Ayesa – Cardiac risk prediction
Ayesa
Sector: Health
Business Case
Improve the triage of patients with non-traumatic pain in the A&E department (ischemic heart disease) to reduce the number of false positives without increasing false negatives.
Objectives
Correctly classify ischemic heart disease in A&E with emphasis on minimising false negatives that mean that someone may be sent home, which may result in ischemic heart disease.
Use case
Input previous patient information (primary, history and medication). Capture information that is collected during the A&E episode in the progress reports. Creation of several models to classify patients. Consensus/weighting algorithm
Infrastructure
On Premise
Technology
Automatic or Deep Learning Text mining
Data
Tabular data, evolutionary reports. Anonymised private datasets.
Resources
Emergency professionals Documentalists Data scientist with NLP and ML knowledge
Difficulties and learning
Clinical data, low quality data, relevant data only in natural language, no risk tolerance, high staff turnover in A&E, etc.
KPIs (business impact and metrics of the model)
Cost savings and savings in hospital resources due to unnecessary admissions. Reduction of false positives.
Funding
This project received funding from the competitive Hazitek research and innovation programme of the Basque Government.
Collaborators, Partners
BioCruces
