Time-Continuous Prognostic Mapping of Bell's Palsy Using Yanagihara Scores and Electroneurography

Bell's palsy impairs social functioning. Conventional prognostic assessment relies on clinical grading and electroneurography (ENoG), but predictive accuracy remains limited. This study introduces a time-continuous probabilistic approach that integrates longitudinal Yanagihara scores and ENoG to enable dynamic, individualized prognostic estimation.

We conducted a retrospective longitudinal study of patients with Bell's palsy treated between April 2013 and March 2024. Eligible patients were aged ≥ 15 years, hospitalized within 1 week of onset, and received high-dose corticosteroids. Exclusion criteria included other etiologies, prior facial palsy, facial nerve decompression, treatment discontinuation, or incomplete data. Recovery was defined as a Yanagihara score ≥ 36, and nonrecovery as failure to achieve this within ≥ 6 months. ENoG values obtained within 4 weeks of onset were stratified into five groups. Multivariable logistic regression was performed using ENoG, sex, age, diabetes, hypertension, dyslipidemia, body mass index, and HbA1C as covariates. Logistic curves were fitted to longitudinal scores, and Monte Carlo/Bayesian procedures were applied to generate recovery probability maps.

Among 128 eligible patients, ENoG emerged as the only significant predictor of recovery. The probability maps provided a novel visualization of individualized prognosis at any time point. This method estimated recovery likelihood even in patients with follow-up intervals, addressing irregular limitations of prior fixed-timepoint models. Conditional mutual information plateaued around 50 days, indicating stabilization of predictive discrimination beyond that period.

Incorporating longitudinal Yanagihara scores obtained during outpatient follow-up enhances recovery prediction accuracy. This practical, dynamic approach facilitates intuitive risk visualization and supports improved prognostic counseling in clinical practice.