Nicolas Wadih Chbat

Development and validation of electronic surveillance tool for acute kidney injury: A retrospective analysis.

INTRODUCTION Timely detection of acute kidney injury (AKI) facilitates prevention of its progress and potentially therapeutic interventions. The study objective is to develop and validate an electronic surveillance tool (AKI sniffer) to detect AKI in 2 independent retrospective cohorts of intensive care unit (ICU) patients. The primary aim is to compare the sensitivity, specificity, and positive and negative predictive values of AKI sniffer performance against a reference standard. METHODS This study is conducted in the ICUs of a tertiary care center. The derivation cohort study subjects were Olmsted County, MN, residents admitted to all Mayo Clinic ICUs from July 1, 2010, through December 31, 2010, and the validation cohort study subjects were all patients admitted to a Mayo Clinic, Rochester, campus medical/surgical ICU on January 12, 2010, through March 23, 2010. All included records were reviewed by 2 independent investigators who adjudicated AKI using the Acute Kidney Injury Network criteria; disagreements were resolved by a third reviewer. This constituted the reference standard. An electronic algorithm was developed; its precision and reliability were assessed in comparison with the reference standard in 2 separate cohorts, derivation and validation. RESULTS Of 1466 screened patients, a total of 944 patients were included in the study: 482 for derivation and 462 for validation. Compared with the reference standard in the validation cohort, the sensitivity and specificity of the AKI sniffer were 88% and 96%, respectively. The Cohen κ (95% confidence interval) agreement between the electronic and the reference standard was 0.84 (0.78-0.89) and 0.85 (0.80-0.90) in the derivation and validation cohorts. CONCLUSION Acute kidney injury can reliably and accurately be detected electronically in ICU patients. The presented method is applicable for both clinical (decision support) and research (enrollment for clinical trials) settings. Prospective validation is required.

Predicting ICU hemodynamic instability using continuous multiparameter trends

Background: Identifying hemodynamically unstable patients in a timely fashion in intensive care units (ICUs) is crucial because it can lead to earlier interventions and thus to potentially better patient outcomes. Current alert algorithms are typically limited to detecting dangerous conditions only after they have occurred and suffer from high false alert rates. Our objective was to predict hemodynamic instability at least two hours before a major clinical intervention (e.g., vasopressor administration), while maintaining a low false alert rate. Study population: From the MIMIC II database, containing ICU minute-by-minute heart rate (HR) and invasive arterial blood pressure (BP) monitoring trend data collected between 2001 and 2005, we identified 132 stable and 104 unstable patients that met our stability-instability criteria and had sufficient data points. Method: We first derived additional physiological parameters of shock index, rate pressure product, heart rate variability, and two measures of trending based on HR and BP. Then we developed 220 statistical features and systematically selected a small set to use for classification. We applied multi-variable logistic regression modeling to do classification and implemented validation via bootstrapping. Results: Area under receiver-operating curve (ROC) 0.83±0.03, sensitivity 0.75±0.06, and specificity 0.80±0.07; if the specificity is targeted at 0.90, then the sensitivity is 0.57±0.07. Based on our preliminary results, we conclude that the algorithms we developed using HR and BP trend data may provide a promising perspective toward reliable predictive alerts for hemodynamically unstable patients.

Systems modeling and simulation applications for critical care medicine

Critical care delivery is a complex, expensive, error prone, medical specialty and remains the focal point of major improvement efforts in healthcare delivery. Various modeling and simulation techniques offer unique opportunities to better understand the interactions between clinical physiology and care delivery. The novel insights gained from the systems perspective can then be used to develop and test new treatment strategies and make critical care delivery more efficient and effective. However, modeling and simulation applications in critical care remain underutilized. This article provides an overview of major computer-based simulation techniques as applied to critical care medicine. We provide three application examples of different simulation techniques, including a) pathophysiological model of acute lung injury, b) process modeling of critical care delivery, and c) an agent-based model to study interaction between pathophysiology and healthcare delivery. Finally, we identify certain challenges to, and opportunities for, future research in the area.

Noninvasive Estimation of Respiratory Mechanics in Spontaneously Breathing Ventilated Patients: A Constrained Optimization Approach

This paper presents a method for breath-by-breath noninvasive estimation of respiratory resistance and elastance in mechanically ventilated patients. For passive patients, well-established approaches exist. However, when patients are breathing spontaneously, taking into account the diaphragmatic effort in the estimation process is still an open challenge. Mechanical ventilators require maneuvers to obtain reliable estimates for respiratory mechanics parameters. Such maneuvers interfere with the desired ventilation pattern to be delivered to the patient. Alternatively, invasive procedures are needed. The method presented in this paper is a noninvasive way requiring only measurements of airway pressure and flow that are routinely available for ventilated patients. It is based on a first-order single-compartment model of the respiratory system, from which a cost function is constructed as the sum of squared errors between model-based airway pressure predictions and actual measurements. Physiological considerations are translated into mathematical constraints that restrict the space of feasible solutions and make the resulting optimization problem strictly convex. Existing quadratic programming techniques are used to efficiently find the minimizing solution, which yields an estimate of the respiratory system resistance and elastance. The method is illustrated via numerical examples and experimental data from animal tests. Results show that taking into account the patient effort consistently improves the estimation of respiratory mechanics. The method is suitable for real-time patient monitoring, providing clinicians with noninvasive measurements that could be used for diagnosis and therapy optimization.

Constrained optimization for noninvasive estimation of work of breathing.

