Simpson T H, Stewart M, Kaltenbach J A
Wayne State University, Detroit, Michigan.
J Am Acad Audiol. 1993 Jul;4(4):258-63.
Audiometric records of 213 employees with low noise exposure were selected from the public domain data base underpinning Draft American National Standards Institute (ANSI) S12.13-1991 Standard Evaluating the Effectiveness of Hearing Conservation Programs. Records were selected from an ANSI S12.13 "control" group demonstrating raw audiometric thresholds in 1-dB steps, presumably resulting from self-recording audiometry. Raw audiometric data were rounded to the nearest 5-dB step, and ANSI S12.13 analysis protocols were applied to both raw and rounded data sets. The magnitude of all proposed ANSI test statistics increased with audiometric step size. Percent worse sequential and percent better or worse sequential protocols were markedly more sensitive to threshold step size than standard deviation of difference thresholds. The direction and magnitude of change in these outcomes supports the assumption that criterion ranges in the draft standard hold artifactual bias against hearing conservation programs (HCPs) collecting audiometric data in 5-dB steps.
从作为美国国家标准学会（ANSI）S12.13 - 1991《听力保护计划有效性评估标准草案》基础的公共数据库中，选取了213名低噪声暴露员工的听力测定记录。这些记录来自ANSI S12.13的“对照组”，其显示的原始听力阈值以1分贝为步长，推测是由于自我记录听力测定法所致。原始听力测定数据被四舍五入到最接近的5分贝步长，并且将ANSI S12.13分析方案应用于原始数据集和四舍五入后的数据集。所有提议的ANSI测试统计量的大小均随听力测定步长的增加而增加。与差异阈值的标准差相比，顺序变差百分比和顺序变好或变差百分比方案对阈值步长明显更敏感。这些结果的变化方向和幅度支持这样一种假设，即标准草案中的标准范围对以5分贝步长收集听力测定数据的听力保护计划（HCPs）存在人为偏差。
