Generating real-world evidence: Using automated data extraction to replace manual chart review.
Bibliographic record
Abstract
e18096 Background: Real world evidence is a valuable resource to help guide clinical care beyond evidence generated from clinical trials, for example safety and effectiveness of novel treatments in special populations. Administrative databases often lack sufficient clinical detail to address gaps in the improvement of patient management and quality of care. Detailed clinical data collection and curation are resource intensive, limiting the ability to generate and maintain large informative cancer databases. Darwen, novel technology developed by Pentavere, enables the automation of data abstraction from unstructured hospital electronic medical records and may eliminate the need for manual chart review. Methods: Health records were identified through an institutional cancer registry from patients with stage IIIB/IV lung cancer (NSCLC or SCLC) diagnosed and treated at the Princess Margaret Cancer Centre between 01/01/2015 and 31/12/2017. Cases underwent automated data extraction including demographics, comorbidities, treatment, concurrent medications and outcomes until 30/06/2018. Agreement with data fields extracted using manual data collection in an external validation set of patients is planned. Results: Of 1210 patients identified, 538 were eligible for analysis. From automated data abstraction, 9.9% were reported to have SCLC, 67.5% adenocarcinoma, 11.2% squamous carcinoma, 28% EGFR mutations, 5.8% ALK fusions and 9.3% tumour PDL1 > = 50%. Of the 304 (56.5%) that received systemic therapy, initial treatment was chemotherapy for 55.6%, targeted therapy in 34.2% and immunotherapy in 10.2%. Additional outcome data and agreement with manually curated data fields will be presented. Conclusions: Automated software to extract clinical data is a powerful new tool to generate and maintain databases that yield high quality real world clinical evidence. This is a critical next step to improve clinical decision making, inform evidence-based practice and improve quality of cancer care.
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 | 0.341 | 0.638 |
| Meta-epidemiology (narrow) | 0.002 | 0.003 |
| Meta-epidemiology (broad) | 0.004 | 0.004 |
| Bibliometrics | 0.035 | 0.025 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.012 | 0.009 |
| Open science | 0.005 | 0.009 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.010 | 0.004 |
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".