Chronic diseases contribute to 71% of deaths worldwide every year, which corresponds to 41 million deaths per year. It has been estimated that among these deaths, 15 million people between the ages of 30 and 69 years die mainly because of cardiovascular disease, cancer, chronic respiratory diseases, or diabetes [1]. Apart from mortality, the consequences of these chronic diseases include a decrease in the quality of life [2,3] and an economic burden for both households and countries [4-6]. The use of information and communication technology for health-related purposes has the potential to mitigate these consequences by offering numerous benefits for disease management, such as facilitating access to health information and helping to increase the understanding of the disease [7]. It is also considered a flexible, low-cost method for patients to obtain information in comparison with face-to-face education sessions [8]. Web-based interventions are an example of information and communication technology that has the potential to educate people living with a specific chronic disease condition and can help to improve their self-care over the long term through education and peer support [8,9]. These web-based interventions can be in a guided format by including features such as electronic counseling (e-counseling) and long-distance monitoring by health care professionals (HCPs) [10] or can be self-guided, defined in this paper as an absence of individual or face-to-face contact between HCPs and the users. Previous studies have investigated the integration of various features (eg, reminders and opportunities for social support) and the design of these web-based interventions. They concluded that the implementation of specific features and degree of adherence to the intervention are key factors in determining their success [11,12]. However, these studies do not distinguish between interventions with one-on-one or in-person contact among users with (guided) and without (self-guided) an HCP. As contact with HCPs or e-consultations can lead to a higher cost per usage and decrease the accessibility of the intervention [13], it is important to understand the inclusion of specific features and evaluation of adherence to these self-guided interventions.

The definition and measurement of adherence to self-guided interventions are still subject to debate [14,15]. Adherence is defined by the World Health Organization as “the extent to which a person’s behaviour – taking medication, following a diet, and/or executing lifestyle changes, corresponds with agreed recommendations from a health care provider” [16]. However, this definition is not adapted in the context of information and communication technology; there is no prescribed dosage that users of specific web-based interventions should be using to have the expected behavioral change [12]. The difficulty in defining adherence to web-based self-guided interventions is further accentuated by the differences in which they have been measured across studies with the use of parameters, such as the number of log-ins, the content viewed, and/or the time spent on the intervention [14].

A deeper understanding of previously published evaluations of self-guided educational interventions is required. The goals of this systematic review are to investigate how web-based intervention features are evaluated to determine their acceptability and to explore how adherence to web-based self-guided interventions are defined and measured. An understanding of the specific features and standardization of the definition of adherence to web-based self-guided interventions can help increase their efficacy and help to develop future web-based interventions for disease management.

To achieve these objectives, a systematic review of studies investigating the acceptability of the included features in web-based educational interventions on chronic health conditions was conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses framework [17]. For this review, chronic health conditions also include chronic diseases. Nine chronic health conditions were selected from a list of common chronic diseases in Canada [18]. The selection of these studies was related to the implication of a web-based educational intervention on patients’ self-management and their commonality across different age groups. Cancer and mental illness were excluded from this systematic review because of the broad variety of disease and treatment methods [19-21]. The selected categories were defined as follows: (1) arthritis, (2) celiac disease, (3) epilepsy, (4) inflammatory bowel disease (including Crohn disease and ulcerative colitis), (5) metabolic disorders (including hypertension, dyslipidemia, atherosclerosis, heart failure, gestational diabetes mellitus [GDM], and type 1 and type 2 diabetes mellitus), (6) multiple sclerosis, (7) overweight and obesity, (8) respiratory disease (including chronic respiratory disease, asthma, and chronic obstructive pulmonary disease [COPD]), and (9) kidney diseases (including end-stage renal disease and nephritis).

