Healthcare fraud is harmful to patients, providers, and taxpayers. Health fraud in the U.S. alone conservatively represents $68 billion annually and could be as high as $230 billion. Thus, timely and effective fraud detection is imperative to improve the quality of care.
Accurate fraud detection that moves beyond traditional rules-based systems can equip healthcare organizations with the ability to target their resources where they can do the most good. The health insurance provider Aetna already uses 350 machine learning (ML) models to combat fraud, and new models are coming out of research centers regularly. Many of these models fit under the umbrella of anomaly detection systems, which target aberrations in large sets of data.
Anomaly Detection in Healthcare
Anomaly detection is all about finding patterns of interest (outliers, exceptions, peculiarities, etc.) that deviate from expected behavior within data. Given this definition, it’s worth noting that anomaly detection is, therefore, very similar to noise removal and novelty detection.
Anomaly detection can be used for a host of medical use cases, such as sepsis prevention, hospital bed allocation optimization, and preliminary radiology and dermatology screenings. Yet fraud detection remains a terrific anomaly detection project for the healthcare sector because it doesn’t influence the medical care directly, and can help improve clinician trust.
There is a clear return on investment (ROI) with successful fraud detection systems, which can demonstrate value. If doctors and technicians do not trust that ML systems can provide value or are capable of improving tried-and-tested methodologies, they are unlikely to integrate them into workflows. This is why communication and transparency are so critical to implementing ML-based systems; the human element and change management are inextricably linked to the success of the project.
Visualization and collaboration may seem like spheres to tackle once the project is complete, but it’s imperative that teams—especially clinical ones— understand the data and its processing from the beginning. Experts are often resistant to change, but their subject matter expertise means that involving them early on can lead to better stronger systems, in addition to ones that will actually be used.