Andersen’s Behavioral Model of Health Services Use has been widely used to analyze factors associated with health service use. The results suggest that health service usage is primarily motivated by individual illness, but the quality and quantity of health service usage vary significantly based on socioeconomic factors, such as income or health-insurance status [21-23]. The model posits 3 dimensions: predisposing factors (eg, age, education), enabling factors (eg, income, hospital density), and need factors (eg, health status) that influence health care usage [24-26]. Examining these dimensions may also be a suitable model when exploratory research is needed due to lack of previous studies on telehealth use among underserved patients with CLBP, as in this study.

This cross-sectional analysis was conducted on the baseline study measures data from the OPTIMUM trial, a multisite pragmatic randomized trial of telehealth-delivered MBSR for patients with CLBP. Data collection occurred between April 2021 and August 2023, with 12-month outcome data collection completed in November 2024. All variables were collected upon enrollment. OPTIMUM was embedded into primary care clinics at 3 health systems (Boston Medical Center, Massachusetts; University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania; and Piedmont Health Services, in partnership with the University of North Carolina [UNC] Chapel Hill). The 3 health systems serve predominantly underserved patients: Boston Medical Center is a safety-net institution serving an urban, low-income, racially and ethnically diverse population. UNC Chapel Hill has 2 sites: an academic center–based family medicine clinic and a federally qualified health center network serving many rural patients and those underrepresented in research. The University of Pittsburgh serves an urban or suburban population. Because we were interested in the factors involved in telehealth use experience for underserved populations, we restricted our analysis to the baseline measurements of participants from the safety-net health setting (Boston Medical Center) and the federally qualified health centers (UNC at Chapel Hill, Piedmont Health Services).

Research staff at each participating clinic approached potentially eligible patients during clinic visits, through telephone outreach, or via clinician referrals. Staff verified eligibility, provided study information, and obtained informed consent. Participants were screened to ensure they met inclusion criteria: age ≥18 years, diagnosis of CLBP, willing and able to provide telephone informed consent, able to speak and read English, absence of markers suggesting a serious condition, and a Pain, Enjoyment of Life and General Activity (PEG) scale score ≥3. Participants were excluded if they had an unexplained fever or weight loss, pregnancy, metastatic cancer, or cohabitation with another study participant.

Patients who met eligibility criteria were randomly assigned to medical group visits plus 90-minute, instructor-led MBSR treatment for 8 consecutive weeks via telehealth or to usual care alone. At the beginning of the MBSR session, for approximately 30 minutes, the primary care provider met briefly and privately with each participant through a telehealth visit (in a breakout room in the HIPAA [Health Insurance Portability and Accountability Act] compliant Zoom platform) to review medical aspects of treatment, focused specifically on the patient’s back pain. Participants were administered several measures at baseline, 8-week, 6-month, and the 12-month follow-up [27]. Further information about OPTIMUM is published elsewhere [27].

Because this analysis focused on understanding telehealth use experience among underserved populations, we restricted the analytic sample to participants enrolled from Boston Medical Center (safety-net setting) and UNC or Piedmont Health Services (federally qualified health centers).

Our outcome of interest in this paper was telehealth use experience as measured at enrollment, that is, telehealth use experience. Telehealth use experience includes the use of telehealth prior to or during the COVID-19 pandemic.

The online baseline data included measures of demographics, back-related pain, treatments, the impact of pain on quality of life, pain beliefs, medical history, health habits, and traumatic experiences. Dimensions of Andersen’s Behavioral Model of Health Services Use and variables are as follows:

This study is reported in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines (Checklist 1) [37] for cross-sectional studies. For descriptive analysis, we assessed the characteristics based on telehealth use experience (had previous telehealth visit or never had telehealth visit). Chi-square or Fisher exact tests were used for categorical variables, and Wilcoxon rank-sum tests were used for continuous variables.

Missing data were minimal (<5% for most variables). The low level of missingness did not warrant multiple imputation, which would have been unlikely to materially change the results. For continuous variables, histograms and Q-Q plots were used to assess normality. Nonnormally distributed variables were analyzed using nonparametric tests (Wilcoxon rank-sum). Categorical variables were analyzed using chi-square or Fisher exact tests depending on cell counts.

Bivariate analyses served as a screening step for multivariable modeling. A liberal threshold (α=.10) was used to avoid excluding potential predictors. Multivariable logistic regression retained variables with P<.05 in unadjusted models. Formal multiple-comparison correction was not applied because the primary goal was to build an adjusted model to identify factors associated with telehealth use experience, rather than to conduct multiple independent hypothesis tests.

The analytic sample was determined by available baseline data from the safety-net and federally qualified health center sites. These sites were selected based on the study objective of understanding telehealth use among underserved populations; thus, sample size was dictated by enrolled participants meeting eligibility criteria within these settings. We report both unadjusted odds ratios (ORs) and adjusted odds ratios (AORs) and 95% CIs as applicable. All analyses were conducted in SAS (version 9.4) with P<.05 considered statistically significant unless otherwise specified.

