Worksite physical activity interventions.

ABSTRACT

BACKGROUND: National objectives for public health have targeted worksite as important settings for interventions to increase physical activity. However, expert reviews reveal no scientific consensus about the effectiveness of worksite interventions for increasing physical activity or fitness. METHODS: We judged the quantity and quality of existing evidence against scientific standards for the internal and external validity of the research design and the validity of measurements. Meta-analytic methods were used to quantify the size of effects expressed as Pearson correlation coefficients (r). Variation in effect was examined in relation to several features of the studies deemed important for implementing successful worksite interventions. Pre-experimental cohort studies were excluded because they are sensitive to secular trends in physical activity. RESULTS: Twenty-six studies involving nearly 9,000 subjects yielded 45 effects. The mean effect was heterogeneous and small, r = 0.11 (95% CI, -0.20 to 0.40), approximating 1/4 S.D., or an increase in binomial success rate from 50% to 56%. Although effects varied slightly according to some of the study features we examined, effects were heterogeneous within levels of these features. Hence, the moderating variables examined did not explain variation in the effects (P > 0.05). The exception was that effects were smaller in randomized studies compared with studies using quasi-experimental designs (P < 0.05). CONCLUSIONS: Our results indicate that the typical worksite intervention has yet to demonstrate a statistically significant increase in physical activity or fitness. The few studies that have used an exemplary sample, research design, and outcome measure have also yielded small or no effects. The generally poor scientific quality of the literature on this topic precludes the judgment that interventions at worksites cannot increase physical activity or fitness, but such an increase remains to be demonstrated by studies using valid research designs and measures.