The realist evaluation methodology was used to gain a comprehensive understanding of why patients use or do not use web-based self-management support interventions [25]. We structured the analysis using the context-mechanism-outcome configuration to identify contextual factors (features of the conditions, eg, patient characteristics, that influence the intervention mechanisms) and potential mechanisms (what and how intervention components are responsible for change) that affect the intervention outcome (Figure 1). The identified mechanisms are described as “potential” as it was beyond the scope of this study to also test their effectiveness. In this study, we used data from 2 previous studies [14,16] that were approved by the medical ethics committee of Arnhem-Nijmegen in the Netherlands (No. 2015-1908 for CVD and 2014-1208 for RA).

Program use was a dichotomous outcome and was divided into nonusers (0 or 1 visit) and users (≥2 visits). The cut-off point between users and nonusers was arbitrarily set at 2 visits because this was seen as a reflection of whether a patient would have had the opportunity to benefit from the program.

We investigated the relations between context, mechanism, and outcome to determine which factors are associated with the use of a web-based self-management support program. Figure 2 summarizes the patient, disease, and program characteristics that influence program use.

A random sample of study participants was interviewed to gain insight into why they did or did not use the web-based self-management program. In the CVD group, 6 nonusers, 6 low users, and 6 high users were interviewed. In the RA group, 4 nonusers, 4 low users, and 13 high users were interviewed. The results were divided into 3 themes: program design, implementation strategies, and BCTs. Table 6 provides quotes that show the barriers and facilitators for program use on the 3 themes: program design, implementation strategies, and BCTs.

Most interviewees were pleased with the program design. However, some experienced difficulties in using the program, and so they did not use it as often. A search function would make it easier to find relevant information. Several users and nonusers stated that they had overlooked parts of the program; for example, 1 participant only used the diaries because he did not know that training modules were available. Another major reason for not using the program were problems with logging in. These observations indicate that ease of use was an important factor for program use among our respondents.

Explanations given by the respondents as to why they did or did not use the program also revealed factors affecting program use. Several respondents stated that they did not participate for their own benefit but rather to facilitate scientific research. Others used the program following advice from their health care professional or because they were curious and wanted to better understand their disease. The biweekly reminders to fill out the diaries in the Coping with RA program helped many respondents to use the diaries.

Comments related to program content were assigned to the relevant BCTs, and some of these BCTs were identified as potential mechanisms affecting program use. The first BCT (providing general information about health behavior) was often mentioned in the interviews. For example, respondents with a long disease history stated that the information was too general. Furthermore, some respondents saw on the overview page that none of the modules contained new or interesting information, and so they did not use the program further. Respondents reported that reliable information was a reason for using the program. The Vascular View program includes a physical activity and nutrition diary (for the self-monitoring of behavior BCT), which was rarely used. One respondent said they had missed a feedback function in the diaries and had already used other, more advanced, mobile apps instead. The pain and fatigue diaries in the Coping with RA program were used more often by respondents (for the self-report of behavior BCT). Patients appreciated the possibility of keeping track of their pain and fatigue and of receiving a graphical overview of their input (the delayed feedback of behavior BCT). Program users also liked the stories and videos of peers (which provided information about peer behavior BCT and modeling BCT). One respondent said that these made her feel recognized and supported and showed her that she was on the right track.

In this realistic evaluation of 2 web-based self-management interventions, we searched for patient, disease, and program characteristics that determine whether patients will use the programs. Regarding the relationship between context and outcome, patient and disease characteristics, younger age, and lower level of education were associated with program use. In addition, 4 quality of life subscales (social function, physical role limitations, vitality, and bodily pain) interacted significantly with the diagnosis group to affect program use. Regarding the relationship between Potential Mechanisms (program characteristics) and outcome, participants indicated that an easy-to-use, clear, and transparent design would motivate them to use the program. Email reminders and recommendations from health care professionals were found to be potential implementation mechanisms for promoting program use. The top five BCT techniques that encouraged interviewees to use the program were (1) tailored information, (2) self-report of behavior, (3) delayed feedback, (4) information about peer behavior, and (5) modeling.

Our findings show that patient and disease characteristics can be used to tailor web-based self-management interventions and, therefore, increase their use. Younger age increased program use in our study, which is in agreement with the results of previous studies. However, in contrast to our finding that a lower level of education increased program use, earlier studies showed that a higher level of education increased program use [18-20,32,33]. Despite this discrepancy, these results show that age and education both influence program use, possibly because they are both related to eHealth literacy. eHealth literacy is the ability to seek, find, understand, and appraise health information from electronic sources and to use this knowledge to address or solve a health problem [34]. Concerns have been raised about a digital divide, which is the gap between patients who are able to use eHealth and those who are not [19]. Our study emphasizes the need to pay attention to these issues, as both age and education are strongly related to eHealth literacy [35], and eHealth literacy is needed to benefit from web-based interventions. Different forms of self-management support should be provided to people with low eHealth literacy.

