Routine laboratory data for predicting 30-day emergency department revisits: the AXIS-2 risk score

Thirty-day emergency department (ED) revisit is a major quality indicator reflecting morbidity and healthcare burden. Laboratory data obtained during outpatient encounters may capture underlying biological stress axes. This study aimed to develop and externally validate an interpretable logistic regression model—the AXIS-2 (Anemia–Inflammation Two-Axis Composite Index)—for predicting 30-day ED revisits using routine laboratory parameters. This retrospective study included 92,386 adult outpatients who visited a tertiary academic hospital between January 2015 and August 2025. Laboratory variables were biologically grouped into two axes: Axis-1 (hematologic indicators of anemia and erythropoiesis: hemoglobin, lymphocyte count, mean corpuscular volume, neutrophil count, platelet count, red cell distribution width, and white blood cell count) and Axis-2 (inflammatory and catabolic markers: C-reactive protein, ferritin, albumin, neutrophil-to-lymphocyte ratio, and platelet-to-lymphocyte ratio). A multivariable logistic regression model was trained using stratified sampling and isotonic calibration. Model performance was assessed in both internal and external test datasets using discrimination (AUC-ROC, AUC-PR), classification (F1 score, MCC, sensitivity, specificity), calibration, and decision curve analysis. The 30-day revisit rate was 30%. Both hematologic (Axis-1) and inflammatory (Axis-2) axes were independent predictors of revisit risk (Axis-1 OR 2.06; Axis-2 OR 2.31; p < 0.001). The model showed excellent discrimination (AUC-ROC 0.921 in the test set; 0.935 in the external set) and balanced classification accuracy (≈85% for both sensitivity and specificity). Negative predictive value reached 95%, and calibration metrics indicated strong concordance between predicted and observed probabilities. Decision curve analysis demonstrated clear net benefit within the 10–40% probability range. High-risk patients also exhibited higher 30-day readmission (≈45%), transfusion (≈15%), and 180-day mortality (≈3%) rates compared with the low-risk group. AXIS-2 is a transparent, laboratory-based model that accurately predicts 30-day emergency revisits and offers interpretable outputs suitable for integration into clinical workflows. Its strong calibration and decision-analytic benefit support its use as a cost-efficient tool for post-discharge monitoring and healthcare resource optimization.