Association of red blood cell distribution width with hospital admission and in-hospital mortality across all-cause adult emergency department visits

Woo Suk Hong, Ákos Rudas, Elijah Bell, Jeffrey N. ChiangJAMIA 2023 · DOI

Abstract

Red blood cell distribution width (RDW), a component of the complete blood count panel routinely ordered as the initial workup for the undifferentiated patient, may be a generalizable biomarker for acuity in the Emergency Department (ED).

Introduction

Emergency Departments (EDs) are high-stakes environments requiring rapid triage to distinguish between patients who can be safely discharged and those requiring urgent admission or at high risk of mortality. While specific biomarkers exist for certain conditions, such as troponin for myocardial infarction, there is a lack of generalizable biomarkers that apply across the entire spectrum of adult ED visits.

Red blood cell distribution width (RDW) measures anisocytosis, or the variation in the size of red blood cells. While previously linked to outcomes in specific conditions like heart failure or sepsis, its utility as a broad-spectrum triage tool for all-cause visits has remained largely unproven. This study investigates whether RDW can serve as a universal indicator of physiological stress and illness acuity.

Methods

Researchers conducted a retrospective analysis of 210,930 adult ED visits at a large academic hospital system (UCLA) between 2013 and 2022. The study utilized two primary modeling approaches:

  • Logistic Regression (LR): To determine the statistical weight of RDW relative to other clinical variables.
  • XGBoost: A high-performance gradient-boosted decision tree algorithm used to capture non-linear relationships and maximize predictive accuracy.

The models incorporated demographics, 27 comorbidity categories, triage vital signs, and standard blood panels (BMP and CBC). Subgroup analyses were performed across three age brackets (<45, 45–65, and >65) and 374 distinct diagnosis categories to ensure the findings were generalizable.

Results

RDW demonstrated a clear and steady gradient associated with clinical severity. Mean RDW values were 13.8% for discharged patients, 15.1% for admitted patients, 15.3% for ICU admissions, and 16.8% for those who died in the hospital.

RDW distributions across clinical outcomes, showing clear gradients between discharge/admission, floor/ICU, and survival/mortality.

The predictive power of RDW was significant across the board. The XGBoost models achieved an Area Under the Curve (AUC) of 0.90 for predicting hospital admission and an impressive 0.96 for in-hospital mortality. Patients with an RDW greater than 15.5 faced a 4.87-fold higher risk of death compared to those with lower values.

RDW distribution across age subgroups for discharged and admitted patients, demonstrating consistency across demographics.

In machine learning models, RDW emerged as one of the most informative features, often carrying more weight than traditional vital signs like heart rate or blood pressure. Its predictive utility remained consistent across nearly all 374 diagnosis categories, from infections to trauma.

Discussion and Conclusion

RDW functions as a "universal stress signal," reflecting systemic physiological disruptions such as inflammation and oxidative stress. While it is highly specific—meaning an elevated RDW is a strong "red flag" for severe illness—it has lower sensitivity, as a normal RDW does not necessarily guarantee a patient is healthy.

Limitations of the study include its retrospective nature and reliance on data from a single geographic region. However, the findings suggest that RDW is a powerful, "hidden in plain sight" biomarker. Integrating RDW into real-time clinical decision support (CDS) tools within Electronic Health Records could help clinicians identify high-acuity patients who might otherwise be overlooked during busy shifts.

Summary updated

Abstract

RDW, a component of the complete blood count panel routinely ordered as the initial workup for the undifferentiated patient, may be a generalizable biomarker for acuity in the ED.
BibTeX
@inproceedings{hong2023association,
  author = {Woo Suk Hong and Ákos Rudas and Elijah Bell and Jeffrey N. Chiang},
  title = {{Association of red blood cell distribution width with hospital admission and in-hospital mortality across all-cause adult emergency department visits}},
  booktitle = {JAMIA},
  year = {2023},
  doi = {10.1093/jamiaopen/ooad053},
}
Figure