In clinical medicine, diagnostic decisions are often taught through the context of algorithms. While algorithms are helpful in interpreting information and triaging patients, our column discusses how algorithmic thinking and anchoring bias, or the propensity to rely on one piece of information to inform diagnosis, increases the effect of racial and gender bias. We explore how these patterns of thinking can lead to increased errors in diagnosing and treating patients of color, with this occurring more frequently in female patients.

Case Presentation 

A 24-year-old female with a past medical history of tubo-ovarian abscess, surgical abortion in 2019 complicated by uterine perforation and recurrent venous thromboembolism (VTE), presented to the emergency department with severe abdominal pain, emesis, and nausea. Her mood was agitated and tearful and on exam and she was reported to be in significant distress. She had presented for similar abdominal pain one month prior and reported increasingly irregular menses and severe pain. She was treated for pelvic inflammatory disease (PID) during that admission but had no improvement in her pain symptoms. She underwent a CT scan of the abdomen/pelvis which showed mild interval increased size of the complex collection within the endometrial cavity concerning for hematometra or pyometra. 

The patient had a history of sexually transmitted infections (STIs) in the past, and her pain symptoms were thought to be from PID-related changes to her cervix and uterus, despite the imaging evidence suggesting pyo- or hematometra as well as ovarian vein thrombosis. Upon further investigation, the patient did not have a recent positive STI screening and was last positive in 2015. The patient had progressive, continuous pain, so specialty consultations were requested from Infectious Disease, Obstetrics and Gynecology, Gastroenterology, and Hematology. All specialists felt that her pain could not be attributed to an etiology related to their specific service. After reviewing the imaging in the setting of her continuing pain, she ultimately underwent a dilation and curettage procedure to remove the hematometra collection. Within 24 hours, the patient’s pain had markedly improved and her overall mood and ability to communicate had also changed. 

Discussion 

We can ask ourselves if our clinical reasoning is affected by omitting the patient’s race above. The patient identifies a black woman. 

Would it have changed the initial misdiagnosis of PID one month previously? In this case, diagnostic error occurred at two different visits at two different institutions. Was this influenced by the patient’s gender, race, and age? 

In our case, multiple issues, including anchoring onto PID and racial and gender bias affecting the interpretation of the patient’s pain and acuity, had led to initial misdiagnosis and delay in treatment. In several studies, pain levels are noted to be interpreted disparately between patients of different races, as well as description and communication styles. For example, one analysis of 1.8 million records concluded that the language used in clinical notes showed that physicians tend to focus less on the pain, emotions, and physical diagnosis of Black patients as opposed to White patients.¹ This phenomenon is directly linked to the undertreatment of Black patients’ pain, which is increased with inclusion of female gender. 

In a study by Beach et al, physicians are more likely to communicate disbelief of the concerns of Black and female patients. The use of judgment words and evidentials (phrases used to convey a lack of credibility) was significantly higher in Black and female patients, with a statistically significant interaction in the number of evidentials in clinical notes about Black female patients.² For example, in our case, clinical notes use language such as “patient reports compliance with PID treatment,” corroborating the suspicion physicians had that her symptoms were again due to possible PID. These studies both support the existence of implicit bias, both racial and gender bias, in how physicians interact with and interpret the symptoms of Black female patients. Lastly, the rates of anchoring bias and the process of triaging patients are different across races. In Boley et al, the process of triaging patients in the emergency room is largely influenced by racial bias. The Black patients, across all chief complaints, were rated as low acuity when compared with White patients with statistical significance.³ In our patient, her symptoms were presumed to be due chronic PID changes in upon presentation despite CT and transvaginal ultrasound evidence of worsening fluid collection. The assumptions made due to her race, gender and age led to a missed diagnosis and delay in treatment, specifically ascribing sexual promiscuity and melodrama to the patient’s behavior.

Conclusion 

Racial bias further worsens the existing healthcare disparities through physician perception of patients, downplaying the complaints of Black patients, and undertreatment of Black female patients’ pain. Our understanding of the role that racial bias plays, especially when first recognizing a patient’s symptoms in the emergency department, is essential to correct the differences in the treatment delivered to Black female patients. In the future, acknowledging and identifying racial and gender bias within clinical algorithmic thinking should be performed at all stages of medical training to promote equitable delivery of care to all patients.