Study on the Effectiveness of Combining Different Hearing Test Techniques in Aiding the Diagnosis of NIHL Using a Cross-sectional Survey

Objective: Conventional audiometry (CA) has limitations in detecting early noise-induced hearing loss (NIHL). This study aims to assess the effectiveness of combining different hearing test techniques in aiding the diagnosis of NIHL by CA. Methods: This cross-sectional study enrolled 138 NIHL patients (diagnosed by CA) and 138 matched controls. The receiver operating characteristic (ROC) curve was used to assess the diagnostic performance of each supplementary diagnostic technique (e.g., extended high-frequency audiometry [EHF], distortion product otoacoustic emissions [DPOAE], the Dichotic Digits Test [DDT], or Bamford-Kowal-Bench Speech-in-Noise Test [BKB-SIN]). A parsimonious model was developed and validated. ROC curves were used to assess the improvement in CA’s diagnostic performance with the parsimonious model. The odds ratio (OR) for the parsimonious model was calculated among noise-exposed groups with normal CA. Results: Significant differences were observed in four audiological measures (EHF, DPOAE, DDT, and BKB-SIN) between case and control groups ( P < 0.05). ROC analysis revealed distinct diagnostic performance among the four tests ( P < 0.05). Multivariate logistic regression identified EHF, DPOAE, and DDT as independent diagnostic factors ( P < 0.01). The parsimonious model (EHF+DPOAE+DDT) demonstrated a comparable goodness-of-fit to the full model (EHF+DPOAE+BKB-SIN+DDT) ( P > 0.05), while it significantly outperformed any single test alone ( P < 0.01). Bootstrap resampling validated its robust stability across subgroups (age, sex, and work experience). The Delong test showed improved diagnostic efficacy in the CA + parsimonious model group (AUC increased by 0.09, P < 0.01) and in CA-defined severity groups (AUC: 0.94–0.96). The multivariate logistic regression analysis showed that, among noise-exposed workers with normal CA, ORs of relevant indicators from the parsimonious model for early NIHL were 1.31 [95% Confidence Intervalat (CI): 1.21–1.47, P < 0.01]. Conclusion: A parsimonious combination of EHF, DPOAE, and DDT is effective for aiding CA in diagnosing early NIHL. More studies are needed for further validation.