In this systematic review, participation levels in health promotion interventions at the workplace were typically below 50%. A large variation in participation levels and determinants of initial participation in worksite health promotion was shown, and except for sex few statistically significant associations with initial participation were found. Female workers had a higher participation than men, but this difference was not observed for interventions consisting of fitness centre programmes. In addition, the review showed that programs that provide (1) incentives, (2) offer a multi-component strategy, (3) focus on multiple behaviours rather than on physical activity only have a higher overall participation level.
A major reason for choosing the worksite as setting for health promotion is the possibility to reach large groups [7
]. It is striking that the differences between participation levels were large, with mainly low participation levels, but also levels up to 64%. The large variation is comparable to the findings of Glasgow and colleagues (1993), who found participation levels ranging from 20% to 76%. The authors noticed that attending a single screening does not require much commitment [9
]. In our review, we included only studies evaluating interventions aimed at physical activity and/or nutrition, and therefore excluded studies evaluating only a single health risk assessment (HRA). The median participation level found in a review on 24 studies by Bull and colleagues (2003) was higher than the median reported in this review (61% versus 34%) [7
]. It is not clear if Bull and colleagues included studies evaluating a HRA.
The findings on determinants of participation are in accordance with the review of Glasgow and colleagues [9
]. The overall view is that female employees are more likely to participate in health promotion programmes than male employees.
After pooling, an overall higher participation level for married employees was found. All other demographic characteristics showed no consistent pattern. Only for age, there appeared to be a trend with a higher participation among younger employees, and lowest participation level among the oldest age group. As mentioned, just few statistically significant associations for health- and work-related determinants were found. Several studies have reported higher participation in smaller worksites albeit without providing quantitative information [34
]. This finding is supported in this review by the included study of Blake and colleagues (1996) [14
]. No pooled ORs were calculated for the health- and work-related determinants due to the large variation in definition of determinants and programmes evaluated.
More than 80% of the studies evaluating a WHPP on nutrition or PA did not report any determinants of non-participants. In 1993, Glasgow and colleagues already recommended that future studies should report participation levels, the number of employees entering the programme, and demographic information [9
]. This information is needed to gain insight in potentially selective participation and external validity. Just few studies included information on educational level and income. Since unhealthy lifestyles are more common among lower socio-economic groups, it is important to get insight in the reach (and effectiveness) in these specific groups. Information on determinants should be an essential aspect of a process evaluation. In the RE-AIM framework for the evaluation of the public health impact of health promotion interventions, the 'reach' dimension is included which is measured by comparing records of participants and complete sample information for a defined population, in this case the worksite [36
]. In the recent CONSORT statements it is emphasized to include information on the eligible participants in order to increase the validity [37
In total, 64 out of 130 (49%) associations between determinants and participation did not reach statistical significance. These null associations may be the result of a small sample size and lack of statistical power, and the presence of another risk factor or confounder [38
]. It is not likely that most null associations are explained by the sample size or confounding, because most studies had sample sizes larger than 500 subjects, and most ORs were calculated by means of univariate analysis. Thus, the lack of a clear health-related selection in participation suggests that WHPPs are able to reach those most-at-risk and, hence, provide a valuable setting.
After stratification of the demographic determinants by programme type, it appeared that fitness centre studies do not suffer from a lower participation among men. Further, no statistically significant differences in demographic determinants were found between programme categories. The finding that fitness centre studies do not favour female workers in comparison with other programme categories, suggests that the content of intervention programmes should be tailored to the population characteristics.
In addition to determinants that may play a role in the uptake of interventions in the context of work settings, several programme characteristics were associated with participation. First, this review and others [39
] suggest that the inclusion of an incentive can have beneficial effects on reach, hence increasing the absolute number of people who engage in health-related activities. Second, the present finding that more multi-component interventions do not decrease the uptake is in itself reassuring. A potential explanation for this finding may be that these interventions offer a large choice for potential participants. It could be hypothesized that multi-component interventions may have bigger participation levels as it matches with a larger array of people, whereas a mismatch is more likely for single components whereby persons may not see the need or be ready to engage in a particular activity. Finally, in this review a fee for participation was not identified as a barrier to participate. The 4 studies reporting on interventions with a fee for participation included 1 very large study [28
]. Excluding this study showed among the remaining 3 studies a lower participation level (participation level: 24.3%; 95% CI: 22.7%–25.8%) as compared to studies not requiring a fee for participation (participation level: 31.7%; 95% CI: 31.5–31.9%). This indicates that the results of the pooled analysis should be interpreted carefully depending on the studies included.
Low participation levels will result in decreased (cost-)effectiveness of intervention programmes on population level and a potentially decreased generalizability of the results [40
]. Implications for raising participation levels in WHPPs are the provision of incentives, or a broad array of programme offers. To what degree these strategies affect also compliance to an intervention programme should be considered.
This systematic review has some limitations. First, the literature search was limited to two electronic databases, with an overlap of 86% of the articles. With just two electronic databases and only English publications included, it is possible that we missed some useful studies. We assume this does not have a major effect on the findings. Second, many interventions are conducted in practice that are not well-evaluated and not published in scientific literature. This review is limited to the published research. Third, 8 out of 30 studies were excluded because they reported only qualitative information on initial participation. Fourth, pooling of all determinants was impossible because of the large heterogeneity in definition of initial participation, in programme components, and measurement of determinants. Finally, due to the limited information provided in studies, the possibility to study the interaction between determinants and programme characteristics was restricted.