Artificial Intelligence Holds Promise in Detecting Home Health Medicare Fraud

A study on artificial intelligence (AI) suggests the technology holds promise for detecting Medicare fraud within home health and hospice. However, leaders within the industry say a measured approach should be taken, given potential shortcomings and pitfalls related to AI.

The research team at Florida Atlantic University programmed computers to predict, classify and flag potentially fraudulent Medicare Part B claims from 2012 to 2015. Fraudulent activities included patient abuse or neglect and billing for services that weren’t delivered.

The team applied algorithms to detect patterns of fraud in the Centers for Medicare & Medicaid Services (CMS) data because “patterns in the data are hidden from us” as humans, said Taghi Khoshgoftaar, Florida Atlantic University director of Data Mining and Machine Learning Lab in the Department of Computer and Electrical Engineering and Computer Science.

The study tested six different machine learning methods to determine which is most accurate in identifying anomalies.

The study focused solely on Medicare Part B claims, but this type of data analytics and machine learning could translate to other areas, including those that affect home health care. For example, the FAU researchers also have used AI on claims for Medicare Part D and DME. They believe they have only scratched the surface and plan to conduct further trials.

“There are a lot of things we can do both from the data side and algorithm side — different ways of looking at data or using different algorithms to better predict fraudulent cases,” Khoshgoftaar told Home Health Care News. “This is very important not just for Medicare fraud, but also for other domains like insurance fraud, financial fraud and transaction fraud.”

Using automated fraud detection could “reduce the cost to the system” and reduce auditors’ workloads, thus preserving more resources for patients, he said.

Ferreting out fraud

CMS already has developed numerous programs and tools to fight claims fraud, which is an ongoing problem within home health care and hospice.

Two years ago, the U.S. Department of Health & Human Services’ Office of the Inspector General (OIG) performed a data analysis and flagged more than 500 home health agencies and 4,500 physicians as having suspicious practices. In an attempt to rein in fraud, CMS has taken sometimes drastic steps. Such actions include moratoria on new providers in certain high-fraud areas, and introducing programs such as the forthcoming Review Choice Demonstration that place burdens on scrupulous providers.

Hospice also is grappling with fraudulent billing and other practices that raise regulatory red flags. Earlier this year, OIG released a report detailing fraud vulnerabilities within the Medicare hospice program. It showed that every year from 2006 to 2016, about three-quarters of hospice beneficiaries did not visit with a hospice physician. OIG also indicated that hospices frequently fail to meet plan of care requirements.

Although CMS already employs data analytics and some forms of machine learning, “they can use AI to advance the tools they are using now,” Khoshgoftaar said. “It’s good to use a tool like AI that is very effective in detecting fraudulent cases.”

Home health care agencies recognize the possibilities for machine learning to improve verification and operating efficiencies.

“AI or any kind of digital review of records is certainly an improvement over the current process, which often involves physically copying pieces of paper and putting in them in fax machines or envelopes,” said Danielle Pierotti, RN, Ph.D., acting president and CEO of industry associations ElevatingHOME & VNAA.

Not a panacea

Machine learning could revolutionize home health care fraud detection, but experts recommend considering the whole picture before rushing to implement the tool.

“We need to be cognizant that every innovation has downsides … that need to be managed,” Pierotti told HHCN. “… [AI] is not a panacea. It doesn’t solve every problem.”

One problem with automated Medicare fraud detection is that simple errors in patient records could be flagged as a fraudulent claim.

“There can be labeling errors,” Khoshgoftaar said. “This is an issue we always have with AI. In our field, we really battle with this.”

The home health care industry is aware of AI’s shortcomings and the need for further testing and development.

“There is a major difference between fraudulent billing and minor claim documentation flaws,” William A. Dombi, president of the National Association for Home Care and Hospice (NAHC), told HHCN.

Computerized predictive modeling could be a program integrity tool, but documentation flaws and patient-specific clinical practices must be accounted for and not treated as fraud, he emphasized.

For machine learning to effectively aid fraud detection, “we have to make sure we have full and complete documentation, which continues to be a challenge in home-based care,” Pierotti said.

Imperfect AI programs could contribute to over-generalizations, which can lead to oversight prejudices and inaccurate conclusions, Dombi said.

“While NAHC strongly supports sensible and targeted program integrity efforts, we must be certain that individualized analysis remains a central feature of anti-fraud efforts,” he said.

Although some elements still need work and greater AI implementation will take years, home health care stakeholders see many possibilities.

“I really welcome researchers into home based care,” Pierotti said. “We need all the help we can get and all the great minds … to assist with general improvements across the board.”

Written by Katie Pyzyk

Photo Credit:

Guest Contributor

Categories:


Companies:

, , ,

By continuing to use the site, you agree to the use of cookies. More Information

The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this. For more information, see our cookie policy

Close