While the most infamous examples of health insurance fraud involve Medicare, private insurers like Blue Cross Blue Shield are also hit with tens of thousands of suspected fraud events every year. However, the evidence on the prevalence of fraud is stunningly weak given the apparently enormous scale of the problem. The most widely used estimates hail from the National Health Care Anti-Fraud Association, which suggests healthcare fraud accounts for between $68 billion and $280 billion a year, which as Donald Simborg noted in JAMIA, is astonishing for both its enormity and huge margin of error. Considering that the Coalition Against Fraud estimates total insurance fraud at $80 billion for all industries, this suggests almost all insurance fraud is health insurance fraud. To put that in perspective, the lower bound estimate is higher than all healthcare research and development combined, which clocks in at less than $50 billion according to the CMS.

No doubt the uncertainty about these estimates owes largely to the fact that fraud, by its nature, is hard to catch. But the inconsistency of the estimates may also stem from differing definitions of fraud. Although we typically think of medical fraud as involving identity theft or intentionally billing for drugs and treatments that were never ordered, medical billing errors may represent a much more common experience, as related by this Wall Street Journal coverage of the $10,000 mystery proceedure that the insurer paid for even though it never happened. Billing errors, which may or may not be intentional, aren't always counted as fraud, and seldom lead to criminal prosecutions.

How common are billing errors? As with fraud in general, estimates are sketchy. The most widely cited figure in the media—cited by Consumer Reports—comes from the The Medical Billing Advocates of America, who claim that an outlandish 80 percent of all medical bills contain errors. I've contacted MBAA about their methods and will update if they send a response. This Wall Street Journal piece cites Stephen Parente's much lower estimates of 30 to 40 percent, though I couldn't find this estimate in his extensive published papers (again, will update if he responds to my inquiry). The cleanest published research I found comes from JAMA, and using Medicare data from the 1980s put the estimate of billing errors much lower at between 20 to 14 percent—and falling over time—with over billing exceeding under billing to the tune of roughly 2 percent of total spending.

This over billing may not be accidental. A different study compared providers' medical databases to their billing databases to uncover discrepancies between the diagnoses made and the treatments billed for various conditions. It found low rates of patients who were prescribed but not billed for treatments—1.1 percent for heart failure and 12 percent for hypertension—but very high rates of patients being billed for things for which there was no corresponding record in the medical database—29.6 percent for heart failure and 26.8 percent for hypertension. This method does not necessarily allow us to estimate the rate of billing errors, and there may be legitimate reasons for the discrepancy, but the asymmetry between the two directions of the discrepancy suggests that many billing errors may be deliberate attempts to gouge insurers, which is a form of fraud often not classified as such.

Therefore, despite the poor quality of the data, we do have reason to suspect that healthcare fraud constitutes a significant cost for the typical health insurance policy holder.