Improving Primary Care Pharmaceutical Use
Measuring study effects and effect size
We focused on the rates of prescribing of pharmaceutical products as the primary outcomes of interest in evaluating the impact of different interventions. However, we also considered other types of outcome, especially when authors had identified them as principal targets of an intervention. Additional provider-oriented measures of effect include improved health provider knowledge about disease and therapy, better diagnostic capability, or improved performance on drug-related tasks such as drug labeling or dispensing. Patient-oriented outcomes were also considered including changes in mortality, duration or severity of illness, and patient understanding about illness or drugs.
In order to be able to draw basic conclusions about relative effectiveness of different intervention strategies, it is useful to have a metric for comparing the size of improvements across different studies. However, there is no simple strategy for comparing relative effectiveness when interventions take place in different health environments, use different intervention approaches, target different audiences and practices, and measure different outcomes. Despite the inherent problems, we developed a systematic process for summarizing relative effect sizes in all interventions included in this review that met our minimum standards for methodological adequacy of study design.
In evaluating each intervention, we focused on the two or three outcome measures that were identified by authors as the principal targets for their intervention. To indicate an intervention=s magnitude of effects, we used the single outcome measure in this group for which the study achieved its largest positive change, e.g., reduction in antibiotic prescribing, increase is percent of cases treated according to a treatment guideline, or reduction in disease-specific child mortality. All outcome measures were converted to a scale where positive change was indicated by positive numbers.
The procedure for calculating effect size depended on whether or not the outcome in question was measured as a percentage (e.g., percent of patients receiving an injection), a mortality rate (e.g., deaths per 1000