Review of Primary Care-Based Physical Activity Intervention Studies

Methods

We conducted electronic MEDLINE searches for the years 1980 to 1998 using Grateful Med and the National Library of Medicine’s PubMed search engines. We also performed computerized searches of psychological abstracts, ERIC and HealthStar databases, and of The Journal of Family Practice’s World WideWeb site (www.jfampract.com). The keywords we used for searching included: physical activity and randomized controlled trials, physical activity counseling and primary care or medical office, exercise interventions and primary care or medical office, and physical activity and health promotion. We limited the search to the English language. We searched bibliographies of selected studies and previous reviews for relevant articles. Finally, we contacted 3 experts in the field of physical activity for leads on unpublished trials. Although we attempted to be inclusive, our literature search may have been biased by the English language-only criterion and by the use of a single searcher.

The following inclusion criteria were used: randomized controlled trial or quasiexperimental study having a comparison group, intervention delivered or initiated in a primary care setting, and reported results on at least 1 measure of physical activity. We excluded studies that focused solely on patients with cardiovascular disease because they are generally considered to comprise a separate body of literature and not be representative of the larger population of patients seen in primary care.¹⁸ We identified 48 articles and retained 15 for the final review. The 33 not selected were not controlled trials, focused on a specific medical condition, or did not take place in a primary care setting. We contacted the authors of the 11 studies from which it was not possible to calculate effect size or odds ratio. Three authors responded with information adequate to perform the calculations.

The methodology rating we used was based on the RE-AIM model. RE-AIM was used rather than alternatives, such as the CONSORT criteria, because of its more balanced emphasis on internal and external validity and its relevance for practice-oriented research. The 6 criteria for rating methodology (and the possible points for each) are as follows:

Study design (0-3). One point was assigned for randomized studies. Two points were given to studies that used a placebo, equal contact control, or alternative treatment. One point was given for use of a comparison condition that controlled for a component of the intervention, usual care, or advice-only control groups.

Analyses (0-1). One point was given for analyses that controlled for potential confounding variables or adjusted for baseline physical activity.

Dependent variable (0-1). One point was given to studies that used previously validated outcome measures or that reported the reliability and validity of the physical activity measure.

Reach (0-1). One point was given to studies reporting on the percentage or representativeness of participants.

Implementation (0-2). One point was given to studies for which the intervention was conducted by clinic staff. An additional point was given if the quality of the intervention implementation was reported.

Attrition (0-2). One point was given if attrition was reported. An additional point was given if the rate of attrition was 10% or less or if an intention-to-treat analysis or imputation of missing data was used.

Results

The Table shows descriptive information and short- and long-term outcomes for the 15 studies we reviewed. It includes the methodology rating, sample size, study design and the physical activity intervention (ie, length and content), outcome variables and length of follow-up, and outcomes presented in terms of effect sizes for continuous variables and odds ratios for dichotomous variables for both short-term (less than 12 months) and long-term (12 months or longer) outcomes. Interrater reliability between the 2 independent raters on the methodology score was rho = 0.85, and disagreements were resolved by conference with the senior author.

Outcomes

Of the 10 studies reporting 0- to 11-month post- intervention outcomes, 7 reported statistically significant physical activity outcomes. Effect sizes ranged from 0.003 to 0.26 and odds ratios from 1.04 to 3.73 (median = 1.88). Of the 7 studies reporting postintervention outcomes at 12 months or longer, 3 reported statistically significant outcomes. The only study among these from which an effect size could be calculated reported d = 0.09, and follow-up odds ratios ranged from 0.09 to 1.39 (median = 1.25). Unfortunately, it is not possible to summarize these outcomes in a straightforward manner, such as percentage of patients who changed from being sedentary to meeting the new guidelines for regular moderate physical activity or number needed to treat. This is because rather than using standard measures of physical activity, the authors of most of the studies developed or selected their own continuous (for which we have presented mean effect size) or dichotomous physical activity measures (for which we have presented odds ratios). What can be said is that well-controlled physical activity studies have generally produced moderate short-term improvements, but these results are often less encouraging at long-term follow-up.