A209 VALIDATION OF A NATURAL LANGUAGE PROCESSING ALGORITHM TO EXTRACT DATA FOR SYSTEM-LEVEL ADENOMA DETECTION RATE CALCULATION
Bibliographic record
Abstract
Patients of endoscopists with lower adenoma detection rates (ADR) are more likely to die from missed colorectal cancers. Measuring ADR is challenging at the system level as pathology results are generally reported in unstructured electronic medical records. Natural language processing (NLP) can be used to extract relevant information from text-based records. At Cancer Care Ontario (CCO), we developed and validated a NLP algorithm to identify colorectal adenomas in unstructured electronic pathology reports available in CCO’s Electronic Mapping Reporting and Coding (eMaRC) data. We identified pathology reports from colonoscopies in eMaRC as those with specimen type ‘biopsy’ and anatomic site ‘colon’. The sampling period was restricted to 2015–16 and patients older than 50 years. From this sampling frame, two random samples of 450 and 1,000 reports were selected as the test and validation sets. Expert clinicians reviewed and classified reports as adenoma or other. The test set was used to develop an NLP algorithm to identify adenomas using Base SAS 9.4. Statistical analyses, including sensitivity (recall), specificity, precision (positive predictive value) and F1 score of the NLP algorithm compared to clinician review were determined. A significant proportion of Ontario colonoscopists, patologists and laboratories were represented in the examined validation sets. The sensitivity of the NLP algorithm was approximately 100% (95 %CI: 98.51–100) and 99.81% (95 %CI: 98.97–100) in the test and validation sets, respectively. Similarly, the specificity was 99.08% (95 %CI: 94.99–99.98) and 100% (95 %CI: 99.21–100). The CCO NLP algorithm was highly accurate in identifying colorectal adenomas in eMaRC data across many institutions in Ontario,. This lays the groundwork to measure ADR at the system-level in Ontario Cancer Care Ontario
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 teacher head, 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".