What Causes Hearing Loss: Advancing the Methods

项目总结/摘要 现有的统计方法不足以对听力测定评估的听力进行质量控制或分析 测量.听力测定评估的听力测量的统计分析具有挑战性， 数据的复杂相关结构，并且因为听力损失的定义通常基于 纯音平均值通常是在三个或四个指定频率处的测量的平均值， 导致个体频率数据中可用的有用信息可能丢失。此外，现在 有可能在诊所外使用廉价的电子设备进行听力测试， 规模流行病学听力研究。在这样做时，统计分析必须考虑到测量误差， 听力测试，以防止偏见的估计协会和因果关系的影响。我们将开发全新的 用于听力数据的质量控制以及用于有效和高效地评估听力损失的方法 关联及其因果关系，同时考虑到多层次的相关性， 听力测定评估的听力数据中多个频率的多变量结果。我们将开发 方法，以纠正测量误差引起的偏差，估计协会和因果关系的影响， 在非临床环境中测量听力结果的研究。我们将应用听力数据分析 以护士健康研究2为基础的听力保护研究方法。用户友好公开 可用软件开发将是所有待开发新方法的核心特征。我们 已经形成了一个跨学科的专家理论和应用统计学家，流行病学家和 听力学家，我们希望有很好的装备，以解决已确定的挑战性问题。

Project Summary / Abstract Existing statistical methods are insufficient for quality control or analysis of audiometrically-assessed hearing measurements. Statistical analysis of audiometrically-assessed hearing measurements is challenging due to the complex correlation structure of the data, and because the definition of hearing loss is typically based on the pure tone average which is typically the average of measurements at three or four specified frequencies, leading to a potential loss of useful information available in the individual frequency data. In addition, it is now possible to perform audiometric testing outside of the clinic with inexpensive electronic equipment in large- scale epidemiologic hearing studies. In doing so, statistical analysis must account for measurement error in the hearing tests to prevent bias in the estimates of associations and causal effects. We will develop entirely novel methods for quality control of hearing data, and for validly and efficiently assessing the exposure-hearing loss associations and their causal relationships, while accounting for the multiple layers of correlation and multivariate outcomes from multiple frequencies in audiometrically-assessed hearing data. We will develop methods to correct for measurement error-induced bias in the estimated associations and causal effects for studies where hearing outcomes are measured in non-clinical settings. We will apply the hearing data analysis methods to the Conservation of Hearing Study based in the Nurses’ Health Study II. User-friendly publicly available software development will be a central feature accompanying all new methods to be developed. We have formed an interdisciplinary team of expert theoretical and applied statisticians, epidemiologists and audiologists, and we expect to be well equipped to solve the challenging problems that have been identified.