5 Everyday AI Applications in Healthcare

What can Artificial Intelligence (AI) do for me? Can it make my hospital more money? Will it help reduce human error and accidents? Is anything they say about AI actually true?

All good questions I assure you. The simple answer is…almost certainly. 

In this blog, we’ll take a look at five specific uses for Artificial Intelligence currently in the Healthcare industry.

1. How AI Can Optimize Supply Chains and Save Money

With the right data and the right team, you can determine when is the opportune time to reorder supplies. An application like this is created, fine-tuned, and deployed by taking into account patient volume, trends, and supply levels to predict when is the best time to make orders. This helps prevent the excess of drugs and supplies and ensures that the right ones are ordered on time and in-stock. 

A more forward-thinking (albeit larger)  investment would be to use RFIDs to track the supply life cycle. Then, AI can be deployed to help with inventory control, price visibility, and if a shipment is what it says it is by where it goes in the hospital. Prevent supply shortages, lower labor and material costs, and the best of all, predict spikes in demand.

In 2017, Propublica estimated that $765 billion a year is thrown out in wasted medical supplies; that amounts to over $1,000 per person per year! As a healthcare consumer, I’m personally appalled at the amount of my money being wasted. As a shareholder in a hospital who no doubt wants to provide quality services but also make money, you seem to be losing a lot of money to the landfills. If a 10th of the money is used to develop a piece of software that can yield a 10x return, well, that sounds like a good long-term investment to me.

2. How AI Can Help Triage Patients

Outpatient surgery clinics can be risky. One method currently being tested is the use of AI to predict the likelihood that a patient will have an adverse reaction on the day of the surgery. Impossible? Quite possible actually. You see AI takes into account not only the health records but the state of the patient on the day. They compare that to similar patients who’ve needed hospitalization after an outpatient surgery. A score is given to your team to determine whether or not you wish to operate. This is one of the more experimental use cases but pretty nifty if it works out. In fact, the domain of personalized medicine is targeted towards many similar applications intended to treat the whole patient by using AI to capture all signals relevant to treatment.

But what about emergency rooms? Well, it’s getting used there too. It’s new but imagine how this could reduce the burden in the ICU department. In a study by American College of Surgeons, a machine learning model triaged with an 82% accuracy when surgeons had only a 70% accuracy rate on their dataset. 

3. How AI Can Diagnose a Patient Better

Doctors must use their eyes multiple times a day to determine from images whether or not an X-ray, MRI, or CAT scan shows if a patient has an illness, but this can sometimes result in a false positive. No doctor on your staff or patient in your care should have to go through the emotional roller coaster that is misdiagnosis. Not only that but short staffed hospitals can use some help with making diagnosis. 

In 2020, NYU Langone Health was able to build an algorithm that was able to identify 12% of misdiagnosed cancer patients. While not quite as accurate as doctors with 10+ years of experience yet, it’s very close. 

Their approach worked best on one specific type of cancer. The images used were also not from a diverse population so when new images were introduced it didn’t perform as well. While AI isn’t going to replace radiologists or pathologists soon, the hospitals that figure out the solution first will no doubt benefit from the multiple eyes of a computer vision model.   

Not enough people at risk for cancer to compel you?

Well in King’s College London in partnership with the University of Strathclyde is trying to determine the pre-eclampsia risk in pregnant women. They are combining a tool that gives a risk score into an integrated system for doctors and midwives to use. Through their research, women’s pregnancy monitoring will improve and become safer. But also, in 2017, the U.S. government determined that follow up care is a $1B problem a year. 

4. How AI Can Help Detect Fraudulent Events

Banking institutions around the world are using AI to prevent cyber thieves from stealing your money. That same technology is currently being used to detect fraudulent medical claims. It’s a tricky problem because AI must learn which human behaviors map to a fraudulent event.

In Nov. 2021, two men were sentenced in Florida for fraudulently claiming $134 million in VA healthcare services. That’s a lot of money for the government and its insurance providers to lose. Fraud-detection algorithms are being used to detect double billing, phantom billing, unbundling, and upcoding to prevent such events from happening. 

Anomaly detection works in banking and it can work in healthcare with the right teams and technology. Such algorithms look for signature behaviors in human activity and learn patterns that may represent fraudulent activity. As the algorithms see more and more data, they are better able to differentiate normal patterns from abnormal or fraudulent ones. 

5. How AI Can Detect Prescription Errors

Can an AI algorithm detect that a larger than normal dose of medicine has been administered to a patient? Yes, it’s possible. With staffing shortages and overworked professionals, data input errors occur. But having a piece of software that can detect anomalies helps flag a potential mistake by offering further review before that medication is administered. This same technology can also – wait for it – detect if a doctor is prescribing higher than normal volumes of medications.

This has the potential to circumvent malpractice lawsuits. It can also prevent doctors from mistakenly prescribing medications that interact with each other negatively or mislabeling medication in the health records (1 mL of Ibuprofen vs. 1,000 mL of Ibuprofen).

In Closing

The AI applications for Healthcare captured in this blog are only a fraction of the total addressable AI use cases. If you’re looking to better leverage the power of AI (especially if you’re in the early stages of your AI journey) we strongly recommend you check out our free guide on How to Implement a Successful AI Strategy at Your Company.

More to explore

Accelerate and automate your data projects with the phData Toolkit

Introducing
Data Coach is our premium analytics training program with one-on-one coaching from renowned experts.