Gender and ethnicity correlate with ED-ICU treatments and outcomes.

Adesh Kadambi BEng^1,2*, Senan Ebrahim PhD^1,3*, Shivam Dubey BA^1,4, Leon Naar MD⁵, Eliza Nguyen MPH^1,6, Hassaan Ebrahim BA^1,7, Michael Chilazi MD^1,8, Jarone Lee MD/MPH⁵, Ali Raja MD/MBA/MPH^3,5

*These authors contributed equally to this work.

Author Affiliations:

1. Hikma Health, San Jose, CA, USA
2. University of Toronto, Toronto, ON, CA
3. Harvard Medical School, Boston, MA, USA
4. University of Waterloo, Waterloo, ON, CA
5. Massachusetts General Hospital, Boston, MA, USA
6. University of South Florida Morsani College of Medicine, Tampa, FL, USA
7. Harvard John F. Kennedy School of Government, Cambridge, MA, USA
8. Johns Hopkins Hospital, Baltimore, MD, USA

Corresponding Author: Senan Ebrahim PhD, [email protected]

Abstract:

Background and Objectives: Appropriately caring for patients from the emergency department (ED) to the intensive care unit (ICU) represents a persistent challenge for physicians. Previous studies have suggested that gender and ethnicity play a significant role in the management and outcomes of ICU patients. In this study, we sought to characterize the influence of gender and ethnicity on the outcomes of critically ill patients.

Methods: In this study, we utilized clinical data from the eICU Collaborative Research Database, consisting of 200,859 ICU stays from 139,367 patients at 208 hospitals in the United States. These data were filtered to our study population by identifying trauma admissions and analgesic administration. A series of chi-squared was conducted with power at 80% and an alpha of 0.05, Bonferroni corrected to 0.00625, for several comparisons. We then applied a gradient boosted tree model for classification and calculated SHAP values for model interpretability.

Results: All-cause mortality rates for ED-ICU trauma patients were significantly different between female (3.78%) and male (5.62%) patients, as measured by a chi-squared test (p = 0.0055). We observed a significantly lower proportion of patients of Native American ethnicity receiving pain medications as compared to other ethnicities. The trauma mortality gradient boosted model performed at an AUROC of 0.73, while the analgesic use model performed at an AUROC of 0.75.

Conclusion: Our analysis suggests that ED-ICU trauma admissions have lower mortality for female patients. We have also identified that analgesia administration for ICU patients is significantly lower for Native American patients, despite higher pain scores. These clinical correlations merit future exploration at both the national and hospital levels to formulate evidence-based treatment guidelines that are equitable and optimized.