Title: Identifying risk factors for late-life dementia based on job characteristics during the working life Project participants: Péter Hudomiet (RAND, PI), Irineo Cabreros (RAND), Michael D. Hurd (RAND), and Susann Rohwedder (RAND) Revised Abstract 01/05/2021 To date, no treatment is available for Alzheimer’s disease and most types of Alzheimer’s disease-related dementias (AD/ADRD), even though this devastating condition affects many older adults. However, there are several known modifiable risk factors for AD/ADRD, which offer hope of finding interventions that may delay its onset. Investigating how job characteristics are related to the risk of AD/ADRD is a promising way of identifying such modifiable risk factors, because individuals spend a substantial portion of their lives working. Differences in job characteristics and work activities, for example, may explain (at least partially) why highly educated individuals face a substantially lower risk of developing dementia than lower educated individuals do. We propose to investigate the relationship between a large number of job characteristics and AD/ADRD using novel estimation methods, and a systematic and reproducible approach. We will use hundreds of occupational job characteristics from the Occupational Information Network (O*NET) database, which can be linked to the nationally representative Health and Retirement Study (HRS). The O*NET is promising for dementia research because it includes many measures describing cognitive activities, such as whether a job requires memorization, critical thinking, mathematical reasoning, or social orientation. The O*NET has hundreds of measures that are often strongly correlated. We will develop statistical methods to optimally process the items into easy-to-use low dimensional measures; and study their explanatory power for dementia. This study has four specific aims. First, we will develop and implement a methodology which we call Occupation-Wide Association Study (OWAS), to derive occupational risk factors of AD/ADRD. OWAS is inspired by the statistical approach employed in Genome-Wide Association Studies, which is used to identify genetic risk factors of various medical conditions. As such, the OWAS methodology will leverage established statistical techniques to handle the complex O*NET data with highly correlated items, and to correct for the number of false positives in multiple hypotheses testing. To complement the data-driven OWAS methodology, our second aim is to construct more detailed job characteristic measures guided by prior medical and social science research, such as quantitative skills, executive cognitive functions, job control, and social orientation. Third, we will examine the explanatory power of the developed scores for various cognitive outcomes in the HRS, such as age-adjusted probabilities of dementia and longitudinal change in cognition. Fourth, we will test how much of the correlation between basic demographic covariates and de...