Background/Purpose: Systemic lupus erythematosus (SLE) is a heterogeneous autoimmune rheumatic disease whose prognosis varies by race and sex. To understand whether the cognitive processes of clinical decision-making influence the diagnosis of SLE, we conducted a factorial survey experiment to evaluate variation in primary care providers’ (PCP) diagnosis of SLE when all aspects of a clinical case are held constant, with the exception of race and sex.

Methods: An online survey was distributed via email from 04/11/2024 to 06/10/2024 to PCPs across the US. Participants were presented with five clinical vignettes, one of which was SLE, where we randomly modified the race (Black/White) and sex (female/male) of the “case” (Fig 1). Primary outcomes were correct text-based responses for SLE diagnosis after initial case presentation (top three initial diagnoses) and after reviewing additional lab results (final diagnosis). We secondarily measured participants’ review time and planned next steps (treatment, referral, tests) as a proxy for certainty. We calculated descriptive statistics for all outcomes (mean, median, proportion) stratified by assigned randomized factor and used chi-square tests to evaluate between group differences with respect to correct diagnoses.

Results: A total of 1036 PCPs (42.6% women, mean age 52.1±12.0 years) completed the survey. Most completed residency ≥11 years ago (76.4%) and indicated their practice focused on patient care (85.6%). At initial case presentation, 63.7% (660/1036) of participants correctly identified SLE within their top three differential diagnoses. These correct initial diagnoses significantly differed according to race and sex, with the highest proportion of correct SLE diagnoses occurring for Black females (Table 1). Overall, most participants elected to order tests (91.8%, 951/1036) and only 7.9% (82/1036) elected to send a referral, of which 95.1% were to rheumatology. Median review time among participants with correct initial diagnoses was longer for White male vignettes (178 s), followed by Black female (154 s), Black male (139 s), and White female (133 s) vignettes. After participants reviewed lab results, the proportion assigning a final diagnosis of SLE (64.0%, 663/1036) remained unchanged from the initial diagnosis. A final diagnosis of SLE occurred more often among female versus male vignettes (Table 1). Infectious arthritis and viral illness were common incorrect diagnoses for Black and White male cases, respectively (Fig 2).

Conclusion: We found that despite identical clinical information, a PCP’s diagnosis of SLE was significantly influenced by the race and sex of the patient. Indeed, the highest accuracy at initial diagnosis occurred for Black females, suggesting cognitive biases in clinical decision-making. These results persisted at the final diagnosis after reviewing additional lab results. Review times also varied by race and sex, with longer times for White males, hinting at varying degrees of diagnostic certainty. Altogether, findings underscore the importance of addressing implicit biases in medical training and practice to ensure equitable diagnostic processes for all patients.

« Back to ACR Convergence 2024

ACR Meeting Abstracts - https://acrabstracts.org/abstract/beyond-the-symptoms-exploring-cognitive-bias-in-lupus-diagnosis-within-primary-care/