A yearlong investigation

A startling finding about cases of doctor sexual misconduct in Georgia prompted The Atlanta Journal-Constitution to undertake a yearlong national examination of the issue.

As is often the case with investigative reporting, this series grew out of other work. Reporter Danny Robbins was examining orders by the Georgia Composite Medical Board for his 2015 stories on prison medical care. In doing so, he saw orders allowing doctors to continue practicing after a finding that they had sexually violated patients.

He compiled those orders, discovering about 70 cases clearly involving sexual misconduct. And in about two-thirds of those cases, he was shocked to find, doctors either didn’t lose their licenses or were reinstated after being sanctioned. That included doctors who had repeatedly crossed the line with patients.

To see if Georgia was an exception, the AJC hired a legal researcher to study laws governing medical practices in every state. Meanwhile, reporters gathered studies and looked for cases around the country, compiled from news reports and other public sources. That work raised questions about the pervasiveness of doctor sexual misconduct. The research, and periodic scrutiny from other news organizations, also suggested that doctors were treated differently from other sexual offenders.

So the AJC decided to examine the system that is supposed to protect patients from doctors who sexually abuse. At first we submitted public records requests to medical boards or other regulatory agencies in every state, seeking databases identifying doctors who had been disciplined and the reasons for their sanctions. Nearly all said they didn’t keep such data, and only a few provided other information addressing our requests.

At that point, our data journalism team wrote computer programs to “crawl” regulators’ websites — a process known as scraping — and obtain board orders. This required building about 50 such programs tailored to agencies across the country. The result: More than 100,000 board documents were collected.

To assist us in identifying those involving sexual misconduct, we then created a computer program based on “machine learning” to analyze each case and, based on keywords, give each a probability rating that it was related to a case of physician sexual misconduct.

We then read all the documents for more than 6,000 cases to determine the nature of each one and board action. We eliminated cases completed before 1999 and duplicate orders when a doctor was licensed in more than one state.

That work set the stage for additional reporting. Over the past year, the project team sought other records, interviewed victims, doctors, regulators and experts, and completed other research. The project team also attempted to contact every doctor named in this investigation.