Abstract Primary Subject area Neonatal-Perinatal Medicine Background Universal head ultrasound (HUS) screening for severe neurological injury (SNI) injury is recommended for infants born < 32 weeks’ gestational age (GA). However, the risk of SNI varies inversely with GA at birth and other known risk factors; therefore targeted screening may be more appropriate. Objectives The objective of the study is to develop a risk-stratified HUS screening protocol for infants born < 32 weeks’ to identify SNI accurately while minimizing resource use. Design/Methods Retrospective cohort study of infants born 23-31 weeks’ admitted to a tertiary NICU between 2011-2017. Patient characteristics were extracted from the Canadian Neonatal Network database. All HUS were individually reviewed by a trained abstractor and grouped based on date of exam relative to birth: ≤ 3 days, 4-7 days, 8-14 days, 28-42 days and 35-42 weeks’ corrected GA. Severe neurological injury was defined as intraventricular hemorrhage grade ≥ 3 and/or periventricular leukomalacia on HUS. Logistic regression models were used to identify perinatal risk factors for SNI and determine the number and timing of HUS with highest diagnostic accuracy. Results Of 651 infants included, 72 (11%) developed SNI. Independent risk factors for SNI were GA <29 weeks (AOR 3.09, 95% CI 1.65-6.08), vasopressors (AOR 2.95, 95% CI 1.24-6.80) and mechanical ventilation on day of admission (AOR 2.22, 95% CI 1.23-4.11). Infants were grouped into three screening groups based on their exposure to these risk factors (Table 1). Diagnostic accuracy of 63 models of combinations of HUS time points were assessed, and a screening protocol was developed based on the specific time points of HUS that maximized diagnostic accuracy (area under the ROC curve >0.9) while minimizing number of HUS for each screening group (Table 2). Using this protocol could reduce the total number of HUS performed by 920 (40%) and median number of HUS per infant from three (IQR 2-4) to 2 (IQR 1-2; p < 0 .001). Conclusion Implementation of a risk factor-based HUS screening protocol may reduce resource use while maintaining high diagnostic accuracy for SNI, and reflects choosing wisely in the NICU.
To develop a head ultrasound (HUS) screening protocol for infants born <32 weeks gestational age (GA) that accurately identifies severe brain injury (SBI) while minimizing resource use.Retrospective cohort study of infants born <32 weeks GA, admitted to a level 3 neonatal intensive care unit between 2011 and 2017. Timing and results of each HUS were reviewed. SBI was defined as intraventricular hemorrhage grade ≥3 and/or periventricular leukomalacia. Logistic regression models were used to identify risk factors and evaluate the predictive value of HUS at different time points during hospitalization.Of 651 included infants, 71 (11%) developed SBI. Risk factors for SBI were GA at birth <29 weeks (adjusted odds ratio (aOR) 2.87, 95% confidence interval (CI) 1.50-5.48), vasopressors on admission (aOR 3.08, 95%CI 1.38-6.88) and mechanical ventilation on admission (aOR 2.50, 95%CI 1.33-4.68). Infants were classified into three risk groups based on these risk factors, and combinations of 1-5 HUS time points were evaluated to determine the optimal number and timing of HUS for each group. The optimal number of screening HUS ranged from 1 for low-risk to 2 for high-risk infants. Adopting a screening protocol using the number and timing of HUS optimized by risk group could reduce the total number of HUS performed by 40% and the median number of HUS per infant from 3 (IQR 2-4) to 2 (IQR 1-3) (p < .01).Implementation of a risk factor-based HUS screening protocol can help reduce resource use while maintaining high sensitivity for detecting SBI.
'%3e%3cg id='b'%3e%3cpath id='a' d='M0-8v8h4z' transform='rotate(18 0 -8)'/%3e%3cuse xlink:href='%23a' transform='scale(-1 1)'/%3e%3c/g%3e%3cuse xlink:href='%23b' transform='rotate(72)'/%3e%3cuse xlink:href='%23b' transform='rotate(144)'/%3e%3cuse xlink:href='%23b' transform='rotate(216)'/%3e%3cuse xlink:href='%23b' transform='rotate(288)'/%3e%3c/g%3e%3c/svg%3e)
