Travis J. Moss

Heart rate dynamics distinguish among atrial fibrillation, normal sinus rhythm and sinus rhythm with frequent ectopy

Atrial fibrillation (AF) is usually detected by inspection of the electrocardiogram waveform, a task made difficult when the signal is distorted by noise. The RR interval time series is more frequently available and accurate, yet linear and nonlinear time series analyses that detect highly varying and irregular AF are vulnerable to the common finding of frequent ectopy. We hypothesized that different nonlinear measures might capture characteristic features of AF, normal sinus rhythm (NSR), and sinus rhythm (SR) with frequent ectopy in ways that linear measures might not. To test this, we studied 2722 patients with 24 h ECG recordings in the University of Virginia Holter database. We found dynamical phenotypes for the three rhythm classifications. As expected, AF records had the highest variability and entropy, and NSR the lowest. SR with ectopy could be distinguished from AF, which had higher entropy, and from NSR, which had different fractal scaling, measured as higher detrended fluctuation analysis slope. With these dynamical phenotypes, we developed successful classification strategies, and the nonlinear measures improved on the use of mean and variability alone, even after adjusting for age. Final models using all variables had excellent performance, with positive predictive values for AF, NSR and SR with ectopy as high as 97, 98 and 90%, respectively. Since these classifiers can reliably detect rhythm changes utilizing segments as short as 10 min, we envision their application in noisy settings and in personal monitoring devices where only RR interval time series may be available.

New-onset Atrial Fibrillation in the Critically Ill*

Objective: To determine the association of new-onset atrial fibrillation with outcomes, including ICU length of stay and survival. Design: Retrospective cohort of ICU admissions. We found atrial fibrillation using automated detection (≥ 90 s in 30 min) and classed as new-onset if there was no prior diagnosis of atrial fibrillation. We identified determinants of new-onset atrial fibrillation and, using propensity matching, characterized its impact on outcomes. Setting: Tertiary care academic center. Patients: A total of 8,356 consecutive adult admissions to either the medical or surgical/trauma/burn ICU with available continuous electrocardiogram data. Interventions: None. Measurements and Main Results: From 74 patient-years of every 15-minute observations, we detected atrial fibrillation in 1,610 admissions (19%), with median burden less than 2%. Most atrial fibrillation was paroxysmal; less than 2% of admissions were always in atrial fibrillation. New-onset atrial fibrillation was subclinical or went undocumented in 626, or 8% of all ICU admissions. Advanced age, acute respiratory failure, and sepsis were the strongest predictors of new-onset atrial fibrillation. In propensity-adjusted regression analyses, clinical new-onset atrial fibrillation was associated with increased hospital mortality (odds ratio, 1.63; 95% CI, 1.01–2.63) and longer length of stay (2.25 d; CI, 0.58–3.92). New-onset atrial fibrillation was not associated with survival after hospital discharge (hazard ratio, 0.99; 95% CI, 0.76–1.28 and hazard ratio, 1.11; 95% CI, 0.67–1.83, respectively, for subclinical and clinical new-onset atrial fibrillation). Conclusions: Automated analysis of continuous electrocardiogram heart rate dynamics detects new-onset atrial fibrillation in many ICU patients. Though often transient and frequently unrecognized, new-onset atrial fibrillation is associated with poor hospital outcomes.

Signatures of Subacute Potentially Catastrophic Illness in the ICU: Model Development and Validation.

