Maria Glymour, SD

Research Interests
- Alzheimer's disease and related causes of cognitive aging and dementia
- Social determinants of health and health equity
- Social policies and health
- Causal inference in social epidemiology and dementia research

My research focuses on how social factors experienced across the lifecourse, from infancy to adulthood, influence cognitive function, dementia, stroke, and other health outcomes in old age. I am especially interested in education and other exposures amenable to policy interventions. The health of current cohorts of elderly individuals in the US reflect a lifetime of social exposures, including educational experiences shaped by major changes in schooling policies. Education is especially interesting because it is such a powerful predictor of health and historically, access to education has frequently been restricted based on race, gender, and other socially enforced criteria. One thread of my research examines how changes in schooling laws and school quality in the 20th century might have influenced the health and cognitive outcomes of current cohorts of elderly, including adults subject to race-based school segregation. Our results suggest that extra schooling has substantial benefits for memory function in the elderly. I have also worked on the influence of "place" on health, for example to understand the excess stroke burden for individuals who grew up in the US Stroke Belt. In a project with colleagues including Drs. Rachel Whitmer, Elizabeth Rose Mayeda, and Paola Gilsanz, we are continuing a unique multi-ethnic cohort of older adults in Northern California, with a wealth of lifecourse biological and social data to offer insight into the reasons for racial/ethnic differences in Alzheimer's and dementia risk (https://rachelwhitmer.ucdavis.edu/khandle).

A separate theme of my research focuses on overcoming methodological problems encountered in analyses of social determinants of health, Alzheimer's disease, and dementia. For many reasons, research focusing on lifecourse epidemiology as well as cognitive aging introduces substantial methodological challenges. Sometimes, these are conceptual challenges, and clear causal thinking can help! Many of these challenges are being addressed in the MELODEM (MEthods in LOngitudinal research on DEMentia) initiative, an international group of researchers focusing on analytic challenges in research on dementia and cognitive aging. MELODEM has working group phone calls on the first and third Thursdays of the month, open to all (https://sites.bu.edu/melodem/). My group works with numerous colleagues on methods to improve measurement, including crosswalking across data sets. For example, in work with Dr. Zeki Al Hazzouri, we are linking data sets with detailed information at different lifecourse periods -- e.g., childhood, early adulthood, and later adulthood -- to better evaluate long-term effects of exposures at specific sensitive ages. In work with Dr. Cathy Schaefer, Ron Krauss, and many others, we are fielding emulated trial designs in the large, diverse Kaiser Permanente Northern California cohort. This setting is exceptional for emulated trial designs because of the large size, long follow-up, and combination of high-quality clinical data plus social and genetic information for large groups of study participants.

I have advocated the use of causal directed acyclic graphs (DAGs) as a standard research tool to represent our causal hypotheses and help elucidate potential biases in proposed analyses. In other cases, the methodological problems require more analytical solutions that have been developed elsewhere in epidemiology or in other disciplines, but are rarely applied to these research questions. Instrumental variables analyses of natural or induced experiments are one promising example. Genetic variations have recently been advanced as possible instrumental variables to estimate the health effects of a wide range of phenotypes, an approach sometimes called “Mendelian Randomization.” Using genetic polymorphisms as instrumental variables could provide a very powerful tool for social epidemiology, but the inferences from such analyses rest on strong assumptions. Thus I am currently working with a team to explore approaches to evaluating the plausibility of those assumptions in applications for social epidemiology.

Students and post-doctoral fellows interested in research collaborations related to my work are welcome to send me an email directly or contact Marta Flores, mdflores@bu.edu, who handles my calendar.