This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
Despite considerable efforts to encourage telehealth use during the COVID-19 pandemic, we witnessed a potential widening of health inequities that may continue to plague the US health care system unless we mitigate modifiable risk factors.
This study aimed to examine the hypothesis that there are systemic differences in telehealth usage among people who live at or below 200% of the federal poverty level. Factors that we consider are age, gender, race, ethnicity, education, employment status, household size, and income.
A retrospective observational study was performed using the COVID-19 Research Database to analyze factors contributing to telehealth inequities. The study period ranged from March 2020 to April 2021. The Office Ally database provided US claims data from 100 million unique patients and 3.4 billion claims. The Analytics IQ PeopleCore Consumer database is nationally representative of 242.5 million US adults aged 19 years and older. We analyzed medical claims to investigate the influence of demographic and socioeconomic factors on telehealth usage among the low-income racial and ethnic minority populations. We conducted a multiple logistic regression analysis to determine the odds of patients in diverse groups using telehealth during the study period.
Among 2,850,831 unique patients, nearly 60% of them were female, 75% of them had a high school education or less, 49% of them were unemployed, and 62% of them identified as non-Hispanic White. Our results suggest that 9.84% of the patients had ≥1 telehealth claims during the study period. Asian (odds ratio [OR] 1.569, 95% CI 1.528-1.611,
Factors that impact telehealth usage include age, gender, race, education, employment status, and income. While low-income racial and ethnic minority communities are at greater risk for health inequities among this group, Hispanic communities are more likely to use telehealth, and non-Hispanic Black patients continue to demonstrate telehealth inequity. Gender, age, and household income contribute to health inequities across gradients of poverty. Strategies to improve health use should consider characteristics of subgroups, as people do not experience poverty equally.
The COVID-19 pandemic offered a glimpse of what could occur if inequities in telehealth usage are not alleviated. The global health emergency led to significant actions by federal and state agencies to mitigate the spread of the virulent contagion. Simultaneously, there were considerable efforts to provide safe access to needed health care services, while minimizing in-person contact among health providers and patients. Public health officials supported measures to decrease telehealth restrictions and increase reimbursement for telehealth services such as store and forward services, remote patient monitoring, and audio only (telephone) services. The Centers for Disease Control and Prevention estimated that telehealth visits increased by 154% in March 2020 when compared to the same time frame in 2019 [
Poverty has a significant impact on health access, usage, and outcomes. The intersectionality of poverty and race magnifies health inequities in the health care and public health systems [
While many studies examine people across socioeconomic statuses and make conclusions about people who live in poverty, this study investigates telehealth use only among people who experience poverty, as these experiences are not equal. In this study, we use federal-level guidelines to conceptualize poverty [
The COVID-19 Research Database Consortium provided data for the study. The consortium, facilitated by Datavant, is a private and public partnership across industries in the United States to facilitate data sharing and promote public health research. The Consortium provided access to Office Ally and Analytics IQ PeopleCore Consumer linked databases. The Office Ally database provided deidentified US claims data from 100 million unique patients and 3.4 billion medical claims. The Analytics IQ PeopleCore consumer database is a nationally representative database of 242.5 million US adults aged 19 years and older. Analytics IQ PeopleCore Consumer data provided deidentified patient-level data including health characteristics, medical care, and social determinants of health to help decision makers better understand their patients. With the linked identifiers (common tokens) provided by COVID-19 Research Database, we combined the Office Ally claims data with Analytics IQ PeopleCore Consumer data, which enabled us to retrieve patient-related information and examine telehealth usage across demographic and socioeconomic indicators. Telehealth in this study was defined as a range of web-based communications including remote monitoring, telephone calls, and videoconferencing.
The COVID-19 Research Database was established in compliance with regulatory standards to protect patient privacy. The COVID-19 Research Database received a waiver of patient consent by the Western Institutional Review Board for the use of Health Insurance Portability and Accountability Act (HIPAA)–certified deidentified data on April 20, 2020. Exemption status was granted by the Western Institutional Review Board for HIPAA-limited data sets and non–HIPAA-covered data on May 14, 2020. This exemption covers all research performed in the COVID-19 Research Database. In addition, researchers with approved study proposals are granted access only to specific data sets that are necessary to answer their research questions. Only deidentified and limited data sets are made available through the database and certified before access was granted. Individual project institutional board approval was not needed.
The study period was from March 2020 to April 2021. To investigate telehealth usage in low-income populations, data were retrieved from claim records of 2,850,831 patients whose household incomes were at or below 200% of the federal poverty level. Telehealth claims were identified by screening for current procedural terminology modifier codes 95, GT, and GQ. The current procedural terminology is a medical code set that uses a uniform language for coding and reporting health services and medical procedures. The modifiers 95, GT, and GQ supplement claim forms by adding extra information about the services provided. In this case, these codes informed us that the services were delivered via telehealth.
The data were aggregated at the patient level to investigate telehealth usage; that is, whether a patient used telehealth during the study period. A patient with ≥1 telehealth claims during the study period was assigned a value of 1 to the dependent variable, otherwise 0. R software (The R Foundation) was used for the analysis. A multiple logistic regression analysis was used to determine the odds of using telehealth among patients in different subgroups during the study period. Categorical variables were created to divide patients into groups by demographic and socioeconomic characteristics. The total number of claims of each patient during the study period was included in the logistic regression analysis to control its potential impact on the dependent variable—telehealth usage. A
We analyzed 2,850,831 unique patients and their claim records. The results indicate that among patients in low-income positions, 9.84% of them had ≥1 telehealth claim during the study period. In comparison, among patients whose incomes are above the low-income levels (200% of the federal poverty level), 12.86% of them had ≥1 telehealth claim during the study period. The total number of claims of each patient during the study period ranged from 1 to 16 (mean 3.75, SD 3.59). Nearly 60% of participants were female, 75% of them had a high school education or less, 49% of them were unemployed, and 62% of them identified as non-Hispanic White. Patient characteristics are summarized in
Description of the patients in this study (N=2,850,831).
