. In the 1960s, the National Aeronautics and Space Administration had begun performing physiologic monitoring over a distance and, in 1972, had launched the Space Technology Applied to Rural Papago Advanced Health Care system to deliver remote health care to the Papago Indian Reservation in Arizona using 2-way microwave transmission. In 1989, with the creation of the World Wide Web, the possibilities for telemedicine expanded, and in 1999, the Centers for Medicare and Medicaid began covering telehealth consultations for patients in underserved areas. The American Recovery and Reinvestment Act of 2009 included telemedicine funding, and by 2020, telemedicine was projected to be a $34 billion industry.

While the utilization of telemedicine was not mainstream for most patients, the coronavirus disease 2019 (COVID-19) pandemic caused an abrupt shift toward this modality of healthcare delivery, as many hospitals and clinics shut down, and social distancing was encouraged. Most healthcare systems shifted toward telemedicine in a seamless fashion, with the assumption that the majority of Americans would be able to avail themselves of remote healthcare. Indeed, the Pew Research Center had reported in 2015 that 84% of Americans used the internet. Still, it remained unclear as to how many people felt comfortable using telemedicine for various tasks, from communicating with a physician to scheduling appointments to filling prescriptions. Were there disparities in terms of computer and internet access that disproportionately put some patients at a disadvantage when it came to accessing healthcare during the pandemic? We sought to answer these questions using data from a national population-based survey conducted in 2018, just before the pandemic hit.

Methods

The National Health Interview Survey (NHIS) is considered the largest source of health information for the American civilian non-institutionalized population. Each year, the Centers for Disease Control and Prevention conducts a face-to-face survey using a complex sampling scheme that is designed to be reflective of the population. In 2018, questions were fielded regarding individuals’ use of telemedicine. In particular, people were asked the following: “During the past 12 months, have you ever used computers for any of the following: (1) look up health information on the internet, (2) fill a prescription, (3) schedule an appointment with a health care provider, (4) communicated with health care provider by email”. In addition, people were asked, “How often do you use a computer?”, “Do you use the internet?”, “Do you send or receive emails?”. Of note, there was no question regarding the use of a smartphone (as opposed to a computer); however, the question pertaining to internet use was regardless of device. We evaluated the impact of various sociodemographic factors on the use of telemedicine before the pandemic. While it would have been ideal to evaluate these metrics in 2019, questions regarding telemedicine usage were not fielded by the NHIS in that year; one would not anticipate dramatic changes between 2018 and 2019. Statistical analyses were conducted using SAS-callable SUDAAN software. As these data are deidentified, publicly available, and analyzed in aggregate, this study was deemed exempt by the Human Investigations Committee of Yale University.

Results

The final response rate for the 2018 NHIS was 53.1%; however, this is reflective of the complex hierarchical sampling scheme the NHIS employs (see details here: https://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/NHIS/2018/srvydesc.pdf). However, of the sample adults approached (from whom the data for this analysis is based), the response rate was 83.9%. Twenty-five thousand forty-nine people, representing 245,842,992 in the population, responded to these questions. Of the responders, 19% stated they used a computer “never or almost never,” 18% stated they did not use the internet within the preceding 12 months, and 25% stated they did not use email. Over the previous 12 months, 55% stated they had looked up health information on the internet, 11% had filled a prescription online, 16% had scheduled a medical appointment on the internet, and 17% had communicated with a healthcare provider by email. Of the 82% of people who stated that they used the internet, only 20% had scheduled a medical appointment online (P < .001) within the previous 12 months, and 14% had filled a prescription online within that time frame (P < .001). Of the 75% who stated they used email, only 22% had communicated with a healthcare provider using this modality within the past year (P < .001).

Sociodemographic factors significantly impacted how often people used a computer (Table I ), whether they used the internet and email (Table II ), and whether they had looked up health information on the internet, filled a prescription online, scheduled a medical appointment on the internet, or communicated with a healthcare provider via email (Table III ). On multivariate analysis, age, race, education, income, and insurance were independent predictors of internet use (P < .001); region was no longer significant (P < .001) (Table IV ).

