Adjustment terms and coefficients of nonlinear regression-based kurtosis-adjusted equivalent sound level method

Background Noise-induced hearing loss (NIHL) is a prevalent occupational health problem in workplace settings, with non-steady noise exposure being particularly widespread. Although kurtosis-adjusted equivalent sound level ( L'_\textEX,\text8\text h ) methods based on linear regression are available to assess hearing damage, the complexity of kurtosis effects necessitates introducing new metrics through nonlinear regression to improve predictive accuracy for hearing loss from non-steady noise exposure.

Objective To examine the adjustment terms \mathrmlg(\beta _N/\beta _G) and coefficients (λ) of L'_\textEX,\text8\text h using nonlinear regression and evaluate the method’s effectiveness in assessing occupational hearing loss associated with non-steady noise exposure.

Methods A cross-sectional study design was employed, enrolling 1034 manufacturing workers and evaluating noise exposure to estimate hearing loss metrics. Quantile regression analysis quantified the proportional contributions of influencing factors for noise-induced permanent threshold shift at 3, 4, and 6 kHz frequencies (NIPTS₃₄₆). L'_\textEX,\text8\text h was computed using multiple linear regression and nonlinear least squares optimization. Paired t-tests analyzed predictive efficacy before and after kurtosis adjustment to evaluate its impact on ISO 1999 NIPTS₃₄₆ predictions. Chow tests compared the similarity of logistic regression curves for high-frequency noise-induced hearing loss (HFNIHL) incidence between non-steady noise ( L'_\textEX,\text8\text h ) and steady noise ( L_\textEX,\text8\text h ), thereby validating the effectiveness of L'_\textEX,\text8\text h .

Results The quantile regression analysis showed that kurtosis was a significant indicator of occupational hearing loss risk from non-steady noise (contribution rate was 27.5%, P<0.05). The linear regression and nonlinear least squares methods obtained six different combinations of coefficients (λ) and adjustment terms \mathrmlg(\beta _N/\beta _G) . Incorporating these L'_\textEX,\text8\text h adjustments into the ISO 1999 predictive model significantly reduced NIPTS underestimation (from 14.2 dB HL with L_\textEX,\text8\text h to 11.5–5.6 dB HL with L'_\textEX,\text8\text h , P < 0.001). The model L'_\textEX,\text8\text h\text,6=L_\mathrmE\mathrmX,8\;\mathrmh+7.9\mathrmlg(\beta _N/10) achieved the most significant improvement, reducing underestimation by 8.7 dB HL. The logistic curve analysis further demonstrated that all L'_\textEX,\text8\text h dose-response relationships for non-steady noise aligned more closely with the steady noise group than L_\textEX,\text8\text h , with L'_\textEX,\text8\text h\text,6 showing the smallest divergence (difference: 4.1%).

Conclusions  L'_\textEX,\text8\text h demonstrates superior efficacy in assessing the risk of occupational hearing loss induced by non-steady noise exposure. The L'_\textEX,\text8\text h metric, constructed by calculating adjustment terms and coefficients through nonlinear regression analysis, exhibits greater effectiveness than linear regression methods in evaluating occupational hearing loss associated with non-steady noise exposure.