Development and validation of a nomogram for predicting neonatal acute kidney injury in very low birth weight infants

Acute kidney injury (AKI) is an independent risk factor associated with mortality among neonates. In this study, we aimed to evaluate the incidence and predictive factors associated with AKI and develop and validate a nomogram to predict neonatal AKI in very low birth weight (VLBW) infants. In this retrospective cohort study, VLBW infants admitted to our neonatal intensive care unit between April 2014 and March 2020 were included. We analyzed the incidence of AKI, as defined by changes in serum creatinine and urine output, and its associated risk factors and outcomes. A multivariate logistic regression analysis with backward elimination was used to build a prediction nomogram for AKI development. Using receiver operating characteristic analyses, we determined cutoff values that provide appropriate sensitivity and specificity. Among the 234 infants included, 91 (38.9%) had AKI. In the multivariate logistic regression analysis, indomethacin use (adjusted odds ratio [OR]: 14.87; 95% confidence interval [CI]: 6.31–35.03; p < 0.001), gestational age (adjusted OR: 0.71; 95% CI: 0.57–0.87; p = 0.004), and small for gestational age (adjusted OR: 0.26; 95% CI: 0.09–0.87; p = 0.03) were significantly associated with the risk of AKI. The incidence of neonatal AKI with indomethacin had a sensitivity of 75.8% (95% CI: 66.1–83.5%) and a specificity of 89.5% (95% CI: 83.4–93.5%). A nomogram to predict neonatal AKI was developed using these three factors. The model exhibited good discrimination with an area under the curve of 0.92 (95% CI: 0.87–0.95). The cutoff values for adequate sensitivity and specificity were 0.18 with 92.3% sensitivity and 64.3% specificity, 0.26 with 85.7% sensitivity and 79.0% specificity, and 0.42 with 76.9% sensitivity and 90.9% specificity. The developed nomogram is a valid and convenient model that can be used to predict the risk of neonatal AKI in VLBW infants. With appropriate adjustment of cutoff values, our three-variable prediction model for neonatal AKI was able to achieve a sensitivity as high as 92.3%. Our model is more valuable as a screening test than a prediction model with indomethacin alone. This nomogram may facilitate early diagnosis and be useful for the management of kidney function in VLBW infants at risk of neonatal AKI.