Machine-learning alerts increase referrals for epilepsy surgery evaluation in a randomized trial
Source check: Source checked. Not independent scientific confirmation.
Finding: At one specialized pediatric epilepsy center, an NLP support system using an SVM and n-gram features screened neurology notes and randomized providers of 284 identified potential candidates to receive an alert or usual care. During a median 24-month follow-up, presurgical evaluation referrals were 9.8% with alerts and 3.1% in controls (adjusted HR 3.21; 95% CI 0.95–10.8; one-sided p=.03); nine alert-group children and no controls underwent surgery. The trial was prospective and randomized, but involved one center, had a wide confidence interval, and used one-sided testing; it does not establish benefit across other settings.
Technology: Natural-language processing with a supervised support vector machine and n-gram features
Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.
- AI contribution
- An NLP/SVM system screened free-text EHR neurology notes for potential surgical candidates and generated provider alerts.
- Test and result
- A prospective randomized trial compared provider alerts with standard care among 300 potential candidates at 14 outpatient clinic sites of one pediatric epilepsy center; presurgical referrals and subsequent surgery were recorded.
- Limitation / still unknown
- Generalizability beyond the specialized center and the effect in other care settings were not established. The referral estimate was imprecise, and the paper reports one-sided p values.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://onlinelibrary.wiley.com/doi/full/10.1111/epi.17629
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 22 Nov 2026. Review scope: Full primary publisher article, including Methods, Results, author affiliations and publication history.
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 4/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.