Using Biomedical Text as Data and Representation Learning for Identifying Patients with an Osteoarthritis Phenotype in the Electronic Medical Record
Bibliographic record
Abstract
IntroductionElectronic medical records (EMRs) are increasingly used in health services research. Accurate/efficient identification of a target population with a specific disease phenotype is a necessary precursor to studying the health of these individuals.
 Objectives and ApproachWe explored the use of biomedical text as inputs to supervised phenotype identification algorithms. We employed a two-stage classification approach to map the discrete, sparse high-dimensional biomedical text data to a dense low dimensional vector space using methods from unsupervised machine learning. Next we used these learned vectors as inputs to supervised machine learning algorithms for phenotype identification.
 We were able to demonstrate the applicability of the approach to identifying patients with an osteoarthritis (OA) phenotype using primary care data from the Electronic Medical Record Administrative data Linked Database (EMRALD) held at ICES.
 ResultsEMRALD contains approximately 20Gb of biomedical text data on approximately 500,000 patients. The unit of analysis for this study is the patient. We were interested in identifying OA patients using solely text data as features.
 Labelled outcome information wass available from a random sample of 7,500 patients. We divided patients into training (N=6000), validation (N=750) and test (N=750) cohorts. We learned low dimensional representations of the input text data on the entire EMRALD corpus (N=500,000). We used learned numeric vectors as inputs to supervised machine learning models for OA classification (N=6,000 training set patients).
 We compared models in terms of accuracy, sensitivity, specificity, PPV and NPV. The best learned models achieved approximately 90\% sensitivity and 80\% specificity. Classification accuracy varied as a function of learned inputs.
 Conclusion/ImplicationsWe developed an approach to phenotype identification using solely biomedical text as an input. Preliminary results suggest our two-stage ML approach has improved operating characteristics compared to existing clinically derived decision rules for OA classification. Future work will explore the generalizability of this methodology to other disease phenotypes.
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 | 0.002 | 0.003 |
| 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 | 0.001 | 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".