Introduction
- NHS England have the ability to audit Heart Failure patients – but only once they present to hospital
- NHS Scotland do not currently have the ability to audit any patients
- Red Star AI have proved their ability to identify all patients with heart failure from their medical records
- Red Star AI can also use those records to assess clinical performance
- Red Star AI have identified the optimisation of prescribing as a promising opportunity for performance improvement
Method
- Data from 2010-21 supplied by NHS Tayside
- NLP applied to echocardiogram reports
- NLP data augmented with other clinical records (bloods, prescribing, demographics, etc)
- Records assessed using statistical, machine learning and other data science techniques
Results
- Red Star AI accurately identified 96.4% of heart failure cases
- Prescribing predicts mortality:
- 6% of the cohort on optimal treatment in the community died – 12% of those on suboptimal died, a 100% increase
- 17% of the cohort on optimal treatment suffering an unscheduled admission died – 46% of those on suboptimal treatment died, a 270% increase
- The majority of patients are not on optimal treatment
- 40 of 98 (41%) of patients reviewed by a cardiologist were considered suitable for uptitration, with Quality of Life (KCCQ) the primary benefit
Discussion
- Red Star AI are developing a Heart Failure auditing capability
- It will tell us who in Scotland has heart failure, and what we know about their condition and standard of care
- In due course it will be used to surface priority patients and report on standards compliance across Scotland