AI App On The Verge of Reforming Cardiac Arrest Treatment


University of Gothenburg researchers have come up with a new app that can enhance the treatment of cardiac arrest by pooling data from thousands of past patients.

The AI app helps doctors witness how numerous patients have fared when it comes to varied approaches for the treatment of cardiac arrest. Doctors can get this information post submitting their own patient data into their web application.

The app happens to be available for free from the website of Gothenburg Cardiac Arrest Machine Learning Studies website. That said, the scientists have stressed the fact that the inferences from the algorithm of the app should be taken into account by people who have the appropriate skills.

AI-based decision support tools have now become more prevalent across many areas of healthcare, but there is still an exhaustive debate among health professionals on how patients as well as care services can benefit from them.

The application makes use of the data from the Swedish Cardiopulmonary Resuscitation Register from numerous and varied cardiac patient cases. The scientists use a progressively advanced form of machine learning to inform clinical production models that help in allowing the app to pinpoint factors which have gone on to affect the previous outcomes.

The algorithms of the app take into account varied factors that concern cardiac arrest, like the period of treatment involved, any previous ill health or medication taken and also the socio-economic status of the patient being treated. As per the researchers, it is going to be many years before AI-based support tools get included in the official recommendation when it comes to cardiac arrest treatment. The doctors, however, are free to use this tool in addition to the new, evidence-based methods.

As per Araz Rawshani, who is the researcher on the project, he, along with several others who are into treating emergency patients with cardiac arrest, have begun using the prediction platform as part of the process that decides on the care level. The solutions from these tools, according to him, often mean that they get confirmation on the views they have already arrived at. It subjects them to not letting the patients be exposed to the treatment, which happens to be painful, and also helps save them the resources for care.

Thus far, the research group has already gone on to publish two decision support tools. The first happens to be a prediction model which is called SCARS-1 which tells if the present case of the patient resembles any previous one. As per the team, this model’s precision is unusually high, with a sensitivity of 95% and a specificity of 89%.

Fredrik Hessulf, one of the doctoral students from the University of Gothenburg, says that the decision support system is one of several blocks in a big puzzle. They have many different factors to take into account while deciding whether to go ahead with cardiopulmonary resuscitation. It indeed happens to be one of the highly demanding treatments that should be given to only those patients who are going to benefit from it and will be able to lead a meaningful life after their hospital stay.

Approximately 55,000 patient cases have been used to inform the algorithm. 393 elements affecting the chances of patients’ survival for 1 month after the cardiac arrest were taken into account by the support tool. There are almost 10 factors out of these that have been found to be very relevant to the survival of patients. Getting a cardiac viable rhythm again post the emergency care admission has been zeroed-upon as the most significant factor.

As far as the second support tool is concerned, it gets data from the survivors of the out-of-hospital cardiac arrest until the time they were discharged. This tool took into account 886 elements across 5098 cases of patient data from the Swedish Cardiopulmonary Resuscitation Register.

This tool can help carers identify which patients are at high risk of another cardiac arrest or even death within a year of the original episode. According to researchers, these factors can also help doctors understand the long-term survival possibilities after cardiac arrest.

The tool’s accuracy is good, as it can predict with around 70% dependability if the patient is going to die or may have yet another cardiac arrest within 12 months. They are looking for success when it comes to the development of the prediction model in order to elevate the precision levels. As of today, it has proven to be a support to doctors in gauging elements that have a tremendous bearing on the survival rate of cardiac arrest patients who have already been discharged from the hospital.