Prediction of hospital length of stay in decompensated cirrhosis using hematologic and inflammatory indices.

OBJECTIVE

To identify predictors of prolonged hospital length of stay (LOS) in decompensated cirrhosis (DC) patients and to develop a prediction model based on hematologic and inflammatory indices.

METHODS

We retrospectively studied 338 hospitalized patients with DC. Admission blood indices reflecting hematologic status (hemoglobin, hematocrit, red blood cell (RBC) count, platelets, mean corpuscular volume, red cell distribution width [RDW]) and inflammation (monocyte-to-lymphocyte ratio [MLR], neutrophil-monocyte-lymphocyte ratio [NMLR], systemic inflammation response index [SIRI]) were compared between patients with LOS ≥ 7 days vs < 7 days. Multivariable logistic regression and machine learning were used to identify independent predictors of LOS ≥ 7 days, with model performance assessed by area under the receiver operating characteristic curve (AUC) on a held-out test set.

RESULTS

Of 338 patients (median age 58 years, 33% female), 52% had LOS ≥ 7 days. Prolonged LOS was associated with more severe cytopenias (lower hemoglobin, hematocrit, RBC count, platelets) and heightened inflammation (higher RDW and inflammatory ratios; all P < 0.05). Nonlinear thresholds were observed. An admission RBC count < 5.2 × 1012/L and hemoglobin > 117 g/L were linked to reduced risk of prolonged stay. A predictive model using these 11 markers achieved moderate discrimination (test AUC 0.72; sensitivity 72%, specificity 58%).

CONCLUSIONS

Admission hematologic and inflammatory markers are independent predictors of prolonged LOS in DC. A machine learning model derived from routinely available low-cost tests provided moderate risk stratification for identifying patients at increased risk and supporting early clinical management.