Combining Human Knowledge and Machine Learning to Increase Retention in Care for HIV Patients

Rayid Ghani
Director of Research, Center for Data Science and Public Policy
The University of Chicago Harris School of Public Policy

This work seeks to advance predictive modeling for implementing healthcare delivery, particularly predicting whether an HIV-positive individual will retain in care. Despite the importance of retaining patients in care little it known about the factors – geographic, economic, social – that lead a patient to drop out of care. There is a great opportunity for innovation to use machine learning to uncover the factors that lead to a person to drop out of care and providing a personalized intervention to bring them back into care.

 

Engineered Extracellular Vesicle-Mediated Delivery of Targeted Nucleases to Inactivate HIV Proviral DNA

Joshua L. Leonard, PhD
Associate Professor of Chemical and Biological Engineering
Northwestern University McCormick School of Engineering

Nuclease proteins that degrade genomically integrated HIV DNA have been developed, but no strategy has yet been identified for delivering such proteins to infected cells in a patient. To potentially address this need, this project will harness a recently-appreciated mechanism by which cells transfer their contents to other cells via the secretion and subsequent uptake of nano-scale extracellular vesicles. If successful, this project could ultimately establish a novel therapeutic strategy for addressing the persistent challenge of eliminating HIV reservoirs.

 

Using Computational Linguistics to Amplify the Effects of Text Messaging-Based HIV Interventions

Carlos Gallo, PhD
Research Assistant Professor of Psychiatry and Behavioral Sciences
Faculty Affiliate, Center for Prevention Implementation Methodology (Ce-PIM)
Northwestern University Feinberg School of Medicine

Mobile and text-based interventions provide unique opportunities to engage adolescent MSM in HIV prevention. However, mHealth interventions suffer from low usage which mitigates their efficacy. This project will use computational linguistics methods to analyze users’ demonstrated linguistic preferences with the goal of improving future mHealth interventions. Mobile and text-based programs with high uptake and adherence have the potential to promote HIV testing, harm reduction, and engagement in health care.

 

 

Towards Implementation of Hybridized HIV Testing Strategies to Optimize Care Continuum Metrics

Moira McNulty, MD
Coggeshall Fellow and Clinical Instructor of Medicine
The University of Chicago 

While testing strategies have been evaluated for yield of new HIV diagnoses, little is known about how these methods should be employed to optimize engagement in downstream care in order to improve outcomes such as linkage, retention, and viral suppression. This project aims to 1) assess the HIV testing event as the entry point into the HIV care continuum and 2) develop a data collection tool to inform the possibility of future integration of HIV testing services in order to encourage the adoption, adaptation, and scale out of effective strategies. We will concurrently develop data collection tools to assess the barriers and facilitators to implementation and integration of complementary HIV testing strategies, as seen by key stakeholders.

 

 

Towards an Effective PrEP Diffusion Model: Identifying Criteria for Peer Change Agent Selection 

Aditya Khanna, PhD
Research Assistant Professor, Biological Sciences Division
The University of Chicago

This pilot proposal aims to develop a data-driven theoretical mechanism to select PCAs who are “concurrent” (across multiple social environments) and/or “durable” (across time), who can then be trained to improve HIV prevention and care continuum outcomes among younger black men who have sex with men (YBMSM; age: 16-35 years) in Chicago, a population that has experienced a rising HIV incidence over the last decade. This work will help identify social change parameters and diffusion thresholds for PrEP uptake, and the results will be used to develop an agent-based model (ABM) for PrEP diffusion in future work. Additionally, we will assess ways of making PCA identification tractable for public health departments that might not have detailed longitudinal network datasets, as the ones we will be using in this study.

 

Image result for Aaruni KhanolkarEvaluation of immune function fitness in vertically-infected HIV+ patients with early ART initiation

Aaruni Khanolkar, MBBS, PhD
Assistant Professor of Pathology
Northwestern University and Lurie Children’s Hospital

Emerging evidence suggests that HIV+ children who initiate ART early in life represent a very valuable cohort to test therapies to attempt a cure. A thorough assessment of immune function in these children constitutes a critical component of this ambitious goal. In this pilot study we will evaluate and compare the functional fitness of the host immune response between two vertically-infected HIV+ cohorts that initiated ART early (at a combined median age of 1 year of life) but differ in viral load suppression measured longitudinally on ART. These assessments will include measurements of HIV-specific T cell polyfunctionality, cytokine-responsive phosphoprotein signatures in bulk T cell subsets and neutralizing antibody responses both before and after immunization against seasonal influenza virus strains.