Table I

Sociodemographic factors associated with computer use

Factor	Frequency of Computer Use, %				P value
	Never or almost never	Some days	Most days	Everyday
Region					<.001
Northeast	16.8%	11.2%	6.0%	66.1%
Midwest	17.7%	9.4%	5.8%	67.1%
South	21.8%	10.7%	5.0%	62.5%
West	17.5%	10.6%	5.9%	66.0%
Age					<.001
≤20	8.9%	10.0%	7.4%	73.7%
21–40	10.7%	10.0%	5.5%	73.8%
41–60	15.9%	10.5%	5.2%	68.3%
61–80	29.7%	11.5%	6.0%	52.7%
>80	66.3%	9.2%	3.3%	21.3%
Race					<.001
White	18.4%	9.4%	5.6%	66.6%
Black	22.6%	16.8%	5.6%	55.0%
Asian	16.0%	9.6%	4.5%	69.8%
Other	31.6%	16.2%	5.4%	46.8%
Education					<.001
<Grade 12	1.3%	14.6%	2.4%	26.3%
High school graduate	0.7%	15.6%	6.0%	48.7%
Some college	11.6%	11.3%	6.9%	70.2%
Bachelors	4.8%	5.3%	5.6%	84.3%
Masters	3.6%	3.4%	5.0%	88.0%
Professional/doctorate	2.8%	2.2%	2.7%	92.3%
Family income					<.001
<PL	39.3%	16.0%	4.8%	39.9%
1–1.99x PL	34.9%	14.8%	4.9%	45.4%
2–3.99x PL	18.8%	11.5%	6.3%	63.4%
≥4x PL	7.0%	6.3%	5.6%	80.6%
Insurance					<.001
Uninsured	31.4%	15.7%	5.2%	47.7%
Medicare	38.6%	12.4%	5.6%	43.4%
Medicaid	29.9%	17.6%	6.3%	46.3%
Private	7.0%	7.5%	5.4%	80.1%
Military	12.0%	8.1%	7.9%	72.0%
Other	23.5%	15.0%	4.8%	56.8%
Gender					.299
Male	19.4%	10.6%	5.3%	64.8%
Female	18.7%	10.4%	5.8%	65.1%

PL, poverty level.

Table II

Sociodemographic factors associated with internet and email use

Factor	Internet Use, %	P value	Email Use, %	P value
Region		.002		<.001
Northeast	82.1%		73.9%
Midwest	83.7%		76.6%
South	79.9%		71.8%
West	83.1%		77.6%
Age		<.001		<.001
≤20	93.7%		83.5%
21–40	92.1%		85.5%
41–60	84.5%		77.0%
61–80	68.7%		61.1%
>80	32.0%		26.9%
Race, %		<.001		<.001
White	82.7%		75.4%
Black	76.8%		69.2%
Asian	84.7%		79.0%
Other	70.5%		59.5%
Education		<.001		<.001
<grade 12	46.8%		35.1%
High school graduate	72.2%		60.5%
Some college	88.7%		81.9%
Bachelors	94.9%		91.5%
Masters	96.0%		93.6%
Professional/doctorate	96.7%		95.2%
Family income		<.001		<.001
<PL	63.8%		53.9%
1–1.99x PL	68.3%		58.5%
2–3.99x PL	81.7%		73.3%
≥4x PL	92.5%		87.6%
Insurance		<.001		<.001
Uninsured	74.7%		63.4%
Medicare	59.9%		51.5%
Medicaid	72.2%		62.0%
Private	93.6%		88.0%
Military	88.6%		82.3%
Other	78.5%		67.3%
Gender		.136		<.001
Male	81.4%		73.3%
Female	82.3%		75.8%

PL, poverty level.

Table III

Sociodemographic factors associated with online healthcare-related activities within the past 12 months

Factor	Look Up Health Information, %	P value	Fill A Prescription, %	P value	Make AnAppointment, %	P value	Email WithProvider, %	P value
Region		<.001		<.001		.039		<.001
Northeast	57.2%		9.7%		16.7%		15.6%
Midwest	58.0%		13.1%		16.4%		16.3%
South	52.4%		10.7%		15.7%		14.5%
West	55.9%		12.5%		19.3%		21.1%
Age		<.001		<.001		<.001		<.001
≤20	51.2%		2.7%		13.2%		6.9%
21–40	62.8%		10.4%		20.6%		16.9%
41–60	58.5%		14.1%		18.1%		19.3%
61–80	46.7%		12.5%		12.7%		16.5%
>80	19.1%		4.5%		3.3%		5.9%
Race		.001		.001		<.001		<.001
White	57.1%		12.5%		17.2%		17.6%
Black	46.7%		6.6%		13.3%		11.3%
Asian	54.6%		9.7%		22.3%		19.1%
Other	41.1%		6.6%		8.2%		7.1%
Education		<.001		<.001		<.001		<.001
<Grade 12	22.3%		2.3%		3.7%		2.6%
High school graduate	40.8%		7.0%		8.5%		8.7%
Some college	59.1%		11.8%		16.3%		16.1%
Bachelors	72.7%		16.6%		26.0%		25.2%
Masters	76.7%		19.7%		32.3%		32.6%
Professional/doctorate	78.0%		19.2%		32.1%		35.1%
Family income		<.001		<.001		<.001		<.001
<PL	39.5%		4.9%		8.8%		6.7%
1–1.99x PL	42.1%		5.3%		7.7%		8.0%
2–3.99x PL	54.1%		10.2%		14.9%		13.8%
≥4x PL	66.3%		16.9%		24.3%		25.0%
Insurance		<.001		<.001		<.001		<.001
Uninsured	42.7%		3.7%		6.2%		4.7%
Medicare	40.7%		10.7%		10.8%		14.3%
Medicaid	45.4%		5.4%		9.0%		8.4%
Private	65.4%		14.0%		22.6%		21.2%
Military	61.0%		20.9%		21.5%		23.0%
Other	52.8%		7.3%		13.9%		9.1%
Gender		<.001		<.001		<.001		<.001
Male	49.4%		9.5%		14.3%		14.3%
Female	60.9%		13.3%		19.3%		18.9%

