Identification of complex metabolic states in critically injured patients using bioinformatic cluster analysis

Identifying TBI Physiological States by Clustering Multivariate Clinical Time-Series Data

Determining clinically relevant physiological states from multivariate time-series data with missing values is essential for providing appropriate treatment for acute conditions such as Traumatic Brain Injury (TBI), respiratory failure, and heart failure. Utilizing non-temporal clustering or data imputation and aggregation techniques may lead to loss of valuable information and biased analyses. In our study, we apply the SLAC-Time algorithm, an innovative self-supervision-based approach that maintains data integrity by avoiding imputation or aggregation, offering a more useful representation of acute patient states. By using SLAC-Time to cluster data in a large research dataset, we identified three distinct TBI physiological states and their specific feature profiles. We employed various clustering evaluation metrics and incorporated input from a clinical domain expert to validate and interpret the identified physiological states. Further, we discovered how specific clinical events and interventions can influence patient states and state transitions.

Radiation-induced multi-organ injury

PURPOSE

Natural history studies have been informative in dissecting radiation injury, isolating its effects, and compartmentalizing injury based on the extent of exposure and the elapsed time post-irradiation. Although radiation injury models are useful for investigating the mechanism of action in isolated subsyndromes and development of medical countermeasures (MCMs), it is clear that ionizing radiation exposure leads to multi-organ injury (MOI).

METHODS

The Radiation and Nuclear Countermeasures Program within the National Institute of Allergy and Infectious Diseases partnered with the Biomedical Advanced Research and Development Authority to convene a virtual two-day meeting titled 'Radiation-Induced Multi-Organ Injury' on June 7-8, 2022. Invited subject matter experts presented their research findings in MOI, including study of mechanisms and possible MCMs to address complex radiation-induced injuries.

RESULTS

This workshop report summarizes key information from each presentation and discussion by the speakers and audience participants.

CONCLUSIONS

Understanding the mechanisms that lead to radiation-induced MOI is critical to advancing candidate MCMs that could mitigate the injury and reduce associated morbidity and mortality. The observation that some of these mechanisms associated with MOI include systemic injuries, such as inflammation and vascular damage, suggests that MCMs that address systemic pathways could be effective against multiple organ systems.

Trans-Omics analysis of post injury thrombo-inflammation identifies endotypes and trajectories in trauma patients

Understanding and managing the complexity of trauma-induced thrombo-inflammation necessitates an innovative, data-driven approach. This study leveraged a trans-omics analysis of longitudinal samples from trauma patients to illuminate molecular endotypes and trajectories that underpin patient outcomes, transcending traditional demographic and physiological characterizations. We hypothesize that trans-omics profiling reveals underlying clinical differences in severely injured patients that may present with similar clinical characteristics but ultimately have very different responses to treatment and clinical outcomes. Here we used proteomics and metabolomics to profile 759 of longitudinal plasma samples from 118 patients at 11 time points and 97 control subjects. Results were used to define distinct patient states through data reduction techniques. The patient groups were stratified based on their shock severity and injury severity score, revealing a spectrum of responses to trauma and treatment that are fundamentally tied to their unique underlying biology. Ensemble models were then employed, demonstrating the predictive power of these molecular signatures with area under the receiver operating curves of 80 to 94% for key outcomes such as INR, ICU-free days, ventilator-free days, acute lung injury, massive transfusion, and death. The molecularly defined endotypes and trajectories provide an unprecedented lens to understand and potentially guide trauma patient management, opening a path towards precision medicine. This strategy presents a transformative framework that aligns with our understanding that trauma patients, despite similar clinical presentations, might harbor vastly different biological responses and outcomes.

Redefining critical illness

Research and practice in critical care medicine have long been defined by syndromes, which, despite being clinically recognizable entities, are, in fact, loose amalgams of heterogeneous states that may respond differently to therapy. Mounting translational evidence-supported by research on respiratory failure due to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection-suggests that the current syndrome-based framework of critical illness should be reconsidered. Here we discuss recent findings from basic science and clinical research in critical care and explore how these might inform a new conceptual model of critical illness. De-emphasizing syndromes, we focus on the underlying biological changes that underpin critical illness states and that may be amenable to treatment. We hypothesize that such an approach will accelerate critical care research, leading to a richer understanding of the pathobiology of critical illness and of the key determinants of patient outcomes. This, in turn, will support the design of more effective clinical trials and inform a more precise and more effective practice at the bedside.

Hierarchical Cluster Analysis Identifies Distinct Physiological States After Acute Brain Injury

BACKGROUND

Analysis of intracranial multimodality monitoring data is challenging, and quantitative methods may help identify unique physiological signatures that inform therapeutic strategies and outcome prediction. The aim of this study was to test the hypothesis that data-driven approaches can identify distinct physiological states from intracranial multimodality monitoring data.

METHODS

This was a single-center retrospective observational study of patients with either severe traumatic brain injury or high-grade subarachnoid hemorrhage who underwent invasive multimodality neuromonitoring. We used hierarchical cluster analysis to group hourly values for heart rate, mean arterial pressure, intracranial pressure, brain tissue oxygen, and cerebral microdialysis across all included patients into distinct groups. Average values for measured physiological variables were compared across the identified clusters, and physiological profiles from identified clusters were mapped onto physiological states known to occur after acute brain injury. The distribution of clusters was compared between patients with favorable outcome (discharged to home or acute rehab) and unfavorable outcome (in-hospital death or discharged to chronic nursing facility).

RESULTS

A total of 1704 observations from 20 patients were included. Even though the difference in mean values for measured variables between patients with favorable and unfavorable outcome were small, we identified four distinct clusters within our data: (1) events with low brain tissue oxygen and high lactate-to-pyruvate ratio-values (consistent with cerebral ischemia), (2) events with higher intracranial pressure values without evidence for ischemia (3) events which appeared to be physiologically "normal," and (4) events with high cerebral lactate without brain hypoxia (consistent with cerebral hyperglycolysis). Patients with a favorable outcome had a greater proportion of cluster 3 (normal) events, whereas patients with an unfavorable outcome had a greater proportion of cluster 1 (ischemia) and cluster 4 (hyperglycolysis) events (p < 0.0001, Fisher-Freeman-Halton test).

CONCLUSIONS

A data-driven approach can identify distinct groupings from invasive multimodality neuromonitoring data that may have implications for therapeutic strategies and outcome predictions. These groupings could be used as classifiers to train machine learning models that can aid in the treatment of patients with acute brain injury. Further work is needed to replicate the findings of this exploratory study in larger data sets.

Big data and predictive analytics in neurocritical care

PURPOSE OF REVIEW

To describe predictive data and workflow in the intensive care unit when managing neurologically ill patients.

RECENT FINDINGS

In the era of Big Data in medicine, intensive critical care units are data-rich environments. Neurocritical care adds another layer of data with advanced multimodal monitoring to prevent secondary brain injury from ischemia, tissue hypoxia, and a cascade of ongoing metabolic events. A step closer toward personalized medicine is the application of multimodal monitoring of cerebral hemodynamics, bran oxygenation, brain metabolism, and electrophysiologic indices, all of which have complex and dynamic interactions. These data are acquired and visualized using different tools and monitors facing multiple challenges toward the goal of the optimal decision support system. In this review, we highlight some of the predictive data used to diagnose, treat, and prognosticate the neurologically ill patients. We describe information management in neurocritical care units including data acquisition, wrangling, analysis, and visualization.

