Improving AFib Registry Accuracy with AI Integration

This case study details the implementation of Lighthouse, an AI-powered abstraction platform by Carta Healthcare, at Lahey Health to enhance the accuracy of their AFib registry. Initially, an experienced abstractor achieved an inter-rater reliability (IRR) score of 92.29 percent, which fell below the required threshold of 98 to 99 percent. This prompted the introduction of Lighthouse, which employs a Hybrid Intelligence model that combines AI analysis with human oversight. Following this integration, the abstractor's IRR score improved significantly to 99.09 percent. The case study highlights how the use of AI can reduce variability in data abstraction while maintaining accountability, ultimately leading to higher data quality and consistency across abstractors. The results underscore the effectiveness of combining human expertise with AI capabilities in clinical data management, suggesting a shift towards Hybrid Intelligence as a future standard in the field.