Atrial Fibrillation (AF) is a common heart problem causing an irregular heartbeat that occurs in many individuals. It may make the heart beat more rapidly and reduce the heart’s ability to pump blood to the body. It allows blood clots to form in the heart, which can cause strokes if they are pumped to the brain. Avoiding AF is important.
People who develop AF whilst in hospital seem to stay longer in the Intensive Care Unit after surgery, are more likely to develop complications and have a higher risk of dying.
About 30-50% of people having heart surgery develop AF shortly after the operation. Different preventative treatments such as beta blockers and amiodarone reduce the likelihood of developing AF, along with an individual’s lifestyle including what they eat and how active they are.
Preventative treatments carry risks, so it is important to identify people most likely to benefit from them (i.e. the benefits are bigger than the risks). Currently there are no good tools (mathematical models) to predict who will get AF after heart surgery. Previous tools are not used in clinical practice, partly because there are weaknesses in how they were developed. For example, some do not include modern data like ultrasound pictures of the heart that are now routine before surgery. A modern reliable prediction model is needed.
We will develop two reliable prediction models to identify which patients are at greatest risk of developing Atrial Fibrillation (AF) following heart surgery. One will predict the risk at assessment prior to surgery, and the second will predict who may develop AF after surgery. (Two models are needed as changes during surgery may alter the risk of AF).
There are three parts to our study:
First, we will produce a list of possible factors that alter the risk of getting AF after heart surgery. We will do this by:
- doing a detailed review of what has been published including medical papers, and any clinical trials,
- seeing if we can identify specific risks (known as risk factors) in a UK general practice database (CALIBER) which includes 90,000 people who had heart surgery,
- asking mathematical and clinical experts, and
- using a modern computer technique (called machine learning) to look for previously unrecognised AF risk factors in a large United States (US) database (the Partners Research Database (PRD), which includes over 30,000 patients) and in the CALIBER database
Using this list, we will use mathematics (standard statistical approaches) and ‘machine learning’ within the PRD to develop prediction models to identify patients at increased risk of AF.
Meanwhile we will work with two large UK NHS heart centres that together do 6000 heart operations a year, and an ongoing UK prospective clinical trial to ensure the risk factor list and when a patient develops AF are documented precisely.
Finally, we will see if our new models work by using it in a new study using information from the two UK NHS heart centres to see if our models predict who will get AF reliably.
