HSRProj will be retired on September 14, 2021, no updates will be made to HSRProj after this date.
Detailed information about this transition can be found on the June 3, 2021 Technical Bulletin post.
If you have questions or suggestions, please contact NLM Customer Service.
Information about ongoing health services research and public health projects
|The handling of missing data induced by time-varying covariates in comparative effectiveness research involving HIV patients|
|Investigator (PI):||Desai, Manisha|
|Performing Organization (PO):||
(Current): Stanford University, School of Medicine, Department of Medicine, Biomedical Informatics Research Division, Quantitative Sciences Unit / (650) 725-1946
|Supporting Agency (SA):||Patient-Centered Outcomes Research Institute (PCORI)|
|Record Source/Award ID:||PCORI/ME-1303-5989|
|Funding:||Total Award Amount: $802,122|
|Award Information:||PCORI: More information and project results (when completed)|
|Abstract:||Advances in antiretroviral therapy (ART) have dramatically reduced mortality from HIV, enabling reclassification of HIV as a chronic condition. Numerous studies suggest that some drugs increase the risk of cardiovascular disease, although findings are inconsistent. Studies differ largely due to methodological choices, including study design, definition of exposure, and approaches to handling missing data. It is crucial to incorporate information on drug exposure and other confounders over time; patients vary their regimens over time and for reasons that may be related to their condition. Thus, in order to not implicate the wrong drug, information on changes in regimen and other factors must be considered. Including information over time complicates the analysis, however. One such complication is the introduction of missing data. Common methods for handling missing data yield misleading descriptions of relationships. Appropriate methods for handling missing data are computationally burdensome; software does not exist for many situations and the analyst must rely on his/her own programming skills to implement specialized techniques. Multiple imputation (MI) is a reasonably accessible and theoretically sound method for handling missing data. Available in mainstream software, its special application is required due to the unique issues posed by time-varying covariates and outcomes that are only partially observed for those individuals who do not experience a cardiovascular event during the observation period. We propose an extensive simulation study to evaluate commonly applied methods to this setting, to investigate the performance of standard MI in this context, and to adapt and evaluate MI methods utilized in a longitudinal setting where the outcome is fully observed to this particular setting. Based on our findings we will develop concrete guidelines on how to use MI in the context of partially observed outcomes and time-varying covariates. We will develop user-friendly open-source software in order to optimize the use of recommended methods and to eliminate lack of software as a barrier to employing missing data methods. Finally, we will illustrate methods considered on data from the U.S. veteran population of HIV-infected individuals using the Veterans Health Administration's rich longitudinal Clinical Case Registry (CCR), the analysis of which motivated this proposal. This work has the potential to greatly impact patients living with HIV. Currently, there is no consensus on which ART agents increase cardiovascular risk. Our proposal will address the implications of methodological choices for handling missing data when conducting comparative effectiveness research in the longitudinal setting. Importantly, the development of guidelines will unify analyses enabling combination of evidence across studies in the form of meta-analyses, and accessibility to software will eliminate barriers to incorporating missing data methods into analyses. More on this project: Montez-Rath ME, Winkelmayer WC, Desai M. Addressing missing data in clinical studies of kidney diseases. Clin J Am Soc Nephrol. 2014 Jul;9(7):1328-35. doi: 10.2215/CJN.10141013. PubMed PMID: 24509298; PubMed Central PMCID: PMC4078963.|
|Record History:||('2019: Project extended to 2019 ',) ('2017: Project extended to 2018',)|