Subspace Clustering of Physiological Data From Acute Traumatic Brain Injury Patients: Retrospective Analysis Based on the PROTECT III Trial

BACKGROUND

With advances in digital health technologies and proliferation of biomedical data in recent years, applications of machine learning in health care and medicine have gained considerable attention. While inpatient settings are equipped to generate rich clinical data from patients, there is a dearth of actionable information that can be used for pursuing secondary research for specific clinical conditions.

OBJECTIVE

This study focused on applying unsupervised machine learning techniques for traumatic brain injury (TBI), which is the leading cause of death and disability among children and adults aged less than 44 years. Specifically, we present a case study to demonstrate the feasibility and applicability of subspace clustering techniques for extracting patterns from data collected from TBI patients.

METHODS

Data for this study were obtained from the Progesterone for Traumatic Brain Injury, Experimental Clinical Treatment-Phase III (PROTECT III) trial, which included a cohort of 882 TBI patients. We applied subspace-clustering methods (density-based, cell-based, and clustering-oriented methods) to this data set and compared the performance of the different clustering methods.

RESULTS

The analyses showed the following three clusters of laboratory physiological data: (1) international normalized ratio (INR), (2) INR, chloride, and creatinine, and (3) hemoglobin and hematocrit. While all subclustering algorithms had a reasonable accuracy in classifying patients by mortality status, the density-based algorithm had a higher F1 score and coverage.

CONCLUSIONS

Clustering approaches serve as an important step for phenotype definition and validation in clinical domains such as TBI, where patient and injury heterogeneity are among the major reasons for failure of clinical trials. The results from this study provide a foundation to develop scalable clustering algorithms for further research and validation.

Challenges and Opportunities in Multimodal Monitoring and Data Analytics in Traumatic Brain Injury

PURPOSE OF REVIEW

Increasingly sophisticated systems for monitoring the brain have led to an increase in the use of multimodality monitoring (MMM) to detect secondary brain injuries before irreversible damage occurs after brain trauma. This review examines the challenges and opportunities associated with MMM in this population.

RECENT FINDINGS

Locally and internationally, the use of MMM varies. Practical challenges include difficulties with data acquisition, curation, and harmonization with other data sources limiting collaboration. However, efforts toward integration of MMM data, advancements in data science, and the availability of cloud-based infrastructures are now affording the opportunity for MMM to advance the care of patients with brain trauma. MMM provides data to guide the precision management of patients with traumatic brain injury in real time. While challenges exist, there are exciting opportunities for MMM to live up to this promise and to drive new insights into the physiology of the brain and beyond.

Three Data-Driven Phenotypes of Multiple Organ Dysfunction Syndrome Preserved from Early Childhood to Middle Adulthood

Multiple organ dysfunction syndrome (MODS) is one of the major causes of death and long-term impairment in critically ill patients. MODS is a complex, heterogeneous syndrome consisting of different phenotypes, which has limited the development of MODS-specific therapies and prognostic models. We used an unsupervised learning approach to derive novel phenotypes of MODS based on the type and severity of six individual organ dysfunctions. In a large, multi-center cohort of pediatric, young and middle-aged adults admitted to three different intensive care units, we uncovered and characterized three distinct data-driven phenotypes of MODS which were reproducible across age groups, where independently associated with outcomes and had unique predictors of in-hospital mortality.

C-reactive protein concentration as a risk predictor of mortality in intensive care unit: a multicenter, prospective, observational study

Background

It is not clear whether there are valuable inflammatory markers for prognosis judgment in the intensive care unit (ICU). We therefore conducted a multicenter, prospective, observational study to evaluate the prognostic role of inflammatory markers.

Methods

The clinical and laboratory data of patients at admission, including C-reactive protein (CRP), were collected in four general ICUs from September 1, 2018, to August 1, 2019. Multivariate logistic regression was used to identify factors independently associated with nonsurvival. The area under the receiver operating characteristic curve (AUC-ROC), net reclassification improvement (NRI), and integrated discrimination improvement (IDI) were used to evaluate the effect size of different factors in predicting mortality during ICU stay. 3 -knots were used to assess whether alternative cut points for these biomarkers were more appropriate.

Results

A total of 813 patients were recruited, among whom 121 patients (14.88%) died during the ICU stay. The AUC-ROC values of PCT and CRP for discriminating ICU mortality were 0.696 (95% confidence interval [CI], 0.650–0.743) and 0.684 (95% CI, 0.633–0.735), respectively. In the multivariable analysis, only APACHE II score (odds ratio, 1.166; 95% CI, 1.129–1.203; P = 0.000) and CRP concentration > 62.8 mg/L (odds ratio, 2.145; 95% CI, 1.343–3.427; P = 0.001), were significantly associated with an increased risk of ICU mortality. Moreover, the combination of APACHE II score and CRP > 62.8 mg/L significantly improved risk reclassification over the APACHE II score alone, with NRI (0.556) and IDI (0.013). Restricted cubic spline analysis confirmed that CRP concentration > 62.8 mg/L was the optimal cut-off value for differentiating between surviving and nonsurviving patients.

Conclusion

CRP markedly improved risk reclassification for prognosis prediction.

Neurocritical Care: Bench to Bedside (Eds. Claude Hemphill, Michael James) Integrating and Using Big Data in Neurocritical Care

The critical care environment drives huge volumes of data, and clinicians are tasked with quickly processing this data and responding to it urgently. The neurocritical care environment increasingly involves EEG, multimodal intracranial monitoring, and complex imaging which preclude comprehensive human synthesis, and requires new concepts to integrate data into clinical care. By definition, Big Data is data that cannot be handled using traditional infrastructures and is characterized by the volume, variety, velocity, and variability of the data being produced. Big Data in the neurocritical care unit requires rethinking of data storage infrastructures and the development of tools and analytics to drive advancements in the field. Preprocessing, feature extraction, statistical inference, and analytic tools are required in order to achieve the primary goals of Big Data for clinical use: description, prediction, and prescription. Barriers to its use at bedside include a lack of infrastructure development within the healthcare industry, lack of standardization of data inputs, and ultimately existential and scientific concerns about the outputs that result from the use of tools such as artificial intelligence. However, as implied by the fundamental theorem of biomedical informatics, physicians remain central to the development and utility of Big Data to improve patient care.

An evaluation of time series summary statistics as features for clinical prediction tasks

Background

Clinical prediction tasks such as patient mortality, length of hospital stay, and disease diagnosis are highly important in critical care research. The existing studies for clinical prediction mainly used simple summary statistics to summarize information from physiological time series. However, this lack of statistics leads to a lack of information. In addition, using only maximum and minimum statistics to indicate patient features fails to provide an adequate explanation. Few studies have evaluated which summary statistics best represent physiological time series.

