Autonomic information flow improves prognostic value of heart rate patterns after abdominal aortic surgery

Abstract

Purpose Complex autonomic communication can be assessed by autonomic information flow (AIF) functions. The objective was to evaluate new complexity measures involving physiologically relevant time scales to predict the length of stay (LOS) in the hospital after abdominal aortic surgery (AAS). Our hypothesis was that AAS disrupts AIF, that restoration of AIF is necessary for recovery from major surgery, and that measures of AIF are superior to conventional heart rate variability (HRV) measures and equivalent to APACHE IV score in predicting LOS. Materials and Methods Twenty-four–hour Holter recordings were analyzed in 94 patients after AAS for standard time, frequency domain, and several complexity measures of different time scales derived from AIF functions. The risk of staying in the hospital for longer than 7 days as a function of HRV measures and APACHE IV score was modeled by logistic regression. The area under the curve (AUC) of receiver operating characteristic with 95% confidence interval as measure of predictive accuracy was calculated and internally cross-validated. Results The long-term dacay of AIF over 100s (LD 100) with cross-validated AUC = 0.67 (0.56-0.79) nearly reached the predictive accuracy of the APACHE IV score with AUC = 0.69 (0.58-0.79). None of the traditional time and frequency domain HRV measures remained in the multivariate models. The LD 100 adjusted for ventilatory support with AUC = 0.70 (0.59-0.81) was equivalent to the APACHE IV score in this patient group. Although the strongest correlation between AIF measures and the APACHE IV score was found for LD 100, r was only −0.37. Conclusions Results confirm the hypothesis that AIF measures characterize pathophysiologic autonomic communication better than traditional HRV measures and may have similar predictive value to the APACHE IV score for LOS. The relative independence of the APACHE IV score and AIF measures suggests that AIF measures could add to clinical risk prediction.