All study procedures were approved by the single Institutional Review Board at the University of Pittsburgh (STUDY20110378). All participants provided informed consent before the study and agreed to the use of their data for research purposes, including secondary analyses. Participation in the OPTIMUM study was voluntary, and participants could decline without consequences. To protect participants’ privacy and confidentiality, only anonymized data were used for this analysis, and results are reported in aggregate so that no individual participant can be identified. Participants were compensated US $30 for completing study measures administered at baseline and follow-up.

This cross-sectional analysis examined the individual- and county-level social vulnerability characteristics of those who had and those who never had a telehealth visit at enrollment among participants from a safety-net hospital and a network of federally qualified health centers in the OPTIMUM trial. We found that 174 of 244 (71.3%) participants had prior experience with telehealth. While a large portion of our sample had previously used telehealth, there were many who had not, thus indicating that while telehealth may be well received by many patients, it remains more accessible to certain groups of patients than others [17,38].

Similar to previous research, we found that females reported a higher experience with telehealth usage compared to males [19]. Our analysis indicated that compared to non-Hispanic White participants, non-Hispanic Black and Hispanic participants had lower odds of telehealth use experience. Given the range of health disparities across the United States, it is reasonable to assume that not all people benefit from using telehealth. In the pre–COVID-19 period, telehealth visits were significantly lower among Hispanic patients than among non-Hispanic White patients and non-Hispanic Black patients [39]. Although telehealth use increased among these groups during the COVID-19 period, disparities persist with reliable internet connectivity, digital literacy, and user comfort with technology among these populations [40,41].

In the adjusted models, we did not find a statistically significant association in our data between internet access and telehealth use experience, which may be explained by factors unique to our sample, such as high number of participants with internet access (228/244, 93.4%) and a large portion of our sample (213/244, 87.3%) having a high school degree or higher, which may have impacted significance. Previous research has found that while internet access is a foundational requirement for telehealth, it is not the only or even the most significant factor determining telehealth use when compared to, or in combination with, education, employment status, and digital literacy [42,43].

Unsurprisingly, lower comfort with Zoom and other virtual meeting platforms was associated with markedly lower odds of telehealth use experience. Prior research has found that the strongest predictor for telehealth use among participants was their comfort with technology rather than access alone [44]. Additionally, we found that in comparison to a high school degree, having less than a high school degree or more than a high school degree was associated with greater odds of previous telehealth use after adjustment. Our findings add to the mixed results of previous studies, which have found either no differences associated with telehealth use or that lower education and income were associated with lower telehealth use [45,46].

Our findings indicated that participants who had 5 or more years of CLBP had higher telehealth use experience, which may have influenced their desire to participate in the telehealth trial. Previous research has shown that patients with chronic pain are more likely to use telehealth because it offers easy access to specialists and reduces the amount of time taken off work to travel to an appointment [47-49].

We found no statistically significant association between ADI, SVI ranking, and the odds of previous telehealth use. Characteristics of our sample may explain possible reasons why ADI and SVI ranking was not significantly associated with previous telehealth. Among 244 participants, a large portion of our sample (n=213, 87.3%) had a high school degree or higher, lived in the metro area (n=222, 90.9%), and had access to the internet at home (n=228, 93.4%). Other research on the association between ADI, SVI, and telehealth usage shows mixed results. While some studies link higher disadvantage (higher ADI or SVI) to lower telehealth usage, others report no significant association between these indices and telehealth, suggesting factors such as age or technology access may play a larger role [50-52].

Our study has several strengths and limitations. The strengths of our study include a large sample size of an understudied population with CLBP. Additional strengths include the use of reliable and valid measurements. A limitation of this study is that all participants voluntarily enrolled in a telehealth trial and were required to have internet access in some form, which may indicate a higher comfort level with technology and limit the generalizability of the findings. Additionally, participants in our sample were more likely to receive a telehealth visit because they lived in communities with higher telehealth use (eg, urban communities served by large health systems).

Another limitation was the comfort with Zoom or virtual meeting platforms measure, which captured participants’ confidence in using these platforms. The measure did not discern whether participants had adequate digital literacy skills, just if they felt confident using those platforms. Other limitations include the following: (1) the cross-sectional nature of the study design, which precludes any ability to draw causal inferences; (2) self-reported variables, introducing the likelihood of recall bias; and (3) the recruitment of most participants from metro areas. Prior research has found that while both metro and rural underserved populations own and use digital health technology, rural residents are less likely to communicate with their health providers using that technology [53].

The results of our study have important implications. First, given the high economic burden of CLBP and the growing emphasis on evidence-based treatments such as MBSR, telehealth offers a promising alternative for medically underserved populations. Factors that served as enablers included home internet access and greater comfort with Zoom. Targeted outreach and culturally tailored engagement strategies are needed to address disparities related to race or ethnicity and employment status. Policymakers and health care stakeholders must take proactive measures to mitigate these disparities, enhance access, and ensure that telehealth fulfills its potential to improve outcomes for all patients with CLBP.

In a safety-net and federally qualified health center context, prior telehealth experience was common among patients with CLBP willing to enroll in a telehealth-delivered MBSR trial. Although certain predisposing factors (race or ethnicity, employment status) were associated with lower odds of telehealth experience, these factors did not prevent trial enrollment. Thus, targeted outreach to improve digital comfort and inclusive engagement strategies may further reduce disparities and support broader implementation of telehealth for CLBP.