Methods

In this paper, we summarize 14 statistics describing the characteristics of physiological time series, including the central tendency, dispersion tendency, and distribution shape. Then, we evaluate the use of summary statistics of physiological time series as features for three clinical prediction tasks. To find the combinations of statistics that yield the best performances under different tasks, we use a cross-validation-based genetic algorithm to approximate the optimal statistical combination.

Results

By experiments using the EHRs of 6,927 patients, we obtained prediction results based on both single statistics and commonly used combinations of statistics under three clinical prediction tasks. Based on the results of an embedded cross-validation genetic algorithm, we obtained 25 optimal sets of statistical combinations and then tested their prediction results. By comparing the performances of prediction with single statistics and commonly used combinations of statistics with quantitative analyses of the optimal statistical combinations, we found that some statistics play central roles in patient representation and different prediction tasks have certain commonalities.

Conclusion

Through an in-depth analysis of the results, we found many practical reference points that can provide guidance for subsequent related research. Statistics that indicate dispersion tendency, such as min, max, and range, are more suitable for length of stay prediction tasks, and they also provide information for short-term mortality prediction. Mean and quantiles that reflect the central tendency of physiological time series are more suitable for mortality and disease prediction. Skewness and kurtosis perform poorly when used separately for prediction but can be used as supplementary statistics to improve the overall prediction effect.

Linking Big Data and Prediction Strategies: Tools, Pitfalls, and Lessons Learned

OBJECTIVES

Modern critical care amasses unprecedented amounts of clinical data-so called "big data"-on a minute-by-minute basis. Innovative processing of these data has the potential to revolutionize clinical prognostics and decision support in the care of the critically ill but also forces clinicians to depend on new and complex tools of which they may have limited understanding and over which they have little control. This concise review aims to provide bedside clinicians with ways to think about common methods being used to extract information from clinical big datasets and to judge the quality and utility of that information.

DATA SOURCES

We searched the free-access search engines PubMed and Google Scholar using the MeSH terms "big data", "prediction", and "intensive care" with iterations of a range of additional potentially associated factors, along with published bibliographies, to find papers suggesting illustration of key points in the structuring and analysis of clinical "big data," with special focus on outcomes prediction and major clinical concerns in critical care.

STUDY SELECTION

Three reviewers independently screened preliminary citation lists.

DATA EXTRACTION

Summary data were tabulated for review.

DATA SYNTHESIS

To date, most relevant big data research has focused on development of and attempts to validate patient outcome scoring systems and has yet to fully make use of the potential for automation and novel uses of continuous data streams such as those available from clinical care monitoring devices.

CONCLUSIONS

Realizing the potential for big data to improve critical care patient outcomes will require unprecedented team building across disparate competencies. It will also require clinicians to develop statistical awareness and thinking as yet another critical judgment skill they bring to their patients' bedsides and to the array of evidence presented to them about their patients over the course of care.

Phenotyping Multiple Organ Dysfunction Syndrome Using Temporal Trends in Critically Ill Children

Multiple organ dysfunction syndrome (MODS) is one of the most common causes of death in critically ill children. However, despite decades of clinical trials, there are no comprehensive approaches to the management of MODS or effective targeted therapies that have consistently improved outcomes. Better understanding the heterogeneity of MODS and characterizing subgroups of MODS patients could improve our understanding of the syndrome and help us develop new management strategies. We analyzed a cohort of 5,297 children with MODS from two children's hospitals and used subgraph-augmented non-negative matrix factorization (SANMF) to identify unique temporal patterns in organ dysfunction across four novel subgroups. We demonstrate that these subgroups are composed of patients with distinct clinical characteristics and are independently predictive of clinical outcomes. Our work suggests that these subgroups represent four relevant phenotypes of pediatric MODS that could be used to identify novel management strategies.

Approaches to Multimodality Monitoring in Pediatric Traumatic Brain Injury

Traumatic brain injury (TBI) is a leading cause of morbidity and mortality in children. Improved methods of monitoring real-time cerebral physiology are needed to better understand when secondary brain injury develops and what treatment strategies may alleviate or prevent such injury. In this review, we discuss emerging technologies that exist to better understand intracranial pressure (ICP), cerebral blood flow, metabolism, oxygenation and electrical activity. We also discuss approaches to integrating these data as part of a multimodality monitoring strategy to improve patient care.

Machine Learning and Artificial Intelligence in Neurocritical Care: a Specialty-Wide Disruptive Transformation or a Strategy for Success

PURPOSE OF REVIEW

Neurocritical care combines the complexity of both medical and surgical disease states with the inherent limitations of assessing patients with neurologic injury. Artificial intelligence (AI) has garnered interest in the basic management of these complicated patients as data collection becomes increasingly automated.

RECENT FINDINGS

In this opinion article, we highlight the potential AI has in aiding the clinician in several aspects of neurocritical care, particularly in monitoring and managing intracranial pressure, seizures, hemodynamics, and ventilation. The model-based method and data-driven method are currently the two major AI methods for analyzing critical care data. Both are able to analyze the vast quantities of patient data that are accumulated in the neurocritical care unit. AI has the potential to reduce healthcare costs, minimize delays in patient management, and reduce medical errors. However, these systems are an aid to, not a replacement for, the clinician's judgment.

Detecting time-evolving phenotypic topics via tensor factorization on electronic health records: Cardiovascular disease case study

OBJECTIVE

Discovering subphenotypes of complex diseases can help characterize disease cohorts for investigative studies aimed at developing better diagnoses and treatments. Recent advances in unsupervised machine learning on electronic health record (EHR) data have enabled researchers to discover phenotypes without input from domain experts. However, most existing studies have ignored time and modeled diseases as discrete events. Uncovering the evolution of phenotypes - how they emerge, evolve and contribute to health outcomes - is essential to define more precise phenotypes and refine the understanding of disease progression. Our objective was to assess the benefits of an unsupervised approach that incorporates time to model diseases as dynamic processes in phenotype discovery.

METHODS

In this study, we applied a constrained non-negative tensor-factorization approach to characterize the complexity of cardiovascular disease (CVD) patient cohort based on longitudinal EHR data. Through tensor-factorization, we identified a set of phenotypic topics (i.e., subphenotypes) that these patients established over the 10 years prior to the diagnosis of CVD, and showed the progress pattern. For each identified subphenotype, we examined its association with the risk for adverse cardiovascular outcomes estimated by the American College of Cardiology/American Heart Association Pooled Cohort Risk Equations, a conventional CVD-risk assessment tool frequently used in clinical practice. Furthermore, we compared the subsequent myocardial infarction (MI) rates among the six most prevalent subphenotypes using survival analysis.

RESULTS

From a cohort of 12,380 adult CVD individuals with 1068 unique PheCodes, we successfully identified 14 subphenotypes. Through the association analysis with estimated CVD risk for each subtype, we found some phenotypic topics such as Vitamin D deficiency and depression, Urinary infections cannot be explained by the conventional risk factors. Through a survival analysis, we found markedly different risks of subsequent MI following the diagnosis of CVD among the six most prevalent topics (p < 0.0001), indicating these topics may capture clinically meaningful subphenotypes of CVD.

