Neurocritical Care Informatics: Translating Raw Data into Bedside Action
()
About this ebook
Health care in the twenty-first century requires intensive use of technology in order to acquire and analyze data and manage and disseminate information. No area is more data intensive than the neurointensive care unit. Despite the massive amount of data, however, providers often lack interpretable and actionable information. This book reviews the concepts underlying the emerging field of neurocritical care informatics, with a focus on integrated data acquisition, linear and nonlinear processing, and innovative visualization in the ICU. Subjects addressed in individual chapters are thus wide ranging and encompassing, for example, multimodal and continuous EEG monitoring and data integration, display of data in the ICU, patient-centered clinical decision support, optimization of collaboration and workflow, and progress towards an “integrated medical environment”. All of the nine chapters have been written by international thought leaders in the field.
Related to Neurocritical Care Informatics
Related ebooks
Clinical Decision Support System: Fundamentals and Applications Rating: 0 out of 5 stars0 ratingsData-Driven Healthcare: Revolutionizing Patient Care with Data Science Rating: 0 out of 5 stars0 ratingsDeveloping Medical Apps and mHealth Interventions: A Guide for Researchers, Physicians and Informaticians Rating: 0 out of 5 stars0 ratingsGlioma Imaging: Physiologic, Metabolic, and Molecular Approaches Rating: 0 out of 5 stars0 ratingsHealthcare Information Technology for Cardiovascular Medicine: Telemedicine & Digital Health Rating: 0 out of 5 stars0 ratingsGenomic Medicine: A Practical Guide Rating: 0 out of 5 stars0 ratingsData Pulse: A Brief Tour of Artificial Intelligence in Healthcare Rating: 0 out of 5 stars0 ratingsCognitive Informatics: Reengineering Clinical Workflow for Safer and More Efficient Care Rating: 0 out of 5 stars0 ratingsChimerism: A Clinical Guide Rating: 0 out of 5 stars0 ratingsPrecision Medicine in Cancer Therapy Rating: 0 out of 5 stars0 ratingsThe Nursing Informatics Implementation Guide Rating: 0 out of 5 stars0 ratingsData Science Project Ideas, Methodology & Python Codes in Health Care Rating: 0 out of 5 stars0 ratingsPediatric Neuro-oncology Rating: 0 out of 5 stars0 ratingsEthics of Innovation in Neurosurgery Rating: 0 out of 5 stars0 ratingsHealth informatics: Improving patient care Rating: 3 out of 5 stars3/5Adult Critical Care Medicine: A Clinical Casebook Rating: 0 out of 5 stars0 ratingsPersonalized and Precision Medicine Informatics: A Workflow-Based View Rating: 0 out of 5 stars0 ratingsPathway Analysis for Drug Discovery: Computational Infrastructure and Applications Rating: 0 out of 5 stars0 ratingsPractical Oncologic Molecular Pathology: Frequently Asked Questions Rating: 0 out of 5 stars0 ratingsHealthcare and Artificial Intelligence Rating: 0 out of 5 stars0 ratingsTarget Volume Delineation for Pediatric Cancers Rating: 0 out of 5 stars0 ratingsAccelerated Path to Cures Rating: 0 out of 5 stars0 ratingsPractical Lymph Node and Bone Marrow Pathology: Frequently Asked Questions Rating: 0 out of 5 stars0 ratingsLogic Synthesis for Genetic Diseases: Modeling Disease Behavior Using Boolean Networks Rating: 0 out of 5 stars0 ratingsPrecision Medicine Rating: 0 out of 5 stars0 ratingsDeep Learning for Medical Decision Support Systems Rating: 0 out of 5 stars0 ratingsValue-Based Approaches to Spine Care: Sustainable Practices in an Era of Over-Utilization Rating: 0 out of 5 stars0 ratingsInterfacing Bioelectronics and Biomedical Sensing Rating: 0 out of 5 stars0 ratingsTelemedicine: Overview and Application in Pulmonary, Critical Care, and Sleep Medicine Rating: 0 out of 5 stars0 ratingsInnovations in Modern Endocrine Surgery Rating: 0 out of 5 stars0 ratings
Medical For You
