A Framework for Patient State Tracking by Classifying Multiscalar Physiologic Waveform Features

Dynamics of ventilatory pattern variability and Cardioventilatory Coupling during systemic inflammation in rats

Introduction: Biometrics of common physiologic signals can reflect health status. We have developed analytics to measure the predictability of ventilatory pattern variability (VPV, Nonlinear Complexity Index (NLCI) that quantifies the predictability of a continuous waveform associated with inhalation and exhalation) and the cardioventilatory coupling (CVC, the tendency of the last heartbeat in expiration to occur at preferred latency before the next inspiration). We hypothesized that measures of VPV and CVC are sensitive to the development of endotoxemia, which evoke neuroinflammation. Methods: We implanted Sprague Dawley male rats with BP transducers to monitor arterial blood pressure (BP) and recorded ventilatory waveforms and BP simultaneously using whole-body plethysmography in conjunction with BP transducer receivers. After baseline (BSLN) recordings, we injected lipopolysaccharide (LPS, n = 8) or phosphate buffered saline (PBS, n =3) intraperitoneally on 3 consecutive days. We recorded for 4-6 h after the injection, chose 3 epochs from each hour and analyzed VPV and CVC as well as heart rate variability (HRV). Results: First, the responses to sepsis varied across rats, but within rats the repeated measures of NLCI, CVC, as well as respiratory frequency (fR), HR, BP and HRV had a low coefficient of variation, (<0.2) at each time point. Second, HR, fR, and NLCI increased from BSLN on Days 1-3; whereas CVC decreased on Days 2 and 3. In contrast, changes in BP and the relative low-(LF) and high-frequency (HF) of HRV were not significant. The coefficient of variation decreased from BSLN to Day 3, except for CVC. Interestingly, NLCI increased before fR in LPS-treated rats. Finally, we histologically confirmed lung injury, systemic inflammation via ELISA and the presence of the proinflammatory cytokine, IL-1β, with immunohistochemistry in the ponto-medullary respiratory nuclei. Discussion: Our findings support that NLCI reflects changes in the rat's health induced by systemic injection of LPS and reflected in increases in HR and fR. CVC decreased over the course to the experiment. We conclude that NLCI reflected the increase in predictability of the ventilatory waveform and (together with our previous work) may reflect action of inflammatory cytokines on the network generating respiration.

Prediction of Postoperative Deterioration in Cardiac Surgery Patients Using Electronic Health Record and Physiologic Waveform Data

BACKGROUND

Postoperative hemodynamic deterioration among cardiac surgical patients can indicate or lead to adverse outcomes. Whereas prediction models for such events using electronic health records or physiologic waveform data are previously described, their combined value remains incompletely defined. The authors hypothesized that models incorporating electronic health record and processed waveform signal data (electrocardiogram lead II, pulse plethysmography, arterial catheter tracing) would yield improved performance versus either modality alone.

METHODS

Intensive care unit data were reviewed after elective adult cardiac surgical procedures at an academic center between 2013 and 2020. Model features included electronic health record features and physiologic waveforms. Tensor decomposition was used for waveform feature reduction. Machine learning-based prediction models included a 2013 to 2017 training set and a 2017 to 2020 temporal holdout test set. The primary outcome was a postoperative deterioration event, defined as a composite of low cardiac index of less than 2.0 ml min-1 m-2, mean arterial pressure of less than 55 mmHg sustained for 120 min or longer, new or escalated inotrope/vasopressor infusion, epinephrine bolus of 1 mg or more, or intensive care unit mortality. Prediction models analyzed data 8 h before events.

RESULTS

Among 1,555 cases, 185 (12%) experienced 276 deterioration events, most commonly including low cardiac index (7.0% of patients), new inotrope (1.9%), and sustained hypotension (1.4%). The best performing model on the 2013 to 2017 training set yielded a C-statistic of 0.803 (95% CI, 0.799 to 0.807), although performance was substantially lower in the 2017 to 2020 test set (0.709, 0.705 to 0.712). Test set performance of the combined model was greater than corresponding models limited to solely electronic health record features (0.641; 95% CI, 0.637 to 0.646) or waveform features (0.697; 95% CI, 0.693 to 0.701).

CONCLUSIONS

Clinical deterioration prediction models combining electronic health record data and waveform data were superior to either modality alone, and performance of combined models was primarily driven by waveform data. Decreased performance of prediction models during temporal validation may be explained by data set shift, a core challenge of healthcare prediction modeling.

