Journal Issue: Unintentional Injuries in Childhood Volume 10 Number 1 Spring/Summer 2000
Eight electronic databases were searched for studies published in any language. The databases included MEDLINE (1966–98), EMBASE (1974–98), Psychinfo (1967–98), Current Contents (wk1–wk25), HealthSTAR (1975–98), Sportdiscus (1949–98), CINAHL (1982–98), and The Cochrane Database of Systematic Reviews (CDSR) (1998, Issue 2). Separate searches of each database were conducted for the following injury categories: (a) bicycle, walk, and motor vehicle; (b) gun, drown, and choke; (c) fall, burn, and electrocution; (d) poisoning and cuts; (e) sports and amputation. Detailed search strategies are available from the corresponding author. When references were retrieved by more than one database, priority in downloading was given to MEDLINE. The journal Injury Prevention from March 1995 to 1998 was searched by hand and four potentially relevant studies were identified. A fifth article was identified through an article found in MEDLINE. Two additional articles were selected based on expert opinion. However, due to time constraints, reference lists from relevant reviews were not examined for potential studies.
Once the search process was complete, two members of the research team screened each citation and available abstract in the database. For all databases, letters, comments, editorials, or articles on sunburn, abuse, and other intentional injuries were excluded. All references to injury in children or adolescents and those that used a study design that included a control group were tagged as potentially relevant. Articles were reviewed and assessed using eligibility criteria established by the team.
A study was included if: (1) it included a control group that did not receive the intervention; (2) the target population was between 0 and 19 years of age (if a greater age range was examined, separate data analysis had to be conducted for the 0 to 19 age group); (3) it examined the effectiveness of a community-based intervention based in a setting such as a recreational center, school, or day-care center and involved any intervention between an individual-based approach and a state or nationwide one; and (4) it reported injury rates or change in an injury-reducing behavior. Three members of the team independently reviewed all potentially relevant articles to determine the eligibility of each document. All disagreements were resolved by consensus.
Pertinent information was extracted from each article by one member of the team using a Data Extraction Form (available upon request). To interpret the study, we attempted to extract the primary outcome of interest, the intervention used, and relevant data about the results. The primary outcome reports the effectiveness of the intervention and is used to determine which sample size to report for this study. If the primary outcome was unclear, the outcome which seemed to be the “most important” to eliminate bias was chosen. For example, if helmet ownership and helmet use were both outcomes, helmet use was selected. In cases where attitude, behavior, and knowledge were all examined, behavior was selected.
Following data extraction, quality assessment was conducted for each randomized controlled trial (RCT) and nonrandomized controlled trial (NRCT) using the Jadad Scale, a validated instrument designed to assess methodological quality.1 Quality is scored between 0 and 5, with 5 being the highest score/quality. Three members of the team calibrated themselves on the instrument, and two independently completed quality assessments of each relevant study. Disagreements were resolved by consensus.
A total of 1,236 citations were identified, of which 349 were duplicates, resulting in 887 articles. The initial screen identified 158 potentially relevant articles. Two articles were reviews, one was a meta-analysis, and the remaining 155 articles contained original data. Of the 155 articles that were potentially relevant, 28 were included in the review; 4 were not discussed further in this review, because they addressed other types of injury prevention (such as gun-playing); 5 were excluded and are discussed in the article by DiGuiseppi and Roberts in this journal issue; 5 were unavailable through the library; 2 were in French; and 111 did not meet the eligibility criteria. More than half of the excluded articles did not contain a control and/or did not use a community-based intervention. Of the 28 studies discussed, 6 were RCTs and 22 were NRCTs.1
Jadad, A.R., Moore, R.A., Carroll, D., et al. Assessing the quality of reports of randomized clinical trials: Is blinding necessary?Controlled Clinical Trials