CONCLUSION

This study demonstrates the potential benefits of using tensor-decomposition to model diseases as dynamic processes from longitudinal EHR data. Our results suggest that this data-driven approach may potentially help researchers identify complex and chronic disease subphenotypes in precision medicine research.

Classifying juvenile onset primary open angle glaucoma using cluster analysis

Aim

To classify unrelated patients with juvenile onset primary open angle glaucoma (JOAG) into clinically useful phenotypes using cluster analysis.

Methods

Out of the 527 unrelated patients with JOAG, the study included 414 patients who had all the phenotypic characteristics required for the study. A cluster analysis was performed to classify the patients based on their iris and angle morphology, age of onset, highest untreated intraocular pressure (IOP), worst mean deviation and greatest vertical cup disc ratio of the worst eye. The iris features were broadly classified into three groups: those with normal iris crypts (NIC), those with prominent iris crypts (PIC) and those with absence of iris crypts. The gonio photographs were graded as normal appearing angle or those with angle dysgenesis in the form of a featureless angle, one with a high iris insertion and an angle with prominent iris processes. Using a hierarchical clustering model and a two-way cluster analysis, the distribution of clusters of JOAG was analysed to obtain a classification of JOAG subtypes.

Results

The four major clusters identified were: Cluster 1 with NIC and normal angles had the lowest untreated IOP and higher age of onset among all clusters. Cluster 2 with NIC and featureless angle was found to be associated with earliest age of onset. Cluster 3 had NIC and either a high iris insertion or prominent iris processes. Cluster 4 was a heterogeneous cluster with maximum number of patients in a group comprising of those with PIC and high iris insertion.

Conclusions

Cluster analysis extracted four subgroups of the JOAG phenotype that have clinical and prognostic significance and can potentially be helpful while evaluating these patients in the clinics.

A conceptual time window-based model for the early stratification of trauma patients

Progress in the testing of therapies targeting the immune response following trauma, a leading cause of morbidity and mortality worldwide, has been slow. We propose that the design of interventional trials in trauma would benefit from a scheme or platform that could support the identification and implementation of prognostic strategies for patient stratification. Here, we propose a stratification scheme based on defined time periods or windows following the traumatic event. This 'time-window' model allows for the incorporation of prognostic variables ranging from circulating biomarkers and clinical data to patient-specific information such as gene variants to predict adverse short- or long-term outcomes. A number of circulating biomarkers, including cell injury markers and damage-associated molecular patterns (DAMPs), and inflammatory mediators have been shown to correlate with adverse outcomes after trauma. Likewise, several single nucleotide polymorphisms (SNPs) associate with complications or death in trauma patients. This review summarizes the status of our understanding of the prognostic value of these classes of variables in predicting outcomes in trauma patients. Strategies for the incorporation of these prognostic variables into schemes designed to stratify trauma patients, such as our time-window model, are also discussed.

Artificial intelligence in neurocritical care

BACKGROUND

Neurocritical care combines the management of extremely complex disease states with the inherent limitations of clinically assessing patients with brain injury. As the management of neurocritical care patients can be immensely complicated, the automation of data-collection and basic management by artificial intelligence systems have garnered interest.

METHODS

In this opinion article, we highlight the potential artificial intelligence has in monitoring and managing several aspects of neurocritical care, specifically intracranial pressure, seizure monitoring, blood pressure, and ventilation.

RESULTS

The two major AI methods of analytical technique currently exist for analyzing critical care data: the model-based method and data driven method. Both of these methods have demonstrated an ability to analyze vast quantities of patient data, and we highlight the ways in which these modalities of artificial intelligence might one day play a role in neurocritical care.

CONCLUSIONS

While none of these artificial intelligence systems are meant to replace the clinician's judgment, these systems have the potential to reduce healthcare costs and errors or delays in medical management.

Machine learning in critical care: state-of-the-art and a sepsis case study

BACKGROUND

Like other scientific fields, such as cosmology, high-energy physics, or even the life sciences, medicine and healthcare face the challenge of an extremely quick transformation into data-driven sciences. This challenge entails the daunting task of extracting usable knowledge from these data using algorithmic methods. In the medical context this may for instance realized through the design of medical decision support systems for diagnosis, prognosis and patient management. The intensive care unit (ICU), and by extension the whole area of critical care, is becoming one of the most data-driven clinical environments.

RESULTS

The increasing availability of complex and heterogeneous data at the point of patient attention in critical care environments makes the development of fresh approaches to data analysis almost compulsory. Computational Intelligence (CI) and Machine Learning (ML) methods can provide such approaches and have already shown their usefulness in addressing problems in this context. The current study has a dual goal: it is first a review of the state-of-the-art on the use and application of such methods in the field of critical care. Such review is presented from the viewpoint of the different subfields of critical care, but also from the viewpoint of the different available ML and CI techniques. The second goal is presenting a collection of results that illustrate the breath of possibilities opened by ML and CI methods using a single problem, the investigation of septic shock at the ICU.

CONCLUSION

We have presented a structured state-of-the-art that illustrates the broad-ranging ways in which ML and CI methods can make a difference in problems affecting the manifold areas of critical care. The potential of ML and CI has been illustrated in detail through an example concerning the sepsis pathology. The new definitions of sepsis and the relevance of using the systemic inflammatory response syndrome (SIRS) in its diagnosis have been considered. Conditional independence models have been used to address this problem, showing that SIRS depends on both organ dysfunction measured through the Sequential Organ Failure (SOFA) score and the ICU outcome, thus concluding that SIRS should still be considered in the study of the pathophysiology of Sepsis. Current assessment of the risk of dead at the ICU lacks specificity. ML and CI techniques are shown to improve the assessment using both indicators already in place and other clinical variables that are routinely measured. Kernel methods in particular are shown to provide the best performance balance while being amenable to representation through graphical models, which increases their interpretability and, with it, their likelihood to be accepted in medical practice.

Predicting ICU readmission using grouped physiological and medication trends

BACKGROUND

Patients who are readmitted to an intensive care unit (ICU) usually have a high risk of mortality and an increased length of stay. ICU readmission risk prediction may help physicians to re-evaluate the patient's physical conditions before patients are discharged and avoid preventable readmissions. ICU readmission prediction models are often built based on physiological variables. Intuitively, snapshot measurements, especially the last measurements, are effective predictors that are widely used by researchers. However, methods that only use snapshot measurements neglect predictive information contained in the trends of physiological and medication variables. Mean, maximum or minimum values take multiple time points into account and capture their summary statistics, however, these statistics are not able to catch the detailed picture of temporal trends. In this study, we find strong predictors with ability of capturing detailed temporal trends of variables for 30-day readmission risk and build prediction models with high accuracy.

