A Clinical Risk Stratification Approach for Sensorineural Hearing Loss in Pediatric Bacterial Meningitis: A Retrospective Cohort Study from Vietnam

Objective: To address the long-term burden of sensorineural hearing loss (SNHL) following pediatric bacterial meningitis and the need for early identification of at-risk children, particularly in resource-limited settings where universal post-meningitis screening is impractical. This study aimed to develop a clinically accessible risk stratification approach, using routinely available variables, to support triage and prioritization of audiological follow-up. Methods: A retrospective cohort study was conducted at Children’s Hospital 1, Ho Chi Minh City, Vietnam. Children aged 1 month to 15 years, who were hospitalized with bacterial meningitis and underwent audiological assessment after recovery, were included. Hearing outcomes were evaluated by auditory brainstem response and/or otoacoustic emission testing ≥4 weeks post-discharge. Demographic, clinical, microbiological, and cerebrospinal fluid variables from early hospitalization were extracted. A parsimonious multivariable logistic regression model using only early-available predictors was constructed to reduce overfitting and enhance clinical use. Model discrimination and calibration were assessed, with internal validation via bootstrap resampling. Predicted probabilities were used to build a pragmatic SNHL risk stratification framework. Results: Among 80 children, 19 (23.8%) developed permanent SNHL. Multivariable analysis identified three independent predictors: Glasgow Coma Scale score <8 at admission, pneumococcal etiology, and cerebrospinal fluid glucose <40 mg/dL. The final three-variable model showed good discrimination and acceptable internal calibration. Predicted probabilities categorized patients into low-, moderate-, and high-risk groups with increasing SNHL rates. Conclusion: SNHL is a frequent and severe complication of pediatric bacterial meningitis. A simple three-variable model based on early clinical, microbiological, and biochemical features may aid early risk stratification and follow-up prioritization. Findings require cautious interpretation due to the limited sample size and retrospective design. External validation is needed prior to clinical application.