Written by Hannah Harp
Spoon Feed
A mobile app in development uses a smartphone camera to predict bilirubin level in neonates and was very sensitive and moderately specific compared to serum bilirubin.
There’s an app for that
Neonatal jaundice is a common source of parental distress, frequent clinic visits, and multiple lab draws. While transcutaneous bilirubin (Tcb) monitoring can be used, it is an expensive piece of equipment that requires frequent calibration and has a high rate of false positives. The study aimed to develop and validate a smartphone-based machine learning (ML) app for neonatal jaundice screening by predicting serum bilirubin (TSB) using skin yellowness indicators from the forehead, sternum, and abdomen. Conducted in Singapore on 546 neonates, it showed strong correlation (Pearson r = 0.84) and agreement with TSB, achieving 100% sensitivity and 70% specificity, compared to 51% sensitivity of Tcb. The area under the ROC curve was 0.89, indicating high diagnostic accuracy. The app offers potential for cost-effective, remote neonatal jaundice screening with further validation required. This study was performed with a population that did not include Fitzpatrick skin types V or VI (darker tones), and was less accurate in skin type I (fair skin).
How will this change my practice?
Once this app is further developed and calibrated for all skin tones, it will be a great day for under-resourced clinics who can’t afford TcB sensors (and the heels of newborns)!
Source
Development and Validation of a Smartphone Application for Neonatal Jaundice Screening. JAMA Netw Open. 2024 Dec 2;7(12):e2450260. doi: 10.1001/jamanetworkopen.2024.50260. PMID: 39661385
