Today, artificial intelligence is used through health care for administrative tasks, such as improving medical coding, and for some clinical use, such as enhancing the reviews of radiologists for diagnostic images.
Here, though, some hospitals and health systems start working on what some experts see as the next step in the development of artificial intelligence in health care: prediction.
In particular, the predictive analyzes that support artificial intelligence receive a lot of attention by managers managers managers and executive analyzes.
One of the experts who believe that the prediction of it is the following boundaries of Amnesty International in the field of health care is Dr. Mento Tirka, a cardiologist in ancient warriors affairs Palo Alto for healthcare, medical director and scientific employee in the seller of their Elywite and professor of medicine at Stanford University.
Turakhia has more than 25 years of experience in patient care, research and experiments results, data science and artificial intelligence, organizing medical devices, and creating and marketing digital health products. It is worth noting that he was the investigator participating in the teacher Study the heart of apples.
Health care news I recently spoke with Turakhia to deeply dive into artificial intelligence predictions, talking about the steps necessary to advance artificial intelligence from her current state to her predictive capabilities, how AI can predict the identification of health conditions and enable preventive care, expand health care and create better results for patients with artificial intelligence, And integrate artificial intelligence predictive into health care information systems.
Q: You say that artificial intelligence needs to advance from its current state to prediction capabilities. What are the steps to get there?
A. More Great early steps in artificial intelligence in health care It was in classification, or identification of patterns, starting with medical photography. Deep learning algorithms are very effective-in many cases, doctors exceed-in determining diagnoses on X-rays, ultrasound, or heart planning. AI can also excel in measurements, such as estimating the left ventricular ventricular part of the heart ultrasound, which is often stressful and vulnerable to human error.
Recently, artificial intelligence is applied to extract diagnoses from electronic health records notes and even through the artificial intelligence of conversation with patients. This is in the classification.
Prediction, however, varies. It focuses on predicting future results instead of identifying current cases. It comes to the use of data available to estimate the risk of developing or experimenting with a clinical event in the future.
For example, even if the ECG ambulance screen does not discover atrial fibrillation today, the data it picks up still can reveal signals that expect AF risks below the line. Likewise, alternatively, biological signs, sleep patterns and activity data – that are often used to track fitness or track sleep – can instead can be analyzed in the heart of the heart in the future.
It comes to the use of vital signs, sleep data and activity not to win fitness or “sleep badges” on your intelligence hour, but to estimate the risk of healing from heart failure.
Access to predictive capabilities requires strong and generalized data sets related to clinical results. Historically, health data – photography, ECG, smart watches, medical records, and medical insurance entry data are all approved independently. By linking these data sources at the patient level, you can get multi -dimensional and longitudinal data that can be used to develop artificial intelligence models to predict the results in that data.
The next wave of artificial intelligence in health care It will turn from the diagnosis of current cases to predicting future health risks, which paves the way for pre -emptive and preventive care.
Q: You believe that artificial intelligence will be used more strongly to determine health conditions and enable preventive care. How is that?
A. AI will enable us to determine the future risks of health conditions and clinical events with more accurately. When I think about developing artificial intelligence, I would like to think about the 2 x 2 matrix: what is easy or difficult for humans for what is easy or difficult for artificial intelligence.
Take the ambulance ECG monitoring again as an example. The first step was to develop strong artificial intelligence to diagnose irregular heartbeat. We have published this in Nature Medicine In 2019 (Hannun Aw et al
The next step is more complicated: Use ECG to predict the future risks of atrial fibrillation. ECG can discover the fine structural and electrical changes in the heart that increases the risk of AF. When combined with continuous ECG data – like 14 days of monitoring – AI can determine the critical patterns that humans may miss. Merging these patterns into the predictive risk model is difficult to calculate humans, but it is possible and easy for artificial intelligence.
From there, artificial intelligence can go further – estimating the risk of future results such as stroke or heart failure, two known conditions due to AF. The development of these predictive capabilities requires linking various data groups and conducting important development work. However, the potential benefits of early intervention and prevention are unusual.
Q: Amnesty International will generate better results for the patient, suggest. How will this happen?
A. Many people may not realize Monitoring from the patient’s data It started more than 30 years ago. In the nineties of the last century, manufacturers of cultivated cardiac ancients – such as heart attack devices and defibrillers – developed systems to monitor the function of the device from remote and discover the irregular heartbeat.
Today, with the progress of the sensor miniaturing, what is required from the day visiting the office now doing at home – even in a smart hour. For example, smart watch algorithms can discover continuous irregular impulses and alert the user to the possibility of atrial fibrillation, allowing early detection. This really happens.
Looking at the future, the incorporation of multiple data flows – ECGS, vital marks, sleep data and more – in longitudinal models will enable artificial intelligence to determine health risks before clinical events occur. For example:
- Predicting the beginning of AF, heart failure, or sleep apnea.
- It is discovered when the heart failure increases, which increases the risk of hospitalization.
In these scenarios, doctors, patients and health systems can take proactive steps – such as confirming diagnoses, starting treatments or controlling medications to reduce hospitalization risk.
Now, in order for this to succeed, Amnesty International must be strong in its performance. This means that the measures of accuracy and prediction – such as positive and negative predictive values - should be high. For example, if only 5 % of all positive AI results are correct, 95 % are false positives, which is very unhelpful and can be harmful.
That is why Artificial intelligence will work better For somewhat common conditions, the identification of diseases or rare events with high accuracy remains very difficult.
Q. You expect hospitals and health systems to integrate artificial intelligence prediction in information systems. How will they do that, and to what end?
A. There are two basic applications of predictive artificial intelligence within hospitals and health systems.
First, at the patient level. This condition is already ongoing. In the care of outpatient clinics, doctors often depend on the basic risk degrees that are a small number of clinical variables. These degrees have limited predictive accuracy.
Artificial intelligence enhances these tools by integrating dozens or even hundreds of data points to create more accurate risk assessments. Even when artificial intelligence is not completely predicting, it can serve as support for the decision, which reduces the inappropriate differences in care. For example, Amnesty International can ensure that patients with atrial fibrillation are receiving anticoagulant treatment as recommended in clinical instructions.
On the pathological side, many companies have developed early warning systems, which threatens life for overwhelming infection. By the time when the septic shock occurs, the time is often late, with death rates of 30-40 %. Studies have shown that septic alert systems not only lead to better results for the patient but also improve the commitment of doctors for treatment protocols.
As a result, the quality of care can also improve.
Second, at the population level. For integrated and value -based health systems, Amnesty International can predict patients At the highest risk of using health care, the emergency room visits and hospital entry are usually. This allows the source interventions to reduce expensive events.
Interestingly, the most effective solutions can be low-tech-such as home visits, regular verification over the phone, and ensure adherence to drugs or family participation support.
Some health systems even explore the “factors” of artificial intelligence or virtual nurses to conduct a remote and monitor the patient. Merging artificial intelligence with the prediction with these tools carries the ability to enhance care, reduce costs and improve results.
Follow Bill Hit coverage on LinkedIn: Bill Seuiki
Email him: bsiwicki@himss.org
Information technology healthcare is a HIMSS media publication