Brain on Fire: My Month of Madness Rating: 4 out of 5 stars4/5The Diabetes Code: Prevent and Reverse Type 2 Diabetes Naturally Rating: 5 out of 5 stars5/5The Little Book of Hygge: Danish Secrets to Happy Living Rating: 4 out of 5 stars4/5Gut: The Inside Story of Our Body's Most Underrated Organ (Revised Edition) Rating: 4 out of 5 stars4/5The Emperor of All Maladies: A Biography of Cancer Rating: 5 out of 5 stars5/5Hidden Lives: True Stories from People Who Live with Mental Illness Rating: 4 out of 5 stars4/5Mating in Captivity: Unlocking Erotic Intelligence Rating: 4 out of 5 stars4/5Living Daily With Adult ADD or ADHD: 365 Tips o the Day Rating: 5 out of 5 stars5/5Women With Attention Deficit Disorder: Embrace Your Differences and Transform Your Life Rating: 5 out of 5 stars5/5"Cause Unknown": The Epidemic of Sudden Deaths in 2021 & 2022 Rating: 5 out of 5 stars5/5Adult ADHD: How to Succeed as a Hunter in a Farmer's World Rating: 4 out of 5 stars4/5Tight Hip Twisted Core: The Key To Unresolved Pain Rating: 4 out of 5 stars4/5The Amazing Liver and Gallbladder Flush Rating: 5 out of 5 stars5/5The White Coat Investor: A Doctor's Guide to Personal Finance and Investing Rating: 4 out of 5 stars4/5ATOMIC HABITS:: How to Disagree With Your Brain so You Can Break Bad Habits and End Negative Thinking Rating: 5 out of 5 stars5/5The Gene: An Intimate History Rating: 4 out of 5 stars4/5Woman: An Intimate Geography Rating: 4 out of 5 stars4/5Super Human: The Bulletproof Plan to Age Backward and Maybe Even Live Forever Rating: 4 out of 5 stars4/5A Letter to Liberals: Censorship and COVID: An Attack on Science and American Ideals Rating: 3 out of 5 stars3/5WomanCode: Perfect Your Cycle, Amplify Your Fertility, Supercharge Your Sex Drive, and Become a Power Source Rating: 4 out of 5 stars4/5This Is Going to Hurt: Secret Diaries of a Young Doctor Rating: 5 out of 5 stars5/5Ghost Boy: The Miraculous Escape of a Misdiagnosed Boy Trapped Inside His Own Body Rating: 4 out of 5 stars4/5I'm So Effing Tired: A Proven Plan to Beat Burnout, Boost Your Energy, and Reclaim Your Life Rating: 3 out of 5 stars3/5The Grieving Brain: The Surprising Science of How We Learn from Love and Loss Rating: 4 out of 5 stars4/5The Hormone Reset Diet: Heal Your Metabolism to Lose Up to 15 Pounds in 21 Days Rating: 4 out of 5 stars4/5
Reviews for Neurocritical Care Informatics
0 ratings0 reviews
Book preview
Neurocritical Care Informatics - Michael De Georgia
© Springer-Verlag GmbH Germany 2020
M. De Georgia, K. Loparo (eds.)Neurocritical Care Informaticshttps://doi.org/10.1007/978-3-662-59307-3_1
1. Information Technology in Critical Care
Michael De Georgia¹, ² , Farhad Kaffashi³ , Frank J. Jacono⁴, ⁵ and Kenneth Loparo⁶, ⁷
(1)
Center for Neurocritical Care, Neurological Institute, University Hospitals Cleveland Medical Center, Cleveland, OH, USA
(2)
Department of Neurology, Case Western Reserve University School of Medicine, Cleveland, OH, USA
(3)
Department of Electrical Engineering and Computer Science, Case Western Reserve University, Cleveland, OH, USA
(4)
Pulmonary, Critical Care, and Sleep Medicine, Department of Medicine, Case Western Reserve University School of Medicine, Cleveland, OH, USA
(5)
Division of Pulmonary, Louis Stoke VA Medical Center, Cleveland, OH, USA
(6)
ISSACS: Institute for Smart, Secure and Connected Systems, Cleveland, OH, USA
(7)
Department of Electrical Engineering and Computer Science, IoT Collaborative, Case Western Reserve University, Cleveland, OH, USA
Michael De Georgia (Corresponding author)
Email: [email protected]
Farhad Kaffashi
Email: [email protected]
Frank J. Jacono
Email: [email protected]
Kenneth Loparo
Email: [email protected]
Keywords
Critical careInformaticsMonitoringData acquisitionData analysisTechnologyComputers
Portions of this chapter have been previously published in De Georgia MA, Kaffashi F, Jacono FJ, Loparo KA, Information Technology in Critical Care: Review of Monitoring and Data Acquisition Systems for Patient Care and Research,
The Scientific World Journal, vol. 2015, Article ID 727694, 9 pages, 2015. https://doi.org/10.1155/2015/727694.