The search method was elaborated with the help of a librarian. The PubMed, Cochrane Library, and EMBASE (Excerpta Medica dataBASE) databases were used to ensure that all articles related to the topic were covered. Keywords (Textbox 1), derived from Medical Subject Headings (MeSH), were searched in the titles or abstracts. The search combined each medical condition with the web-based, education, and intervention terms. A full list of the search methods is included in Multimedia Appendix 1. If the clinical trial protocol was available, the corresponding author’s name and the study title were further searched on these databases to find the relevant publications. The reference lists of the selected articles were also screened to capture potential articles. The screening and selection of the articles were performed independently by 2 reviewers (LFX and AI), and consensus was reached through a discussion to ensure agreeability. A third researcher (ASB) was consulted for a nonunanimous discussion for the selection of the articles.

Inclusion criteria were as follows: (1) the study included a web-based educational intervention designed for people living with this health condition (eg, transfer of knowledge to this population), (2) the intervention aimed to improve clinical outcomes defined as the result of a health care intervention, which includes a change in clinical laboratory values (eg, level of blood glucose, blood lipid profile), lifestyle behavior (eg, improvement in eating habits and level of physical activity), use of health care system (eg, use of emergency department and length of hospitalization), and quality of life [22] related to an existing chronic health condition, (3) no in-person or one-to-one contact with an HCP within the intervention, (4) only contacted the research team for technical support or an introductory meeting during the intervention (to ensure the pragmatism of the study results [23] and limit the impact of these contacts on the adherence to the intervention), (5) the included population is ≥14 years old (age cutoff where people can make their own health care decisions in Quebec, Canada [24]), (6) the articles (published or in-press, to have a full portrait of the intervention and have peer-reviewed evidence) were published between January 1, 2005, and June 15, 2020, in English or in French, (7) the articles are fully available to the researchers, and (8) no restriction on the design of the study but only original research was included.

Studies corresponding to any of the following criteria were excluded from this systematic review: (1) the intervention is for family members or HCPs only, (2) the intervention has only a purpose of prevention/assessment/screening aftercare, (3) the web-based intervention included a live session or personalized e-counseling, (4) the intervention consisted of only emails, discussion forums, and/or recording functions, (5) the study explicitly stated an inclusion of participants with severe depression, and (6) the primary target outcome was related to mental health.

For each study, the following information was collected and compared: the year of publication, country where the study took place, study design, targeted chronic health conditions, primary clinical outcomes, age group of the population, sample size, intervention given to the experimental and control groups, and length of the intervention.

In this study, a feature is defined as any functionality within a web-based educational intervention other than text-based educational modules, supporting users to have a better learning or navigation experience or to improve clinical outcomes. The term feature and functionality are used interchangeably for this review. Both analyses of acceptability of the features and adherence to the intervention were based on reported information contained in the articles or the complete protocol cited from the selected articles. The method for evaluating features and their acceptability on the outcomes of the intervention are discussed. The measurement and criteria used to evaluate adherence to the intervention were collected and compared between studies.

All the data were collected from information within the articles, the related published supplementary documents, or the cited references. If >1 article reported the same intervention and outcomes but had different sample sizes, then articles stating results of the acceptability of the features or adherence to the intervention were reported. If none or all the articles reported these details, the latest publication was analyzed. However, information related to the acceptability of the features and adherence was collected from all related articles. If 2 interventions within the same study correspond to the inclusion criteria of this review, the intervention with the highest number of features was analyzed. The data from each study were then grouped into themes. EndNote X9.2 for Macintosh was used to regroup the articles.

The searches on the PubMed, Cochrane Library, and EMBASE databases resulted in 4091 potentially eligible articles (Figure 1). The titles and abstracts were reviewed, resulting in 390 articles. The titles and abstracts of potential articles from the reference list of the selected articles were also reviewed (n=34). After reading the full articles, 20 studies were selected.

The study characteristics are presented in Table 1. The identified articles included 6 areas of chronic health conditions: arthritis (n=1) [25], celiac disease (n=1) [26], metabolic disorders other than weight issues (n=8; metabolic syndrome [27], cardiac condition [28], hypertension [29], type 2 diabetes [30-33], and GDM [34]), multiple sclerosis (n=1) [35], overweight and obesity (n=7 studies) [13,36-41], and respiratory diseases (n=2; asthma [42] and COPD [43]). No study related to epilepsy, inflammatory bowel disease, or chronic kidney disease was found. The primary clinical outcomes were mainly related to changes in weight [13,36-41]. The studies were predominantly conducted in the United States [13,27,28,31,33,39-41,43] and Australia [25,26,34,36-38,42]. All the selected studies were randomized controlled trials, except for the study by Hutchesson et al [36], which was a pre-post design, and the study by Umapathy et al [25], which had a quasi-experimental design. All selected articles were published in English.

