Statistics and Methodology Research
We are developing and disseminating statistical methods and evaluation approaches to improve our understanding of how health care services and systems can improve the lives of individuals with mental illness and addiction.
Featured Recent Research
Combining randomized trials and non-experimental data to examine treatment effect heterogeneity
Randomized trials offer unbiased treatment effect estimates but are generally underpowered to examine treatment effect heterogeneity. In contrast, data sources such as electronic health records offer large samples and information on the use of services in practice, but suffer from confounding. Dr. Stuart and colleagues are working on methods for combining data sources to understand treatment effect heterogeneity, including applications in mental health and substance use. Learn more:
Policy evaluation methods
Multiple Center researchers are involved in mental health and substance use policy evaluations. These evaluations often utilize data from multiple locations over time, and compare locations or groups that experienced a policy change to those that did not. Numerous methodological questions come up, such as how to best utilize individual patient data (e.g., from health insurance claims data) for these analyses, and how to account for variation in effects across groups and time, as well as how to best integrate qualitative and quantitative information.
- Using difference-in-differences in mental health services research
- Ready to roll? Practical guidance on whether and when to aggregate data in health policy evaluation
- Moving beyond the classic difference-in-differences model: A simulation study comparing statistical methods for estimating effectiveness of state-level policies
- Methodological challenges and proposed solutions for evaluating opioid policy effectiveness
- Protocol: Mixed-methods study to evaluate implementation, enforcement, and outcomes of U.S. state laws intended to curb high-risk opioid prescribing
Generalizing results from randomized trials to populations of interest
One way to make research results more relevant for policy making is to ensure that the subjects in studies reflect the individuals in the population that may be affected by any policy change. Through grants from the National Institutes of Health, Dr. Stuart and Dr. Mojtabai and colleagues have a body of research examining how well results from randomized trials may carry over to target populations of interest, including methodological work and applications, such as to the NIDA Clinical Trials Network.
Estimating causal effects in non-experimental settings
Dr. Stuart and colleagues, including students, are developing and assessing statistical methods for better estimating causal effects in non-experimental settings. These methods have been applied to estimate the effects of behavioral therapy after suicide attempt, alternative health care financing models, and nursing home report quality report.
- The use of propensity scores in mental health services research
- Using marginal structural models in mental health services research
- Using mixed methods to inform causal inference
- Propensity score methods for observational studies with clustered data: A review
- Generalizing Observational Study Results: Applying Propensity Score Methods to Complex Surveys
- Matching Methods for Causal Inference: A Review and a Look Forward
- John McGready interviews Dr. Liz Stuart