Sometimes hospitalists say negative things about their patients in the electronic health record (EHR). But would you guess that they may be more than twice as likely to do so when the patient is Black?
This latest indication of racism in health care comes from an analysis of nearly 18,500 patients with 40,113 history and physical notes, written from January 2019 to October 2020 at the University of Chicago. Researchers used machine learning to identify the terms “nonadherent,” “aggressive,” “agitated,” “angry,” “challenging,” “combative,” “noncompliant,” “confront,” “noncooperative,” “defensive,” “exaggerate,” “hysterical,” “unpleasant,” “refuse,” and “resist” in the EHR.
These words were more common in the records of patients who were Black versus White. Similar increases in their use were seen for patients who had Medicare or Medicaid versus private insurance, and were unmarried versus married, as well as whether the encounter was in the hospital rather than outpatient, according to the results, which were published online by Health Affairs on Jan. 19.
Two of the study authors, ACP Medical Student Member Michael Sun and Monica Peek, MD, MPH, a professor of medicine, both at the University of Chicago, recently provided perspective on their findings.
Q: Were you surprised by the results?
A: Mr. Sun: To be quite frank, I don't think any of the coauthors were very surprised by our findings. It was, in part, a study to explore data which might validate what we knew to be true in experience. … What was surprising was to see the effect of the different variables, [such as] … Medicare or Medicaid versus employer-based insurance. Certainly we were somewhat surprised to see that different encounters—inpatient versus outpatient versus the ED—had a role as well. And then also the effect of the COVID-19 pandemic: We actually saw a lower adjusted odds for encounters during that time period having a negative descriptor.
Q: What do you think might explain the reduction in negative language during the pandemic?
A: Mr. Sun: Unfortunately, we can't really point to any one thing that drove this finding. At least for me, the takeaway is that changing provider language is not impossible or is certainly more attainable than our previous practices might suggest.
Dr. Peek: I would add to that people have been talking about the dual pandemics of COVID and racism, this racial reckoning, happening simultaneously. So despite the additional drain on the workforce because of the pandemic, which actually predisposes physicians to become burned out and use cognitive shortcuts like implicit bias because of the extreme workload, we found that physicians were documenting fewer negative descriptors about Black patients during this time. When George Floyd was murdered, the entire world stopped and took noted of the plight of African Americans. And I think that may have shifted, even temporarily, how people were perceiving this population and … it may have manifested in how physicians were documenting their notes about African Americans.
Q: Why do you think inpatient encounters were more likely to show this negative bias?
A: Mr. Sun: Outpatient versus inpatient, there's the longitudinal relationship, there's a lot more personal investment. You often have at least a little bit more time with the patient and a lower-pressure setting. You might have a more empathetic interpersonal encounter. That's one idea. The other idea is that as a hospitalist or in an inpatient ward, we have language that we often default to using, such as “This patient was noncompliant with x medication” … portraying bias that we don't realize. We're not saying that hospitalists are intentionally or overtly using these words in a biased way, but … we're trying to de-normalize this.
Dr. Peek: When you get to know patients over time, you really get to know them as individuals in all their humanity, you're less likely to make generalized assumptions about who they are based on what social group they belong to. For hospital medicine, one [takeaway] is thinking about ways to enhance the continuity-of-care relationship. Some hospitals, like the University of Chicago, are playing around with different models of care for inpatient medicine. For example, for people who are frequently hospitalized, having an identified hospitalist that takes care of them when they're hospitalized and when they're in the outpatient setting to increase their continuity of care … kind of like the good old days.Q: Are there any other strategies that you think should be considered as solutions to this issue?
Mr. Sun: Everyone has the ability to think about the language that we're using. We hope that this study really prompts reflection on our day-to-day practices. And then there's also peer-to-peer intervention, giving feedback to others. … Within my medical education, we had workshops and practice runs writing up patient notes, and then I hit the wards, and residents and attendings are always giving me feedback on how to make the note better. Having a stronger stance on what language we should be using, what things we should consider when writing our notes, at the education level is going to be really important. … This would [address] the symptom of our language being in the medical record, but not the underlying bias, not the underlying ways in which we are perceiving our minoritized patients differently.
Q: How do you tackle that underlying bias?
A: Dr. Peek: We're seeing now a lot of efforts to really try and address racism structurally within health care institutions—not just cultural competency or diversity, but racism. It requires a shift in your thinking. When we see a person or an entire population of people who have essentially been run through the machine of structural inequities and tossed out at the other end, we are like, “Oh, look at this individual, they look a mess,” instead of us thinking, “What challenges they must have had to endure.” … It's so much easier to see patients in the state that they're in and to blame those individuals, as opposed to having a more structured analysis of this society that we live in. … We need a better sense of how we can equip all providers with a better understanding about how structural inequities create poor health.
Q: You mentioned peer feedback as a strategy. How difficult is that, particularly as a trainee?
A: Mr. Sun: When I hear my residents or my attendings use this language or worse language, I feel conflicted. I'm not going to bring up this study, and I'm not going to bring up how I feel about these day-to-day words every single day, because they impact my evaluation. There is always a power dynamic to be considered. As younger physicians or medical students, we really need these leaders or attendings who have been doing this work for a long time to help advocate … and make this thinking and our values a lot more accepted on the wards, and not just in the classroom.
Q: Do you think hospital notes being open to patients has had or will have any impact on this?
A: Mr. Sun: There's a paper that we cited that looked at the effect of the implementation of an open notes policy at one institution. Analyzing physician language for how relatable and accessible it is, in terms of nonmedical speak or the way that they're describing their patients, they largely didn't see a difference. … I think it is worthwhile to note that we do pay very close attention to our language in terms of medical malpractice, or when patients do bring forward concerns, such as weight, having preferences about terms like obesity. … Clearly it's possible for us to change, it's just a matter of which interventions or what can we really do proactively. … You can imagine that patients who have full access to their hospital records may feel very negatively about the words that we're using on a day-to-day basis. I hope that we can start to consider what we say to our colleagues, but also now indirectly to the patient.