Douglas Levy

Douglas Levy

Associate Professor, Medicine, Harvard Medical School
Core Faculty of the Mongan Institute Health Policy Center, Massachusetts General Hospital

Dissertation Title:  "Investigating Policies to Reduce Tobacco Use and Harm: Findings and Methods"

This dissertation consists of three chapters, each a separate paper, that taken together examine policies aimed at reducing the prevalence of tobacco use in the United States as well as methodologies used in public health research.

In the first paper, I used Monte Carlo simulations to analyze how adding coverage for smoking cessation interventions to a private health insurance plan would affect health expenditures within the plan and to determine whether insurers have internal incentives to offer smoking cessation benefits. I also examined how adding such a benefit would affect overall medical expenditures and Medicare expenditures for a cohort of enrollees to determine whether there are societal benefits associated with insurance coverage of smoking cessation interventions. I find that private insurers do not have internal incentives to offer smoking cessation benefits, but there are sufficient societal gains to warrant policies encouraging their provision.

In the second paper, written with Ellen Meara, we analyzed the effect of the 1998 Master Settlement Agreement (MSA) on maternal smoking and birth weight outcomes. The MSA resulted in an immediate 22% increase cigarette prices. Previous literature predicted such an increase would reduce maternal smoking by 10-20% and reduce low birth weight outcomes by 2-4%. Using birth certificate data and time-series analysis, we find only a 2.2% reduction in maternal smoking prevalence and a 1.8% reduction in low birth weight outcomes following the MSA, though there were larger effects for teenage mothers. We conclude that price changes may not be as promising a method for reducing maternal smoking as earlier predicted.

The third paper is work done together with James O'Malley and Sharon-Lise Normand. We extended a published method for reducing bias in clinical trials where there are missing data and treatment non-compliance to include adjustment for continuous covariates. Monte Carlo simulation methods were used to compare the performance of our covariate-adjusted estimators to previously established estimators under a variety of circumstances. Our estimators had the smallest mean squared errors and the least bias under the broadest range of circumstances. We illustrated our methods using data from a clinical trial comparing antipsychotic drugs.


Graduation Year