1.1 Computers in the ICU
Clinical information systems are now common in most hospitals. These systems have evolved along several parallel lines beginning, not surprisingly, in 1946 with the introduction of the Electronic Numerical Integrator and Computer (ENIAC), the first general-purpose computer. The size of a room and weighing in at 27 tons, ENIAC was developed to calculate missile trajectories for the US Army [1]. Five years later, International Business Machines (IBM) introduced the first commercially available computer, the Engineering Research Associates 1103. Because of their exorbitant cost, the use of early computers was limited to large corporations to help manage accounting. In the 1960s, academic institutions followed suit and began developing computer systems to streamline their growing business operations. A decade later, hospitals began to invest in EMR systems including the problem-oriented medical record (POMR) at the University of Vermont [2], the computer-stored ambulatory record (COSTAR) at Harvard [3], health evaluation through logical processing (HELP) at the University of Utah [4], and The Medical Record (TMR) at Duke University [5]. While Indiana’s Regenstrief record was one of the first systems for both the inpatient and outpatient settings [6], these early EMRs were rarely connected to the real-time data-intensive environment of the ICU. This was a world unto itself.
Shubin and Weil are credited with introducing the computer to the ICU in 1966 for the purpose of automatically collecting vital signs from the bedside monitor [7]. By connecting an IBM 1710 computer through an analog-to-digital converter to bedside devices, they were able to collect arterial and venous pressure, heart rate, temperature, and urinary output. This had been done before in the operating room, though not easily. Using a mechanical contraption, McKesson recorded tidal volumes, fraction of inspired oxygen, and blood pressure in 1934 [8]. The development of the computer (and particularly the microprocessor), however, made this onerous task much easier. Other early applications of computers in medicine included one of the first clinical decision support systems to aid in the diagnosis of hematologic disorders by Lipkin and colleagues [9], systems for respiratory monitoring by Stacy and Peters [10], and automation of blood transfusion after cardiac surgery by Sheppard and colleagues [11].
At the same time, microprocessors also began to appear in bedside medical devices. Integrating digital signal processors into devices led to improvements in the quality of their output data. Mechanically controlled ventilators were replaced with microprocessor-driven ones resulting in better programming and alarm capabilities. Coronary care units were established following the introduction of synchronous direct current cardioversion and, in 1967, when prophylactic lidocaine administration was shown to decrease the risk of ventricular fibrillation. The ability to intervene in ventricular arrhythmias after myocardial infarction spurred the development of continuous electrocardiographic (ECG) monitoring of the heart.
This experience with computers in academic institutions inspired vendors to offer commercial ICU computer systems such as the patient data management system (PDMS) from Hewlett-Packard; however, adoption was slow because the primitive user interfaces and complex menus were not suited to the fast pace of the ICU [12]. In the 1980s, automatic collection of heart rate and blood pressure became more advanced with data being presented in graphical displays that mimicked the familiar bedside flow sheet. The architecture also evolved from the locally contained model to the client/server model in which a workstation in the ICU (the client) interacted with a central computer housing patient data (the server) via a local area network (LAN). Navigational tools became more user friendly though the analytical capabilities remained limited [13]. Links to the fledgling hospital EMR systems were being made beginning with the computer system that handled admissions, discharges, and transfers (ADT) so that patient demographic data could be readily accessed.
In parallel to the ICU, computers were also being introduced in the 1980s into the operating room. Picking up where McKesson left off, Gravenstein and colleagues introduced computerized anesthesia records in 1986 [14], which allowed for more reliable collection, storage, and presentation of data during the perioperative period as well as provided basic record keeping functions (thus in their infancy, such systems were called anesthesia record keepers
). Still, as in the ICU, data from medical devices were rarely integrated with the other vital signs.
In the 1990s, ICU systems improved significantly with increased clinical functionality, more powerful databases, and Internet access. Web-based software used Web browsers to display the user interface and simple queries of cumulative patient data were supported. Vendors migrated the technology that had been developed for the OR, the ability to record and present continuous patient data as well as provide links to physician notes, nursing documentation, and laboratory and imaging data from the evolving EMR systems, thus creating large enterprise systems, now broadly referred to as clinical information systems (see Table 1.1 for a timeline of the development of computers in the ICU).
