Search

Intensive care medicine is complex and resource-demanding. A critical and common challenge lies in inferring the underlying physiological state of a patient from partially observed data. Specifically for the cardiovascular system, clinicians use observables such as heart rate, arterial and venous blood pressures, as well as findings from the physical examination and ancillary tests to formulate a mental model and estimate hidden variables such as cardiac output, vascular resistance, filling pressures and volumes, and autonomic tone. Then, they use this mental model to derive the causes for instability and choose appropriate interventions. Not only this is a very hard problem due to the nature of the signals, but it also requires expertise and a clinician’s ongoing presence at the bedside. Clinical decision support tools based on mechanistic dynamical models offer an appealing solution due to their inherent explainability, corollaries to the clinical mental process, and predictive power. With a translational motivation in mind, we developed iCVS: a simple, with high explanatory power, dynamical mechanistic model to infer hidden cardiovascular states. Full model estimation requires no prior assumptions on physiological parameters except age and weight, and the only inputs are arterial and venous pressure waveforms. iCVS also considers autonomic and non-autonomic modulations. To gain more information without increasing model complexity, both slow and fast timescales of the blood pressure traces are exploited, while the main inference and dynamic evolution are at the longer, clinically relevant, timescale of minutes. iCVS is designed to allow bedside deployment at pediatric and adult intensive care units and for retrospective investigation of cardiovascular mechanisms underlying instability. In this paper, we describe iCVS and inference system in detail, and using a dataset of critically-ill children, we demonstrate its use and power to identify bleeding, distributive states, and cardiac dysfunction, in isolation and in combination.Author summaryA common challenge clinicians face across different disciplines is estimating the hidden physiological state of a patient based on partially observed data. Here we describe iCVS (inferring Cardio-Vascular States): a dynamical mechanistic model of the cardiovascular system. We developed iCVS with a translational goal in mind, showing high explanatory power, its inference relies only on routinely available signals, and enables the identification of various clinically important shock states. We demonstrate the use of the model on a dataset that was collected in a pediatric intensive care unit.

Background and objectivesChildren with congenital heart disease (CHD) are predisposed to rapid deterioration in the face of common childhood illnesses. When they present to their local emergency departments (ED) with acute illness, rapid and accurate diagnosis and treatment is crucial to recovery and survival. Previous studies have shown that ED physicians are uncomfortable caring for patients with CHD and there is a lack of actionable guidance to aid in their decision making. To support ED physicians' key decision components (sensemaking, anticipation, and managing complexity) when managing CHD patients, a Clinical Decision Support System (CDSS) was previously designed. This pilot study evaluates the effect of this CDSS on ED physicians' decision making compared to usual care without clinical decision support.MethodsIn a pilot scenario-based simulation study with repeated measures, ED physicians managed mock CHD patients with and without the CDSS. We compared ED physicians' CHD-specific and general decision-making processes (e.g., recognizing sepsis, starting antibiotics, and managing symptoms) with and without the use of CDSS. The frequency of participants' utterances related to each key decision components of sensemaking, anticipation, and managing complexity were coded and statistically analyzed for significance.ResultsAcross all decision-making components, the CDSS significantly increased ED physicians' frequency of “CHD specific utterances” (Mean = 5.43, 95%CI: 3.7–7.2) compared to the without CDSS condition (Mean = 2.05, 95%CI: 0.3–3.8) whereas there was no significant difference in frequencies of “general utterances” when using CDSS (Mean = 4.62, 95%CI: 3.1–6.1) compared to without CDSS (Mean = 5.14 95%CI: 4.4–5.9).ConclusionA CDSS that integrates key decision-making components (sensemaking, anticipation, and managing complexity) can trigger and enrich communication between clinicians and enhance the clinical management of CHD patients. For patients with complex and subspecialized diseases such as CHD, a well-designed CDSS can become part of a multifaceted solution that includes knowledge translation, broader communication around interpretation of information, and access to additional expertise to support CHD specific decision-making.

Background and ObjectivesChildren with congenital heart disease (CHD), have fragile hemodynamics and can deteriorate due to common childhood illnesses and the natural progression of their disease. During these acute periods of deterioration, these children often present to their local emergency departments (ED) where expertise in CHD is limited, and appropriate intervention is crucial to their survival. Previous studies identified that determining the appropriate intervention for CHD patients can be difficult for ED physicians, particularly since key components of effective decision making are not being met. Although key components of effective decision making for ED physicians have been identified, they have yet to be transformed into actionable guidance. We used decision centered design (DCD) to translate key components of decision making into decision requirements and associated design concepts, that we subsequently incorporated into a prototype clinical decision support system (CDSS).MethodsUsing framework analysis, transcripts from Critical Decision Method interviews of CHD experts and ED physicians were inductively coded to identify key decision requirements for ED physicians that are currently not well supported, and their associated design concepts. A design workshop was held to refine the identified key decision requirements and design concepts as well as to sketch information that would satisfy the identified requirements. These were iteratively incorporated into a prototype CDSS.ResultsThree decision requirements: (1) distinguish the patient's unique physiology based on their unique cardiac anatomy, (2) explicitly consider CHD specific differential diagnoses to allow a more structured reflection of diagnosis, and (3) select CHD appropriate interventions for each patient, were identified. These requirements along with design concepts and information needs identified through the design workshop were incorporated into the CDSS prototype.ConclusionWe identified key decision requirements and associated design concepts, that informed the design of a CDSS to provide actionable guidance for ED physicians when managing CHD patients. Meeting ED physicians' decision components with a CDSS requires the translation of their key decision requirements in its design. If not, we risk creating designs that interfere with clinician performance.