METHODS

We study physiological measurements and medications from the Multiparameter Intelligent Monitoring in Intensive Care II (MIMIC-II) clinical dataset. Time series of each variable are converted into trend graphs with nodes being discretized measurements of each variable. Then we extract important temporal trends by applying frequent subgraph mining on the trend graphs. The frequency of a subgraph is a good cue to find important temporal trends since similar patients often share similar trends regarding their pathophysiological evolution under medical interventions. Important temporal trends are then grouped automatically by non-negative matrix factorization. The grouped trends could be considered as an approximate representation of patients' pathophysiological states and medication profiles. We train a logistic regression model to predict 30-day ICU readmission risk based on snapshot measurements, grouped physiological trends and medication trends.

RESULTS

Our dataset consists of 1170 patients who are alive 30 days after discharge from ICU and have at least 12 h of data. In the dataset, 860 patients were not readmitted and 310 were readmitted, within 30 days after discharge. Our model outperforms all comparison models, and shows an improvement in the area under the receiver operating characteristic curve (AUC) of almost 4% from the best comparison model.

CONCLUSIONS

Grouped physiological and medication trends carry predictive information for ICU readmission risk. In order to build predictive models with higher accuracy, we should add grouped physiological and medication trends as complementary features to snapshot measurements.

Unraveling the complexities of invasive multimodality neuromonitoring

Acute brain injuries are a major cause of death and disability worldwide. Survivors of life-threatening brain injury often face a lifetime of dependent care, and novel approaches that improve outcome are sorely needed. A delayed cascade of brain damage, termed secondary injury, occurs hours to days and even weeks after the initial insult. This delayed phase of injury provides a crucial window for therapeutic interventions that could limit brain damage and improve outcome. A major barrier in the ability to prevent and treat secondary injury is that physicians are often unable to target therapies to patients' unique cerebral physiological disruptions. Invasive neuromonitoring with multiple complementary physiological monitors can provide useful information to enable this tailored, precision approach to care. However, integrating the multiple streams of time-varying data is challenging and often not possible during routine bedside assessment. The authors review and discuss the principles and evidence underlying several widely used invasive neuromonitors. They also provide a framework for integrating data for clinical decision making and discuss future developments in informatics that may allow new treatment paradigms to be developed.

EventThread: Visual Summarization and Stage Analysis of Event Sequence Data

Event sequence data such as electronic health records, a person's academic records, or car service records, are ordered series of events which have occurred over a period of time. Analyzing collections of event sequences can reveal common or semantically important sequential patterns. For example, event sequence analysis might reveal frequently used care plans for treating a disease, typical publishing patterns of professors, and the patterns of service that result in a well-maintained car. It is challenging, however, to visually explore large numbers of event sequences, or sequences with large numbers of event types. Existing methods focus on extracting explicitly matching patterns of events using statistical analysis to create stages of event progression over time. However, these methods fail to capture latent clusters of similar but not identical evolutions of event sequences. In this paper, we introduce a novel visualization system named EventThread which clusters event sequences into threads based on tensor analysis and visualizes the latent stage categories and evolution patterns by interactively grouping the threads by similarity into time-specific clusters. We demonstrate the effectiveness of EventThread through usage scenarios in three different application domains and via interviews with an expert user.

Translational approaches to coagulopathy after trauma: Towards targeted treatment

Mitchell J. Cohen discusses why trauma care must go beyond restoring perfusion to target disorders of inflammation and coagulation in severely injured patients.

Clot Formation Is Associated With Fibrinogen and Platelet Forces in a Cohort of Severely Injured Emergency Department Trauma Patients

INTRODUCTION

Anticoagulation, fibrinogen consumption, fibrinolytic activation, and platelet dysfunction all interact to produce different clot formation responses after trauma. However, the relative contributions of these coagulation components to overall clot formation remain poorly defined. We examined for sources of heterogeneity in clot formation responses after trauma.

METHODS

Blood was sampled in the emergency department from patients meeting trauma team activation criteria at an urban trauma center. Plasma prothrombin time of 18 s or longer was used to define traumatic coagulopathy. Mean kaolin-activated thrombelastography (TEG) parameters were calculated and tested for heterogeneity using analysis of means. Discriminant analysis and forward stepwise variable selection with linear regression were used to determine if prothrombin time, fibrinogen, platelet contractile force (PCF), and D-dimer concentration, representing key mechanistic components of coagulopathy, each contribute to heterogeneous TEG responses after trauma.

RESULTS

Of 95 subjects, 16% met criteria for coagulopathy. Coagulopathic subjects were more severely injured with greater shock and received more blood products in the first 8 h compared with noncoagulopathic subjects. Mean (SD) TEG maximal amplitude (MA) was significantly decreased in the coagulopathic group (57.5 [SD, 4.7] mm vs. 62.7 [SD, 4.7], t test P < 0.001). The MA also exceeded the ANOM predicted upper decision limit for the noncoagulopathic group and the lower decision limit for the coagulopathic group at α = 0.05, suggesting significant heterogeneity from the overall cohort mean. Fibrinogen and PCF best discriminated TEG MA using discriminant analysis. Fibrinogen, PCF, and D-dimer were primary covariates for TEG MA using regression analysis.

CONCLUSIONS

Heterogeneity in TEG-based clot formation in emergency department trauma patients was linked to changes in MA. Individual parameters representing fibrin polymerization, PCFs, and fibrinolysis were primarily associated with TEG MA after trauma and should be the focus of early hemostatic therapies.

Data collection and interpretation

Patient monitoring is routinely performed in all patients who receive neurocritical care. The combined use of monitors, including the neurologic examination, laboratory analysis, imaging studies, and physiological parameters, is common in a platform called multi-modality monitoring (MMM). However, the full potential of MMM is only beginning to be realized since for the most part, decision making historically has focused on individual aspects of physiology in a largely threshold-based manner. The use of MMM now is being facilitated by the evolution of bio-informatics in critical care including developing techniques to acquire, store, retrieve, and display integrated data and new analytic techniques for optimal clinical decision making. In this review, we will discuss the crucial initial steps toward data and information management, which in this emerging era of data-intensive science is already shifting concepts of care for acute brain injury and has the potential to both reshape how we do research and enhance cost-effective clinical care.

In Search of a Cure for Sepsis: Taming the Monster in Critical Care Medicine

In spite of over half a century of research, sepsis still constitutes a major problem in health care delivery. Although advances in research have significantly increased our knowledge of the pathogenesis of sepsis and resulted in better prognosis and improved survival outcome, sepsis still remains a major challenge in modern medicine with an increase in occurrence predicted and a huge socioeconomic burden. It is generally accepted that sepsis is due to an initial hyperinflammatory response. However, numerous efforts aimed at targeting the proinflammatory cytokine network have been largely unsuccessful and the search for novel potential therapeutic targets continues. Recent studies provide compelling evidence that dysregulated anti-inflammatory responses may also contribute to sepsis mortality. Our previous studies on the role of regulatory T cells and phosphoinositide 3-kinases in sepsis highlight immunological approaches that could be explored for sepsis therapy. In this article, we review the current and emerging concepts in sepsis, highlight novel potential therapeutic targets and immunological approaches for sepsis treatment and propose a biphasic treatment approach for management of the condition.

