Presence of Negative Descriptors in Notes about Patients with Sickle Cell Disease Compared to Other Stigmatized Patient Populations

Introduction

Clinician bias can affect attitudes about patients and impact clinical decision-making. Managing sickle cell disease (SCD) can be challenging due to clinical uncertainty and time pressures, leading to use of cognitive shortcuts or implicit biases. These may manifest in physician notes to the detriment of patient care. Almost all patients with SCD in the U.S. are Black, and many require intravenous opioid medication to manage acute pain episodes. It is not fully understood to what degree each of these three intersecting factors – race, chronic pain, and opioid use – contribute to clinician bias toward patients with SCD.

Methods

This cross-sectional study compares the notes of patients with SCD to four comparison groups. This study builds upon prior work in which natural language processing was used to determine instances of negative descriptors in patient medical records (Sun et al., Health Affairs, 2022). Using ICD-10 codes for SCD, chronic pain, and opioid use disorder (OUD), we identified 38,816 notes across 17,791 patients for this analysis. We compared the presence and type of negative descriptors in the notes of patients with SCD to four comparison groups without SCD: 1) Black patients, 2) patients with chronic pain, 3) patients with OUD, and 4) non-Black patients without chronic pain or OUD (“counterfactual comparisons”). We also compared the first three comparison groups to the fourth group. We used multilevel multivariable logistic regression models to determine the odds of having a negative descriptor in a medical note, comparing patients with SCD to each comparison group. Multilevel modeling accounted for notes clustered within patients. All models adjusted for age, sex, marital status, insurance, encounter type, and comorbidities.

Results

Patients with SCD were younger, less commonly married, more commonly on Medicaid, and had less severe comorbid medical conditions than comparison groups. Counterfactual comparisons were more commonly married and on private insurance than other comparison groups. Of the 243 patients with SCD, 45% (111) also had a chronic pain diagnosis, and 35% (84) had an OUD diagnosis.

The most frequently used negative descriptors among patients with SCD were “refused”, “not compliant”, and “not adherent”. The negative descriptors with the largest effect sizes were “angry”, “not compliant”, and “aggressive”. Patients with SCD had higher odds of negative descriptors in notes relative to Black patients (aOR 2.33, 95% CI 1.60-3.38) and to counterfactual comparisons (aOR 5.65, 95% CI 3.28-9.74), all without SCD. By contrast, patients with SCD had lower odds of negative descriptors relative to patients with OUD (aOR 0.49, 95% CI 0.27-0.88), and they had no difference in negative descriptors relative to patients with chronic pain. Among patients with OUD, negative descriptors did not differ between those with and without SCD. Black patients (aOR 2.25, 95% CI 1.83-2.77) and patients with OUD (aOR 2.58, 95% CI 1.18-5.61), all without SCD, had higher odds of negative descriptors than counterfactual comparisons, but at a smaller magnitude compared to patients with SCD. Patients with SCD had higher odds of the negative descriptor “refused” being in their notes relative to Black patients (aOR 3.20, 95% CI 2.03-5.04), patients with chronic pain (aOR 1.82, 95% CI 1.03-3.22), and counterfactual comparisons (aOR 6.93, 95% CI 3.42-14.0), but there was no difference relative to patients with OUD.

Conclusions

The most frequently used negative descriptors for patients with SCD focused on physician expectations regarding treatment (e.g., “not adherent”), although racially charged terms such as “angry” and “aggressive” had large effect sizes. Negative descriptors were more common for patients with SCD than for Black patients without SCD and counterfactual comparisons. In contrast, the odds of negative descriptors did not differ between patients with SCD and those with chronic pain. Additionally, among patients with OUD, the odds of negative descriptors were comparable between patients with SCD and those without SCD. This suggests that the stigma associated with chronic pain disorders and OUD may be major contributors to the use of negative language about SCD patients, even though the majority of SCD patients have neither chronic pain nor OUD. Further work is needed to better understand the types of clinician bias towards patients with SCD and how to best intervene.