Abstract
Population aging in the US and its increase in racial/ethnic diversity has resulted in a growing body of literature aimed at measuring health disparities among minority older adults. Disparities in health outcomes are often evaluated using self-reported measures and, to attend to linguistic diversity, these measures are increasingly being used in languages for which they were not originally developed and validated. However, observed differences in self-reported measures cannot be used to infer disparities in theoretical attributes, such as late-life depression, unless there is evidence that individuals from different groups responded similarly to the measures-a property known as measurement invariance. Using data from the Positive Minds-Strong Bodies randomized controlled trial, which delivered evidence-based mental health and disability prevention services to a racially/ethnically diverse sample of minority older adults, we applied invariance tests to two common measures of anxiety and depression (the GAD-7 and the HSCL-25) and two measures of level of functioning (the Late-Life FDI and the WHODAS 2.0) comparing four different languages: English, Spanish, Mandarin, and Cantonese. We found that these measures were conceptualized similarly across languages. However, at the item-level symptom burden, we identified a non-negligible number of symptoms with some degree of differential item functioning. Spanish speakers reported more worry symptoms and less somatic symptoms for reasons unrelated to their psychological distress. Mandarin speakers reported more feelings of restlessness, and both Mandarin and Cantonese speakers reported no interest in things more often for reasons unrelated to their psychological distress. Mandarin and Cantonese speakers were also found to consistently report more difficulties performing physical activities for reasons unrelated to their level of functioning. In general, invariance tests have been insufficiently applied within psychological research, but they are particularly relevant as a prerequisite to accurately measure health disparities. Our results highlight the importance of conducting invariance testing, as we singled out several items that may require careful examination before considering their use to compare symptoms of psychological distress and level of functioning among ethnically and linguistically diverse older adult populations.