Enhancing data quality for cross-national harmonization: Assessment of cognitive function in the CHARLS HCAP by language, literacy, and visual impairment

阿尔茨海默病和相关痴呆症(ADRD)是全球公共卫生和政策的重大挑战。这个 NIA通过资助协调认知评估来投资解决全球ADRD负担 礼宾(HCAP)美国健康和退休及其国际合作伙伴研究的子研究。HCAP 是一项创新，旨在首次实现对老年人认知功能的可比测量 世界各地的成年人。中国健康与退休纵向研究中的HCAP是第一 部署在2018年CHARLS的第四波中，面向60岁的≥参与者。中国是一个幅员辽阔、多元化的国家。 它的人口中至少有九种语言和七种普通话方言。文盲(35.3%) 未矫正视力损害(患病率为18.3%)在老年人群中也很常见。 然而，为CHARLS HCAP准备和管理的采访材料假设在#年流畅 口语和书写标准普通话，没有系统地进行调整，以说明文盲或视觉障碍 减损。如果认知测试项目的表现受到这些无关特征的影响，而不是 被试者的真实认知水平，那么差异项目功能(DIF)被认为存在于该认知 测试项目。虽然DIF可能会导致有偏见的认知分数，但我们有工具可以在给定假设的情况下调整 这样的DIF可以改善认知测量。此补充应用程序的目标是去污 CHARLS HCAP认知测试组合中因口语或方言而产生的测量差异，书面 识字能力和视觉能力来自真正潜在的认知功能水平。父R01(R01AG070953)旨在 在五个不同的国家和地区统计协调HCAP措施和测试国家一级的DIF (美国、墨西哥、英国、南非和印度)。本补充申请是 父母资助的统计机制和科学范围。我们建议1)识别和调整潜力 CHARLS HCAP中认知功能的测量因语言或方言、识字和视觉的差异 能力；以及2)在CHARLS前一波的认知功能测量中识别和调整潜在的DIF (波1、2、3)，按语言或方言、识字和视觉能力。进行DIF分析和调整后 根据DIF所需的认知分数，我们将检查年龄、性别、教育程度、 城乡居民DIF调整后的认知分数与非DIF调整后的认知分数在某些方面会有所不同 这表明在DIF调整后，由于语言或方言、文盲和视力障碍而导致的偏见减少 得分。成功完成这项拟议补充资料将有助于提高现有CHARL的质量 认知数据，并将为内容、管理和面试官培训提供建议 认知测试电池在未来的CHARLS HCAP计划浪潮中。我们预计这项工作将使我们能够 建议在采取认知措施的同时收集更多信息，以提高数据质量并提供帮助 解释，如语言、读写能力和视觉能力的客观测量。

Alzheimer’s disease and related dementias (ADRD) are a major global public health and policy challenge. The NIA has invested in addressing the global ADRD burden through funding the Harmonized Cognitive Assessment Protocol (HCAP) sub-studies of the US Health and Retirement and its International Partner Studies. The HCAP is an innovation that aims to allow, for the first time, comparable measurement of cognitive function among older adults around the world. The HCAP in the China Health and Retirement Longitudinal Study (CHARLS) was first deployed in Wave 4 of CHARLS in 2018 to participants aged ≥60 years. China is a large, diverse country with at least nine languages and seven dialects of Mandarin spoken amongst its population. Illiteracy (35.3% prevalence) and uncorrected visual impairment (18.3% prevalence) are also common in the older population. However, interview materials for the CHARLS HCAP were prepared and administered assuming fluency in spoken and written standard Mandarin and were not systematically adapted to account for illiteracy or visual impairment. If performance on a cognitive test item is affected by these extraneous features rather than a respondent’s true level of cognition, then differential item functioning (DIF) is said to be present for that cognitive test item. While DIF can lead to biased cognitive scores, we have tools that can, given assumptions, adjust for such DIF to improve measurement of cognition. The goal of this supplemental application is to decontaminate measurement differences in the CHARLS HCAP cognitive test battery due to spoken language or dialect, written literacy, and visual ability from true underlying levels of cognitive function. The parent R01 (R01AG070953) aims to statistically harmonize HCAP measures and test for DIF at the country-level across five different countries (US, Mexico, England, South Africa, and India). The present supplemental application is an extension of the statistical machinery and scientific scope of the parent grant. We propose to 1) identify and adjust for potential DIF in measurement of cognitive function in CHARLS HCAP due to language or dialect, literacy, and visual ability; and 2) Identify and adjust for potential DIF in measurement of cognitive function in prior waves of CHARLS (Waves 1, 2, 3), by language or dialect, literacy, and visual ability. After performing DIF analysis and adjusting cognitive scores as needed for DIF, we will examine whether the associations of each of age, sex, education, and urban/rural residence with DIF-adjusted cognitive scores will differ from the non-DIF-adjusted scores in ways that suggest reduction of bias due to language or dialect, illiteracy, and vision impairment in the DIF-adjusted scores. Successful completion of this proposed supplement will help improve the quality of existing CHARLS cognitive data, and will produce recommendations for the content, administration, and interviewer training for the cognitive test battery in future planned waves of CHARLS HCAP. We anticipate that this work will enable us to recommend additional information to collect alongside cognitive measures to enhance data quality and assist interpretations, such as objective measurement of language, literacy, and visual ability.