New technologies and ballooning data sets allow us to model problems and test solutions in ways inconceivable just a decade ago. Nowhere is this more apparent than in the efforts of MIT researchers working to improve human health—as these examples show.

EVALUATING CARE MANAGEMENT
Joseph Doyle, Erwin H. Schell Associate Professor of Management in applied economics / Amy Finkelstein PhD ’01, John and Jennie S. MacDonald Professor of Economics
Five percent of the US population accounts for more than half of our health care expenditures. The Camden Coalition of Healthcare Providers developed a care management program that uses a data-driven “hotspotting” process to identify these high-needs patients and to provide them with medical management and assistance in accessing social programs. With support from the Abdul Latif Jameel Poverty Action Lab (J-PAL), Doyle, Finkelstein, and colleagues have partnered with the New Jersey organization on a randomized evaluation of its program. The study will generate evidence on whether the Camden approach is effective—as measured by hospital readmissions, health care utilization, participation in social programs, earnings, employment, and mortality.

MATH AND PHYSIOLOGY
Thomas Heldt PhD ’04, W. M. Keck Career Development Professor of Biomedical Engineering; faculty at IMES
Heldt develops modeling techniques to understand physiological signals, such as arterial blood pressure or cerebral blood flow velocity waveforms. For example, his group is refining an algorithm that estimates intracranial pressure, which is important to measure in a variety of neurological conditions, including traumatic brain injury, hydrocephalus, or hemorrhagic stroke. “We try to incorporate our understanding of the underlying physiology into the analysis tools,” Heldt recently told MIT News. “We can represent the physiology mathematically through differential equations or algebraic equations, and then interpret the data within the context of those mathematical relationships.”

4 BATCHES
Philippe Rigollet, associate professor of mathematics; core member, Center for Statistics; affiliate member, Broad Institute
The design of ethical clinical trials must compromise between caution (testing one subject at a time) and efficiency (all of them simultaneously). The former optimizes outcomes for the most patients but is impractical; the latter maximizes the speed of research, but sacrifices potential benefits to a large segment of study enrollees. A solution is to group subjects into smaller batches: the FDA currently recommends four batches, but their size is not precisely regulated. Rigollet and collaborators developed a mathematical technique for sizing each batch. They have demonstrated that by applying their rule, four batches can yield patient outcomes comparable to the most cautious procedure of testing subjects one at a time, while presenting almost no loss of efficiency in relation to current standards.

1 IN 4 DEATHS
Laurie Boyer, associate professor of biology and biological engineering / Manolis Kellis ’99, MNG ’99, PD ’03, PhD ’03, professor of computer science; faculty at CSAIL and Broad Institute
Heart disease is the leading cause of death in adults, accounting for one in four deaths in the US. Genetic variants from genome-wide association studies can reveal new therapeutic targets, but their discovery requires impractically large patient groups. To speed up this search, researchers from the Boyer and Kellis labs analyzed known genetic markers that increase heart disease risk, and found that they show common epigenetic signatures. Those signatures enabled them to identify several additional genetic contributors to heart disease with much smaller cohorts than would otherwise be necessary. The researchers plan to apply this strategy to help reveal new therapeutic targets in diverse inherited diseases, including Alzheimer’s, schizophrenia, and type 2 diabetes.

COMMUNITY-BASED DATA
Mariana Arcaya, assistant professor of urban studies and planning
The nonprofit creators of the Healthy Neighborhoods Equity Fund have partnered with an MIT team, led by Arcaya, on a longitudinal study of how its investments in transit-oriented development projects affect residents’ health. Focusing on three Boston-area communities experiencing interventions and six control communities, the MIT researchers partnered this summer with community members to field surveys on topics from health to violence and displacement. This fall, the residents will help the researchers analyze the survey results, which will be integrated with findings from a statewide All-Payer Claims Database. Says Arcaya, “We want to combine the power of big data with rich, qualitative data that is coproduced and co-owned with community-based organizations.”

1 IN 7 WOMEN
Regina Barzilay, Delta Electronics Professor of Electrical Engineering and Computer Science; faculty at CSAIL
One out of seven women will be diagnosed with breast cancer—and two years ago, Barzilay became one of them. Now she is leveraging the tools of her field, natural language processing, to advance cancer research. Many physicians’ observations and clinical findings are still recorded in free form, with variable wordings. This valuable information is much harder to incorporate into research than the more structured yet limited data sets produced by clinical trials. Collaborating with oncologists from Massachusetts General Hospital, Barzilay employs machine-learning methods operating over free text and images to improve models of disease progression and clinical decision making.

THE PATH OF A SNEEZE
Lydia Bourouiba, Esther and Harold E. Edgerton Assistant Professor of civil and environmental engineering; associate faculty at IMES
What did high-speed cameras teach Bourouiba and colleagues about humble sneezes and coughs? They revealed the form our mucosalivary fluid takes as it launches from our bodies, morphing within a second from a balloon of fluid into a spray of droplets. Such insights could help to map the spread of infections in close quarters. According to Bourouiba, there are clear limits to the accuracy of data about transmission routes acquired via traditional low-tech methods such as surveying people about their movements and interactions: “We are trying to have direct measurements and experimentally validated mechanistic models of contamination risks to root disease control and prevention in the physical sciences.”

Topics

Share your thoughts

Your email address will not be published. Required fields are marked *