Heart failure is the leading cause of death and disability in the United States, costing healthcare systems worldwide more than $30 billion annually. Current approaches to treatment are limited by crude clinical assessments of the disease. In a new study, Yale researchers have successfully used big data methods to improve prediction of heart failure patient survival. They also described data-driven categories of patients that are distinct in their response to commonly used therapies.
This innovative approach, detailed in the Journal of the American Heart Association, could lead to better care for this incurable chronic condition, the researchers said.
Led by Drs. Tariq Ahmad and Nihar Desai, both assistant professors in Yale’s Section of Cardiovascular Medicine, the research team analyzed health data from a large registry of more than 40,000 patients. The researchers used a statistical “machine-learning” technique to first predict outcomes for the patients one year after diagnosis. They also applied cluster analysis methods to sort the patients into four clinically recognizable categories with different responses to commonly used medications.
The big-data methods vastly outperformed currently used measures of heart failure, and had better prediction of risk than previously published prediction models, Ahmad said. The research team also used entirely data-driven methods to group patients into distinct clusters that responded differently to medical therapies.
As a final step, the researchers used their findings to develop a predictive online tool that could be integrated into electronic health records in healthcare systems. Their long-term goal is to apply these advanced analytic strategies to improve research and to provide personalized care for heart-failure patients as well as “enhanced intelligence” to clinicians at the bedside, Ahmad said.
Source: Read Full Article
