463 - Validation of the STARZ Scoring System in Predicting Acute Kidney Injury in Very Low Birth Weight Neonates in the United States

Publication Number: 2451.463

RIYA KALRA, METROHEALTH SYSTEM, PARMA HEIGHTS, OH, United States; Gwen Weagraff, University of Florida, Mentor, OH, United States; Akshat Ritesh Shah, The MetroHealth System, Cleveland, OH, United States; Anna Rybinska-Campbell, MetroHealth, Cleveland, OH, United States; Deepak Kumar, Case Western Reserve University School of Medicine, Cleveland, OH, United States; Prabhu S. parimi, Case Western Reserve University School of Medicine, Cleveland, OH, United States

Background: Acute kidney injury (AKI) in very low birth weight (VLBW) neonates contributes substantially to morbidity, mortality, and prolonged hospitalization. Early identification of infants at risk is critical to enable timely preventive strategies. The STARZ score, developed and validated in India, predicts neonatal AKI within the first week of life, but its performance in U.S. populations remains unknown.

Objective: This study aimed to evaluate the predictive reliability of the STARZ score in a U.S. cohort of VLBW infants.

Design/Methods: A single-center, retrospective cohort study was conducted on 272 inborn VLBW (≤1500 g) preterm (≤32 weeks) infants admitted to the MetroHealth Medical Center NICU (2019-2023). Infants with congenital anomalies or surgical interventions in the first 7 days were excluded. STARZ scores were calculated from clinical and biochemical parameters, and AKI was defined using modified KDIGO criteria. Predictive accuracy was assessed by sensitivity, specificity, predictive values, and area under the ROC curve (AUC).

Results: AKI occurred in 26 infants (9.6%), primarily Stage 1 (81%). The mean gestational age and birth weight were 28.3 ± 2.7 weeks and 1040 ± 310 g. A STARZ score ≥ 31.5 was observed in 47% of the cohort. The model demonstrated good discrimination with an AUC = 0.806 (95% CI 0.722-0.889), sensitivity 84.6%, specificity 55.7%, positive predictive value 16.8%, and negative predictive value 97.2% (Table 1, Figure 1). Proportion of infants who developed AKI increased directly with increasing STARZ score and the majority (95%) of infants with AKI had STARZ score ≥ 31.5 (Figure 2). Infants with higher STARZ scores were significantly more likely to require positive-pressure ventilation, have gestational age < 28 weeks, sepsis, cardiac disease, elevated creatinine, or require furosemide/inotropes (p < 0.001). AKI was associated with longer NICU stay (104 ± 82 vs. 74 ± 46 days; p < 0.004) and higher mortality (19.2% vs. 2.4%; p < 0.005).

Conclusion(s): The STARZ score demonstrated strong predictive ability for AKI in VLBW infants in a U.S. NICU, showing excellent sensitivity and negative predictive value. However, its modest specificity and low positive predictive value suggest that refinement, potentially including additional perinatal factors or a higher cut-off threshold, may improve accuracy for this population. This study provides the first U.S.-based validation of the STARZ model and highlights its potential as a practical early risk stratification tool for neonatal AKI.

Predictive capability amongst study VLBW cohort (Comparative predictive data from the original STARZ study added)
ROC: Receiver Operating Characteristic; VLBW: Very Low Birth Weight

Area under the ROC curve for AKI among VLBW neonates
Receiver operating characteristic (ROC) curve for the scoring system. Area under the ROC curve = 0.806 (0.722-0.889); p value < 0.001

Proportion of VLBW neonates with AKI across different STARZ score
The figure shows the relationship between the proportion of VLBW infants with AKI and the STARZ score. Among infants with AKI (n= 26, 9.6%), 95% had STARZ score ≥ 31.5 (red dotted line)