[Abstract quality assessment of articles from the Annales de Dermatologie]

Abstract

Introduction: Article's abstracts are an important part of the publication, widely available in electronic databases. We assessed the quality of abstracts in the Annales de Dermatologie. The main objective was to compare abstract quality in 3 periods in the past decade. In this Journal, structured abstracts are required since 1993. The secondary objective was to compare structured and non structured abstract quality.

Material and methods: Three periods were assessed: 1991-92, 1996 and 2000. We selected the abstracts of original studies, excluding non focused descriptive case series. We used a criteria scale elaborated by Narine et al. Thirty criteria were assessed for each abstract, divided into 8 categories: purpose, research design, setting, subjects, intervention, outcome measurement, results and conclusion. If applicable to the study, a criterion was rated 1 (present) or 0 (absent). The final score for each abstract was defined as the ratio of positive answers among rated criteria. Mean scores for each period were compared using analysis of variance (mean +/- standard deviation). The temporal trend was calculated by simple linear regression. The mean scores of structured and unstructured abstracts were also evaluated and then compared.

Results: For each period, chronologically, 8, 17 and 18 (total 43) abstracts were evaluated. Mean scores for each period were 0.72 +/- 0.20 (1991-92); 0.69 +/- 0.12 (1996) and 0.83 +/- 0.08 (2000). These 3 scores were significantly different (P=0.006), with a trend toward increasing (Regression coefficient R(2)=0.136; P=0.015). Research setting obtained the lowest score (0.4). Structured (0.72 +/- 0.20; n=35) and non structured abstracts (0.76 +/- 0.12; n=8) were not significantly different.

Discussion: Abstracts of articles published in the Annales de Dermatologie contained on average 3/4 of the required informative items. The trend toward better scores may be explained by an increased attention of the editorial board as well as authors and reviewers. The quality could be improved by systematically mentioning the research setting.