Trend: An Important Step, but Not Enough

Des Jarlais et al. provide an important service by publishing the TREND statement, which lists data reporting recommendations for behavioral and public health interventions.1 We commend the authors for addressing such issues as information on the target population, recruitment criteria, methods of imputing missing data, comparison of the study population to the target population of interest, and testing of causal pathways. These criteria, although important, are not enough to realize the TREND group’s purpose of creating generalizable knowledge. The goal of public health interventions is to make a difference on a population level (or in a representative sample or segment of a specified population). Reporting the TREND criteria will improve the quality of the literature, but additional criteria related to external validity are also needed. We offer suggestions based on recent literature reviews of behavior change studies.2–6 Few studies reported on the representativeness or impact of setting-level factors and intervention staff, which are critical to an understanding of moderating variables and external validity. To address this problem, we recommend the addition of the following criteria to TREND: Report the percentage of eligible settings (e.g., schools, worksites, community organizations) that participated and compare characteristics of those participating with characteristics of those declining (or with characteristics of a specified target population of settings). Report the percentage of intervention staff members in these settings who participate in delivering the intervention; compare characteristics of those participating with characteristics of those declining (or with characteristics of a specified target population of clinicians); and compare differences in intervention delivery and outcomes among those who participate. Report costs of the intervention, such as capital outlay and staff hours, and, where feasible, report on more sophisticated economic outcomes. Report long-term results and the extent to which settings sustain or modify the program after the formal study has been completed. The TREND group stated that nonrandomized designs are needed to strengthen the evidence-based public health practice literature. The advantages of nonrandomized designs over randomized controlled trials include being less expensive to conduct and not requiring agreement to randomization for participation. Nonrandomized designs provide greater opportunity than randomized controlled trials to obtain cost estimates in limited-resource environments and to evaluate implementation conducted by typical staff members. The TREND criteria can benefit the public health community as the CONSORT7 criteria have benefited medicine. In response to the TREND group’s call for feedback, we offer these suggestions to address issues critically important for the translation of research to practice.