There’s been a lot of media coverage recently about AI, how it works and whether it will have a negative impact on the workforce. Mastercard® Healthcare Solutions develops responsible AI tools that empower workforces to be more efficient, accurate and decisive. When it comes to healthcare fraud, waste and abuse (FWA), the collaboration of humans and AI is necessary to identify the many nuances that determine legitimate claims.
The complexities of healthcare billings require skilled investigators who identify the nuances of claims, codes, and collusion. They have expertise in medical modalities, the newest drugs, and the most common fraud schemes, and juggle vast amounts of data for millions of claims.
In this article, we will explore the role of AI as an assistive tool that transforms claims triage and empowers the members of your special investigation unit (SIU) to be a strong, productive team.
AI’s role in detecting healthcare FWA
Imagine being able to help your team work better. You would enhance their performance, help them meet their quotas, and increase their workplace satisfaction.
Medicare alone processes more than 4.5 million claims a day. It’s nearly impossible for humans to detect every case of suspected FWA without the help of technology. Using an AI and machine learning solution to triage healthcare billings has a positive impact on workflow. A key advantage is the increased speed and accuracy of assessments, resulting in fact-based, high-quality decisions. Two aspects of AI, supervised learning and unsupervised learning, are at play.
Supervised learning uses intelligence from past healthcare claims to train models on billing behavior patterns. Models rapidly approve claims that match the criteria for approved billings and flags those that don’t. Models can be trained to automatically deny the obviously bad claims and flag for review before payments are made.
Unsupervised learning examines claims that don’t match a model’s training. This advantage over legacy rules-based solutions provides transaction scores based on the probability of being incorrect or fraudulent. AI focuses investigators’ work by flagging these anomalies for manual investigation. SIUs feel empowered knowing their work is targeted on truly questionable claims.
AI and machine learning have the tremendous ability to sort positive and negative outcomes; AI sorts recognized and anomalous behaviors, and manual reviews confirm or refute incorrect or fraudulent billings. By simplifying and speeding the triage process, investigators can focus on high-value, suspicious transactions.
Empowering healthcare SIUs with AI
FWA is more than fraud and may include billing errors such as incorrect or changed procedure codes and unnecessary procedures by over-cautious practitioners. Incorrect procedure records cost the U.S. healthcare system millions of dollars a year, amongst the estimated $430 billion annual losses to FWA.
Examples of coding errors include patients being billed 45 minutes for short consultations, and unbundled services, where practitioners bill for each component of a procedure, leaving one or more services off the list as justification for not using the single bundled code for the overall procedure.
AI also picks up on subtle fraud, such as a practitioner billing for examining both of a patient’s legs when only one was examined (modifier fraud).
Aided by AI models that have been trained to accurately flag FWA and reduce false positives, SIUs can focus on complex fraud or billing issues. Improved efficiencies and increased FWA identification potentially save healthcare payers millions of dollars annually.
Investigators are empowered to follow up on flagged claims and providers, knowing that preliminary review identified the anomalies, preventing wasted follow-up time and the tedious scanning of thousands of lines of data.
The essence of human expertise
When evaluating the potential for AI to cause job displacement, it might be useful to recall the popularization of pocket calculators. A new concept in the 1970s, calculators weren’t allowed for use in educational settings. More recently, educators questioned if students with low math grades were too reliant on calculators.
A 2019 study of college students showed interesting results. While calculus students with the highest grades used calculators, they were also the students who understood the fundamentals of math. The authors concluded that, “the findings raise doubts about any substantial long-term effects on college mathematics performance of calculator use in high school.”
Not unlike the calculator, AI is a tool that performs complex mathematical functions but cannot replace the essence of human expertise. Only an experienced healthcare fraud investigator recognizes the intricate nuances of healthcare modalities, billings, and myriad procedural codes. Both an art and a science, complex claims determination requires the human experience.
Our brains have unique reasoning, using intuition to pick up nuances that machines can’t learn. For example, not all patients have routine medical examinations, resulting in different stages of disease when diagnosed. Furthermore, every human is unique, with individual behaviors, symptoms, presentations of disease, pain thresholds, and more. Only human insight can determine if further investigation is required.
Embracing AI to enhance FWA detection
As AI use becomes more popular in business settings, the world is asking questions like “Does it make independent solutions I’d disagree with?” and “Is AI going to replace me?”
AI is a superior tool that optimizes workflows by triaging good, bad, and suspicious claims. Rather than replacing humans, it focuses on the most complex and costly fraud and erroneous billings for investigation. The entire SIU will be able to function more smoothly and team leads will be able to allocate resources where and how ever needed.
The AI-human partnership is certain to transform claims processing, saving the healthcare industry billions of dollars in fraud, waste, and abuse as this revolutionary technology continues to proliferate.