[Prognostic factors and construction of a predictive model for patients with refractory low-frequency sudden...

Objective: To investigate the prognostic factors for patients with refractory low-frequency sudden sensorineural hearing loss, and develop a predictive model. Methods: Patients with refractory low-frequency sudden sensorineural hearing loss who were admitted to the PLA General Hospital from January 2009 to December 2023 and did not respond to standard outpatient treatment were retrospectively enrolled. Based on treatment outcomes, patients were divided into a recovery group and a non-recovery group. Clinical data were compared between the two groups. Multivariate logistic regression analysis was used to identify prognostic factors, and a nomogram was constructed. The predictive performance of the model was evaluated using the receiver operating characteristic (ROC) curve, Hosmer-Lemeshow test, and decision curve analysis to assess its discrimination, calibration, and clinical utility. Results: A total of 211 patients (89 males and 122 females) aged (45.8±12.0) years were included. There were 68 and 143 patients in the recovery group and non-recovery group, respectively. Compared to the non-recovery group, the recovery group exhibited younger age, shorter onset-to-treatment time, lower 1 000 Hz hearing thresholds, milder hearing loss, and higher proportions of headache and hypotension/low intracranial pressure (all P<0.05). Multivariate logistic regression analysis identified that hypertension (OR=3.380, 95%CI: 1.166-9.794, P=0.025), delayed treatment (8-21 days: OR=3.306, 95%CI: 1.469-7.440, P=0.004;>21 days: OR=11.722, 95%CI: 4.805-28.599, P<0.001), and impaired 1 000 Hz hearing (OR=3.001, 95%CI: 1.476-6.099, P=0.002) were risk factors for poor prognosis. The presence of headache (OR=0.335, 95%CI: 0.123-0.912, P=0.032) and hypotension/low intracranial pressure (OR=0.214, 95%CI: 0.060-0.767, P=0.018) were protective factors. The area under the ROC curve (AUC) was 0.822 (95%CI: 0.762-0.881). The calibration curve demonstrated good agreement with the ideal line, and the Hosmer-Lemeshow test showed good model fit (χ²=3.917, P=0.789). The decision curve shows that the predictive model has a positive net benefit across almost all threshold probability ranges. Conclusions: Hypertension, delayed medical consultation, and 1 000 Hz hearing impairment are independent risk factors for poor prognosis in patients with low-frequency sudden sensorineural hearing loss, whereas headache and hypotension/low intracranial pressure are protective factors. The predictive model developed based on these factors provides a reliable tool for prognostic assessment in clinical practice.