Medibank tightens fraud detection with analytics

Automatically scans data for known patterns.

Medibank has uplifted its fraud investigation capabilities with technology that allows it to more easily recognise “patterns of malpractice” and share the data with law enforcement.

The health insurer, which has around 3.7 million members, has switched on IBM i2 enterprise insight analysis software within its internal payment integrity program.

The program houses an investigations team of around 22 including administrative staff, lead investigator Jim Hilliard told IBM’s Think 2018 conference.

“My role is for complex matters – anything that’s going to be referred to law enforcement – but we also do detection, early identification, and look for things within the business that we can fix to stop fraud occurring,” Hilliard said.

“The other side of the payment integrity program is more an educational thing for our [health service] provider network where we use data to identify outliers, speak to them, and we’ll generally get behavioural change out of that.”

Hilliard – who has been with Medibank since late 2015, and was formerly with Victoria Police specialising in organised crime – said that when he arrived, pattern recognition and analysis was mostly performed in spreadsheets and relied on the institutional knowledge of investigators.

While fraud instances were being recognised, the process was overly manual, leaving investigators time-poor.

“We’ve already got business rules in relation to some of our systems and how we identify risky memberships, risky claims and all those sorts of things,” Hilliard said.

“But the identification was coming late so we were having some [financial] leakage before we identified it. It was a huge issue.”

In addition, handing off a complex internal investigation to law enforcement was challenging because the format of the information wasn’t conducive to it being easily digestible or understood.

“When I had a complex matter that had 150 memberships that were all linked by an IP or bank accounts, providing that to police in a spreadsheet was problematic to get them to understand the actual scenario that there was likely one offender, not 150,” Hilliard said.

“What we wanted to achieve was to have a visual picture that we could provide to law enforcement.”

In his past life with Victoria Police, Hilliard had used visual intelligence and investigative analysis software from i2, which became part of IBM back in 2011.

However, he said the tool “wasn’t being used to anywhere near its full potential”, and it was that potential he wanted to now unlock at Medibank.

Hilliard called IBM to discuss the problems, the company’s investigative operations, as well as its range of underlying data sources, and ultimately to build a model that could be used to run a proof-of-concept using a subset of live data.

That flushed out enough use cases to get the project over the line, and the production system was put together over a few weeks by IBM and partner Fujitsu, with most of the time spent understanding the source data systems that would feed the i2 model.

i2 ingests daily data extracts from core business applications like Medibank’s claims system, payments system, and a data warehouse, and provides a means to visually correlate all those data points.

Investigators are able to query i2 to see what they already know about a particular person or provider.

The system is also designed to identify and alert investigators to “behavioural patterns of interest” in the data that are known to be potential indicators of fraud.

These may include immediate claims after a waiting period ends, the maxing out of claim amounts, and impossible distance-time claims.

Flags were triaged by investigators who were then able to assess whether they warranted further attention or action.

Hilliard said the deployment of i2 had been a “game-changer” at Medibank and paid for itself within the first fortnight of production use.

“i2 shows things really quickly and the money out the door [due to fraud is] significantly reduced,” he said.

Instances of fraud that previously took weeks or months to correlate can now be done in a matter of minutes, he said.

Hilliard cited an example where a pre-i2 convicted fraudster reappeared after being released from custody.

“We were actively looking for her at this stage because we suspected that she was going to come back,” he said. “We used i2 to find her [again].”

The system’s outputs had also generated “positive feedback from law enforcement around the country”, since Medibank could simply share the i2 file with police that also had instances of the i2 software.

Medibank is now looking at further enhancements to the system, including the inclusion of more data sources, and expanding the types of alerts that i2 produces.

“There’s a lot more things we see i2 can do for us on the cards,” Hilliard said.

Ry Crozier attended IBM Think 2018 in Las Vegas as a guest of IBM.

Original Post