Abstract
OBJECTIVE: To develop and evaluate the predictive value of a simplified lung ultrasound (LUS) method for forecasting respiratory support in term infants. METHODS: This observational, prospective, diagnostic accuracy study was conducted in a tertiary academic hospital between June and December 2023. A total of 361 neonates underwent LUS examination within 1 h of birth. The proportion of each LUS sign was utilized to predict their respiratory outcomes and compared with the LUS score model. After identifying the best predictive LUS sign, simplified models were created based on different scan regions. The optimal simplified model was selected by comparing its accuracy with both the full model and the LUS score model. RESULTS: After three days of follow-up, 91 infants required respiratory support, while 270 remained healthy. The proportion of confluent B-lines demonstrated high predictive accuracy for respiratory support, with an area under the curve (AUC) of 89.1% (95% confidence interval [CI]: 84.5-93.7%). The optimal simplified model involved scanning the R/L 1-4 region, yielding an AUC of 87.5% (95% CI: 82.6-92.3%). Both the full model and the optimal simplified model exhibited higher predictive accuracy compared to the LUS score model. The optimal cut-off value for the simplified model was determined to be 15.9%, with a sensitivity of 76.9% and specificity of 91.9%. CONCLUSIONS: The proportion of confluent B-lines in LUS can effectively predict the need for respiratory support in term infants shortly after birth and offers greater reliability than the LUS score model.