The remotely delivered workshops were offered throughout an 8-county region in upstate New York. The region is largely rural, with an average population density of 55 residents per square mile (ie, 21 residents/km²), compared to 238 residents per square mile (ie, 92 residents/km²) in New York, excluding New York City [27]. Similar to other rural areas in the northeast, the population is predominantly White, non-Hispanic. Nearly 30% of the general population is at least 60 years old.

Participants who enrolled in 1 of 2 workshop delivery modes (VC or CC) between April and December 2020 and consented to have their data deidentified and shared for research purposes are included in this analysis. To be considered “enrolled,” an individual had to complete the baseline survey and register and receive materials for the workshop.

The implementation, delivery, and evaluation of the CDSME workshops within our health care system’s service region was originally developed as a quality improvement project and was determined to be exempt from ongoing oversight by the Mary Imogene Bassett Hospital's institutional review board. Participants provided written consent to attend the workshop and to have their deidentified data collected for research and evaluation purposes. Participants received a US $30 gift card to a local grocery store as compensation for completing data collection activities. Individuals who did not consent to have their data collected and hence not included in the analyses, were still able to participate in the workshop.

Sociodemographic data (age, gender, education, marriage status, and insurance status) were collected either through paper or electronic questionnaires at baseline. In addition to sociodemographic questions, participants were asked to complete a single-item, self-rated health question [28]. All variables were dichotomized after data collection. Process measures included enrollment information (including workshop delivery mode), attendance, completion status, and reasons for not being able to attend all sessions of the workshop.

The analyses described below were used to answer the following questions: what are the characteristics of participants enrolled in each mode of remote workshop? What are the differences between completers and noncompleters across workshop delivery modes? What are the differences between completers and noncompleters within each delivery mode?

Descriptive statistics were used to characterize the entire sample, as well as for enrollment in each delivery mode. Bivariate analyses were conducted using chi-square tests of association to determine which characteristics were significant between groups for enrollment, an overall comparison of completers versus noncompleters, and which factors influence the likelihood of completion within each delivery mode. Based on these findings, multivariate analyses using logistic regression were conducted to explore which characteristics were independent predictors of enrolling in or completing a particular delivery mode. In all analyses, noncompleters included individuals who failed to show up for the workshop (eg, “no shows”) as well as those who showed up for fewer than 4 sessions.

Data were analyzed with SPSS (version 27; IBM Corp). Four outliers were identified in SPSS and were removed from the data set prior to analyses.

Between April and December 2020, a total of 183 individuals who consented to share their deidentified data enrolled in a VC or small group CC self-management program workshop (2 additional individuals enrolled in a workshop but did not consent to share their data). The majority of the participants across workshops were older (mean age 58.6 years, median 60.0 years), female (n=143, 78.1%), married or partnered (n=93, 52.5%), and self-rated their health as good or better (n=155, 85.6%). Participant demographic characteristics are summarized in Table 1. There were significant differences in age and education level among users of the 2 different delivery modes at enrollment. Those who enrolled in the VC workshop were 3.44 times more likely to be aged 60 years or younger and 2.24 times more likely to be a college graduate (Table 1). Gender, marriage status, self-reported health, and Medicaid status were not significant indicators of program mode enrollment (P=.19, .76, .45, and .11, respectively).

In the multivariate analysis, age and education remained significant predictors of enrollment. The logistic regression model was significant overall (χ²₆=22.3, P<.001). Those who were 60 years old or younger were significantly likelier to enroll in VC than those older than 60 years (odds ratio [OR] 3.16, 95% CI 6.26; P<.001). College graduates were also significantly likelier to enroll in VC than those without college education (OR 2.02, 95% CI 1.00-4.09; P=.05).

There were significant differences between program completers and noncompleters in terms of age, education, and self-reported health (Table 2). Completers likelier to be older than 60 years (OR 2.76, 95% CI 1.51-5.02), have a college education (OR 2.50, 95% CI 1.34-4.68), and report poorer general health (OR 4.60, 95% CI 1.65-12.81) than noncompleters.

