Automatic assessment of adverse drug reaction reports with interactive visual exploration
Bibliographic record
Abstract
A large number of adverse drug reaction (ADR) reports are collected yearly through the spontaneous report system (SRS). However, experienced experts from ADR monitoring centers (ADR experts, hereafter) reviewed only a few reports based on current policies. Moreover, the causality assessment of ADR reports was conducted according to the official approach based on the WHO-UMC system, a knowledge- and labor-intensive task that highly relies on an individual's expertise. Our objective is to devise a method to automatically assess ADR reports and support the efficient exploration of ADRs interactively. Our method could improve the capability to assess and explore a large volume of ADR reports and aid reporters in self-improvement. We proposed a workflow for assisting the assessment of ADR reports by combining an automatic assessment prediction model and a human-centered interactive visualization method. Our automatic causality assessment model (ACA model)-an ordinal logistic regression model-automatically assesses ADR reports under the current causality category. Based on the results of the ACA model, we designed a warning signal to indicate the degree of the anomaly of ADR reports. An interactive visualization technique was used for exploring and examining reports extended by automatic assessment of the ACA model and the warning signal. We applied our method to the SRS report dataset of the year 2019, collected in Guangdong province, China. Our method is evaluated by comparing automatic assessments by the ACA model to ADR reports labeled by ADR experts, i.e., the ground truth results from the multinomial logistic regression and the decision tree. The ACA model achieves an accuracy of 85.99%, a multiclass macro-averaged area under the curve (AUC) of 0.9572, while the multinomial logistics regression and decision tree yield 80.82%, 0.8603, and 85.39%, 0.9440, respectively, on the testing set. The new warning signal is able to assist ADR experts to quickly focus on reports of interest with our interactive visualzation tool. Reports of interest that are selected with high scores of the warning signal are analyzed in details by an ADR expert. The usefulness of the overall method is further evaluated through the interactive analysis of the data by ADR expert. Our ACA model achieves good performance and is superior to the multinomial logistics and the decision tree. The warning signal we designed allows efficient filtering of the full ADR reports down to much fewer reports showing anomalies. The usefulness of our interactive visualization is demonstrated by examples of unusual reports that are quickly identified. Our overall method could potentially improve the capability of analyzing ADR reports and reduce human labor and the chance of missing critical reports.
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 | 0.006 | 0.021 |
| Meta-epidemiology (narrow) | 0.003 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.008 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
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; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
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".