A structured electronic database search of PubMed, PsycINFO, and Web of Science was conducted for Internet intervention studies published from January 1, 1995, through February 8, 2009. The following search terms were used: "Internet" or "Web" or "online" and "health promotion" or "health education" or "health communication" or "health planning" or "prevention" or "intervention" or "behavio* change" or "behavio* modification." The search was limited to the interventions among adults (18 years and older) and English-language peer-reviewed publications. This search strategy was optimized for all consulted databases.

A study was eligible for inclusion if it described an Internet intervention that aimed at the primary prevention of physical chronic diseases among the general public from the age of 18. Relevant behaviors included physical activity, nutrition, weight management, smoking cessation, alcohol consumption, or a combination of these behaviors. Furthermore, the Internet interventions needed to be developed for use among the general public. Next, objective quantitative exposure measures (eg, number of log-ins, number of pages visited, completion of the entire intervention or parts of the intervention, time spent on the intervention website, number of visits to the intervention) needed to be reported. Finally, studies evaluating an intervention in a randomized controlled trial (RCT), a quasi-experimental design, or describing use of an intervention only in a single group study could be included.

The selection of studies took place in 3 phases based on title (author WB), abstract (authors WB and WK), and full publication (WB and WK). Title and abstract screening were done blinded for author, journal, and date of publication. If in doubt about suitability of a study in one phase, the study was included in the next phase. Disagreements on inclusion in the third phase were discussed with a third reviewer (author AO) until consensus was reached.

Data from the included studies were extracted by a team of reviewers and then verified and tabulated for this review by WB, WK, and AO. Based on a standardized extraction form, descriptive key elements and objective exposure measures of all included studies were summarized and presented in tables (Multimedia Appendix 1 and Multimedia Appendix 2). For this extraction, we relied on the information about the study and intervention provided in the published literature (ie, the selected publication, publications that evaluated and reported on the same intervention [see Table 1 for applicable studies], and references to additional design papers or appendices).

Potential exposure-increasing methods and strategies that have been found to be effective in previous studies were divided into the following categories: (1) interactive behavior change strategies, which include methods and strategies delivered in an interactive format (eg, tailored feedback, goal setting tools, action planning tools, or self-monitoring tools); (2) interactive elements, which include elements of the program that are more for fun to improve the attractiveness of the intervention or to provide the option for more information (eg, quizzes, searchable databases, or audio/video); (3) peer support (eg, forum, bulletin board, or chat); (4) counselor support (eg, ask-the-expert, email/phone contact, or counselor-led chat sessions); (5) email and/or phone contact, which may include email/phone messages providing intervention content (eg, personal feedback or newsletters) or email/phone prompts to remind users to revisit the intervention; (6) update of the information on the intervention website, which include, for example, new tools, information, or news; (7) intervention incentives, which refer to incentives that are related to using the Internet intervention and not related to taking part in a study.

For consistency and comparability among studies, the taxonomy of Abraham and Michie was used for the description of interactive behavior change strategies [34]. Within computer-tailored feedback, various types of feedback can be distinguished, such as feedback on performance, cognitive constructs, barrier identification and solutions, and cognitive and behavioral processes. In this study, we considered tailored feedback as one interactive behavior change strategy. Feedback on progress was included separately as this kind of feedback can only be given during a revisit.

Due to the significant heterogeneity between the studies and the variation in the reported exposure measures, the data could not be pooled for quantitative analysis. Therefore, qualitative, systematic, descriptive analyses were performed. This method has been proven to be suitable for systematic reviews [35].

To gain insight into which intervention characteristics may result in better exposure, the studies were listed in a matrix, linking the potential exposure-promoting intervention elements with the outcome measures (Tables 3 and 4). The objective exposure measures used in the different studies were very diverse and presented in different statistics (see Multimedia Appendix 1). Therefore, only those objective exposure measures that are used frequently and presented in the same statistic value are presented in Tables 3 and 4. In Tables 3 and 4, a division was made between interventions that offered fewer than 3 versus more than or equal to 3 interactive behavior change strategies, and that offered interactive elements (yes vs no), peer support (yes vs no), counselor support (yes vs no), email/phone contact (yes vs no), update of the intervention website (yes vs no), and intervention incentive (yes vs no). From this matrix, patterns could emerge indicating that the existence of certain intervention characteristics could result in more exposure to the intervention. Criteria for determining that an exposure-promoting element is probably related to an exposure outcome were that at least 50% of the Internet interventions that included the specific exposure-promoting element should be in the highest exposure category and that the number of studies in the highest category differed substantially (at least 35% difference) from the number of interventions without that element in the highest category of exposure. Only when there was a good balance in the number of interventions that did or did not have a specific exposure-promoting element, inferences about a relation between exposure promoting elements and exposure could be made.

The initial cross-database search yielded 7764 unique publications (Figure 1). After reviewing titles, abstracts, and full publications, 70 publications describing 64 studies were eligible for inclusion in the review (see Table 1). In total, 192 publications were excluded based on abstract and full publication. The most common reason for exclusion in this phase was that a publication did not describe an Internet intervention aimed at the primary prevention of physical chronic diseases (n = 112). Other publications were excluded because they focused on persons below the age of 18 (n = 11), were not targeted at the general public as end users (n = 3), or did not describe the evaluation of an Internet intervention (n = 37). Finally, 29 publications were excluded, as they did not include objective exposure outcome measures.