Identification of Clinically Relevant Groups of Patients Through the Application of Cluster Analysis to a Complex Traumatic Brain Injury Data Set

In neurological intensive care units (NICUs) we are collecting an ever increasing quantity of data. These range from patient demographics and physiological monitoring to treatment strategies and outcomes. The BrainIT database is an example of this type of rich data source. It contains validated data on 264 patients who suffered traumatic brain injury (TBI) admitted to 22 NICUs in 11 European countries between March 2003 and July 2005 [1, 6].

Topological data analysis for discovery in preclinical spinal cord injury and traumatic brain injury

Data-driven discovery in complex neurological disorders has potential to extract meaningful syndromic knowledge from large, heterogeneous data sets to enhance potential for precision medicine. Here we describe the application of topological data analysis (TDA) for data-driven discovery in preclinical traumatic brain injury (TBI) and spinal cord injury (SCI) data sets mined from the Visualized Syndromic Information and Outcomes for Neurotrauma-SCI (VISION-SCI) repository. Through direct visualization of inter-related histopathological, functional and health outcomes, TDA detected novel patterns across the syndromic network, uncovering interactions between SCI and co-occurring TBI, as well as detrimental drug effects in unpublished multicentre preclinical drug trial data in SCI. TDA also revealed that perioperative hypertension predicted long-term recovery better than any tested drug after thoracic SCI in rats. TDA-based data-driven discovery has great potential application for decision-support for basic research and clinical problems such as outcome assessment, neurocritical care, treatment planning and rapid, precision-diagnosis.

Trending autoregulatory indices during treatment for traumatic brain injury

Our goal is to use automatic data monitoring for reliable prediction of episodes of intracranial hypertension in patients with traumatic brain injury. Here we test the validity of our method on retrospective patient data. We developed the Continuous Hemodynamic Autoregulatory Monitor (CHARM), that siphons and stores signals from existing monitors in the surgical intensive care unit (SICU), efficiently compresses them, and standardizes the search for statistical relationships between any proposed index and adverse events. CHARM uses an automated event detector to reliably locate episodes of elevated intracranial pressure (ICP), while eliminating artifacts within retrospective patient data. A graphical user interface allowed data scanning, selection of criteria for events, and calculating indices. The pressure reactivity index (PRx), defined as the least square linear regression slope of intracranial pressure versus arterial BP, was calculated for a single case that spanned 259 h. CHARM collected continuous records of ABP, ICP, ECG, SpO2, and ventilation from 29 patients with TBI over an 18-month period. Analysis of a single patient showed that PRx data distribution in the single hours immediately prior to all 16 intracranial hypertensive events, significantly differed from that in the 243 h that did not precede such events (p < 0.0001). The PRx index, however, lacked sufficient resolution as a real-time predictor of IH in this patient. CHARM streamlines the search for reliable predictors of intracranial hypertension. We report statistical evidence supporting the predictive potential of the pressure reactivity index.

Social determinants of health inequalities: towards a theoretical perspective using systems science

A systems approach offers a novel conceptualization to natural and social systems. In recent years, this has led to perceiving population health outcomes as an emergent property of a dynamic and open, complex adaptive system. The current paper explores these themes further and applies the principles of systems approach and complexity science (i.e. systems science) to conceptualize social determinants of health inequalities. The conceptualization can be done in two steps: viewing health inequalities from a systems approach and extending it to include complexity science. Systems approach views health inequalities as patterns within the larger rubric of other facets of the human condition, such as educational outcomes and economic development. This anlysis requires more sophisticated models such as systems dynamic models. An extension of the approach is to view systems as complex adaptive systems, i.e. systems that are 'open' and adapt to the environment. They consist of dynamic adapting subsystems that exhibit non-linear interactions, while being 'open' to a similarly dynamic environment of interconnected systems. They exhibit emergent properties that cannot be estimated with precision by using the known interactions among its components (such as economic development, political freedom, health system, culture etc.). Different combinations of the same bundle of factors or determinants give rise to similar patterns or outcomes (i.e. property of convergence), and minor variations in the initial condition could give rise to widely divergent outcomes. Novel approaches using computer simulation models (e.g. agent-based models) would shed light on possible mechanisms as to how factors or determinants interact and lead to emergent patterns of health inequalities of populations.

Early hemostatic responses to trauma identified with hierarchical clustering analysis

BACKGROUND

Trauma-induced coagulopathy is a complex multifactorial hemostatic response that is poorly understood.

OBJECTIVES

To identify distinct hemostatic responses to trauma and identify key components of the hemostatic system that vary between responses.

PATIENTS/METHODS

A cross-sectional observational study of adult trauma patients at an urban level I trauma center emergency department was performed. Hierarchical clustering analysis was used to identify distinct clusters of similar subjects according to vital signs, injury/shock severity, and comprehensive assessment of coagulation, clot formation, platelet function, and thrombin generation.

RESULTS

Among 84 total trauma patients included in the model, three distinct trauma clusters were identified. Cluster 1 (N = 57) showed platelet activation, preserved peak thrombin generation, plasma coagulation dysfunction, a moderately decreased fibrinogen concentration and normal clot formation relative to healthy controls. Cluster 2 (N = 18) showed platelet activation, preserved peak thrombin generation, and a preserved fibrinogen concentration with normal clot formation. Cluster 3 (N = 9) was the most severely injured and shocked, and showed a strong inflammatory and bleeding phenotype. Platelet dysfunction, thrombin inhibition, plasma coagulation dysfunction and a decreased fibrinogen concentration were present in this cluster. Fibrinolytic activation was present in all clusters, but was particularly increased in cluster 3. Trauma clusters were most noticeably different in their relative fibrinogen concentration, peak thrombin generation, and platelet-induced clot contraction.

CONCLUSIONS

Hierarchical clustering analysis identified three distinct hemostatic responses to trauma. Further insights into the underlying hemostatic mechanisms responsible for these responses are needed.

A Chemical Engineer's Perspective on Health and Disease

Chemical process systems engineering considers complex supply chains which are coupled networks of dynamically interacting systems. The quest to optimize the supply chain while meeting robustness and flexibility constraints in the face of ever changing environments necessitated the development of theoretical and computational tools for the analysis, synthesis and design of such complex engineered architectures. However, it was realized early on that optimality is a complex characteristic required to achieve proper balance between multiple, often competing, objectives. As we begin to unravel life's intricate complexities, we realize that that living systems share similar structural and dynamic characteristics; hence much can be learned about biological complexity from engineered systems. In this article, we draw analogies between concepts in process systems engineering and conceptual models of health and disease; establish connections between these concepts and physiologic modeling; and describe how these mirror onto the physiological counterparts of engineered systems.

Inducible protein-10, a potential driver of neurally controlled interleukin-10 and morbidity in human blunt trauma

OBJECTIVE

Blunt trauma and traumatic spinal cord injury induce systemic inflammation that contributes to morbidity. Dysregulated neural control of systemic inflammation postinjury is likely exaggerated in patients with traumatic spinal cord injury. We used in silico methods to discern dynamic inflammatory networks that could distinguish systemic inflammation in traumatic spinal cord injury from blunt trauma.

DESIGN

Retrospective study.

SETTINGS

Tertiary care institution.