Disease burden can be both mental and physical and is another possible factor related to the use of web-based self-management support programs. Patients with RA have a lower physical quality of life and experience more pain than those with CVD. Individuals with episodic or deteriorating diseases such as RA have different self-management support needs than those with stable chronic diseases [36,37]. Patients with CVD have reported fewer self-management support needs than those with other chronic diseases because their disease as a smaller impact on their live [38]. These variations in the perceived burden of disease can affect the motivation to change. A higher perceived disease burden has been associated with a higher perception of the necessity for treatment, which increases adherence to treatment [39]. We have shown this in this study; in the RA group, users rated their physical quality of life as higher than nonusers, whereas in the CVD group, users rated their physical quality of life as lower than nonusers. This suggests that a certain level of burden is needed to feel urgency and to be motivated to use a web-based self-management support program. However, a web-based intervention might not be sufficient when the disease burden is too high. In such cases, face-to-face support from health care professionals is recommended [23].

The study participants provided some recommendations for an effective web-based self-management support program. These recommendations included being easy to use, providing appropriate reminders, tailoring information to the user, allowing patients to self-report their behavior and receive delayed feedback, and providing information about peer behavior and modeling. These results are in line with those of a Delphi study that identified new information and the possibility of monitoring personal progress as important factors promoting the use of an eHealth self-management intervention [40]. In addition, previous research has shown that peer support and email or phone contact increase the use of eHealth interventions [10]. These observations suggest that adding an interactive component to our Vascular View and Coping with RA programs, which allows users to communicate with peers and health care professionals, may promote program use. Counselor support has been found to be important for program use in previous studies [10], and our participants stated that interaction would have stimulated them to use the program. The role of health care professionals should never be underestimated, especially as blended care (a combination of eHealth interventions and face-to-face consultations with a health care professional) increases the use of eHealth interventions, including more program components [41].

Our results emphasize that one program will not be suitable for every patient. Self-management programs should be tailored to patients’ individual needs. It should also be noted that not all patients can use and benefit from web-based interventions. The validated Self-Management Screening (SeMaS) questionnaire can help identify potential barriers to self-management and can help health care professionals determine their patients’ support needs [42]. The factors affecting program use identified in this study were in accordance with the components of the SeMaS, including age, education, disease burden (both low and high disease burden can be barriers to self-management), computer skills, and social support. The SeMaS can help health care professionals to choose appropriate interventions and to decide which patients would benefit from a web-based self-management support intervention [43].

Given the complexity of web-based self-management interventions, realistic evaluations can reveal what makes an intervention work, which a simple cause-and-effect relationship between an intervention and its outcome may not be able to do. This is especially important for eHealth interventions because dropout and nonuse rates are high [17]. The aim of a realistic evaluation is to determine what works for whom, in what circumstances, and why. We tried to answer these questions by analyzing what patient and disease characteristics influence program use (context) and by describing what program characteristics influence program use (potential mechanisms; Figure 2). Unfortunately, we were not able to analyze the interaction between context and mechanism and how this affects the outcome. This should be addressed in future research to further improve the tailoring and effectiveness of eHealth interventions.

The findings of our realistic evaluation should be considered in the context of several limitations. The principal limitation was that we used retrospective data collected in 2 separate studies. However, both studies were conducted by the same research group and had the same study design. It was already decided in the development phase that the data would be merged for an overarching study; however, we could not include more questions about factors related to program use in the questionnaires. The interviews were conducted to retrieve patients’ experiences, not to identify program characteristics that influence program use. Vascular View and Coping with RA were developed based on BCTs, and most of these were unobtrusively included in the program. In addition, Vascular View and Coping with RA applied different implementation strategies that could have influenced program use, making the programs harder to compare. Therefore, the program characteristics identified in this study are potential mechanisms and should be tested in future research. Another limitation was that physical comorbidity and time since diagnosis were measured using a questionnaire. Although this provides insight into patients’ experiences, the self-reporting of clinical variables is not always reliable. The last limitation of this study was that we only included participants who had access to the internet and an email address. This may have biased our results by excluding people with a very low level of eHealth literacy. However, the internet is easily accessible in the Netherlands (97% of households have access to the internet [44]), so most people would have been able to participate.

This realistic evaluation identified contexts and potential mechanisms, in the form of patient, disease, and program characteristics, that are associated with the use of web-based self-management support programs. Our results emphasized the importance of (1) tailoring interventions to patients’ needs (depending on age, education, and program characteristics) to increase program use and (2) considering whether all patients can use eHealth interventions (depending on disease burden and eHealth literacy) and providing alternative self-management support when needed. These results are a first step toward improving the tailoring and use of web-based self-management support programs. Future research into the interaction between patient and program characteristics and how this affects program use should be conducted to improve the tailoring of participants to program components.