Sharon-Lise T. Normand is the principal investigator on this project. Jennifer Grandfield is the project contact: 617-432-3287.

A team of statisticians from Harvard School of Public Health and Harvard Medical School are collaborating in research to develop methodology for the analysis of multiple informants or multiple assessment data to measure mental health outcomes or risk factors in community and service-based samples.  In outcomes research it is common to collect multiple outcomes in order to characterize treatment effectiveness or to evaluate the quality of care delivered by health providers. Typically, these outcomes are modeled individually rather than multivariately. Statistical problems arise when the outcomes are noncommensurate—that is, they are measured on different scales, such as binary and continuous.

One focus of the research involves the development and application of multivariate models for the analysis of noncommensurate. Likelihood-based approaches (such as factorization and latent variable approaches) as well as quasi-likelihood approaches (generalized estimating equations) are developed in the context of no missing data and missing data. The methods are applied to assess the quality of care following a crisis for schizophrenia patients and using national data to evaluate quality of care delivered by acute care hospitals.