J. Tod Fetherling
Chief Data Officer
“Artificial intelligence has rocketed to a must-have capability in hospitals and health systems across the country. It's past time for health leaders to act.”
Artificial Intelligence (AI) means a lot of different things to different people. In general, we believe healthcare has defined Artificial Intelligence as the ability to use computers in place of human functions. In particular, the specific uses cases are areas where automation and robotics can be used to improve the efficiency and effectiveness of the patient experience or hospital operations.
According to Forbes Magazine in 2018, “John McCarthy first coined the term artificial intelligence in 1956 when he invited a group of researchers from a variety of disciplines including language simulation, neuron nets, complexity theory and more.” From this group of scientist and technologist the field has evolved and accelerated due to computer technology and human curiosity about how to do more with less.
Merriam-Webster defines artificial intelligence as:
We find there is wide continuum of definitions within the healthcare industry depending on the sophistication of the technology team at the organization. Some organizations are years into the technology related to radiology and very sophisticated image matching algorithms who are reviewing past chest x-rays for potential mis-reads by radiologists. Other organizations have already transcended this process and AI reads the image in advance of the radiologist and makes recommendations for the reader. The radiologist is still going to read the image. If the technology can prompt for potential issues and/or create the document/notes, then the human assisted technology read of an Xray image should become faster and better.
There are multiple use cases in revenue cycle. Technology is being used to be able to submit claims for payments more rapidly by identifying potential errors based on past learnings. Emergency patients are being scanned upon admission and the technology will determine if you might be willing to make a cash payment today at a discount based on your address or health insurance plan. Then the payment can be collected upon exit of the ED.
When it comes to mathematical calculations, computers perform far better than humans. When it comes to subjective matters, humans still perform better than computers. However, computers are learning how to be empathetic, consider more variables, and run multiple scenarios based on time, resources, and outcomes. The field of science is related to Supervised and Non-Supervised Machine Learning.
Supervised Machine Learning is most of the deployed solutions in healthcare today. Some might even reduce it to really good programming. As a programmer who learned on Fortran77, the basic concept of If/Then loops is what has powered computer programs for decades. IF/THEN is used extensively today by Amazon. If customer a buys a grill, then they will also want a grill cover (85% likelihood) and a new spatula (45% likely). 95% of customer who buy a grill, buy a grill cover during checkout (recommendation link). For the non-buyers of the grill cover, we are going to store that information in their profile and present it again on their next visit to the store.
Unsupervised Machine Learning is the concept where the computer themselves find and run the above logic without prompting by the programmers. It’s what we all want and fear. Let’s stay focused on the wants for right now. We will address the fear issues later. We are further away from a fully unsupervised machine learning program. Someone still has to ask the right question in order to start and drive building the logic required to fully extract the value from artificial intelligence. However, the current pace of innovation in this field of study is quickening. All things are possible.
Healthcare Executives are cautiously excited about implementing AI in both clinical and financial matters as it relates to healthcare.
Let’s revisit our scenario above about practical applications available to us today in supervised machine learning in health. Wouldn’t it be great if the physician was presented with a full risk profile of the patient in advance of the patient / physician office visit. The patient is presenting with a cough today. The physician might also ask a series of questions related to the risk profile of the patients based on the aggregation of all patient/person connected records (biometrics from their Fitbit, all of the claims data both at our system and outside of our system, genetic profiles, and the patient’s own observations about their own health). So in the Grill example above, the Physician might say, I see that you have not had a CT scan and since you are a previous smoker and getting a little older, let's order this test and rule out a few other potential issues. I also see that you have had 2-3 injections for knee pain. Is that getting better or worse?
Most people want technology to help them live healthier, happier lives. If we can do that in a responsible and responsive environment, then I believe we will see great advancements in the adoption of artificial intelligence in healthcare.
From a Population Health perspective, many executives believe AI will help them quickly identify potential patients at risk and this will be used for patient acquisition. There are 50+ trigger events where Perception Health is already using this technology as part of CARE Platform.
Pricing and Transparency may represent the next frontier for AI in healthcare. Think about a Walmart Shopping Bot for Healthcare. The bot will tell you where the best price for an MRI or CT Scan is nearest to your location.
Care Teams is another huge opportunity for Artificial Intelligence. When it comes to Diabetes, what is the best team of physicians to either treat diabetes after diagnosis or better yet, who does the best job of preventing Diabetes from progressing. Shouldn’t we be focused on making sure the patients who are pre-diabetic stay there or even reverse the underlying lifestyle changes which causes the disease state in the first place.
In short there are many potential use cases for the use of artificial intelligence to benefit the patient, the provider, and the payer. Shouldn’t we all be working together to improve the care of the communities we serve?