Objectives: Patients in ICUs are susceptible to subacute potentially catastrophic illnesses such as respiratory failure, sepsis, and hemorrhage that present as severe derangements of vital signs. More subtle physiologic signatures may be present before clinical deterioration, when treatment might be more effective. We performed multivariate statistical analyses of bedside physiologic monitoring data to identify such early subclinical signatures of incipient life-threatening illness. Design: We report a study of model development and validation of a retrospective observational cohort using resampling (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis type 1b internal validation) and a study of model validation using separate data (type 2b internal/external validation). Setting: University of Virginia Health System (Charlottesville), a tertiary-care, academic medical center. Patients: Critically ill patients consecutively admitted between January 2009 and June 2015 to either the neonatal, surgical/trauma/burn, or medical ICUs with available physiologic monitoring data. Interventions: None. Measurements and Main Results: We analyzed 146 patient-years of vital sign and electrocardiography waveform time series from the bedside monitors of 9,232 ICU admissions. Calculations from 30-minute windows of the physiologic monitoring data were made every 15 minutes. Clinicians identified 1,206 episodes of respiratory failure leading to urgent unplanned intubation, sepsis, or hemorrhage leading to multi-unit transfusions from systematic individual chart reviews. Multivariate models to predict events up to 24 hours prior had internally validated C-statistics of 0.61–0.88. In adults, physiologic signatures of respiratory failure and hemorrhage were distinct from each other but externally consistent across ICUs. Sepsis, on the other hand, demonstrated less distinct and inconsistent signatures. Physiologic signatures of all neonatal illnesses were similar. Conclusions: Subacute potentially catastrophic illnesses in three diverse ICU populations have physiologic signatures that are detectable in the hours preceding clinical detection and intervention. Detection of such signatures can draw attention to patients at highest risk, potentially enabling earlier intervention and better outcomes.

Dynamic Analysis of Cardiac Rhythms for Discriminating Atrial Fibrillation From Lethal Ventricular Arrhythmias

Background—Implantable cardioverter-defibrillators (ICDs), the first line of therapy for preventing sudden cardiac death in high-risk patients, deliver appropriate shocks for termination of ventricular tachycardia (VT)/ventricular fibrillation. A common shortcoming of ICDs is imperfect rhythm discrimination, resulting in the delivery of inappropriate shocks for atrial fibrillation (AF). An underexplored area for rhythm discrimination is the difference in dynamic properties between AF and VT/ventricular fibrillation. We hypothesized that the higher entropy of rapid cardiac rhythms preceding ICD shocks distinguishes AF from VT/ventricular fibrillation. Methods and Results—In a multicenter, prospective, observational study of patients with primary prevention ICDs, 119 patients received shocks from ICDs with stored, retrievable intracardiac electrograms. Blinded adjudication revealed shocks were delivered for VT/ventricular fibrillation (62%), AF (23%), and supraventricular tachycardia (15%). Entropy estimation of only 9 ventricular intervals before ICD shocks accurately distinguished AF (receiver operating characteristic curve area, 0.98; 95% confidence intervals, 0.93–1.0) and outperformed contemporary ICD rhythm discrimination algorithms. Conclusions—This new strategy for AF discrimination based on entropy estimation expands on simpler concepts of variability, performs well at fast heart rates, and has potential for broad clinical application.

Heart rate dynamics preceding hemorrhage in the intensive care unit.

Occult hemorrhage in surgical/trauma intensive care unit (STICU) patients is common and may lead to circulatory collapse. Continuous electrocardiography (ECG) monitoring may allow for early identification and treatment, and could improve outcomes. We studied 4,259 consecutive admissions to the STICU at the University of Virginia Health System. We collected ECG waveform data captured by bedside monitors and calculated linear and non-linear measures of the RR interbeat intervals. We tested the hypothesis that a transfusion requirement of 3 or more PRBC transfusions in a 24 hour period is preceded by dynamical changes in these heart rate measures and performed logistic regression modeling. We identified 308 hemorrhage events. A multivariate model including heart rate, standard deviation of the RR intervals, detrended fluctuation analysis, and local dynamics density had a C-statistic of 0.62. Earlier detection of hemorrhage might improve outcomes by allowing earlier resuscitation in STICU patients.

Classification of cardiac rhythm using heart rate dynamical measures: validation in MIT–BIH databases ☆

BACKGROUND Identification of atrial fibrillation (AF) is a clinical imperative. Heartbeat interval time series are increasingly available from personal monitors, allowing new opportunity for AF diagnosis. GOAL Previously, we devised numerical algorithms for identification of normal sinus rhythm (NSR), AF, and SR with frequent ectopy using dynamical measures of heart rate. Here, we wished to validate them in the canonical MIT-BIH ECG databases. METHODS We tested algorithms on the NSR, AF and arrhythmia databases. RESULTS When the databases were combined, the positive predictive value of the new algorithms exceeded 95% for NSR and AF, and was 40% for SR with ectopy. Further, dynamical measures did not distinguish atrial from ventricular ectopy. Inspection of individual 24hour records showed good correlation of observed and predicted rhythms. CONCLUSION Heart rate dynamical measures are effective ingredients in numerical algorithms to classify cardiac rhythm from the heartbeat intervals time series alone.

