Skip to Main Content
Center for Methods in Implementation and Prevention Science (CMIPS)
YSPH Home

INFORMATION FOR

HIV/AIDS Implementation Science Research

  1. To improve the evaluation of the causal effects of specific interventions that comprise a ‘package.’ A challenge in assessing the impact of combinations of strategies for reducing HIV incidence is the assessment of which specific interventions were essential, helpful, or ineffective within the package. Using a triangulation of causal methods to estimate the direct effects of package components, we will apply marginal structural models, g-causal methods, principal stratification and instrumental variable methods to estimate causal direct and total effects.
  2. To augment prevention trial designs to adjust for bias due to loss to follow-up Refusal to test and selective loss to follow-up can bias effect estimates. Through a validation study to be conducted in Dar es Salaam, Tanzania, we will assess for the first time the suitability of interviewer identity as a selection variable required for valid application of Heckman selection models for adjusting for this potential bias in effect estimates.
  3. To develop advanced statistical methods for cost-effectiveness estimation and stepped wedge designs Semi-parametric survival data analysis methods will provide improved estimates of comparative effectiveness ratios (CER) and their uncertainty. It will be critical to demonstrate not only the effectiveness but also the cost-effectiveness of combination interventions. New evaluations that grow out of the CRHPTs will benefit from new methodology and software for stepped wedge designs, which simultaneously address the need for rigorous randomized trials and for rapid roll-out of promising large-scale interventions.
  4. To develop and validate novel network methods to improve the evaluation of large combination prevention clinical trials. Based on social linkages in cell phone data, we will investigate the attenuating effect of cross-community mixing on CRHPT effect estimates, and use this information in analyses of study results to help better evaluate the effectiveness of HIV prevention interventions. Investigation of the extent to which the network of cell phone calls serves as a proxy for sexual networks may reveal a low cost approach to quantifying and corrected for mixing and allocation biases.