These differences remained significant in the multivariate analyses. The logistic regression model was significant overall (χ²₆=22.5, P<.001). Workshop completers were likelier be older than 60 years (OR 3.10, 95% CI 1.53-6.31; P=.002), be college graduates (OR 2.52, 95% CI 1.22-5.21; P=.01), and report poorer health (OR 4.26, CI 1.30-13.99; P=.02). In addition, gender emerged as a significant predictor, with females being likelier to complete a workshop than males (OR 2.40, 95% CI 1.03-5.62; P=.04).

When the data were stratified by workshop type (CC and VC), the CC logistic regression model (χ²₆=16.7, P=.01) revealed that Medicaid status was the only independent predictor of completing a CC workshop. Those not enrolled in Medicaid were likelier to complete the program than those enrolled in Medicaid (OR 4.17, 95% CI 1.16-15.07). For the VC logistic regression model (χ²₆=24.8, P<.001), age, gender, education, and self-reported health remained independent predictors of program completion, in that, those who were older than 60 years (OR 5.56, 95% CI 1.81-17.10), female (OR 4.24, 95% CI 1.16-15.43), college graduates (OR 2.77, 95% CI 1.01-7.56), and those who reported poor or fair health (OR 7.00, 95% CI 1.44, 34.07) were likelier to complete the workshop (Table 3).

The purpose of these analyses was to gain a better understanding of the characteristics of those who enroll in and complete novel remote delivery modes of CDSME workshops that were developed in response to the onset of the COVID-19 pandemic. Our findings suggest that younger individuals were likelier to enroll in the VC platform than older individuals. However, older individuals were likelier to successfully complete any self-management workshop, regardless of delivery mode. When controlling for all variables within the multivariate models, not being enrolled in Medicaid significantly predicted CC workshop completion, while being older, female, a college graduate and rating one’s health as fair or poor were all independent predictors for completing a VC workshop.

The lower rate of digital technology use among older populations has been well documented [29], and the disparity has been highlighted during the COVID-19 pandemic [30,31]. However, of those who did enroll, those who were older than 60 years were likelier to complete the program, suggesting that once engaged with the remote delivery platform, older adults are able and willing to complete the workshops. Other researchers have demonstrated that self-management workshop completion rates are much higher among older cohorts [20]. Thus, while the VC delivery mode may initially engage those who are younger, we found it difficult to retain younger individuals in this program. A possible explanation for this finding is that younger and middle-aged adults are more likely engaged in the workforce and with family obligations and are hence unable to commit to a 6-week program, even if the program is offered in the evening [7,32]. This could have been particularly true during the first months of the pandemic when many parents suddenly found themselves essentially home-schooling their children. Another possible reason is that participants may be likelier to drop out if they have difficulty identifying with others in the group (eg, most participants are female, older, and retired) [33,34].

Self-management programs have typically struggled with recruiting and retaining men [35]. Our analyses are in agreement with nationwide studies, in that, nearly 80% of the participants were female. Women were also likelier to complete the program overall. However, it is interesting to note that there was no difference in completion rates between genders for the CC delivery mode. A possible explanation is that the CC mode allows for some level of anonymity; therefore, men may not feel they have to conform as much to masculine help-seeking behaviors. The CC group size is also smaller, and fewer participants per session have been shown to increase retention rates among men enrolled in the self-management programs [35].

The analyses revealed that education level was a significant factor for enrolling in and completing the VC workshop. This finding raises questions about the utility of the VC workshop for individuals who do not have a college education, particularly if one of the delivery modes demonstrates superior health-related outcomes in future randomized controlled trials. The availability of a non–internet-based remote option for those who either did not have internet access or who prefer not to use the internet is important from a health equity standpoint. However, at this time, it is unclear if the CC delivery mode has similar outcomes as the VC platform, particularly since it is shorter in time and attention.