Background and aims: Children with congenital heart disease (CHD) are at risk of deterioration in the face of common childhood illnesses, and their resuscitation and acute management is often best achieved with the guidance of CHD experts. Access to such expertise may be limited outside specialty heart centers and the fragility of these patients is cause for discomfort among many emergency medicine physicians. An understanding of the differences in macrocognition of these clinicians could shed light on some of the causes of discomfort and facilitate the development of a sociotechnological solution to this problem. Methods: Cardiac intensivists (CHD experts) and pediatric emergency medicine physicians (non-CHD experts) in a major academic cardiac center were interviewed using the critical decision method. Interview transcripts were coded deductively based on Klein’s macrocognitive framework and inductively to allow for new or modified characterization of dimensions. Results: While both CHD-experts and non-CHD experts relied on the macrocognitive functions of sensemaking, naturalistic decision making and detecting problems, the specific data and mental models used to understand the patients and course of therapy differed between CHD-experts and non-CHD experts. Conclusion: Characterization of differences between the macrocognitive processes of CHD experts and non-CHD experts can inform development of sociotechnological solutions to augment decision making pertaining to the acute management of pediatric CHD patients.

A firm concept of time is essential for establishing causality in a clinical setting. Review of critical incidents and generation of study hypotheses require a robust understanding of the sequence of events but conducting such work can be problematic when timestamps are recorded by independent and unsynchronized clocks. Most clinical models implicitly assume that timestamps have been measured accurately and precisely, but this custom will need to be re-evaluated if our algorithms and models are to make meaningful use of higher frequency physiological data sources. In this narrative review we explore factors that can result in timestamps being erroneously recorded in a clinical setting, with particular focus on systems that may be present in a critical care unit. We discuss how clocks, medical devices, data storage systems, algorithmic effects, human factors, and other external systems may affect the accuracy and precision of recorded timestamps. The concept of temporal uncertainty is introduced, and a holistic approach to timing accuracy, precision, and uncertainty is proposed. This quantitative approach to modeling temporal uncertainty provides a basis to achieve enhanced model generalizability and improved analytical outcomes.

Background and ObjectivesMachine Learning offers opportunities to improve patient outcomes, team performance, and reduce healthcare costs. Yet only a small fraction of all Machine Learning models for health care have been successfully integrated into the clinical space. There are no current guidelines for clinical model integration, leading to waste, unnecessary costs, patient harm, and decreases in efficiency when improperly implemented. Systems engineering is widely used in industry to achieve an integrated system of systems through an interprofessional collaborative approach to system design, development, and integration. We propose a framework based on systems engineering to guide the development and integration of Machine Learning models in healthcare.MethodsApplied systems engineering, software engineering and health care Machine Learning software development practices were reviewed and critically appraised to establish an understanding of limitations and challenges within these domains. Principles of systems engineering were used to develop solutions to address the identified problems. The framework was then harmonized with the Machine Learning software development process to create a systems engineering-based Machine Learning software development approach in the healthcare domain.ResultsWe present an integration framework for healthcare Artificial Intelligence that considers the entirety of this system of systems. Our proposed framework utilizes a combined software and integration engineering approach and consists of four phases: (1) Inception, (2) Preparation, (3) Development, and (4) Integration. During each phase, we present specific elements for consideration in each of the three domains of integration: The Human, The Technical System, and The Environment. There are also elements that are considered in the interactions between these domains.ConclusionClinical models are technical systems that need to be integrated into the existing system of systems in health care. A systems engineering approach to integration ensures appropriate elements are considered at each stage of model design to facilitate model integration. Our proposed framework is based on principles of systems engineering and can serve as a guide for model development, increasing the likelihood of successful Machine Learning translation and integration.

