Artificial intelligence, or AI, might be a major staple of science-fiction storytelling, but it's also becoming an invaluable tool for the healthcare industry. Its power isn't in controlling robots or threatening the human race, but in processing massive quantities of data that would be too much for a human data engineer to handle.
One of the most recent applications emerging for this technology is in the prevention of healthcare fraud. How can AI help keep patient information safe, and why is this new technology necessary in today's challenging times?
The threat of healthcare fraud
Healthcare fraud is a growing problem. Protected patient data is 10 times more valuable than credit card information, primarily because while people monitor their credit reports, they don't pay very close attention to their medical records. Hackers might get one or two large purchases out of stolen credit card data, but they can milk medical information for months before the patient notices anything is amiss.
Healthcare fraud, including the kind that impacts Medicaid and Medicare, is one of the most common forms of False Claims Act violations. This law, first passed in 1863, helps protect the government from false spending claims.
Legislators updated it in 1986, and since then, the program has recovered more than $35 billion. Healthcare fraud doesn't only look like someone stealing a patient's private information. It comes in many forms, from kickbacks to physicians and pharmacists from pharmaceutical companies, or knowingly selling defective test kits or medical devices. That's where AI might prove to be invaluable.
AI and fraud prevention
The healthcare industry generates incredible amounts of data every day, from patient diagnoses to diagnostic imagery stored digitally on hospital servers. It's more data than a single human statistician — or even a team of them — could hope to sort through.
AI and machine learning systems can sort, categorize and analyze vast amounts of information in a fraction of the time. AI systems also tend to be more accurate because they remove the possibility of human error.
These systems are so much more efficient than human analysts because the constructs of biology don't hold them back. When programmed correctly, they have nearly limitless memory and processing capability, something not even the smartest human brain can manage. Instead of looking at recent data, an AI system can comb through years of patient history and related data in seconds, finding patterns a human analyst might overlook.
Detecting anomalies in healthcare data
AI is already benefiting several different industries, slowly integrating itself and becoming an invaluable tool. How will this kind of automated data analysis work in healthcare?
The technology will work in much the same way — analyzing data, looking for patterns and trying to find specific criteria that could indicate the presence of healthcare fraud. These could include anything from looking for documentation to verify a patient received the service they paid for, or systems programmed to look for the kind of evidence that indicates a simple procedure is getting billed as something more complex, a fraud behavior known as upcoding.
The challenge lies in creating a system that understands and can process the human element without making errors. Ask 100 different doctors to enter data into a system, and they'll do it in 100 different ways. Each professional has a digital fingerprint.
An AI system needs to be able to tell the difference between one person's unique way of entering data and authentic fraud behavior. Additionally, these systems will need to be as secure as possible to protect patient information on networks that, ideally, will span the globe.
Not a substitute for the human element
While AI might be a game-changer when it comes to detecting healthcare fraud, it isn't the only solution. The purpose of an AI system should be to complement current fraud detection processes, rather than to replace them. Computer systems are no substitute for human analysis.
While AI can detect obvious signs of fraud, it will likely flag anything that seems out of place. That includes things like typos — a natural mistake that occurs when you've got human doctors and nurses putting in information manually — and other errors. While correcting these will ensure the patients are getting the care they need, they aren't necessarily fraud.
AI will be a valuable tool for preventing healthcare fraud as more data makes its way into the industry, but it isn't foolproof.
It should never replace analysts who understand human error and will prevent the system from flagging things as fraud that might not be anything more than a typo entered by an overworked and overtired nurse or doctor.
Kayla Matthews is a MedTech journalist and writer. Her work has also been featured on Medical Economics, HIT Consultant, HealthIT Outcomes and Health IT Answers. To read more from Kayla, please visit her blog here.
