I am attending the American Bar Association’s 28th Annual National Institute on Health Care Fraud this week.  Yesterday, I had the opportunity to speak on a panel discussing data mining and other health care fraud enforcement trends with a great group of lawyers – AUSA Carolyn Bell, Kevin Napper, and Andrew Feldman.

The topic of data mining is timely because it has been widely reported that the Government is using real time data analytics to identify potential targets of criminal investigations.  The Government is collecting electronic billing information from health care providers across the country and centralizing the information in one location.  OIG has employed data analytics to constantly review and analyze the data to identify any “spikes” or “outliers.”  The Government refers to this information as “reliable” and “unfiltered.”

We discussed the use of data analytics in the Melgen case.  AUSA Bell was one of three prosecutors that worked on the case.  The Melgen case involved the prosecution of a South Florida ophthalmologist for health care fraud.  Dr. Melgen was reported to be on of the highest individual doctor billers for Medicare.  In the trial, the Government used charts that summarized Dr. Melgen’s billing data and compared it to his peers.   The charts showed huge differences in billings and claims submitted by Dr. Melgen as compared to his “peers.”  For instance, one chart showed that Dr. Melgen collected $57,371,547 for one procedure, while the National median collections for other providers for the same procedure was $3,030,041.

The Government introduced the charts as Rule 1006 evidence.  This was heavily disputed through motions in limine.  The defense argued that the charts should come in through a medical statistician under Rule 702. The Court ultimately ruled that “to the extent the government seeks to present comparison data without laying a proper foundation that the data provides a valid basis for comparison. This, however, does not preclude the government from presenting evidence of the percentages of Defendant’s patients that received various treatments and services based on a review of Defendant’s records, and from inquiring of its expert witnesses what types of percentages would they expect to see in their experience. See United States v. Chibber, 741 F.3d 852, 854-58 (7th Cir. 2017).”

Everyone on the panel was in agreement that defense attorneys will need to hire statisticians to assist in the defense of health care fraud prosecutions.  The statistician will need to review the data to identify flaws in the Government’s analysis and in order to explain why anomalies exist.

More importantly, the statistician will need to attack the Government’s “statistically significant sample” of patients.  Because it would be impossible to review all patient records in these data driven health care fraud prosecutions, the Government picks a “statistically significant sample” of patients to review and present to the jury.  The Government will extrapolate and claim that the sample is representative of all of the provider’s patients. The defense needs the statistician to establish that the Government sample is statistically deficient and not representative of all the patients.