Construction of a recurrence risk prediction model for benign paroxysmal positional vertigo

To investigate factors associated with recurrence in patients with benign paroxysmal positional vertigo (BPPV), develop a recurrence risk prediction model, and further evaluate the incremental value of vestibular evoked myogenic potential (VEMP) parameters in improving model predictive performance.

In this retrospective study, patients with BPPV treated with standardized canalith repositioning maneuvers at Haicang Hospital of Xiamen between January 2022 and December 2024 were included. Demographic data, comorbidities, self-rating anxiety scale (SAS) scores, pittsburgh sleep quality index (PSQI) scores, VEMP parameters, and follow-up information were extracted. Univariable and multivariable logistic regression analyses were conducted to determine factors associated with recurrence, and a recurrence prediction model was developed accordingly. Model discrimination, calibration, and clinical utility were evaluated using receiver operating characteristic (ROC) curve analysis, area under the receiver operating characteristic curve (AUC), bias-corrected calibration curves, the Hosmer-Lemeshow goodness-of-fit test, and decision curve analysis (DCA).

A total of 257 patients with BPPV were included, of whom 64 (24.9%) experienced recurrence during follow-up. Multivariable logistic regression analysis identified diabetes mellitus, hypertension, abnormal ocular vestibular evoked myogenic potential (oVEMP), PSQI score, and SAS score as independent factors associated with BPPV recurrence. The model incorporating these variables showed good discrimination, with an AUC of 0.796 (95% CI, 0.734-0.857), which was significantly higher than that of the model excluding oVEMP abnormality (AUC = 0.736; 95% CI, 0.669-0.804; P = 0.021). The model demonstrated good calibration according to the Hosmer-Lemeshow test (P = 0.270). DCA indicated a favorable net benefit across threshold probabilities ranging from 0.20 to 0.60. Internal validation using bootstrap resampling yielded an optimism-corrected AUC of 0.777 (95% CI, 0.722-0.843).

Diabetes mellitus, hypertension, PSQI score, SAS score, and abnormal oVEMP are associated with the recurrence of BPPV. The prediction model developed from these factors exhibited good discrimination and calibration. This model may facilitate risk‑stratification and inform individualized management considerations for patients with BPPV.