"Sick or not sick?" A mixed methods study evaluating the rapid determination of illness severity in a pediatric emergency department

Objectives: Experienced physicians must rapidly identify ill pediatric patients. We evaluated the ability of an illness rating score (IRS) to predict admission to a pediatric hospital and explored the underlying clinical reasoning of the gestalt assessment of illness. Methods: We used mixed-methods to study pediatric emergency medicine physicians at an academic children's hospital emergency department (ED). Physicians rated patients' illness severity with the IRS, anchored by 0 (totally well) and 10 (critically ill), and shared their rationale with concurrent think-aloud responses. The association between IRS and need for hospitalization, respiratory support, parenteral antibiotics, and resuscitative intravenous (IV) fluids were analyzed with mixed effects linear regression. Area under the curve (AUC) receiver operator characteristic (ROC) curve and test characteristics at different cut-points were calculated for IRS as a predictor of admission. Think-aloud responses were qualitatively analyzed via inductive process. Results: A total of 141 IRS were analyzed (mean 3.56, SD 2.30, range 0-9). Mean IRS were significantly higher for patients requiring admission (4.32 vs. 3.13, p<0.001), respiratory support (6.15 vs. 3.98, p = 0.033), IV fluids (4.53 vs. 3.14, p < 0.001), and parenteral antibiotics (4.68 vs. 3.32, p = 0.009). AUC for IRS as a predictor of admission was 0.635 (95% CI: 0.534-0.737). Analysis of 95 think-aloud responses yielded eight categories that describe the underlying clinical reasoning. Conclusions: Rapid assessments as captured by the IRS differentiated pediatric patients who required admission and medical interventions. Think-aloud responses for the rationale for rapid assessments may form the basis for teaching the skill of identifying ill pediatric patients.