STA 290 Seminar: Donald Edwards (U. S Carolina / UCSF)

Statistics Seminar: STA 290

Thursday, October 18th, 2012 at 4.10pm, MSB 1147 (Colloquium Room) Refreshments 3:30pm, prior to seminar in MSB 4110 (Statistics Lounge)

Speaker: Donald Edwards (Dept. Statistics, Univ. South Carolina / Dept. Epidemiology & Biostatistics, UC San Francisco)

Title: Statistics and Healthcare Fraud

Abstract: Healthcare fraud is a serious American problem: it has recently been featured on CBS’ Sixty Minutes and the NBC Evening News; loss was estimated at $69 billion in 2007 (, more than was later spent on the GM/Chrysler bailouts and the BP oil crisis combined.

Statistical methods, featured in the final volume of 2011’s American Statistician, are center-stage in the battle against healthcare billing abuse. Aggressive model-based methods aid in detecting suspicious provider (physician, hospital, power wheelchair supplier, etc.) billing behavior. Once a provider is chosen for audit, the statistical methods become very conservative: a population of payments made over a fixed time period is isolated, a probability sample selected, and a design-based method used to obtain a lower 90% confidence bound for the total amount overpaid for the period. The provider has the right to challenge the statistical methods used in the audit, and often does so: statisticians working in this arena should be ready to defend every decision they make, under oath, against an aggressive and well-funded team of provider lawyers and self-declared statistical “experts”. Hence, healthcare fraud is also a battleground for the science of statistics.

This talk will discuss how statistical methods are currently used and abused in the fight against healthcare fraud, and some new ideas for improvement, particularly count-based extrapolation methods which do not depend on the Central Limit Theorem.