Shaw-Shepherd Research Collaborative
There is growing interest in using observational data from electronic health records (EHR) and other readily available data sources as cost-effective resources to support medical research. These types of data are subject to error. Spurious or unrecognized associations driven by unvalidated outcomes and exposures could misguide clinical applications and in turn prove harmful to patients. Dr. Pamela Shaw and Bryan Shepherd co-lead two research grants focused on novel statistical methods and study designs to address error-prone data.
Grants Funding
NIH R01 Grant:
Statistical Methods for Correlated Outcome and Covariate Errors in Studies of HIV/AIDS
Principal Investigators: Pamela Shaw and Bryan Shepherd (NIH R01-AI131771)
For this project, we will create novel statistical methods for estimation and study design to reduce or eliminate bias caused by correlated errors in failure-time outcomes and associated covariates. The proposed methods are motivated by and applied to electronic health record data from studies of co-morbidities, particularly cancers, among HIV-infected patients. Open sources tools will be developed to allow researchers to implement these methods and designs.
(link: https://reporter.nih.gov/search/Ibm4KEU3bkOmwLFIWzg44Q/project-details/10107734)
PCORI Methods Grant:
Statistical Methods and Designs for addressing Correlated Errors in Outcomes and Covariates in Studies using Electronic Health Record Data
Principal Investigators: Pamela Shaw and Bryan Shepherd
We will develop new methods and to improve inference in studies that are reliant on error-prone observational data, such as large electronic health records data. To ensure data are of high enough quality to support scientific investigations, data are audited/validated. We will develop a general class of statistical methods can be used in conjunction with data validation and audits to save money, improve estimation, and guide data quality decisions. Specific project goals include: 1) develop novel statistical methods that reduce or eliminate bias caused by correlated errors in time-to-event outcomes and covariates, 2) design optimal multi-wave validation/audit strategies and develop an audit/validation design tool kit, and 3) apply methods and designs to a study of risk factors for early childhood obesity and asthma.