Predictive Analytics Aid Fraud Recovery; Some Call for More Efforts
February 26, 2015 in News
Although CMS’ predictive analytics fraud detection system identified or prevented about $210.7 million in health care fraud in a single year, some experts say the agency could be doing more to improve the system, Modern Healthcare reports.
Background on CMS’ Fraud Detection Tools
The 2010 Small Business Jobs Act appropriated $100 million for CMS to adopt predictive analytics tools to prevent Medicaid and Medicare fraud. CMS used the funding to hire developing teams, headed by IBM and Northrop Grumman.
As of last year, CMS said it used four analytics tools:
- Anomaly models, which alert officials based on factors that appear improbable;
- Predictive models, which compare charges with a fraud profile and raise suspicion;
- Rules-based models, which automatically flag certain charges; and
- Social networking models, which alert officials based on a provider’s associations.
The tools are yielding fewer investigations, but the amount of money gathered from each investigation has increased, according to CMS data. For example, CMS in a 2012 report found that the analytics system generated 536 investigations and assisted 511 existing ones to yield $115.4 million between 2011 and 2012. In comparison, the agency’s 2014 report showed that the system during the second implementation year generated 469 investigations and assisted 348 existing ones to yield $210.7 million in savings.
Experts Split on How To Improve System
However, some experts say CMS could do more to improve system.
Stephen Parente, a professor of health finance at the University of Minnesota, said the $210.7 million is a “fraction of what’s possible.” Parente co-authored a 2012 paper that estimated CMS’ Medicare Part B program could save $18.1 billion annually if it adopted predictive analytics tools similar to the credit card industry’s real-time fraud tools.
In addition, Parente said CMS should focus on stopping payments automatically rather than investigating fraud after it has occurred.
A source on the House Energy Commerce committee said that the 21st Century Cures initiative contains a placeholder for a measure on predictive analytics. The source said that committee members are considering what legislative action would help to bolster CMS’ use of the technology.
Meanwhile, Andrew Asher, a senior fellow at Mathematica Policy Research, cautioned that health care data are more complex than credit card information.
Asher said the adoption of fraud technology has been slow and that more could be done. However, he added there are “significant analytical and technical challenges to get this right,” adding that it is “really critical to have a high level of accuracy.”
He noted that false positives also create a problem, as they can be costly for providers and investigators.
In addition, Asher said that a completely automatic system “may not be a feasible strategy for a long, long time” (Tahir, Modern Healthcare, 2/25).