EDITOR’S PERSPECTIVE

The potential of artificial intelligence to improve patient safety: a scoping review

Artificial intelligence (AI) represents a valuable tool that could be used to improve the safety of care. Major adverse events in healthcare include: healthcare-associated infections, adverse drug events, venous thromboembolism, surgical complications, pressure ulcers, falls, decompensation, and diagnostic errors. The objective of this scoping review was to summarize the relevant literature and evaluate the potential of AI to improve patient safety in these eight harm domains. A structured search was used to query MEDLINE for relevant articles. The scoping review identified studies that described the application of AI for prediction, prevention, or early detection of adverse events in each of the harm domains. The AI literature was narratively synthesized for each domain, and findings were considered in the context of incidence, cost, and preventability to make projections about the likelihood of AI improving safety. Three-hundred and ninety-two studies were included in the scoping review. The literature provided numerous examples of how AI has been applied within each of the eight harm domains using various techniques. The most common novel data were collected using different types of sensing technologies: vital sign monitoring, wearables, pressure sensors, and computer vision. There are significant opportunities to leverage AI and novel data sources to reduce the frequency of harm across all domains. We expect AI to have the greatest impact in areas where current strategies are not effective, and integration and complex analysis of novel, unstructured data are necessary to make accurate predictions; this applies specifically to adverse drug events, decompensation, and diagnostic errors.

Multimodal tensor-based method for integrative and continuous patient monitoring during postoperative cardiac care

Patients recovering from cardiovascular surgeries may develop life-threatening complications such as hemodynamic decompensation, making the monitoring of patients for such complications an essential component of postoperative care. However, this need has given rise to an inexorable increase in the number and modalities of data points collected, making it challenging to effectively analyze in real time. While many algorithms exist to assist in monitoring these patients, they often lack accuracy and specificity, leading to alarm fatigue among healthcare practitioners. In this study we propose a multimodal approach that incorporates salient physiological signals and EHR data to predict the onset of hemodynamic decompensation. A retrospective dataset of patients recovering from cardiac surgery was created and used to train predictive models. Advanced signal processing techniques were employed to extract complex features from physiological waveforms, while a novel tensor-based dimensionality reduction method was used to reduce the size of the feature space. These methods were evaluated for predicting the onset of decompensation at varying time intervals, ranging from a half-hour to 12 h prior to a decompensation event. The best performing models achieved AUCs of 0.87 and 0.80 for the half-hour and 12-h intervals respectively. These analyses evince that a multimodal approach can be used to develop clinical decision support systems that predict adverse events several hours in advance.

INSMA: An integrated system for multimodal data acquisition and analysis in the intensive care unit

Modern intensive care units (ICU) are equipped with a variety of different medical devices to monitor the physiological status of patients. These devices can generate large amounts of multimodal data daily that include physiological waveform signals (arterial blood pressure, electrocardiogram, respiration), patient alarm messages, numeric vitals data, etc. In order to provide opportunities for increasingly improved patient care, it is necessary to develop an effective data acquisition and analysis system that can assist clinicians and provide decision support at the patient bedside. Previous research has discussed various data collection methods, but a comprehensive solution for bedside data acquisition to analysis has not been achieved. In this paper, we proposed a multimodal data acquisition and analysis system called INSMA, with the ability to acquire, store, process, and visualize multiple types of data from the Philips IntelliVue patient monitor. We also discuss how the acquired data can be used for patient state tracking. INSMA is being tested in the ICU at University Hospitals Cleveland Medical Center.

Periodicity: A Characteristic of Heart Rate Variability Modified by the Type of Mechanical Ventilation After Acute Lung Injury

We present a novel approach to quantify heart rate variability (HRV) and the results of applying this approach to synthetic and original data sets. Our approach evaluates the periodicity of heart rate by calculating the transform of Relative Shannon Entropy, the maximum value of the RR interval periodogram, and the maximum, mean values, and sample entropy of the autocorrelation function. Synthetic data were generated using a Van der Pol oscillator; and the original data were electrocardiogram (ECG) recordings from anesthetized rats after acute lung injury while on biologically variable (BVV) or continuous mechanical ventilation (CMV). Analysis of the synthetic data revealed that our measures were correlated highly to the bandwidth of the oscillator and assessed periodicity. Then, applying these analytical tools to the ECGs determined that the heart rate (HR) of BVV group had less periodicity and higher variability than the HR of the CMV group. Quantifying periodicity effectively identified a readily apparent difference in HRV during BVV and CMV that was not identified by power spectral density measures during BVV and CMV. Cardiorespiratory coupling is the probable mechanism for HRV increasing during BVV and becoming periodic during CMV. Thus, the absence or presence of periodicity in ventilation determined HRV, and this mechanism is distinctly different from the cardiorespiratory uncoupling that accounts for the loss of HRV during sepsis.