This paper presents a technique for noninvasive estimation of respiratory muscle effort (also known as work of breathing, WOB) in mechanically ventilated patients. Continual and real-time assessment of the patient WOB is desirable, as it helps the clinician make decisions about increasing or decreasing mechanical respiratory support. The technique presented is based on a physiological model of the respiratory system, from which a cost function is constructed as the sum of squared errors between model-based airway pressure predictions and actual measurements. Quadratic programming methods are used to minimize this cost function. An experimental example on animal data shows the effectiveness of the technique.

Clinical Knowledge-Based Inference Model for Early Detection of Acute Lung Injury

Acute lung injury (ALI) is a devastating complication of acute illness and one of the leading causes of multiple organ failure and mortality in the intensive care unit (ICU). The detection of this syndrome is limited due to the complexity of the disease, insufficient understanding of its development and progression, and the large amount of risk factors and modifiers. In this preliminary study, we present a novel mathematical model for ALI detection. It is constructed based on clinical and research knowledge using three complementary techniques: rule-based fuzzy inference systems, Bayesian networks, and finite state machines. The model is developed in Matlab®’s Simulink environment and takes as input pre-ICU and ICU data feeds of critically ill patients. Results of the simulation model were validated against actual patient data from an epidemiologic study. By appropriately combining all three techniques the performance attained is in the range of 71.7–92.6% sensitivity and 60.3–78.4% specificity.

A comprehensive cardiopulmonary simulation model for the analysis of hypercapnic respiratory failure

We developed a new comprehensive cardiopulmonary model that takes into account the mutual interactions between the cardiovascular and the respiratory systems along with their short-term regulatory mechanisms. The model includes the heart, systemic and pulmonary circulations, lung mechanics, gas exchange and transport equations, and cardio-ventilatory control. Results show good agreement with published patient data in case of normoxic and hyperoxic hypercapnia simulations. In particular, simulations predict a moderate increase in mean systemic arterial pressure and heart rate, with almost no change in cardiac output, paralleled by a relevant increase in minute ventilation, tidal volume and respiratory rate. The model can represent a valid tool for clinical practice and medical research, providing an alternative way to experience-based clinical decisions.

Real-time noninvasive estimation of intrapleural pressure in mechanically ventilated patients: A feasibility study

A method for real-time noninvasive estimation of intrapleural pressure in mechanically ventilated patients is proposed. The method employs a simple first-order lung mechanics model that is fitted in real-time to flow and pressure signals acquired non-invasively at the opening of the patient airways, in order to estimate lung resistance (RL), lung compliance (CL) and intrapleural pressure (Ppl) continuously in time. Estimation is achieved by minimizing the sum of squared residuals between measured and model predicted airway pressure using a modified Recursive Least Squares (RLS) approach. Particularly, two different RLS algorithms, namely the conventional RLS with Exponential Forgetting (EF-RLS) and the RLS with Vector-type Forgetting Factor (VFF-RLS), are considered in this study and their performances are first evaluated using simulated data. Simulations suggest that the conventional EF-RLS algorithm is not suitable for our purposes, whereas the VFF-RLS method provides satisfactory results. The potential of the VFF-RLS based method is then proved on experimental data collected from a mechanically ventilated pig. Results show that the method provides continuous estimated lung resistance and compliance in normal physiological ranges and pleural pressure in good agreement with invasive esophageal pressure measurements.

Bayesian tracking of a nonlinear model of the capnogram.

Capnography, the monitoring of expired carbon dioxide (CO2 ) has been employed clinically as a non-invasive measure for the adequacy of ventilation of the alveoli of the lung. In combination with air flow measurements, the capnogram can be used to estimate the partial pressure of CO2 in the alveolar sacs. In addition, physiologically relevant parameters, such as the extent of CO2 rebreathing, the airway dead space, and the metabolic CO 2 production can be predicted. To calculate these parameters, mathematical models have been previously formulated and applied to experimental data using off-line optimization procedures. Unfortunately, this does not permit online identification of the capnogram to detect changes in the physiological model parameters. In the present study, a Bayesian method for breath-by-breath identification of the volumetric capnogram is presented. The method integrates a model of CO2 exchange in the lungs, which is nonlinear due to the nature of human tidal breathing, with a particle filtering algorithm for estimation of the model parameters and changes therein. In addition, this allowed for a dynamic prediction of the unmeasured alveolar CO2 tension. The method is demonstrated using simulations of the capnogram. The proposed method could aid the clinician in the interpretation of the capnogram

Transient respiratory response to hypercapnia: Analysis via a cardiopulmonary simulation model

In recent years, our group has developed a comprehensive cardiopulmonary (CP) model that comprises the heart, systemic and pulmonary circulations, lung mechanics and gas exchange, tissue metabolism, and cardiovascular and respiratory control mechanisms. In this paper, we analyze the response of the model to hypercapnic conditions and hence focus on the chemoreflex control mechanism. Particularly, we have enhanced the peripheral chemoreceptor model in order to better reflect respiratory control physiology. Using the CO2 fraction in the inspired air as input to the CP model, we were able to analyze the transient response of the system to CO2 step input at different levels, in terms of alveolar gas partial pressures, tidal volume, minute ventilation and respiratory frequency. Model predictions were tested against experimental data from human subjects [1]. Results show good agreement for all the variables under study during the transient phases and low root mean square errors at steady state. This indicates the potential for the model to be used as a valid tool for clinical practice and medical research, providing a complementary way to experience-based clinical decisions.