Roee Gutman

Image
Roee Gutman
Assistant Professor of Biostatistics
Mike Cohea/Brown University
Coffee is good for you; coffee is bad for you. Media reports on health issues often seem contradictory. Roee Gutman is developing statistical methods to make causal inference more accurate and defensible. “That’s my area. Understanding causality from observational studies.”

Pretty much anyone who follows the news has had this frustration. One week the newspaper reports that drinking a little red wine could help you live longer; the next week, it seems to say exactly the opposite.

That Roee Gutman finds such apparent contradictions to be silly is hardly remarkable. But Gutman is rare in his capacity to do something about it. As a biostatistician, his research focuses on developing techniques to make a more accurate, or at least defensible, “causal inference” from epidemiological studies.

“We are not trying to find out if drinking alcohol is correlated with living longer,” said Gutman, assistant professor of biostatistics in the Program in Public Health at Brown. “We actually are interested in finding out if alcohol is making you live longer.”

But sometimes, either because of flawed methods or lacking data, researchers infer cause without making a valid comparison between truly similar groups. If people who are sick drink less alcohol and die earlier, then a study had better compare people in similar health if its authors are to offer conclusions about drinking and longevity.

“The idea is to try to create groups that are similar,” Gutman said. “That’s my area. Understanding causality from observational studies.”

That was a focus of his thesis at Harvard, completed this year. It was clearly promising work. Last year he won a student paper award from the Health Policy Statistics Section of the American Statistical Association for his work on another major section of the thesis: a statistical method for linking information from two different data sets when they are related but not explicitly overlapping.

Gutman needed to do this to answer a morbid but important policy question: What are the end-of-life medical costs associated with different diseases? To even begin to find answers, he had to stitch together a Medicare database, which provides cost information but not a cause of death, and vital statistics mortality files, which provide a cause of death, but not cost information. The key is to come as close as possible to determining which data pertains to the same person.

The databases don’t include names (that would be too easy and would violate patient privacy) but they do include time, date and place of death, race, gender, and age. In a small town, only so many 89-year-old Hispanic women will die at 3:55 p.m. on a given day. But in big cities, data often overlaps. Gutman’s thesis work entailed coming up with a reasonable way of positing the most likely linkages between the databases so that he could get on to the work of determining what Medicare spends in the last six months of life for various terminal ailments.

Gutman’s interest in medicine as the application of his statistics talent runs deep. His father, Haim, is a surgical oncologist. Gutman attended one of his father’s operations as a child and admired his dad’s mission, if not his sometimes upredictable hours. Gutman is now following in his father’s footsteps, in his own way.

“At some point I thought that I might become a doctor, or at least he thought I might become a doctor, but he’s actually pretty happy that there is a statistician who can do the analysis for him,” Gutman said.

The father-son pair published together in the journal Cancer in 2002. The subject was the occurrence of clotting problems in melanoma patients treated with Interferon. They have since worked together on other publications related to cancer treatment in octogenarians, and melanoma of the extremities, and are still collaborating on other projects.

As a statistician deeply interested in medicine, Gutman is a natural fit for the new biostatistics department at Brown. He said people here have seemed interested in his causal inference work as well as his research on linking records from different data sets.

Now that he’s here, he’s interested in tackling questions including how to properly assign credit for a therapeutic effect when patients have used more than one therapy (e.g., two or more different drugs) during the course of their illness. Another one is how to monitor the performance and effectiveness of doctors over time.

But however good a fit Gutman is, there was another reason to stay in New England after Harvard. Despite being an Israeli native, he’s mad for the region’s teams.

“I won’t have to change my allegiance to the Red Sox, Celtics, and Patriots,” he said.

Browse New Faculty by Year