In the selected studies, 19 included an adult population (age 18 years) [13,25,27-34,37-43] and 1 included an adolescent/adult population aged 16 years [26]. The sample size varied from 22 to 456 participants. Seventeen studies included both genders [13,25-33,35,37,39-43]. The intervention length ranged from 8 weeks to 12 months, and in 1 article, the length was not specified [34].

The web-based interventions are summarized in Table 1.

The main features included in the web-based educational intervention and their acceptability are summarized in Table 2. None of these main features were identified in the studies by Noh et al [32] and Sainsury et al [26].

Only 8 studies (8/20, 40%) discussed the acceptability of the features. Acceptability was evaluated based on feedback from the users [33,36,38], their association with clinical outcomes [13,31,33,38,40,41,43], or device (eg, pedometer) adherence [43]. The features that reported positive, negative, or mixed acceptability in the studies are presented with a “+,” “−,” or “±” symbol in Table 2.

Adherence to the intervention was mentioned in 15 studies (75% of the eligible studies, 15/20) using different terms (eg, engagement, use of intervention, retention rate). The rate was reported in 4 studies. The parameters used to measure adherence to the intervention are summarized in Textbox 2.

A decrease in the use of the intervention throughout the study was observed in 6 studies [26,30,36,40,41,43]. For the length of a 16-week intervention, the percentage of log-ins in the study by Hansel et al [30] decreased by one-third in the final month. Moy et al [43] reported a similar decrease in the number of log-ins with time (from 6.8 per month in the first month to 3.0 per month at 12 months). A decrease in the use of the features was also observed, such as the number of opened newsletters [36], answered quizzes [36], and the use of the discussion forum [43]. A similar decrease in the frequency of monthly log-ins was observed in the study by Tate et al [40]. Although this decrease seemed to be progressive with time, Thomas et al [41] reported that it mainly occurred midintervention, 3 months from the beginning. Hutchesson et al [36] also observed that some features (eg, discussion forum and goal settings) had poor usage throughout the intervention and Morgan et al [38] reported that <50% of their participants complied with self-monitoring instructions. However, based on the general use of the intervention (eg, 7 weeks of submission of self-reporting data and weekly log-ins during the 3 months of the intervention), Morgan et al [38] qualified a retention rate of 41% as high. The term retention rate was also used by Sainsbury et al [26] and was measured with the use of the intervention. It was shown that 49.5% of the participants completed 4 of the 5 learning modules, but the authors considered this as a poor retention rate. Kessel et al [35] related the high dropout level (9/20, only 45% of the participants completed the intervention) to the absence of individual support, lack of feedback, and technical challenges. Bosak et al [27] explained that participants with better adherence had increased self-efficacy, but no additional information was provided.

This systematic review highlights the use of specific features in the design of web-based self-guided interventions for people with chronic health conditions and reports on the evaluation of their acceptability. Previous researchers have investigated the importance of features included in guided web-based interventions for people with chronic diseases on their success rate (eg, adherence to the intervention and transfer of health-related information) [11,12]. However, limited data were found on the functionalities of self-guided web-based educational interventions. In-person and one-on-one interactions with an HCP might increase the adherence and use of a web-based intervention [47] but that can also increase the cost of the intervention [13]. Therefore, it is important to investigate the characteristics of web-based interventions. This review demonstrated that goal setting, self-monitoring, and feedback were the most common features. The acceptability of the different features was measured based on the comments collected from users, their influence on clinical outcomes, or device (eg, pedometer) adherence. The use of personalized features with feedback (eg, quizzes) was positively reported. The negative acceptability of the features was mainly related to technical issues and the choice of discussion topics for the intervention. This review also showed that the evaluation of adherence to the intervention was inconsistent among the studies, which limited comparison. A clear definition and measurement of adherence to web-based interventions is lacking.