Of the 64 included studies, 39 were performed in the United States, 6 in Australia, 6 in the United Kingdom and Ireland, 4 in the Netherlands, 2 in Belgium, 2 in Canada, 2 in Norway, and 1 in Switzerland. In all, 12 studies described in 14 publications (hereafter, the number of publications referenced may exceed the number of studies to which they refer) targeted physical activity [16,28,36-47], 4 targeted nutrition (eg fruit, vegetable, or saturated fat consumption) [48-53], 14 targeted weight management (eg, weight loss/reduction or weight maintenance/control) [21,54-66], 18 targeted smoking cessation [22,31,32,67-82], 9 targeted alcohol reduction [83-92], and 7 targeted multiple behaviors [23,93-98]. Most studies had an RCT design and 14 studies were observational one-group studies evaluating use of the Internet intervention. The length of the interventions varied from a one-time visit to 18 months with multiple visits. The majority of the Internet interventions were explicitly informed by one or more behavioral theories. The social cognitive theory [99], the transtheoretical model in total [100], or the stages of change concept from this model only [100] were used most often. A more detailed description of the study characteristics can be found Multimedia Appendix 2.

Multimedia Appendix 2 shows that the number of study participants ranged from 32 to 67,324 with an overall mean of 3367 participants and a median of 408. The mean age varied from 32 to 52 years and the percentage of female participants ranging from 2% to 100%. The percentage of participants with education at a level higher than high school (if reported) varied from 41% to 100%.

Table 2 lists the potential exposure-improving methods and strategies used in the interventions. If two or more Internet interventions were described in one publication, the most extended intervention or the intervention that delivered the content mostly through the Internet is taken into account.

Multimedia Appendix 1 provides a more detailed description of the methods and strategies applied in each Internet intervention.

The provision of tailored feedback (eg, on performance, cognitive constructs, barrier identification and solutions, and cognitive and behavioral processes) was the most often used behavior change strategy across the behaviors except for nutrition and weight management interventions. Goal setting was offered more often in physical activity interventions; action/activity planning was most often used in the physical activity and smoking cessation interventions and self-monitoring in the physical activity and weight management interventions. Feedback on progress was most often used in the multiple behavior interventions, followed by physical activity. The majority of the interventions in all behavioral domains included interactive elements such as quizzes, searchable databases or libraries, heart rate/BMI calculator, and website links, with less use of these elements in weight management and smoking cessation interventions. Peer support was most often used in the weight management, smoking cessation, and alcohol consumption interventions, while counselor support was most common in the weight management interventions, followed by the smoking cessation interventions. Email/phone contact was frequently used in most interventions except for the alcohol consumption and multiple behavior interventions. Regular updates of the intervention website or provision of an incentive for using the intervention were not often used, but when they were, they were used most in the weight management, nutrition and PA interventions.

A large variety of exposure measures were used in the included studies (see Multimedia Appendix 1). The frequency of visits by means of log-in rates was the most commonly used exposure outcome measure (n = 33) although the way in which the data were presented was not consistent across studies as different statistics were used (eg, mean or median). There were also several studies that did not present log-in rates but did present the percentage of users that revisited the intervention (n = 9). Other often used outcome measures were how many people landed on the website, which was mostly registered by “hits” on the website (n = 10), the number of visitors that accessed the program content (n = 24), the number of pages visited (n = 6), completion of the first visit or module (n = 13), and completion of the whole intervention (n = 8). Furthermore, use of intervention methods and/or strategies were also presented as exposure measures, such as use of specific intervention components (interactive behavior change strategies and interactive elements [n = 26], use of peer support [n = 12], and use of counselor support [n = 10]).

In Tables 3 and 4, the studies are listed in matrices combining the objective outcome measures that were mostly presented and the potential exposure-promoting elements. Of all the potential exposure-promoting elements listed in Tables 3 and 4, indications were found for peer support, counselor support, email and/or phone contact with visitors, and updates of the intervention website to be related with more exposure. The provision of peer and counselor support appears to have had a positive influence on the time visitors spent on the website. This can be deduced from the finding that at least 50% of the studies evaluating interventions that included peer or counselor support were listed in the higher category of average time spent on the website compared with the lower percentage of studies evaluating interventions that did not include peer or counselor support, and that the difference in number of interventions listed in the higher category was at least 35%. Both email/phone contact with visitors and updates of the intervention website were related to more average log-ins on the intervention websites, indicated by the higher number of studies on interventions that included these elements listed in the higher average log-in categories, as compared with interventions without these elements.

There are some limitations to this review study that need to be mentioned. The search strategies were limited to include only peer-reviewed English language publications. Therefore, we could have missed important “gray literature” and publication in languages other than English. Next, for this review we relied on the information that was provided in the published literature regarding the description of the intervention and identification of potentially exposure-promoting methods and strategies. Some of the intervention descriptions were very brief, and even the more extensive descriptions available in the literature may not always have been complete. Therefore, we may have missed some of the potential exposure-promoting elements that an intervention contained. In addition, this review can be qualified as a qualitative review as the extracted data from the included studies were summarized and not statistically pooled, which limits the strength of the conclusions that can be drawn. Finally, the used cutoff points for making a ranking within the categories of potential exposure-promoting interventions elements (ie, < 3 or ≥ 3 interactive behavior change strategies, and yes vs no interactive elements) may not have been sensitive enough to detect differences in exposure.

The studies included in this review showed that in the Internet interventions currently available, a wide variety of potentially exposure-improving methods and strategies were used. These methods and strategies were markedly different for the healthy lifestyle behaviors that were studied. Also, a large variety of objective exposure outcome measures were used and there was a lack of consistency in exposure measures reported. Peer support, counselor support, email/phone contact with visitors through sending intervention content and prompts and updates of the intervention website were indicated to result in a longer visit and more log-ins on the website, respectively. More research is needed to gain insight into how intervention characteristics can be used to improve exposure to Internet interventions. More accurate and consistent description of intervention content and more consistency in the report of objective exposure outcomes are recommended. This will enable researchers to better assess associations between intervention characteristics and exposure to health behavior change Internet interventions in the future.