. A Comparison of the Prevalence of Dementia in the United States in 2000 and 2012. JAMA Intern Med. 2017 Jan 1;177(1):51-58. PubMed.


Please login to recommend the paper.


  1. This study adds to the wave of evidence suggesting a decline in the risk of dementia in high-income countries in the past 25 years. This new study is important because it is based on a nationally representative sample from the United States. We had two studies suggesting a decline in the incidence (or risk) of dementia in the United States, one in Framingham, Massachusetts, and one in Olmsted County, Minnesota. Incidence is a more direct measure of risk than prevalence because prevalence is influenced by incidence and survival. However, the Framingham and Olmsted County studies were based on local communities with limited representation of demographic minorities. Thus, the study by Langa et al. provides useful complementary data.

    The study was well-conducted and well-described. The limitations are clearly recognized (e.g., use of short cognitive scales without a clinical examination; changes over time in the percent of use of proxy respondents; use of self-reported data).  

    Three important observations were made:  

    1) The prevalence declined even though the cardiovascular risk profile worsened (e.g., prevalence of hypertension, diabetes, and obesity).  These findings suggest that the increase in prevalence of cardiovascular risk factors was probably counterbalanced by an improvement in treatment (better treatment of hypertension and diabetes).  The increasing trends in late-life obesity and overweight need further study.  

    2)  The relative decrease over 12 years was 24.1 percent for dementia (2.8/11.6 percent) and 11.3 percent (2.4/21.2 percent) for cognitive impairment-no dementia (similar to mild cognitive impairment). The relative decline appears to be greater for dementia than for cognitive impairment-no dementia. This difference, if confirmed, could be useful in interpreting the trends.    

    3) A large segment of the decline remains to be explained. The OR of 0.69 (after age and sex adjustment) only increased to 0.82 after accounting for all of the factors considered. Therefore other yet unknown factors must have played a role.

    View all comments by Walter A. Rocca
  2. This paper from the Health and Retirement study estimates the prevalence of dementia in 2012 compared to 2000, finds a decrease in dementia prevalence, and tries to examine possible causes.

    The strengths of the study are the large samples of ~10,000 persons in each wave and a more diverse sample compared to the Framingham, CFAS, and Rotterdam studies, which included predominantly Caucasians. The findings were similar to those we observed in the Framingham Heart Study (FHS) with lower age-adjusted prevalence of dementia and increased education in more recent years. Further, risk of dementia associated with certain conditions (here, heart disease-stroke was not specifically assessed) was lower in the second epoch than the first.

    One limitation of the study is that a brief cognitive battery was used to categorize persons as possibly having dementia, although this battery was validated against a more detailed assessment in the ADAMS subsample. Further, data on whether or not someone had vascular risk factors was determined only by history and not by direct evaluation. Finally, the survey methods changed between 2000 and 2012, with more home visits and proxy interviews in the second phase. However, the last should, if anything, be expected to increase the percentage of persons with true dementia who were identified as being demented (increased sensitivity). 

    Overall, the HRS data reinforce the FHS findings that age-specific risk of dementia may be decreasing and that changes in vascular risk factors did not explain this change. 

    View all comments by Sudha Seshadri
  3. This analysis comes from a nationally representative and highly respected study that has added measures relevant to dementia identification in recent years. It is very carefully conducted research. These findings are very much in line with findings from another highly respected U.S. study (Framingham). Whilst its design is different from our own in the U.K., the findings are very similar to ours in the Cognitive Function and Ageing Study I and II published a couple of years ago, and also are in line with some of those we synthesized in the European review of studies across time. This study suggests, again, that a substantial drop in prevalence across time has occurred, such that at any given age an older person now is less likely to meet criteria for dementia than a similarly aged person a decade and more ago, leading to a reduction in age-specific prevalence. The possible reasons are not fully known but, using the wide and robust evidence on known risk for dementia as well as changes in profiles of population health (dramatic reductions in stroke and heart attacks), it seems likely that improvements across the whole life course, including education, nutrition, reduction in smoking, better management of vascular risk and diseases, all are likely to have made a contribution.

    In other words there have been major intergenerational influences on health, and dementia is not exempt from such changes despite its very close association with age. The underlying neurobiology of dementia is likely to be a complex interplay of age, genetic risk for specific protection or risk trajectories (large, small, and interactions), life course risk/protection behaviors such as education and smoking, compensatory or reserve factors, and other exposures. We know that education is associated with compensation so that more people with higher education given a particular level of Alzheimer’s-type neuropathology can die without developing dementia, compared to those with lower levels of education. Thus we continue to learn the lesson that late-life dementia is not straightforward, nor is it likely to be amenable to simple solutions. We need to understand the influences that have led to the changes observed and also to ensure that policy makers who can influence the life course of the population's brain health are aware that policies and practice many decades ago seem to be playing out now, and current policies should be assessed for their potential to maintain and continue such positive trends.

    These findings are not likely to be the same in lower- and middle-income countries where life course experiences are very different across the generations.  

    View all comments by Carol Brayne

Make a Comment

To make a comment you must login or register.

This paper appears in the following:


  1. U.S. Dementia Rates Fall