Table 1.1
Timeline of computers in the ICU
1.2 Clinical Information Systems
Several clinical information systems are commercially available today for the ICU and competition among vendors is intensifying. Frost & Sullivan has estimated that the annual US market for emergency, perioperative, and intensive care software solutions is currently approximately $842.2 million and is expected to reach $1.3 billion in 2015 [15]. No one company has a dominant share of the market and several have evolved over the last decade, through various acquisitions of smaller participants, to offer broad end-to-end platforms (GE Healthcare: Legacy Products [16]). For example, GE’s Centricity™ Critical Care system from GE HealthCare (Chalfont St. Giles, UK) is the culmination of the acquisitions of, among others, Marquette Medical Systems, a leading manufacturer of patient monitors; Instrumentarium, a manufacturer of mechanical ventilators and anesthesia equipment; and iPath, the basis of the Operating Room Management Information System. For the EMR side, GE also acquired MedicaLogic, a leading provider of outpatient digital health records, and IDX Systems, primarily a practice management and billing system. IDX was written utilizing MUMPS (Massachusetts General Hospital Utility Multi-Programming System), which currently forms the basis for EpicCare (Epic Systems Corporation, Verona, WI) and Veterans Health Information Systems and Technology Architecture (VistA). GE’s Centricity™ Critical Care is based on HL7 V3 and automatically collects data from monitors, ventilators, and medical devices and displays it in spreadsheets reminiscent of the typical paper ICU chart. Data are collected from medical devices through device interfaces that connect with GE’s Unity Interface Device (ID) network.
Philips Healthcare (Andover, MA) also has long history in the ICU with the introduction of the patient data management system in the early 1970s under the Hewlett-Packard brand. In the 1980s this became CareVue, and the most recent iteration is the IntelliVue Clinical Information Portfolio (ICIP) Critical Care. Like GE’s Centricity™, Philips’ ICIP Critical Care also evolved through a series of acquisitions (Philips: History and Timeline [17]). These included Agilent Technologies Healthcare Solutions Group, a leader in patient monitoring and critical care information management; Witt Biomedical Corporation, a leader in hemodynamic monitoring in catheterization laboratories; and Emergin, a developer of alarm management software. In 2008, Philips acquired TOMCAT Systems Ltd., a company that offers software for the collection of cardiac data and, that same year, acquired VISICU Inc., a provider of tele-ICU technology. On the EMR side, Philips also partnered in 2004 with Epic in order to provide end-to-end integration with electronic medical records. As with GE’s Centricity™, Philips’ ICIP Critical Care supports automatic (or manual) documentation of parametric data with time resolutions up to every 5 min. Philips Information Support Mart interfaces with the ICIP and provides a relational database that warehouses clinical information such as lab results, text notes, medications, and patient demographics that can be queried with special scripts (see MIMIC II below).
While modern clinical information systems do provide end-to-end platforms for the ICU, they remain limited in terms of functionality and high-resolution physiological data acquisition. The underlying actual waveform signals, for example, are not acquired or stored by either the GE’s Centricity™ system or Philips’ ICIP system. This is an important limitation of most commercially available enterprise systems today and is the result of a trade-off between the memory requirements of capturing all of the data (including the underlying waveform morphology) versus capturing only sufficient information to be useful clinically. Currently, no standards have been defined as to where that balance lies. Philips has developed its own proprietary solution for automated waveform data acquisition called the Research Data Exporter (RDE). This solution is restricted to research applications, does not acquire the data at the native sampling rate of the signal, and limits the number of waveforms that can be exported.
1.3 Medical Device Interoperability and Device Data Integration
Central to the growth of critical care has been the proliferation of monitoring technology and stand-alone medical devices. For example, a typical critically ill patient may undergo frequent or continuous monitoring of dozens of physiological parameters including blood pressure, heart rate, respiratory rate, systemic arterial oxygen saturation, tidal volume, peak inspiratory pressures, and body temperature. Patients with brain injury may undergo additional neuromonitoring of intracranial pressure (ICP), continuous electroencephalography (EEG), brain tissue oxygen tension, cerebral blood flow, and microdialysis parameters. An enormous amount of data is generated reflecting dynamic and complex physiology, dynamics that can only be understood by data integration and context. Most of these parameters, however, are generated from stand-alone devices that do not easily integrate with one another. Some connect directly into the bedside monitor, but many others do not (or do so incompletely meaning that not all the data is captured electronically). A lack of functional medical device interoperability is one of the most significant limitations in healthcare today. For example, more than 90% of hospitals recently surveyed by HIMSS use six or more types of medical devices, and only about a third integrate them with one another or with their EMRs (medical devices landscape: current and future adoption, integration with EMRs, and connectivity [18]).