PATIENTS

Twenty-one severely injured thoracocervical traumatic spinal cord injury patients and matched 21 severely injured blunt trauma patients without spinal cord injury.

INTERVENTION

None.

MEASUREMENTS AND MAIN RESULTS

Serial blood samples were obtained from days 1 to 14 postinjury. Twenty-four plasma inflammatory mediators were quantified. Statistical significance between the two groups was determined by two-way analysis of variance. Dynamic Bayesian network inference was used to suggest dynamic connectivity and central inflammatory mediators. Circulating interleukin-10 was significantly elevated in thoracocervical traumatic spinal cord injury group versus non-spinal cord injury group, whereas interleukin-1β, soluble interleukin-2 receptor-α, interleukin-4, interleukin-5, interleukin-7, interleukin-13, interleukin-17, macrophage inflammatory protein 1α and 1β, granulocyte-macrophage colony-stimulating factor, and interferon-γ were significantly reduced in traumatic spinal cord injury group versus non-spinal cord injury group. Dynamic Bayesian network suggested that post-spinal cord injury interleukin-10 is driven by inducible protein-10, whereas monocyte chemotactic protein-1 was central in non-spinal cord injury dynamic networks. In a separate validation cohorts of 356 patients without spinal cord injury and 85 traumatic spinal cord injury patients, individuals with plasma inducible protein-10 levels more than or equal to 730 pg/mL had significantly prolonged hospital and ICU stay and days on mechanical ventilator versus patients with plasma inducible protein-10 level less than 730 pg/mL.

CONCLUSION

This is the first study to compare the dynamic systemic inflammatory responses of traumatic spinal cord injury patients versus patients without spinal cord injury, suggesting a key role for inducible protein-10 in driving systemic interleukin-10 and morbidity and highlighting the potential utility of in silico tools to identify key inflammatory drivers.

In silico modeling: methods and applications to trauma and sepsis

OBJECTIVES

To familiarize clinicians with advances in computational disease modeling applied to trauma and sepsis.

DATA SOURCES

PubMed search and review of relevant medical literature.

SUMMARY

Definitions, key methods, and applications of computational modeling to trauma and sepsis are reviewed.

CONCLUSIONS

Computational modeling of inflammation and organ dysfunction at the cellular, organ, whole-organism, and population levels has suggested a positive feedback cycle of inflammation → damage → inflammation that manifests via organ-specific inflammatory switching networks. This structure may manifest as multicompartment "tipping points" that drive multiple organ dysfunction. This process may be amenable to rational inflammation reprogramming.

A Systems Engineering Perspective on Homeostasis and Disease

Engineered systems are coupled networks of interacting sub-systems, whose dynamics are constrained to requirements of robustness and flexibility. They have evolved by design to optimize function in a changing environment and maintain responses within ranges. Analysis, synthesis, and design of complex supply chains aim to identify and explore the laws governing optimally integrated systems. Optimality expresses balance between conflicting objectives while resiliency results from dynamic interactions among elements. Our increasing understanding of life’s multi-scale architecture suggests that living systems share similar characteristics with much to be learned about biological complexity from engineered systems. If health reflects a dynamically stable integration of molecules, cell, tissues, and organs; disease indicates displacement compensated for and corrected by activation and combination of feedback mechanisms through interconnected networks. In this article, we draw analogies between concepts in systems engineering and conceptual models of health and disease; establish connections between these concepts and physiologic modeling; and describe how these mirror onto the physiological counterparts of engineered systems.

Altering physiological networks using drugs: steps towards personalized physiology

Background

The rise of personalized medicine has reminded us that each patient must be treated as an individual. One factor in making treatment decisions is the physiological state of each patient, but definitions of relevant states and methods to visualize state-related physiologic changes are scarce. We constructed correlation networks from physiologic data to demonstrate changes associated with pressor use in the intensive care unit.

Methods

We collected 29 physiological variables at one-minute intervals from nineteen trauma patients in the intensive care unit of an academic hospital and grouped each minute of data as receiving or not receiving pressors. For each group we constructed Spearman correlation networks of pairs of physiologic variables. To visualize drug-associated changes we split the networks into three components: an unchanging network, a network of connections with changing correlation sign, and a network of connections only present in one group.

Results

Out of a possible 406 connections between the 29 physiological measures, 64, 39, and 48 were present in each of the three component networks. The static network confirms expected physiological relationships while the network of associations with changed correlation sign suggests putative changes due to the drugs. The network of associations present only with pressors suggests new relationships that could be worthy of study.

Conclusions

We demonstrated that visualizing physiological relationships using correlation networks provides insight into underlying physiologic states while also showing that many of these relationships change when the state is defined by the presence of drugs. This method applied to targeted experiments could change the way critical care patients are monitored and treated.

A survey analysis of acoustic trauma related to MR scans

PURPOSE

The maximum limit of MR scanner noise and necessity of ear protection is defined in the IEC standard (IEC60601-2-33) of MR safety. With improvements in MR scanner performance, pulse sequences generating higher scanning noise have been used clinically. In this study, we investigated the factors significantly related to potential acoustic trauma cases (PATC) after MR examinations. To consider the future direction for MR safety and prevention of acoustic trauma, issues related to noise generation by MR scanners and acoustic trauma were systematically reviewed.

METHODS

A statistical analysis was performed using the data set from a survey (n=974) conducted in 2010 by the JSMRM safety committee. Hierarchical clustering analysis was used to extract the characteristics of the responders. With this classification as a reference, tests of independence and a residual analysis were employed to evaluate the factors related to PATC.

RESULTS

No significant relationship was observed between the ear protection policy and the incidence or the reported outcome of PATC. While the two main clusters out of the six clusters extracted were associated with who reported the PATC and the confirmation process of the acoustic noise level of MR scanners, no cluster was associated with the frequency of PATC. An absence of PATC was significantly less reported (p=0.03) and more PATC was reported (p=0.04) by facilities with 3T MR systems.

DISCUSSION

Although the total frequency was 4 cases, it should be noted that persistent hearing disturbances are a possible consequence of MR examinations. Neither the condition of the subjects nor the ear protection method was significantly related to the probability of PATC, suggesting the difficulty of predicting the potential risk of acoustic trauma. It is recommended to more systematically follow up PATC cases and clarify the risk factors.

Prognostic physiology: modeling patient severity in Intensive Care Units using radial domain folding

Real-time scalable predictive algorithms that can mine big health data as the care is happening can become the new "medical tests" in critical care. This work describes a new unsupervised learning approach, radial domain folding, to scale and summarize the enormous amount of data collected and to visualize the degradations or improvements in multiple organ systems in real time. Our proposed system is based on learning multi-layer lower dimensional abstractions from routinely generated patient data in modern Intensive Care Units (ICUs), and is dramatically different from most of the current work being done in ICU data mining that rely on building supervised predictive models using commonly measured clinical observations. We demonstrate that our system discovers abstract patient states that summarize a patient's physiology. Further, we show that a logistic regression model trained exclusively on our learned layer outperforms a customized SAPS II score on the mortality prediction task.