Education level and Medicaid status are often used as proxies for socioeconomic status (SES) [36]. The digital divide is well documented with those of lower SES having less access to and lower usage of web-based services [37,38]; therefore, the abovementioned finding is not surprising. While a nationwide evaluation of the in-person CDSME workshops did not examine differences in education between completers and noncompleters, a Canadian study by McGowan found that education level was not a significant factor for completing in-person self-management workshops [13]. Our in-person workshop findings also indicated no significant difference in education level between completers and noncompleters [39,40].

A mixed methods investigation on access to, use of, and benefits from digital health services also found that being older and having less education affected access to digital health services, as well as living in a lower-SES area. Having less education was also associated with less use of digital health services [41]. However, the largest factors that influenced use were trust in digital health services, eHealth literacy, and confidence in using them. While further research is needed to understand the determinants of trusting digital health services, it is clear that a trusted provider’s recommendation to utilize the service may increase the level of trust in digital services, especially among underserved populations [41].

In this analysis, not being enrolled in Medicaid predicted completion of the program through the CC delivery mode, but did not predict program completion through the VC delivery mode or enrollment through either program delivery mode, which was an unexpected finding. Recently, investigators have attempted to understand the relationship between SES and engagement in self-management programs. A review by Hardman et al [24] on the moderating effect of SES on self-management interventions found that there was some influence of SES on attrition rates, but the lack of high-quality research made it difficult to draw conclusions. Hardman et al [24] recognized the heterogeneity of low-SES groups, and that appropriate interventions or recruitment methods depend largely on the context of that group.

A scoping review of scoping reviews also illustrated the complexities of engaging individuals in web-based health services, recognizing how an individual’s culture and perceived effectiveness of telehealth technologies is intertwined not only with digital literacy but also the social and structural determinants of health [42]. It is likely that there are other factors that we did not measure, which are stronger predictors of completion of the VC delivery mode (eg, digital literacy and trust in technology). From a health equity standpoint, we recommend further research to explore the complex relationships among education, insurance status, SES, the social and structural determinants of health, and remote workshop completion status.

Finally, it is important to recognize the effect that self-reported health status may have on the odds of workshop completion. The overwhelming majority of participants rated their health as good or better; nonetheless, over 80% of those reporting fair or poor health completed the program, compared to 47% of those reporting good or better health. While our sample size was small, leading to very large CIs, this finding does illustrate 2 important points. First, it was difficult to recruit those who rate their health as fair or poor. This finding is in line with that of other research indicating that individuals with poorer self-reported health perceive fewer benefits from digital health services [43]. However, our data show that once these individuals are engaged, they are likely to complete the program. Therefore, exploring ways to engage those who are facing poor health is critical. To do this, program implementers must acknowledge and account for the many social determinants of health, including cultural and environmental determinants, which can impact health outcomes [24,41,43,44].

There are several limitations that must be acknowledged in this analysis. First, the analysis was a service evaluation and was not based on an experimental design. Therefore, it is impossible to determine any causal relationships. Second, external constraints on program delivery dictated how often each mode of workshop was offered, which could have added additional bias into our findings. We also had a very small sample of individuals who reported less than good health, which may affect the accuracy of those findings. In addition, our sample population was from one particular region of New York, and the results are not generalizable to other areas. Finally, the evaluation took place during the first 10 months of the pandemic, which likely influenced health-seeking behaviors. As individuals have settled into the peripandemic era, remote workshop usage patterns may have shifted.

Despite these limitations, our analysis demonstrates strong signals that certain demographic groups are less likely to initially engage with web-based CDSME workshops, particularly those who are older and less educated. In addition, completion of web-based workshops is likelier for those who have a college education, as well as those who are older and female. The traditional CC workshop appeared more accessible to participants. As health care resources rapidly become more digitized, health educators and practitioners must remain aware of this potential disparity, and future work needs to focus on understanding the nuances underlying these disparities and how we can effectively work with our patients and communities to overcome them. Remote program offerings, including those that are independent or asynchronous, can remove barriers that prevent individuals with chronic disease from engaging with in-person self-management workshops, such as transportation issues, childcare concerns, fatigue, and pain [7,45]. However, equivalent, high-quality service that does not rely on a digital platform must continue to be offered to promote health care equity.