Characteristic | Patients, n (%) | |
|
||
|
No | 2,570,252 (90.16) |
|
Yes | 280,579 (9.84) |
|
||
|
Female | 1,696,378 (59.50) |
|
Male | 1,154,453 (40.50) |
|
||
|
≥65 | 954,908 (33.50) |
|
45-64 | 1,014,759 (35.60) |
|
18-44 | 881,164 (30.91) |
|
||
|
High (bachelor’s degree or higher) | 712,750 (25.00) |
|
Low (high school or less) | 2,138,081 (75.00) |
|
||
|
Unemployed | 1,395,440 (48.95) |
|
Part-time | 667,659 (23.42) |
|
Full-time | 787,732 (27.63) |
|
||
|
Non-Hispanic White | 1,768,493 (62.03) |
|
Asian | 55,161 (1.93) |
|
Non-Hispanic Black | 422,415 (14.82) |
|
Hispanic | 579,641 (20.33) |
|
Other | 25,121 (0.88) |
As shown in
We carried out a simple linear regression analysis to investigate the impact of household income on telehealth use within each household size. The dependent variable is the percentage of patients who used telehealth at each income level, and the independent variable is income level. Our linear regression analysis of the patients by household size suggests that within the low-income population, income is a contributor to telehealth usage in households of >2 people. Among 1-2–person households, the association of household income and telehealth usage is insignificant (
Summary of odds ratios from logistic regression analysis.
Characteristic | Odds ratio (95% CI) | ||
|
|||
|
45-64 | 1.123 (1.108-1.138) | <.001 |
|
18-44 | 1.324 (1.304-1.345) | <.001 |
|
|||
|
Male | 0.875 (0.867-0.883) | <.001 |
|
|||
|
High school or less | 0.953 (0.944-0.962) | <.001 |
|
Part-time | 1.067 (1.053-1.081) | <.001 |
|
Full-time | 1.148 (1.133-1.164) | <.001 |
|
Asian | 1.569 (1.528-1.611) | <.001 |
|
Non-Hispanic Black | 0.994 (0.981-1.006) | .32 |
|
Hispanic | 1.612 (1.596-1.628) | <.001 |
|
Other | 1.296 (1.242-1.352) | <.001 |
Summary of the percentage of patients having telehealth visits by household income.
Our study examined telehealth use among people whose household incomes were at or below 200% of the federal poverty level, and we found that 9.84% of the study sample used telehealth services. Previous studies have examined mean household income across all income levels and found a positive relationship between income and telehealth use [
Among people in low-income positions, there were racial and ethnic differences, but they were different from those reported in studies that examine income more broadly. When all income levels are considered, studies suggested that telehealth use among non-Hispanic White patients was greater than that in non-Hispanic Black and Hispanic patients [
The influence of age, gender, and education level on telehealth use is similar across gradients of poverty. Individuals older than 60 years or men are less likely to use telehealth services [
Enhancement of digital inclusion supports a reduction in inequities by addressing issues that are specific to subgroups of people [
While this study is not an exhaustive examination of factors that influenced telehealth use, we did consider key factors that may contribute to inequities in usage. The study used the COVID-19 Research Database and is subject to the limitations of administrative databases. In the Office Ally database, the validity of the data is reliant upon the facilities to report accurate data and code visits correctly. The Analytic IQ PeopleCore Consumer database relies on the accuracy of consumer reporting. This study did not consider contextual factors such as the availability of providers who used telehealth, residential segregation, and the lack of a racial and ethnic minority workforce. Future studies should consider these factors and the variety of cultural perspectives in communities. Focus groups of patients and providers in these and other communities may help explore additional information not captured in surveys and claims data that explicate attitudes and challenges with telehealth access and use. Future work could parse out the influence of sociodemographic characteristics on the type of visits used by this population. Such information could be used to develop community-specific programs that facilitate telehealth access either through education or access to technology equipment.
Our study concludes that among people whose incomes are below the federal poverty threshold, Hispanic and Asian patients were more likely to use telehealth than non-Hispanic White and -Black patients. Patients who are employed full-time, female, aged between 18 and 44 years, and had completed a bachelor’s degree were more likely to use telehealth. Income is positively associated with telehealth usage in 3- to 10-person households. As we seek to promote telehealth usage, it is imperative that we consider the socioeconomic and demographic factors among subgroups of people who experience poverty. Due to the long-standing challenges in the US health care system, inequities have the potential to become entrenched in our society unless we take decisive action to address these challenges. By focusing on communities in low-income positions, we provide professionals and decision makers with additional insight to promote public health in an increasingly digital society. The tragic events of COVID-19 the pandemic show us that we need to bolster the public health infrastructure and take a more meaningful and targeted approach to health equity concerns.
Health Insurance Portability and Accountability Act
odds ratio
We thank the COVID-19 Research Database Consortium for access to and their support for the study. This research is also supported by the Bill and Melinda Gates Foundation. However, the contents of this paper are solely the responsibility of the authors and do not necessarily represent the official views of the Bill and Melinda Gates Foundation and the COVID-19 Research Database Consortium.
The data sets generated and/or analyzed during the current study are not publicly available due to the Consortium’s restrictions and governance policies. The database can be accessed by academic, scientific, and medical researchers from its website [
CW contributed to the conception and the design of the study, interpretation of the results, and drafting and revising of the manuscript. DS contributed to the design of the study, and interpretation, analysis, and revision of the data. All authors approved the final version of the manuscript.
None declared.