Nonlinear analysis of mechanical ventilator weaning data




Mitchell, De Ann Fisher

Journal Title

Journal ISSN

Volume Title



Expeditious mechanical ventilator weaning is a major post-operative goal, yet no one test to determine readiness to wean has proven clearly superior in all patients. Many tests use respiratory rate in assessing readiness to wean. The problem was the unreliability of mechanical ventilator weaning predictors. Prior weaning studies have used linear analysis methods exclusively. The most precise measure of respiratory frequency was sought for future studies of respiratory variability and ventilator weaning. This study examined the relationship between recurrence analysis measures of respiratory rate and interbreath intervals to determine if interbreath intervals were a more precise, information-rich variable than the smoothed respiratory rate displayed on the ventilator. Nonlinear dynamics was the theoretical framework, suited for study of the complex physiological variable, respiration. Nonlinear analysis is essential to understand systems that change irregularly over time, have output unequal to input due to feedback, and are not suited to linear methods. Recurrence analysis revealed dynamical behaviors not detectable by standard linear methods. The sample was 12 adult patients in an urban hospital's intensive care, on ventilation more than 72 hours, and undergoing a weaning trial. Two sets of respiratory data were collected from the ventilator of each patient. Recurrence quantification analysis of each of 24 time series yielded five recurrence measures: percent recurrence, percent determinism, entropy, maxline, and trend. This descriptive, exploratory, and methodological study examined the relationships between recurrence analysis measures of respiratory rate data and recurrence analysis measures of interbreath interval data. The Spearman's rho comparison test revealed only nonsignificant, weak associations. The Bland-Altman test of agreement between methods revealed plots with 22 recurrence measures as outliers beyond the wide limits of agreement. The study demonstrated that recurrence measures of respiratory rate displayed on the mechanical ventilator's microprocessor and recurrence measures of inter breath intervals collected from the ventilator's analog output port are not significantly related, nor were they in complete agreement. The two variables were different measures of respiratory functioning. Interbreath intervals were a precise, information-rich variable for studying ventilator weaning, respiratory variability, and the possible addition of variability to mechanical ventilators.



Mechanical ventilator, Recurrence analysis, Respiration