Our review identified 7 features that were most commonly included in the selected studies (Table 2). Other features such as the use of a pedometer, rewards, adaptation of the website intervention for smartphones, and technical support were also observed but less frequently used. On the basis of the results of this paper, we categorized the included features under the following 3 categories: personalization, interaction, and support. Personalization refers to a function tailored to the individual needs of each participant and can be changed throughout the intervention based on the user’s experience and progress [12]. Goal setting and self-monitoring have this characteristic by adjusting to the needs and progress of the user. The interactive features facilitated the engagement of the participants, increased learning retention [36], and provided a sense of community [31]. These characteristics were found in features such as quizzes, feedback, reminders, and online communities. They allowed an interaction between the intervention and participants and encouraged the users to return to the intervention [27,34,36]. Feedback and reward features correspond to both categories by personalizing the feedback report and varying the amount of rewards or type of written encouragement given to the participants based on the individual’s progress [31,36]. Other features not included in these 2 categories were providing support and reducing the technical barriers of the intervention.

Web-based educational interventions have been shown to be cost-effective compared with traditional face-to-face formats [48-51] and can reduce the production of physical materials (eg, printed documents) [52]. However, the cost related to the development of web-based educational interventions is still significant [52]. Creation of web-based educational modules can be classified into 3 levels: (1) basic content with text, graphics, simple audio, video, and test questions, (2) level 1 content with 25% interactive content (exercise, audio, video, and animations), and (3) level 2 content with highly interactive features (eg, adding game, avatars, custom interactions, and competitions) [53]. According to a study published in 2010, the average number of working hours to produce 1 hour of finished training associated with each of these levels is at least 79, 184, and 490 hours, respectively [53], and the average cost in US dollars is $10,054, $18,583, and $50,371, respectively [53]. Other factors such as the addition of new content and interactive features will further increase the cost [53]. Therefore, it is important to consider the choice of the features and their evaluation to minimize the cost and distribute the financial resources effectively. Our systematic review highlights that features are not frequently evaluated, with only 8 studies (8/20, 40%) reporting on the evaluation of some of the features used. In addition, the negative acceptability of a feature on the user’s experience, clinical outcomes, or device adherence was shown to be related to a lack of responding to the population’s needs, low human contact, and technical difficulties.

Our systematic review had some limitations. The search terms were selected based on MeSH terms; however, other important keywords could have been included. Exclusion of these important keywords might decrease the level of comprehensiveness of the search results. All the qualitative analyses were based on the content of the articles; the omission of information within the published articles might have led to a different interpretation of the results. For example, authors might only have listed the major features in their intervention instead of providing a complete list of all the available features. Only 8 studies (8/20, 40%) reported the acceptability of the features on the clinical outcome, users’ experience, or device adherence, which is a limitation for extrapolating the conclusions of the interventions. The articles included in this review were only selected from 3 databases, limited to published or in-press articles in English and French. In addition, to ensure a higher level of effectiveness in the results, this review also excluded self-guided interventions having individual contact between participants and research professionals during the study for reasons other than technical support or introductory sessions. Therefore, the results of this review might have limited external validity and cannot be applied to all web-based self-guided interventions or specific to any of the selected disease categories.

In conclusion, this systematic review investigated features included in 20 self-guided web-based educational interventions focusing on the self-management of chronic health conditions. It demonstrated the positive implication of specific features related to personalization and interactivity in the interventions on clinical outcomes, users’ experience, or device adherence. However, only a few studies reported the acceptability of the included features; therefore, future research is needed to gain a greater understanding of the roles that each feature plays on the use of web-based interventions. The results of this systematic review provide evidence on the choice and implementation of specific features for future web-based health education interventions, highlighting the importance of understanding the needs of the target population and the need to incorporate more human contact and reducing technical barriers for the effectiveness of self-guided web-based interventions. Moreover, this study also found poor consensus related to the definitions and measurements of adherence in self-guided interventions used to target chronic health conditions. A method for evaluating the level of adherence is proposed in this review but requires future studies for its validation.