MedReadr: Development and Evaluation of an In-Browser, Rule-Based Natural Language Processing Algorithm to Estimate the Reliability of Consumer Health Articles
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Introduction: The internet is a major source of medical information for patients, yet the quality of online health content remains highly variable. Existing assessment tools are often labor-intensive, invalidated, or limited in scope. We developed and validated MedReadr, an in-browser, rule-based natural language processing (NLP) algorithm that automatically estimates the reliability of consumer health articles for patients and providers. Methods: Thirty-five consumer medical articles were independently assessed by two reviewers using validated manual scoring systems (QUEST and Sandvik). Interrater reliability was evaluated with Cohen's κ, and metrics with κ > 0.6 were selected for model fitting. MedReadr extracted key features from article text and metadata using predefined NLP rules. A multivariable linear regression model was trained to predict manual reliability scores, with internal validation performed on an independent set of 20 articles. Results: < 0.05). Key predictive features included currency and reference scores, sentiment polarity, engagement content, and the frequency of provider contact, intervention endorsement, intervention mechanism, and intervention uncertainty phrases. Conclusion: MedReadr demonstrates that structural reliability scoring of online health articles can be automated using a transparent, rule-based NLP approach. Applied to English-language articles from mainstream search results on common medical conditions, the tool showed strong agreement with validated manual scoring systems. However, it has only been validated on a narrow scope of content and is not designed to analyze search results for specific questions or detect misinformation. Future research should assess its performance across a broader range of web content and evaluate whether its integration improves patient comprehension, digital health literacy, and clinician-patient communication.
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.
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 | 0.009 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| 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 it