After Breach, a Health System Chooses Disclosure Over Delay

Archi4 min read

breachsecurityhealthcareincident

A 48-Hour Window and an Incomplete Picture#

The first sign was a single failed login at 2:14 AM. By 6 AM, the security team at Midstate Health Network had traced the failure to a vendor account that should not have had database access. The account belonged to a scheduling contractor, and the login came from an IP address in a country where the contractor had no offices. Logs showed the account had been active since 11 PM the previous night, pulling rows from a table of 1.2 million patient records.

The breach was stopped at 10:14 PM that evening when an engineer revoked the vendor's credentials. But that left the organization with a harder problem. The network had 48 hours to decide what to tell the public, and the forensic review would take at least four days.

State law in this case required notice within 72 hours of confirmation. The board's general counsel read the statute aloud during an emergency meeting at 7 AM. The phrase in dispute was "upon confirmation." The security team had confirmed that an attacker had accessed the database. They had not confirmed which records had been copied, whether the attacker had reached backups, or whether the credential theft was part of a wider campaign. The counsel's advice: if the network waited for the full investigation, it risked violating the law and losing control of the message. If it disclosed too early, it might admit to more harm than actually occurred.

The board had three options. The first was to issue a short notice within the hour, using the language from the law and little else. The second was to wait until the forensic firm produced a preliminary report, expected in 36 hours, and then release a detailed timeline. The third was to notify only affected patients directly and post a short statement on the website, avoiding a press release.

Each option had a cost. A short notice would likely trigger news coverage with more questions than answers. A detailed report would arrive after the 72-hour window had closed. A patient-only notice would satisfy the law in some states but not others, and the network operates across three states with different deadlines.

The discussion turned to the one fact everyone agreed on. The vendor account had queried the patient table 847 times over a 14-hour period. Each query returned a subset of fields: name, birth date, address, and the last four digits of a health insurance ID. No clinical notes, no lab results, and no Social Security numbers were in the table. The table itself was a scheduling export used for appointment reminders.

The decision came at 9:40 AM. The board chose the second option, but with a modification. The network would hire a second forensic firm to run an independent check of the first firm's findings. That work would start immediately, and the network would release a joint statement from both firms at the same time as its own notice. The goal was to make the eventual disclosure comprehensive rather than piecemeal. The risk was that the preliminary report would take longer than 36 hours.

At 2 PM, the first forensic firm reported that the attacker had used a stored credential inside a shared spreadsheet. The spreadsheet had been uploaded to a file-sharing site in 2019 by a former employee. The credential had never been rotated. The firm also found that the attacker had tried to access a second database containing billing records, but the attempt failed because the vendor account lacked permission. That second database held 400,000 payment card numbers. It was not touched.

The independent firm confirmed the findings at 6 PM. The board's public notice went out at 8:12 PM, 46 hours after the initial discovery. The statement said that a scheduling vendor account had been compromised, that the exposed table contained 1.2 million rows of personal information, and that the network had no evidence of copying beyond the 847 queries. It also said the billing database was not reached.

The sentence that took the longest to write was one phrase: "The queries appear to have been automated and systematic." The board argued over "appear" for an hour. The security team said the pattern was consistent with an automated script, but they could not prove it because the logs did not record the script itself. The final wording was "the queries were repeated in a regular pattern," a sentence that satisfied everyone and no one.

The aftermath is still unfolding. The vendor, a scheduling company with 60 employees, said in its own statement that it had reset all credentials and would no longer store them in cloud spreadsheets. The state attorney general's office opened an inquiry under the breach notification law. The network's insurance carrier will cover most of the forensic costs, according to the board minutes released after the notice.

For other organizations, the decision made by Midstate is a case study in a common problem. The law says disclose, but it does not say when the facts are complete enough to disclose. Choosing to wait for a second opinion added 10 hours to the timeline. It also gave the public a report that two independent firms had signed. That single paragraph, buried in the notice, turned out to be the difference between a one-day story and a two-week investigation by local media.

The network's systems are back online. The scheduling contractor is no longer a contractor. The spreadsheet with the stored credential was taken down within an hour of the discovery. A new incident response plan now includes a calendar reminder to rotate vendor credentials every 90 days.

The breach did not involve ransomware, though the attacker's command and control server was later linked to a known ransomware group. The group never sent a demand. The network's decision was never about paying. It was about how much to say, and when.

More articles