HealthChain Pilot Results: Cardiology AI Predictive Model

January 19, 2026

This pilot project aimed to design, implement, and validate an automated AI-driven solution to extract and structure clinically relevant data from unstructured electronic health records in the context of Transcatheter Aortic Valve Implantation (TAVI). The primary goal was to reduce the manual data extraction burden on clinical teams while enabling scalable generation of real-world evidence to support outcome-based decision-making. The pilot successfully deployed a three-agent large language model pipeline to transform routine clinical narratives into analysis- ready variables and demonstrated its operational feasibility in a real hospital setting. Using the automatically generated dataset, a survival prediction model was implemented as a proof of concept, confirming that the extracted data preserved meaningful clinical signal aligned with established risk factors. Despite challenges related to data heterogeneity and integration, the pilot delivered actionable insights, validated the technical approach, and aligned with institutional digital health strategy, establishing a strong foundation for future scale-up and broader clinical and research adoption.

The Challenge, the Solution and the Co-creation Process

The pilot project addressed a critical healthcare challenge in cardiology: the lack of structured, integrated clinical information for managing patients with severe aortic stenosis eligible for TAVI. In Portugal, approximately 25,000 individuals over 75 suffer from symptomatic severe aortic stenosis, with about 4,500 qualifying for TAVI due to high surgical risk. Long waiting times (up to six months) and fragmented clinical data hinder timely and effective care. Clinicians face difficulties in monitoring patients across pre-, peri-, and post-intervention stages due to the absence of structured data and decision-support tools. The management of waiting lists is a challenge in itself.

The solution to address the challenge proposed by the healthcare organisation (HO) involved developing and integrating an AI-powered predictive model into the hospital’s electronic health records system. This tool structures clinical data and enables continuous, standardized monitoring of patients throughout the TAVI pathway. Co-created through close collaboration between cardiologists, data scientists, software developers, and hospital IT teams, the model provides predictive insights that support decision-making, prioritization, and resource planning. This innovation not only improves clinical outcomes and workflow efficiency but also facilitates data export for performance indicators and research, addressing both clinical and administrative needs.

Pilot Overview, Results, and Lessons Learned

This pilot marked the initial release phase of an AI predictive model designed to support decision-making in the treatment of severe aortic stenosis, specifically for patients eligible for TAVI. The pilot leveraged retrospective cardiology data to train and validate the model in a relevant environment.

For patients, the model promises earlier identification and prioritization for TAVI, reducing waiting times, and supporting improved outcomes. For clinical staff, it provides structured, accessible information across the care continuum, aiding in diagnosis, monitoring, and treatment planning. For the HO, it offers improved workflow efficiency and a foundation for robust performance indicators and clinical research.

Several key challenges emerged during the pilot. The primary barrier was data access and integration as clinical data was spread across three different sources: the HO’s internal systems, an external cardiology database provider, and manually maintained spreadsheets within the cardiology clinical service. This fragmentation complicated and delayed the data extraction process. Security protocols for accessing hospital data added further delays, while obtaining data from the external provider proved bureaucratic.

Key lessons learned include:

Early Data Access Planning: Engaging with IT and data custodians at the outset is critical. Initiating access approvals early can prevent delays in later development stages.
- Centralized Data Management: Future projects would benefit from centralized or harmonized data sources. Establishing shared data protocols and infrastructure early can streamline integration and improve data quality.
- Structured Data Preparation: The use of unstructured data (e.g., free text, medical images) highlighted the need for standardized processes in preparing diverse data types for AI input.

Future Plans

Following the pilot, the HO plans to continue using and further developing the AI predictive model, recognizing its alignment with strategic goals of enhancing patient-centered care and improving the measurement of health outcomes. The solution supports a broader institutional shift toward data-driven clinical decision- making and integrated digital health tools.

Ongoing collaboration between the hospital and Promptly will be sustained through new projects aimed at expanding the model’s application beyond cardiology. Future phases will replicate the methodology for other departments and clinical conditions, creating a scalable framework for structured data integration and predictive analytics across the hospital. This expansion will enhance care coordination, outcome tracking, and clinical efficiency at a broader institutional level.

From Promptly’s perspective, the next steps involve scaling the solution for use in additional hospitals and HO. Commercialization efforts are underway to offer the model as a modular, adaptable platform that supports AI-driven data analysis and outcome measurement in diverse clinical settings. The focus will also include refining the model based on real-world feedback and extending its capabilities to accommodate new data types and clinical pathways.

By building on the success and lessons of this pilot, the solution is well-positioned to contribute significantly to digital transformation efforts in healthcare, both within the current hospital and across the wider health system.

Would you like to know more, or are you interested in the solution?

Website: https://promptlyhealth.com/en

Disclaimer: HealthChain project is funded by the European Union. Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or European Innovation Council and SMEs Executive Agency (EISMEA). Neither the European Union nor the granting authority can be held responsible for them.