Deciding which intervention to implement requires accurate information about how patients in the intended population will respond to the interventions under consideration. Clinical trials are almost always conducted with a specific target population in mind, but data collected is rarely a random sample of that population for various reasons, including the difference between participants who agree to participate in a clinical trial and those who refuse, or the clinical setting or geographic location of the clinical trial.
To address this important limitation, methods that accurately transport or generalize trial results from one or more clinical trials to the target population of interest have been developed. The practical utility of these methods is limited by the lack of methods for handling missing data or systematic missing data, both within trial and within target. Systematic missing data occurs when one of the trials and/or the target does not collect information on all covariates.
This application proposes to develop identifiability conditions and estimators that account for both within-trial and systematic missing data when transporting or generalizing treatment effects to a target population. Finite sample performance of the developed estimators will be evaluated using extensive simulations. The methods developed will be used to transport treatment effects from six sexual risk prevention trials conducted among vulnerable adolescents at risk of HIV acquisition. The six trials include three active interventions (family-based, conventional knowledge and skills, and emotion regulation) designed to reduce sexual risk behaviors among HIV-vulnerable adolescents.
- In progress; Recruitment not applicable
- Improving Methods for Conducting Patient-Centered Outcomes Research
- Accelerating PCOR and Methodological Research
The state where the project originates, or where the primary institution or organization is located.
- Rhode Island