Goldilocks and the three data breach estimates
Estimate in haste, repent in leisure?
Over on PHIprivacy.net, I recently reported on a breach in Jersey City involving patient records stolen from a shed behind a doctor’s office. The first media report, on NJ.com, said Dr. Nisar A. Quraishi told police that 40,000 patients’ records had been stolen. At 40,000, that incident would qualify as the second largest breach in New Jersey involving patient information since HHS started posting breaches in September 2009.
Two days later, in follow-up coverage, NJ.com reported that the doctor had revised his estimate of the number of patients affected downwards to 5,000 – 10,000. But by then, most sites that had picked up the original story had moved on or did not go back and update their blog entries or reports. As a result, a Google search of the doctor’s name returns results that show the 40,000 figure in a headline or summary, and not the lower estimate.
So was this self-inflicted reputation harm? Yes. Of course even 5,000 is a lot of patient medical records to be left in a shed behind an empty office, but 5,000-10,000 is not as bad as 40,000, is it?
Today, the situation got even more confusing, as HHS added the incident to its public breach tool. It seems the doctor informed HHS that 20,000 patients were affected.
So which is it: 20,000 patients or 5,000-10,000 patients? Because HHS’s breach tool doesn’t indicate when the report was filed with them, we don’t know whether the 20,000 estimate is more current and more accurate than the 5,000-10,000 estimate, or whether the latter is more accurate. On some level, all of the reports are suspect because the doctor’s failure to provide an exact number might suggest to some that he didn’t maintain a master index of the stored records that would tell him exactly how many patients had records stored in the shed.
Of course, it may not matter anyway in terms of reputation harm if all most people remember is a report that 40,000 patients had their medical information (and in many cases, SSN) stolen because a doctor left paper records going back to 1982 in a shed behind an office he hadn’t used or been to in months.
In any event, while it is frustrating when entities don’t disclose the number affected by a breach, rushing to disclose a number and then having to revise it may be damaging to your reputation. If you overestimated, people may not see the revised lower figure. If you underestimated, people may accuse you of initially trying to downplay the breach when the higher figures come out. And either way, some people will be upset that you had no idea how much data you were retaining and will find you less trustworthy as a result.
So unless you’re sure you have a fairly accurate estimate or unless a state law requires you to provide a number before you’re ready to, why not give yourself a little bit of time to make the determination?