Search Filters to Identify Automation in HEOR: An Umbrella Review of Performance and Overlap
Bibliographic record
Abstract
Background: Search strategies used to identify evidence on automation in Health Economics and Outcomes Research (HEOR) often lack sensitivity and specificity, resulting in information overload or missed studies. This umbrella review evaluated and compared the performance and overlap of search filters commonly used to retrieve automation-related HEOR evidence. Methods: Systematic literature reviews (SLRs) and search filters focusing on any form of automation in HEOR were included. Searches (January 01, 2023–July 03, 2024) were conducted in EMBASE, ISSG Resource, and Google Scholar. Subject headings, search terms, and performance metrics were extracted. Reference lists were cross-checked. Screening was performed by one reviewer, with 20% verified by a second reviewer. The PRESS checklist was used to assess search strategy quality. The protocol was registered with the Open Science Framework (OSF). Results: Seven SLRs and one standalone filter, reporting 11 search strategies, met the inclusion criteria. HEOR relevance was defined by studies applying search filters in contexts of SLRs, indirect treatment comparisons, and economic modeling. Included SLRs retrieved between 5-273 studies. PubMed was the most frequently searched database. Commonly used subject headings included “artificial intelligence,” “deep learning,” “machine learning,” and “natural language processing,” with “artificial intelligence” the most frequent free-text term. Inclusion rates varied: title/abstract (1%–8%), full-text (27%–86%), and final inclusion (0.13%–2.31%). Time to include one study ranged from 0.6-8 hours. Conclusions: Considerable variability in search filter performance was observed, causing lower specificity and inefficient evidence retrieval. Standardized, high-performing search strategies are needed to enhance efficiency and reliability in identifying automation-related HEOR evidence.
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 | 0.438 | 0.707 |
| Meta-epidemiology (narrow) | 0.006 | 0.004 |
| Meta-epidemiology (broad) | 0.022 | 0.023 |
| Bibliometrics | 0.096 | 0.054 |
| Science and technology studies | 0.005 | 0.008 |
| Scholarly communication | 0.014 | 0.019 |
| Open science | 0.008 | 0.014 |
| Research integrity | 0.009 | 0.003 |
| Insufficient payload (model declined to judge) | 0.012 | 0.003 |
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".