Reviewer agreement trends from four years of electronic submissions of conference abstract

Abstract

BACKGROUND: The purpose of this study was to determine the inter-rater agreement between reviewers on the quality of abstract submissions to an annual national scientific meeting (Canadian Association of Emergency Physicians; CAEP) to identify factors associated with low agreement. METHODS: All abstracts were submitted using an on-line system and assessed by three volunteer CAEP reviewers blinded to the abstracts' source. Reviewers used an on-line form specific for each type of study design to score abstracts based on nine criteria, each contributing from two to six points toward the total (maximum 24). The final score was determined to be the mean of the three reviewers' scores using Intraclass Correlation Coefficient (ICC). RESULTS: 495 Abstracts were received electronically during the four-year period, 2001-2004, increasing from 94 abstracts in 2001 to 165 in 2004. The mean score for submitted abstracts over the four years was 14.4 (95% CI: 14.1-14.6). While there was no significant difference between mean total scores over the four years (p = 0.23), the ICC increased from fair (0.36; 95% CI: 0.24-0.49) to moderate (0.59; 95% CI: 0.50-0.68). Reviewers agreed less on individual criteria than on the total score in general, and less on subjective than objective criteria. CONCLUSION: The correlation between reviewers' total scores suggests general recognition of "high quality" and "low quality" abstracts. Criteria based on the presence/absence of objective methodological parameters (i.e., blinding in a controlled clinical trial) resulted in higher inter-rater agreement than the more subjective and opinion-based criteria. In future abstract competitions, defining criteria more objectively so that reviewers can base their responses on empirical evidence may lead to increased consistency of scoring and, presumably, increased fairness to submitters.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 20 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.