Gross Alden L, Liu Ying, Zhang Yuan S, Zhao Yaohui, Li Chihua, Meijer Erik, Lee Jinkook, Kobayashi Lindsay C
Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, 616 N. Wolfe St., Baltimore, MD, 21205, USA.
Center On Aging and Health, Johns Hopkins University, 2024 E. Monument St., Baltimore, MD, 21205, USA.
Aging Clin Exp Res. 2025 May 12;37(1):146. doi: 10.1007/s40520-025-03039-y.
The potentially biasing impacts of low language fluency, illiteracy, and sensory impairments on cognitive test performance are unknown, which may have implications for understanding their roles in cognitive decline and dementia.
We investigated effects of these features on cognitive test item completion and performance among older adults in China, a multilingual country with high prevalence of illiteracy and sensory impairment.
We used cognitive test data from the Harmonized Cognitive Assessment Protocol of the China Health and Retirement Longitudinal Study conducted in 2018 (N = 9755, age 60 + years). We first tested associations of fluency in spoken Mandarin, literacy, and sensory impairment (hearing and vision) with missingness of cognitive items. We then tested for differential item functioning (DIF) in observed cognitive items by these features.
We observed high levels of missing data in most cognitive test items - on average 13% and as high as 65%. Low fluency in spoken Mandarin, illiteracy, and impairments in hearing and vision were each associated with greater odds of missingness on nearly all tests. Partly because of differential missingness, there was minimal evidence of DIF by these features in items in which we expected a priori to find DIF (e.g., repetition of a spoken phrase among those with hearing impairment). Several cognitive test items exhibited statistically significant DIF, however there was minimal evidence of meaningful DIF.
Differential missingness in cognitive items by spoken language, literacy, and sensory impairments is potentially more of an inferential threat than measurement differences in test items.
语言流利程度低、文盲以及感官障碍对认知测试表现的潜在偏倚影响尚不清楚，这可能对理解它们在认知衰退和痴呆中的作用具有重要意义。
在中国这个文盲和感官障碍患病率高的多语言国家，我们调查了这些特征对老年人认知测试项目完成情况和表现的影响。
我们使用了2018年中国健康与养老追踪调查的统一认知评估协议中的认知测试数据（N = 9755，年龄60岁及以上）。我们首先测试了普通话流利程度、识字能力和感官障碍（听力和视力）与认知项目缺失之间的关联。然后，我们通过这些特征测试了观察到的认知项目中的差异项目功能（DIF）。
我们观察到大多数认知测试项目中存在大量缺失数据——平均为13%，高达65%。普通话流利程度低、文盲以及听力和视力障碍在几乎所有测试中都与更高的缺失几率相关。部分由于差异缺失，在我们先验预期会发现DIF的项目中（例如，听力障碍者重复口语短语），几乎没有证据表明这些特征存在DIF。然而，几个认知测试项目表现出统计学上显著的DIF，但几乎没有证据表明存在有意义的DIF。
口语、识字能力和感官障碍导致的认知项目差异缺失可能比测试项目中的测量差异更具推断性威胁。