Risk stratification of ICU patients using topic models inferred from unstructured progress notes

We propose a novel approach for ICU patient risk stratification by combining the learned "topic" structure of clinical concepts (represented by UMLS codes) extracted from the unstructured nursing notes with physiologic data (from SAPS-I) for hospital mortality prediction. We used Hierarchical Dirichlet Processes (HDP), a non-parametric topic modeling technique, to automatically discover "topics" as shared groups of co-occurring UMLS clinical concepts. We evaluated the potential utility of the inferred topic structure in predicting hospital mortality using the nursing notes of 14,739 adult ICU patients (mortality 14.6%) from the MIMIC II database. Our results indicate that learned topic structure from the first 24-hour ICU nursing notes significantly improved the performance of the SAPS-I algorithm for hospital mortality prediction. The AUC for predicting hospital mortality from the first 24 hours of physiologic data and nursing text notes was 0.82. Using the physiologic data alone with the SAPS-I algorithm, an AUC of 0.72 was achieved. Thus, the clinical topics that were extracted and used to augment the SAPS-I algorithm significantly improved the performance of the baseline algorithm.

Discretization of continuous features in clinical datasets

BACKGROUND

The increasing availability of clinical data from electronic medical records (EMRs) has created opportunities for secondary uses of health information. When used in machine learning classification, many data features must first be transformed by discretization.

OBJECTIVE

To evaluate six discretization strategies, both supervised and unsupervised, using EMR data.

MATERIALS AND METHODS

We classified laboratory data (arterial blood gas (ABG) measurements) and physiologic data (cardiac output (CO) measurements) derived from adult patients in the intensive care unit using decision trees and naïve Bayes classifiers. Continuous features were partitioned using two supervised, and four unsupervised discretization strategies. The resulting classification accuracy was compared with that obtained with the original, continuous data.

RESULTS

Supervised methods were more accurate and consistent than unsupervised, but tended to produce larger decision trees. Among the unsupervised methods, equal frequency and k-means performed well overall, while equal width was significantly less accurate.

DISCUSSION

This is, we believe, the first dedicated evaluation of discretization strategies using EMR data. It is unlikely that any one discretization method applies universally to EMR data. Performance was influenced by the choice of class labels and, in the case of unsupervised methods, the number of intervals. In selecting the number of intervals there is generally a trade-off between greater accuracy and greater consistency.

CONCLUSIONS

In general, supervised methods yield higher accuracy, but are constrained to a single specific application. Unsupervised methods do not require class labels and can produce discretized data that can be used for multiple purposes.

Computational approaches for translational clinical research in disease progression

Today, there is an ever-increasing amount of biological and clinical data available that could be used to enhance a systems-based understanding of disease progression through innovative computational analysis. In this article, we review a selection of published research regarding computational methods, primarily from systems biology, which support translational research from the molecular level to the bedside, with a focus on applications in trauma and critical care. Trauma is the leading cause of mortality in Americans younger than 45 years, and its rapid progression offers both opportunities and challenges for computational analysis of trends in molecular patterns associated with outcomes and therapeutic interventions.This review presents methods and domain-specific examples that may inspire the development of new algorithms and computational methods that use both molecular and clinical data for diagnosis, prognosis, and therapy in disease progression.

Linear and nonlinear heart rate variability indexes in clinical practice

Biological organisms have intrinsic control systems that act in response to internal and external stimuli maintaining homeostasis. Human heart rate is not regular and varies in time and such variability, also known as heart rate variability (HRV), is not random. HRV depends upon organism's physiologic and/or pathologic state. Physicians are always interested in predicting patient's risk of developing major and life-threatening complications. Understanding biological signals behavior helps to characterize patient's state and might represent a step toward a better care. The main advantage of signals such as HRV indexes is that it can be calculated in real time in noninvasive manner, while all current biomarkers used in clinical practice are discrete and imply blood sample analysis. In this paper HRV linear and nonlinear indexes are reviewed and data from real patients are provided to show how these indexes might be used in clinical practice.

Computational and systems biology in trauma and sepsis: current state and future perspectives

Trauma, often accompanied by hemorrhage, is a leading cause of death worldwide, often leading to inflammation-related late complications that include sepsis and multiple organ failure. These secondary complications are a manifestation of the complexity of biological responses elicited by trauma/hemorrhage, responses that span most, if not all, cell types, tissues, and organ systems. This daunting complexity at the patient level is manifest by the near total dearth of available therapeutics, and we suggest that this dire condition is due in large part to the lack of a rational, systems-oriented framework for drug development, clinical trial design, in-hospital diagnostics, and post-hospital care. We have further suggested that mechanistic computational modeling can form the basis of such a rational framework, given the maturity of systems biology/computational biology. Herein, we briefly summarize the state of the art of these approaches, and highlight the biological insights and novel hypotheses derived from these approaches. We propose a rational framework for transitioning through the currently fragmented process from identification of biological networks that are potential therapeutic targets, through clinical trial design, to personalized diagnosis and care. Insights derived from systems and computational biology in trauma and sepsis include the centrality of Damage-Associated Molecular Pattern molecules as drivers of both beneficial and detrimental inflammation, along with a novel view of multiple organ dysfunction as a cascade of containment failures with distinct implications for therapy. Finally, we suggest how these insights might be best implemented to drive transformational change in the fields of trauma and sepsis.

Use of models in identification and prediction of physiology in critically ill surgical patients

BACKGROUND

With higher-throughput data acquisition and processing, increasing computational power, and advancing computer and mathematical techniques, modelling of clinical and biological data is advancing rapidly. Although exciting, the goal of recreating or surpassing in silico the clinical insight of the experienced clinician remains difficult. Advances toward this goal and a brief overview of various modelling and statistical techniques constitute the purpose of this review.

METHODS

A review of the literature and experience with models and physiological state representation and prediction after injury was undertaken.

RESULTS

A brief overview of models and the thinking behind their use for surgeons new to the field is presented, including an introduction to visualization and modelling work in surgical care, discussion of state identification and prediction, discussion of causal inference statistical approaches, and a brief introduction to new vital signs and waveform analysis.

CONCLUSION

Modelling in surgical critical care can provide a useful adjunct to traditional reductionist biological and clinical analysis. Ultimately the goal is to model computationally the clinical acumen of the experienced clinician.

Sepsis: from pattern to mechanism and back

Sepsis is a clinical entity in which complex inflammatory and physiological processes are mobilized, not only across a range of cellular and molecular interactions, but also in clinically relevant physiological signals accessible at the bedside. There is a need for a mechanistic understanding that links the clinical phenomenon of physiologic variability with the underlying patterns of the biology of inflammation, and we assert that this can be facilitated through the use of dynamic mathematical and computational modeling. An iterative approach of laboratory experimentation and mathematical/computational modeling has the potential to integrate cellular biology, physiology, control theory, and systems engineering across biological scales, yielding insights into the control structures that govern mechanisms by which phenomena, detected as biological patterns, are produced. This approach can represent hypotheses in the formal language of mathematics and computation, and link behaviors that cross scales and domains, thereby offering the opportunity to better explain, diagnose, and intervene in the care of the septic patient.