Preventing heart disease, one of the leading causes of death worldwide, requires advanced forecasting tools to predict cardiovascular health outcomes. To accelerate innovation in this space, the U.S. Food and Drug Administration (FDA), the Veterans Health Administration (VHA) and collaborating partners launched a global competition to develop and evaluate AI and machine learning models to improve care for U.S. veterans.

More than 300 participants from 16 countries registered for the Veterans Cardiac Health and AI Model Predictions (V-CHAMPS) Challenge, including ZS. Our predictive AI model was named a finalist in Phase 1 and surpassed expectations when validated on synthetic and real-world data (RWD) in Phase 2, earning second place. The ZS team combined diverse expertise to build a high-performing, transparent and explainable AI model.

We sat down with the team to learn more about their innovative solution and what drove their success. “We’re proud to contribute to scientific progress in chronic heart failure and thrilled that our work will help improve patient outcomes,” said Prerna Goel, a ZS data scientist and V-CHAMPS team member.

Where passion changes lives

The team included ZSers from across the globe: Pranava Goundan, Qin Ye, Abhinav Bansal, Sagar Madgi, Prakash Prakash, Subhrajit Samanta, Sayan Chakraborty, Hitashu Kanjani, Prerna Goel, Shubhankar Thakar, Vijay Boda and Xiaoying Shi. Each brings a passion for advancing healthcare using AI and RWD—from developing algorithms for identifying undiagnosed patients, accelerating clinical trials and predicting disease progression to uncovering barriers to treatment.

“Sayan is a tech and AI geek. Prerna has deep RWD, machine learning (ML) and healthcare domain expertise. Vijay and Hitashu are adept ML engineers. I’m proficient at designing end-to-end AI and ML systems and putting the pieces of puzzles together. Our skill sets added up to form a natural coalition,” said Shubhankar Thakar, ZS data scientist.

Designing a transparent and impactful solution

The ZS team focused on building clinically meaningful AI models that captured the recency, frequency, occurrence and variability of critical medical events in patient histories. Key elements of the approach included:

• Intelligent feature discovery: ZS’s cloud-based, automated tool uses genetic algorithms to generate cross-sectional features that assess event aggregation and recency
• Purposeful algorithm experimentation and optimization: The team followed best ML practices to ensure their models were transparent, credible and generalizable
• Enhanced interpretability: By using techniques like Shapley Additive Explanations (SHAP), the team explained feature importance and directional impact
• Focused research and data wrangling: A comprehensive literature review helped the team define novel and known risk factors, cohort criteria and model benchmarks

“Our drive to create real change in healthcare fueled our success,” Prerna said. “We pushed boundaries in predictive AI and stayed focused on improving outcomes for veterans. That shared purpose made all the difference.”

Advancing veteran health with AI

ZS congratulates the team for this outstanding achievement—a model designed to enhance cardiovascular care for veterans. We look forward to continuing collaboration with the FDA, VHA and other partners to share our learnings from the challenge and to develop scalable, real-world solutions.

Looking ahead, ZS will explore ways to enhance these algorithms by incorporating broader patient representation, expanding risk factor analysis and aligning perspectives from pharma, payers, providers and patients.

Follow us on social media to see how ZSers around the world are using AI, data and science to make healthcare more accessible, inclusive and effective.

Life at ZS

Learn what it’s like to be a ZSer:  LinkedIn, Instagram and Facebook 

View ZS’s open roles.