PL, poverty level.

Table IV

Multivariable analysis of sociodemographic factors associated with internet use

Factor	OR (95% CI)	P value
Region		.350
Northeast	Referent
Midwest	1.12 (0.93–1.35)
South	1.02 (0.85–1.23)
West	1.15 (0.95–1.39)
Age		<.001
≤20	Referent
21–40	0.45 (0.29–0.69)
41–60	0.16 (0.10–0.25)
61–80	0.07 (0.04–0.11)
>80	0.02 (0.01–0.03)
Race		<.001
White	Referent
Black	0.71 (0.60–0.84)
Asian	0.67 (0.54–0.83)
Other	0.63 (0.47–0.85)
Education		<.001
<Grade 12	Referent
High school graduate	2.70 (2.34–3.12)
Some college	7.32 (6.27–8.55)
Bachelors	14.84 (12.13–18.16)
Masters	20.17 (15.05–27.03)
Professional/doctorate	29.36 (18.97–45.44)
Family income		<.001
<PL	Referent
1–1.99x PL	1.31 (1.11–1.55)
2–3.99x PL	1.95 (1.64–2.32)
≥4x PL	3.39 (2.78–4.13)
Insurance:		<.001
Uninsured	Referent
Medicare	0.94 (0.76–1.16)
Medicaid	0.96 (0.77–1.20)
Private	1.97 (1.62–2.40)
Military	1.42 (0.85–2.39)
Other	1.03 (0.54–1.95)
Gender		<.001
Male	Referent
Female	1.40 (1.26–1.56)

CI, confidence level; OR, odds ratio; PL, poverty level.

Discussion

These data highlight systemic inequities in the population that could have a significant impact on the delivery of telehealth services. With the recent pandemic, there was a near-ubiquitous switch to healthcare delivery via telemedicine. Underpinning this, however, was the assumption that patients had access to a computer and to the internet. Our data from the 2018 NHIS demonstrate that a significant proportion of the US population was not well-equipped for this transition—with nearly 20% of people stating they “never or almost never” used a computer, and 18% who stated they did not use the internet. Older minority populations, who were of lower-income, less well-educated, and without insurance, were least likely to report using the internet, and these disparities persisted regardless of the geographic region of residence.

Our data do have some limitations. For example, the data we present come from the 2018 NHIS survey, arguably over a year before the pandemic, and there may be concerns that the pandemic caused a remarkable increase in the use of digital technology. While the Health Information National Trends Survey (HINTS), which was conducted between 2017 and 2019, did show an increase in utilization over this time period, there remained ∼15% of the population in their study in 2019 who still did not have internet access, mirroring our findings. Arguably, while we evaluated whether people “used” a computer or the internet, this does not necessarily imply “access”; in other words, it is possible that they did not use the internet before the pandemic but still had access to it, and were then able to avail themselves of telemedicine when forced to do so. However, it is likely that some of these individuals did not have access to devices or the internet at all, and this fueled some of the disparities that were seen. While the NHIS did not specifically ask about smartphone use, it did inquire about internet use, and presumably, those who used their smartphone to access the internet would have been captured in these questions. Still, the fact that the NHIS captured data over 25,000 individuals in a survey designed to be representative of the entire US civilian non-institutionalized population is a tremendous strength of the study and provides insight into the population’s use of technology in healthcare just before the pandemic and their readiness for the dramatic transition that occurred.

The concept that these populations had poor computer and internet access was not previously unknown. In a study of access to e-health tools from 1999−2002, Hsu et al demonstrated that while access and use of e-health services grew rapidly over these 3 years, a “digital divide” persisted among non-white patients and those of low socioeconomic status. Yang et al, using data from the HINTS study between 2017 and 2019, similarly found that patients on Medicaid were significantly less likely to report having access to the internet than non-Medicaid patients, although access increased for both groups across the 3 years of the study (69.7% vs 84.1%, P < .001 for 2017; 75.2% vs 86.8%, P < .001 for 2019). These data echo our findings that 72.2% of Medicaid recipients stated they had used the internet within the preceding 12 months.