External validation in an intermediate unit of a respiratory decompensation model trained in an intensive care unit

Background. Preventing urgent intubation and upgrade in level of care in patients with subclinical deterioration could be of great utility in hospitalized patients. Early detection should result in decreased mortality, duration of stay, and/or resource use. The goal of this study was to externally validate a previously developed, vital sign‐based, intensive care unit, respiratory instability model on a separate population, intermediate care patients. Methods. From May 2014 to May 2016, the model calculated relative risk of adverse events every 15 minutes (n = 373,271 observations) for 2,050 patients in a surgical intermediate care unit. Results. We identified 167 upgrades and 57 intubations. The performance of the model for predicting upgrades within 12 hours was highly significant with an area under the curve of 0.693 (95% confidence interval, 0.658–0.724). The model was well calibrated with relative risks in the highest and lowest deciles of 2.99 and 0.45, respectively (a 6.6‐fold increase). The model was effective at predicting intubation, with a demonstrated area under the curve within 12 hours of the event of 0.748 (95% confidence interval, 0.685–0.800). The highest and lowest deciles of observed relative risk were 3.91 and 0.39, respectively (a 10.1‐fold increase). Univariate analysis of vital signs showed that transfer upgrades were associated, in order of importance, with rising respiration rate, rising heart rate, and falling pulse‐oxygen saturation level. Conclusion. The respiratory instability model developed previously is valid in intermediate care patients to predict both urgent intubations and requirements for upgrade in level of care to an intensive care unit.

IMPACT OF CARDIOGENIC SHOCK WITH OR WITHOUT MECHANICAL CIRCULATORY SUPPORT ON SHORT-TERM AND LONG-TERM SURVIVAL IN HEART FAILURE HOSPITALIZATIONS

Background: The impact of cardiogenic shock (CS) on survival in patients hospitalized for acute decompensated heart failure (ADHF) in the present day coronary care unit (CCU) with current options for mechanical circulatory support (MCS) is of particular interest. Methods: We evaluated outcomes in

THE IMPACT OF INCIDENT ATRIAL FIBRILLATION IN THE INTENSIVE CARE UNIT

Atrial fibrillation (AF) is a common arrhythmia that affects the critically ill, but the incidence and impact of new-onset AF in the intensive care unit (ICU) are poorly characterized. We hypothesized that new-onset AF is associated with poorer outcomes. We analyzed 73 patient-years of continuous

Computers in White Coats: How to Devise Useful Clinical Decision Support Software.

Critical Care Medicine www.ccmjournal.org 1449 The rate of alarms in clinical care is, well, alarming. Some quality control would be gratefully accepted. In this issue of Critical Care Medicine, Chen et al (1) from the critical care redoubt in Pittsburgh report on a combined effort by clinicians and data scientists—results made possible only through fusion of substantial clinical elbow grease and large-scale computing—to address the fundamental challenge of telling true alarms from false. Four clinicians each spent 100 hours (!) inspecting the 973 alarm records analyzed. The result—a computerized decision as to whether the alarm was true or not—promises a clinically significant reduction in alarm frequency with no clinically significant increase in risk to patients. This is not the only article on this topic, but we feel this major work stands apart by pounding home the notion that the hardest part of computer programming to support clinical decisions has nothing to do with computers. We are all witness to the torrent of new computer algorithms and smartphone apps intended to provide clinical decision support in the care of patients, from multi-lead electrocardiograms to colposcopy. There is little question that they provide additional information to bedside practitioners, but there are so many algorithms and apps now—how do we know which will help? And, with so much promise in this field, many data scientists and app developers are turning toward healthcare— what criteria must they satisfy to develop a quality product? One vision is that decision support should synthesize new and nonobvious information that can point the clinician in unanticipated directions or serve as a tiebreaker when clinical scenarios are ambiguous. Increasingly, crossdisciplinary teams like Chen et al (1) at Pittsburgh and Carnegie-Mellon gather together with this kind of goal. Now what do they do? We—from left to right, a clinical cardiologist, a mathematician, and a new-generation hybrid model—have watched and worked in this field for as many as 15 years. Here is our suggested prescription.