Abstract
AbstractEchocardiographic assessment of left ventricular function is crucial in NICU. The study aimed to compare the accuracy and agreement of global longitudinal strain (GLS) with conventional measurements. Real-life echocardiograms of neonates receiving intensive care were retrospectively reviewed. Shortening fraction (SF), ejection fraction (EF) and S’ measurements were retrieved from health records. GLS was calculated offline from stored images. The association with stroke volume indexed for body weight (iSV) was evaluated by regression analysis. The diagnostic ability to identify uncompensated shock was assessed by ROC curve analysis. Cohen's κ was run to assess agreement. 334 echocardiograms of 155 neonates were evaluated. Mean ± SD gestational age and birth weight were 34.5 ± 4.1 weeks and 2264 ± 914 g, respectively. SF, EF, S’ and GLS were associated with iSV with R2 of 0.133, 0.332, 0.252 and 0.633, (all p < .001). Including all variables in a regression model, iSV prediction showed an adjusted R2 of 0.667, (p < .001). GLS explained 73% of the model variance. GLS showed a better ability to diagnose uncompensated shock (AUC 0.956) compared to EF, S’ and SF (AUC 0.757, 0.737 and 0.606, respectively). GLS showed a moderate agreement with EF (κ = .500, p < .001) and a limited agreement with S’ and SF (κ = .260, p < .001, κ = .242, p < .001). GLS was a more informative index of left ventricular performance, providing the rationale for a more extensive use of GLS at the cotside.
Publisher
Springer Science and Business Media LLC