Association for Public Policy Analysis & Management

We estimate trends in age-specific rates of dementia in the U.S. between 1998 and 2014, using a large population representative sample from the Health and Retirement Study (HRS). To advance our understanding of what underlies the trends we analyze composition effects and within-group trends by gender, race, education, and lifetime earnings. Dementia affects many older adults, and it is a very costly disorder, primarily because many individuals who live with dementia require help with activities of daily living. The risk of dementia substantially increases with age. Assuming no change in age-specific rates, the prevalence of dementia will increase substantially, potentially tripling by 2050, reaching 130 million worldwide. However, there is some recent evidence that age-adjusted dementia prevalence may be declining in developed countries, possibly due to rising levels of education and better treatment of key cardiovascular risk factors. Any change in these age-specific rates has important implications for projected prevalence and associated costs, such as nursing home demand in the future. Most of the evidence in the literature about trends in dementia is based on small community studies that are not representative of the population. We use the HRS to estimate trends in age-specific rates of dementia because it has several unique features for the analysis: 1) it is a large, nationally representative, longitudinal dataset with a wide range of information about health status, cognitive abilities, SES, demographics, and other variables; 2) it has good coverage of the population with dementia because it follows individuals into nursing homes; 3) if a sample member cannot complete an interview the HRS conducts an interview with a proxy such as spouse; 4) a subset of HRS members (N=856) were administered a clinical assessment for dementia within the Aging, Demographics and Memory Study (ADAMS). We use Markov Chain Monte Carlo to estimate a latent variable statistical model of dementia. The main identification of the model comes from the clinical diagnosis of dementia from ADAMS to calibrate the cognitive variables that are available in the entire HRS, such as word recall or self-rated memory. The main identification of trends in dementia comes from observed trends in the HRS cognitive variables. Our preliminary findings suggest that the prevalence of dementia declined in the last 20 years, but the rate of decline was considerably smaller than what others have found. We are currently improving our model by incorporating more data and estimating prevalence rates over time by detailed demographic groups. We anticipate that differential changes in prevalence within subpopulations will be informative about the causes of the trends. For example, partially due to changes in the compulsory schooling laws in the first half of the 20^th century, the level of education substantially increased across the cohorts observed in the HRS. We will investigate the co-movement of the level of education and dementia across these cohorts to gain insights about the causal effect of education on the probability of dementia.