Race-Neutral Risk Calculator Still Disadvantages Black Patients
Discipline: Data Science
Subcategory: Physiology and Health
Session: 4
Room: Farragut North
Amaya McNealey - Georgia Institute of Technology
Co-Author(s): Lauren Steimle, Georgia Institute of Technology, Atlanta,GA; Gian-Gabriel Garcia, Georgia Institute of Technology, Atlanta,GA
There has been a longstanding concern in healthcare about the appropriateness of using race and ethnicity in predictive algorithms to facilitate prognosis, diagnosis, and treatment. For example, in maternal health, the use of a risk calculator for predicting the likelihood of a successful vaginal birth after Cesarean (VBAC), which is more favorable than a repeat Cesarean, prompted criticism due to its use of race/ethnicity as a predictor. The calculator disproportionately recommended that women identifying as Black and Hispanic have repeat Cesarean deliveries, which are associated with a higher risk of complications. To mitigate these concerns in “VBAC 1.0” (the original calculator developed in 2001), “VBAC 2.0” was then created in 2021 and replaced the use of race/ethnicity with an indication of chronic hypertension. However, while the statistical performance of VBAC 2.0 was similar to VBAC 1.0, it is unclear to what extent VBAC 2.0 changes risk scores for vulnerable populations. It is also unclear how these updated predictions affect the likelihood of recommending a trial of labor after cesarean (TOLAC) which is thought to be a more favorable outcome for patients if eligible. To compare the effects of both calculators, we conducted a secondary analysis of pregnant patients who had a prior low transverse delivery at Grady Memorial Hospital in Atlanta, GA. We used both the VBAC 1.0 and 2.0 calculators to separately estimate the predicted likelihood of a successful VBAC for each individual based on information received at their first prenatal visit within a gestational age of less than 21 0/7 weeks. We then examined whether the estimated probability of successful VBAC from the 1.0 and 2.0 calculators would be favorable for recommending TOLAC when using a decision threshold of 0.6, as recommended by the American College of Obstetricians and Gynecologists. 451 individuals met the inclusion criterion. The cohort was 84.7% non-Hispanic Black, 7.5% Hispanic, and 7.8% were neither Black nor Hispanic. Of the 451 patients, only 8 (4.7%) of the 169 patients who would have previously been recommended a C-section were now recommended a TOLAC. Moreover, only 4 (3.2%) of the 130 Black patients who would have been recommended a C-section previously were now recommended a TOLAC. Meanwhile, 118 patients (41%) of the 282 patients who previously would have been recommended a TOLAC were now recommended a C-section. The VBAC 2.0 calculator not only seems to limit the likelihood of a TOLAC recommendation in comparison with the original calculator but also seems to exacerbate this limitation for Black women. Our results emphasize the need for further research into when and how to consider race/ethnicity in clinical support tools and AI/ML models.
Funder Acknowledgement(s): N/A
Faculty Advisor: Gian-Gabriel Garcia, giangarcia@gatech.edu
Role: I identified and formulated the problem statement in addition to conducting the literature review. I worked closely with clinicians to effectively clean my data and understand the patients with my cohort, specifically concerning their comorbidities and obstetrical history. I also applied the risk calculators to my cohort and determined their risk score, in addition to conducting an analysis of the results.