Agreement and diagnostic performance of a smartphone-based bilirubin assessment tool for screening neonatal...

Neonatal hyperbilirubinemia is common and may lead to severe neurological injury if not detected and treated promptly. Total serum bilirubin (TSB) remains the reference standard, while transcutaneous bilirubin (TcB) is commonly used for non-invasive screening. Smartphone-based bilirubin assessment may provide a supplementary approach for neonatal jaundice screening.

This study aimed to evaluate the agreement, correlation, and diagnostic performance of a self-developed mobile application for bilirubin measurement compared with TSB and TcB in screening neonatal hyperbilirubinemia.

This prospective diagnostic study was conducted in the neonatal department of Anhui Provincial Hospital, China, between March 2025 and January 2026. Eligible neonates undergoing clinically indicated TSB testing for jaundice before phototherapy were included. TSB, TcB, and mobile app-derived bilirubin measurements were obtained during the same assessment session. Using TSB as the reference standard, agreement between mobile app-derived bilirubin and TSB was assessed using Bland-Altman analysis, and correlation was evaluated using linear regression. Multivariable logistic regression models adjusted for gestational age, postnatal age, and birth weight were used to evaluate diagnostic performance, and receiver operating characteristic curves were generated from predicted probabilities.

Among 121 neonates, 46 were preterm and 75 were term infants. In the overall cohort, mobile app-derived bilirubin did not differ significantly from TSB, with a mean difference of -2.42 μmol/L and 95% limits of agreement from -30.57 to 25.73 μmol/L. Agreement was lower among preterm infants than among term infants. Mobile app-derived bilirubin was positively correlated with TSB (R ² = 0.776, P < 0.001). In the adjusted logistic regression model, mobile app-derived bilirubin was independently associated with neonatal hyperbilirubinemia (odds ratio = 1.047, 95% confidence interval: 1.019-1.081, P = 0.002). The mobile application achieved an area under the curve of 0.936, with sensitivity of 0.912 and specificity of 0.908, while TcB achieved an area under the curve of 0.937, with sensitivity of 0.882 and specificity of 0.931.

Mobile application-derived bilirubin showed good agreement with TSB and diagnostic performance comparable to TcB under standardized conditions. Owing to its non-invasive, convenient, and low-cost nature, this smartphone-assisted approach has potential as a supplementary tool for neonatal hyperbilirubinemia screening, pending further technical optimization and multicenter home-based validation.