Pulse oximetry signals: chaotic analysis in the evaluation of neonatal illness severity

Background: Assessment of neonatal illness severity is important for quality of care control, unbiased NICU performance comparisons, management studies and rationale use of resources. Emerging evidence indicates that the analysis of heart rate dynamics may provide valuable practical diagnostic and prognostic information. Here, we tested the hypothesis that a reduced dynamics of pulse rate (PR), and perfusion index (PI) is associated with higher illness severity in very low birth weight (VLBW) infants with histologically documented chorioamnionitis (HCA), a leading cause of neonatal morbidity and mortality. Methods: Two VLBW population with and without HCA were examined [HCA+: gestational age: 27.1 ± 2.3 wk, birth weight 932±300 g, SNAP-II: 39 (29–45), n=15; HCA-: gestational age: 27.4 ± 0.9 wk, birth weight 935±200 g, SNAP-II: 13 (5–17), n=15]. Relative Lempel-Ziv complexity (LZ), Lyapunov largest exponent (ëi), correlation dimension (CD) and Hurst exponent (H) were determined for 1000 to 10,000 data set points (Chaos Data Analyzer Pro, The Academic Software Library, North Carolina State University, Raleigh, NC, USA) for PR and PI recorded at 4-s intervals (Masimo RS, Masimo Co., Irvine, CA, USA) for each infant. The predictive accuracy of chaotic analysis was calculated using a receiver operating characteristics (ROC) curve analysis. Results: HCA+ infants exhibited significantly decreased LZ (PI-LZ: 0.26±0.13 vs. 0.446±0.07, p<0.0001; PR-LZ: 0.18±0.09 vs. 0.41±0.22, p=0.0008), ëi (PR-ëi: -0.14± 0.24 vs. 0.44 ±0.09, p<0.0001; PI-ëi: 0.21±0.36 vs. 0.42±0.10, p=0.038) and CD (PI-CD: 1.6±0.9 vs. 2.8±0.7, p=0.0003; PR-CD: 1.0±1.2 vs. 2.7±1.5, p=0.0019), as well as increased H (PR-H: 0.36±0.06 vs. 0.21±0.06, p