Reliability of the Peer-Review Process for Adverse Event Rating
Bibliographic record
Abstract
BACKGROUND: Adverse events are poor patient outcomes caused by medical care. Their identification requires the peer-review of poor outcomes, which may be unreliable. Combining physician ratings might improve the accuracy of adverse event classification. OBJECTIVE: To evaluate the variation in peer-reviewer ratings of adverse outcomes; determine the impact of this variation on estimates of reviewer accuracy; and determine the number of reviewers who judge an adverse event occurred that is required to ensure that the true probability of an adverse event exceeded 50%, 75% or 95%. METHODS: Thirty physicians rated 319 case reports giving details of poor patient outcomes following hospital discharge. They rated whether medical management caused the outcome using a six-point ordinal scale. We conducted latent class analyses to estimate the prevalence of adverse events as well as the sensitivity and specificity of each reviewer. We used this model and Bayesian calculations to determine the probability that an adverse event truly occurred to each patient as function of their number of positive ratings. RESULTS: The overall median score on the 6-point ordinal scale was 3 (IQR 2,4) but the individual rater median score ranged from a minimum of 1 (in four reviewers) to a maximum median score of 5. The overall percentage of cases rated as an adverse event was 39.7% (3798/9570). The median kappa for all pair-wise combinations of the 30 reviewers was 0.26 (IQR 0.16, 0.42; Min = -0.07, Max = 0.62). Reviewer sensitivity and specificity for adverse event classification ranged from 0.06 to 0.93 and 0.50 to 0.98, respectively. The estimated prevalence of adverse events using a latent class model with a common sensitivity and specificity for all reviewers (0.64 and 0.83 respectively) was 47.6%. For patients to have a 95% chance of truly having an adverse event, at least 3 of 3 reviewers are required to deem the outcome an adverse event. CONCLUSION: Adverse event classification is unreliable. To be certain that a case truly represents an adverse event, there needs to be agreement among multiple reviewers.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.611 | 0.858 |
| Meta-epidemiology (narrow) | 0.001 | 0.002 |
| Meta-epidemiology (broad) | 0.003 | 0.003 |
| Bibliometrics | 0.008 | 0.005 |
| Science and technology studies | 0.003 | 0.005 |
| Scholarly communication | 0.005 | 0.005 |
| Open science | 0.004 | 0.006 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.003 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; the direct Gemma label and the distilled Codex classifier agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".