Background and ObjectivesChildren with congenital heart disease (CHD) are at risk of deterioration in the face of common childhood illnesses, and their resuscitation and acute treatment requires guidance of CHD experts. Many children with CHD, however, present to their local emergency departments (ED) with gastrointestinal and respiratory symptoms that closely mimic symptoms of CHD related heart failure. This can lead to incorrect or delayed diagnosis and treatment where CHD expertise is limited. An understanding of the differences in cognitive decision-making processes between CHD experts and ED physicians can inform how best to support ED physicians when treating CHD patients.MethodsCardiac intensivists (CHD experts) and pediatric emergency department physicians (ED physicians) in a major academic cardiac center were interviewed using the critical decision method. Interview transcripts were coded deductively based on Schubert and Klein's macrocognitive frameworks and inductively to allow for new or modified characterization of dimensions.ResultsIn total, 6 CHD experts and 7 ED physicians were interviewed for this study. Although both CHD experts and ED physicians spent a lot of time sensemaking, their approaches to sensemaking differed. CHD experts reported readily recognizing the physiology of complex congenital heart disease and focused primarily on ruling out cardiac causes for the presenting illness. ED physicians reported a delay in attributing the signs and symptoms of the presenting illness to congenital heart disease, because these clinical findings were often non-specific, and thus explored different diagnoses. CHD experts moved quickly to treatment and more time anticipating potential problems and making specific contingency plans, while ED physicians spent more time gathering a range of data prior to arriving at a diagnosis. These findings were then applied to develop a prototype web-based decision support application for patients with CHD.ConclusionThere are differences in the cognitive processes used by CHD experts and ED physicians when managing CHD patients. An understanding of differences in the cognitive processes used by CHD experts and ED physicians can inform the development of potential interventions, such as clinical decision support systems and training pathways, to support decision making pertaining to the acute treatment of pediatric CHD patients.

OBJECTIVES Physiologic signals are typically measured continuously in the critical care unit, but only recorded at intermittent time intervals in the patient health record. Low frequency data collection may not accurately reflect the variability and complexity of these signals or the patient's clinical state. We aimed to characterize how increasing the temporal window size of observation from seconds to hours modifies the measured variability and complexity of basic vital signs. DESIGN Retrospective analysis of signal data acquired between April 1, 2013, and September 30, 2015. SETTING Critical care unit at The Hospital for Sick Children, Toronto. PATIENTS Seven hundred forty-seven patients less than or equal to 18 years old (63,814,869 data values), within seven diagnostic/surgical groups. INTERVENTIONS None. MEASUREMENTS AND MAIN RESULTS Measures of variability (SD and the absolute differences) and signal complexity (multiscale sample entropy and detrended fluctuation analysis [expressed as the scaling component α]) were calculated for systolic blood pressure, heart rate, and oxygen saturation. The variability of all vital signs increases as the window size increases from seconds to hours at the patient and diagnostic/surgical group level. Significant differences in the magnitude of variability for all time scales within and between groups was demonstrated (p < 0.0001). Variability correlated negatively with patient age for heart rate and oxygen saturation, but positively with systolic blood pressure. Changes in variability and complexity of heart rate and systolic blood pressure from time of admission to discharge were found. CONCLUSIONS In critically ill children, the temporal variability of physiologic signals supports higher frequency data capture, and this variability should be accounted for in models of patient state estimation.

As part of the PhysioNet/Computing in Cardiology Challenge 2020, we developed a model for multilabel classification of 12-lead electrocardiogram (ECG) data according to specified cardiac abnormalities. Our team, LaussenLabs, developed a novel classifier pipeline with 6 core features (1) the addition of r-peak, p-wave, and t-wave features that were input into the model along with the 12-lead data, (2) data augmentation, (3) competition metric hacking, (4) modified WaveNet architecture, (5) Sigmoid threshold tuning, and (6) model stacking. Our approach received a score of 0.63 using 6-fold cross-validation on the full training data. Unfortunately, our model was unable to run on the test dataset due to time constraints, therefore, our model's final test score is undetermined.

Objective Storage of physiological waveform data for retrospective analysis presents significant challenges. Resultant data can be very large, and therefore becomes expensive to store and complicated to manage. Traditional database approaches are not appropriate for large scale storage of physiological waveforms. Our goal was to apply modern time series compression and indexing techniques to the problem of physiological waveform storage and retrieval. Approach We deployed a vendor-agnostic data collection system and developed domain-specific compression approaches that allowed long term storage of physiological waveform data and other associated clinical and medical device data. The database (called AtriumDB) also facilitates rapid retrieval of retrospective data for high-performance computing and machine learning applications. Main results A prototype system has been recording data in a 42-bed pediatric critical care unit at The Hospital for Sick Children in Toronto, Ontario since February 2016. As of December 2019, the database contains over 720,000 patient-hours of data collected from over 5300 patients, all with complete waveform capture. One year of full resolution physiological waveform storage from this 42-bed unit can be losslessly compressed and stored in less than 300 GB of disk space. Retrospective data can be delivered to analytical applications at a rate of up to 50 million time-value pairs per second. Significance Stored data are not pre-processed or filtered. Having access to a large retrospective dataset with realistic artefacts lends itself to the process of anomaly discovery and understanding. Retrospective data can be replayed to simulate a realistic streaming data environment where analytical tools can be rapidly tested at scale.