Beyond computer utilization and internet use, we evaluated whether people engaged in telehealth activities, such as looking up health information on the internet, filling a prescription online, making an appointment online, or emailing with their doctor. Across all of these activities, we found disparities based on age, gender, race, and socioeconomic factors. Yang et al evaluated 3 similar composite variables: Online Patient-Provider Communication (OPPC), which included metrics of whether patients used electronic means to communicate with their doctor, Social Media for Health Information (SMHI), which included whether patients used the internet to participate in an online forum or support group or watch a health-related video on YouTube, and Buy or Refill Medication (BRM), which assessed whether patients bought or refilled prescriptions online. In 2018, they found that Medicaid recipients had a lower score for OPPC and BRM than non-Medicaid patients (P < .001 and P = .08 for OPPC and BRM, respectively) but a higher rate of SMHI (P = .01). In our study, we found that those who were uninsured or had Medicaid insurance were less likely to engage in any of these behaviors, although we analyzed multiple types of health insurance rather than the dichotomous variable of Medicaid versus non-Medicaid.

While we found significant disparities in terms of computer, internet, and telehealth use between racial groups, this has not been a universal finding. Kontos et al, for example, in their analysis of online adults, found no differences in the use of various telehealth services by race. On the other hand, Jacobs and Ellis found that Black patients were less likely to have at least 1 computer, tablet, or mobile device (57.4% vs 82.2% for Whites, P < .001), and that while telemedicine increased during the pandemic across all groups, the rate of rise was lowest among Black patients (P = .007).

Several studies found, similar to us, that those of lower socioeconomic status were less likely to use telehealth services. For example, Kontos et al found that those with a high school education were significantly less likely to communicate with a doctor or doctor’s office by email compared to those with a college degree (odds ratio [OR]: 0.46; 95% confidence interval [CI]: 0.29–0.72), and those earning less than $20,000 per year were roughly one-third as likely to have bought medicines or vitamins online compared to those earning more than $75,000 per year (OR: 0.34; 95% CI: 0.12–0.95). Chang et al found that those living in high social vulnerability index areas were significantly more likely to lack reliable internet or an appropriate device to engage in telehealth activities. While we were not able to assess the impact of rural versus urban location on the availability of telehealth resources, some authors have found that this is another significant factor impacting telemedicine use; others, however, have not found this to be the case.

The COVID-19 pandemic resulted in a dramatic increase in telemedicine visits, with some reporting a 20-fold increase in the same. While some found that this rapid expansion eliminated many pre-existing disparities, , others found that these were only exacerbated. , , Certainly, for some patients, for whom access to healthcare was difficult due to transportation issues, taking time off work, finding childcare, etc., access may have improved with the shift to telemedicine. However, for the most vulnerable populations who lack computer and internet access, the inability to see their physicians in a face-to-face manner due to the near-universal switch to telemedicine may have, in fact, worsened disparities in terms of access.

Clearly, there may have also been provider-based factors that may have played a role as well in terms of which were able to offer telemedicine options and the mechanisms for this (eg, by phone alone versus video), and patient perceptions regarding the security and confidentiality of telemedicine options may have played a role in uptake. There may be particular cultural groups for whom technology is verboten (eg, certain Amish groups), but we were unable to analyze this in the NHIS dataset.

In conclusion, the lessons learned from the pandemic have highlighted the socioeconomic disparities that affect health care provision and underscored the need in the modern era to be digitally connected. Efforts to expand telemedicine access and reimbursement, along with plans for a wider provision of broadband internet coverage, promise to help improve access, although it is unlikely that this alone will be sufficient. Our findings that nearly two-thirds of octogenarians and over one-third of people with incomes below twice the poverty line report “never or almost never” using a computer are significant. There have been some creative attempts to address these issues. For example, emergency physicians at the University Health Network in Toronto initiated a program called PHONE-CONNECT (Phones for Healthier Ontarians iN EDs–COVID Needs met by Cellular Telephone), in which they provided phones to patients as a point-of-care intervention to overcome the “digital health inequity.” Others have suggested “telemedicine booths” in community settings that may be able to improve access. While these creative solutions may be part of the answer to improving access, ultimately, it behooves us to be cognizant of the impact of sociodemographic factors on healthcare access–whether in person or through telemedicine—and proactively seek universal remedies, particularly for the most vulnerable among us.

Funding/Support

The authors have received no funding for this work.

Conflict of interest/Disclosure

The authors have no relevant conflicts of interest to disclose.