In 2003, straight out of college, Janet Campbell started her first real job as a software developer at Epic, one of the country's biggest producers of electronic health care records. At a desk in the company's Verona, Wisconsin headquarters, she worked on a straightforward project: Building a feature that could restrict diagnostic codes to patients of a specific sex. That way, a clinic could get an alert if a provider tried to bill for a cervical exam, for example, in a patient marked "male."
Tinkering with a section of code about a year into the job, Campbell found herself fixated on the field doctors used to document patients’ sex. "This is weird," she thought. It had just three values: male, female, and unknown.
If they wanted to, clinics could work with Epic to add more choices to the list. But Campbell, thanks to the gender studies classes she took alongside her comp sci coursework, saw a deeper problem: That one field was doing too much. What if, instead of containing the patient's legal sex or sex assigned at birth, it also contained their gender identity? All of those data points were important in a health care setting. And for transgender people and other gender non-conforming patients, she realized, they often don’t match.
Many clinical settings fail in caring for transgender people. According to a 2015 report, 33 percent of transgender people surveyed had at least one negative health care experience in the past year related to their gender identity. Part of the problem is the reluctance of clinicians to simply ask—many don't know how to talk about gender identity, or fear offending patients. And that one-size-fits-all question in Epic’s electronic health care record—the type of system used in nearly 80 percent of outpatient clinics—certainly wasn’t helping matters. So Campbell, now vice president of patient experience at Epic, set out to change it.
Not long after she noticed the patient sex field, Campbell offered to give her team at Epic an educational presentation on gender variation—part of a regular series of workplace talks. She was one of the most junior people in the room, surrounded by other developers, mid-level managers, and a division manager. She still has the slides she presented that day. "Epic is almost completely unable to deal with this level of complexity," reads one, "and neither is the medical profession, for that matter."
The first customer to reach out for help correctly documenting gender identity was probably a user from a University of Wisconsin-affiliated clinic who contacted customer support in 2005, about a year after Campbell's presentation. "I’m sure other customers have had the same question,” reads the support log entry. They had.
In particular, health care workers at LGBTQ-centric organizations like Boston’s Fenway Health had started agitating for more accurate sexual orientation and gender identity data collection. Adding them would help provide better care to LGBTQ patients, including getting them the right preventive screenings, risk assessments, and behavioral health care. In 2014, more than 150 health care and advocacy organizations signed a letter requesting a multi-part question for gender identity in the electronic health care records that providers must use if they want federal funding.
Two years earlier, Campbell had come to a similar conclusion. In January 2012, she posted an internal wiki page containing her imaginings on the design changes needed for a two- or three-step sex and gender question. Meanwhile, a handful of other employees, several themselves transgender or gender non-binary, formed a workgroup to help customers collect gender identity information.
But external pressure to actually build the new features didn’t come until 2015. That’s when the Office of the National Coordinator of Health Information Technology released new regulations: Any outpatient clinic receiving federal incentive payments for using a government-certified electronic health care record—as 78 percent of them do—would have to use software that collects sexual orientation and gender identity information by 2018. Federally Qualified Health Centers, which receive more federal funding, had an even earlier deadline, in 2016. Suddenly, companies like Epic had to make big changes.
At Epic, the small workgroup exploded into a 25-person troupe—they called themselves the Volunteer Army. Changing their product to collect new data was not as simple as it sounds. "It's like the Y2K of the health record," says Campbell.
The code referenced the original patient sex field in hundreds of places. It was in code creating the patient header, a section with each person’s name and demographics. But it also popped up in less obvious places, like the chunks of code calculating normal ranges of blood tests that vary between genetic males and females. It even appeared when generating genetic pedigrees, with males depicted as squares and females as circles. Starting in early 2016, the Epic team had to comb through each appearance, figuring out what information was actually relevant: sex assigned at birth, legal sex, or gender identity.
Other electronic medical record manufacturers were similarly overwhelmed by the task. Rachel Miller, who oversees customer experience at the mental-health-focused Foothold Technology, says the code base used to write their sex data collection tool is older and less flexible than newer areas of the program. Huge amounts of data are attached to the field—and losing any of them while remapping to new values is a developer's worst nightmare. "Having to clean it up is a bigger bite than adding something new," says Miller.
Even with the Volunteer Army on the case, Epic is still cleaning its code to ensure gender identity is displayed consistently and correctly. But in June 2016, it went ahead and released its two-item gender identity question as a special update—sort of like a Windows patch—to its clinical customers.
That doesn't mean everyone is ready to use the new feature, though.
When the company started teaching its customers how to use the new fields, a new set of problems bubbled up. Several members of the workgroup found that customers were using wildly different practices to collect patients' gender identity. "The registrar would look at the patient and fill out whichever value they thought was correct," says Campbell. "Or maybe they'd put what was on the driver's license, or if it was a children's organization, the birth certificate."
That had potential to result in patients being called the wrong name or pronoun—not an inconsequential event. For transgender patients, "being misgendered is very distressing," says Adrian Daul, an emergency medicine physician in Atlanta, "and can be a deterrent from getting care in the first place." Furthermore, electronic health care records are often used as a medical census of sorts. Not documenting transgender status results in being undercounted—a problem when it comes to setting research, policy, and funding priorities.
Epic is working to change those practices. Back in 2013, the workgroup released a strategy handbook to educate clients on the best ways to capture and use gender identity information. And as the software has changed, so has the handbook. It's more than a technical guide to structuring workflows: In a subtle way, it's something of a primer on transgender cultural competence. After an introduction highlighting health disparities in sexual minorities, sections gently explain the difference between gender identity and sex, and suggest that health care workers consider which providers need to know about a patient.
Campbell hasn't heard of any health care organizations rejecting the changes. But she has heard, "We're not sure we're ready for this yet." She suspects resistance to turning on the functionality stems mostly from a lack of confidence. Alex Keuroghlian, assistant professor at Harvard Medical School and director of the National LGBT Health Education Center at The Fenway Institute, is somewhat less sanguine. He’s encountered "clinicians saying, 'I treat everybody with respect—I treat everybody the same way.'" That's not appropriate, he says: "The reality is, to treat this particular subpopulation with respect, you can't treat them in exactly the same way."
With the exception of Federally Qualified Health Centers, collecting the new data is still optional. Which isn’t all bad: "If you try to have the technology force something an organization is not ready for, it's going to result in a poorer patient experience," says Campbell. If a patient is misgendered during a visit despite communicating preferred pronouns, for example, the negative impact may be worse than it would've been.
So institutions are turning on the new questions—slowly. In Atlanta, Adrian Daul has been asking his Atlanta hospital to start sexual orientation and gender identity data collection in its version of Epic for about a year. He says the decision has been delayed by disagreement over who should collect the data from the patient: Some feel it should be the doctor's responsibility, while others feel it should be done by registration personnel. "Everyone has to be on board at the hospital in order to be collecting this stuff and then using it in a sensitive way," he says.
As for the rest of Epic’s customers? Campbell estimates only about 10 to 20 percent have switched on the feature, and another 20 percent or so have expressed interest.
Turning on gender identity functionality isn't hard, she says. "Becoming an organization that can both ask that responsibly and then deal with the information respectfully and responsibly," though, "is a much harder challenge."