In contrast to the plug-and-play
world of consumer electronics, most acute care medical devices are not designed to interoperate. Most devices have data output ports (analog, serial, USB, and Ethernet) for data acquisition, but there is no universally adopted standard that facilitates multimodal data acquisition and synchronization in a clinical setting; each one often uses its own communication protocol to transfer its data. The development and adoption of medical device standards to improve interoperability is ongoing. Although ISO/IEEE 11073 and ASTM F-2761 (Integrated Clinical Environment, ICE) are two applicable standards, the former has not been widely adopted and the latter is still relatively new (2009). Many groups are tackling the problem of interoperability on their own by developing the hardware and software interfaces that enable device connectivity. Connecting with analog data ports requires appropriate hardware interfaces, analog-to-digital (A/D) converters, and filters to eliminate aliasing due to a mismatch between sampling rate and frequency of the signal being acquired. It also requires that the data be properly scaled to the voltage range of the A/D converter (microvolts to millivolts) to maximize the resolution. Digital data is available from some devices through connection to serial (RS-232 or USB) or Ethernet (802.3) ports or using wireless (e.g., 802.11b/g or Bluetooth™) communications. Although these approaches provide the opportunity to individually interface with a variety of devices in the ICU, a system that provides comprehensive, cross-manufacturer medical device integration for the care of a single critically ill patient at the bedside is not available.
Several third party systems have recently emerged specifically to help facilitate this comprehensive data acquisition and integration. For example, Bedmaster XA™ (Excel Medical Electronics, Jupiter, Florida) is a product that can be used to collect medical device data with access through the hospital local area network. First introduced in 1999 to assist clinicians by automatically acquiring a patient’s vital signs from a GE/Marquette patient monitor, the current system works with both GE and Philips patient monitors acquiring parameter data (such as vital signs) from every 5 s to every hour. DataCaptor™ (Capsule Technologie, Paris, France) is another similar product that can be used to collect medical device data.
Time synchronization of the data is a critical feature for multimodal data acquisition from different devices and monitors. Without a master clock
ensuring that all the values and waveforms acquired at the same time line up
exactly in synch, interpreting the information and understanding the interrelationships is difficult, if not impossible. There are two issues. First, when data is being acquired from different devices, each with its own internal clock, the time stamps of data acquired simultaneously can all be different. Time synchronization is therefore necessary when simultaneous analog and digital data streams are acquired in order to align the data. Second, even when acquiring data from a single patient monitor, time drifting from natural degradation, daylight savings time, or incorrect adjustments made by the clinical staff need to be corrected. In the Bedmaster XA™ system, time synchronization is managed by the unity time feature that insures the accuracy of the time clocks on all GEMS devices connected to the unity network. Unity time functions in conjunction with an NTP (Network Time Protocol) server, as specified by the medical facility. Time clocks on all GEMS patient monitors connected to the unity network are automatically reset to the NTP server at a time interval selected by the hospital.
1.4 Data Acquisition and Integration Systems Developed for Research
Commercial off-the-shelf products do not support high-resolution physiological data acquisition, archiving, or annotation with bedside observations for clinical applications. Such systems have been developed in academic settings though mainly for clinical research. Because they are not open source, most of these systems are not readily available. This has resulted in considerable duplication of effort in software development for acquiring and archiving physiological data.
There have been a variety of efforts ranging from developing and testing of new mathematical and analytical tools to hardware/software solutions for patient data acquisition, archiving, and visualization. For example, Tsui and colleagues developed a system for acquiring, modeling, and predicting ICP in the ICU using wavelet transform analysis for feature extraction and recurrent neural networks to compute dynamic nonlinear models [19]. Smielewski and colleagues from Cambridge University have developed the intensive care monitoring (ICM+) system, a configurable software based on MATLAB™ (The Mathworks, Natick, MA) that allows real-time acquisition, archiving, and analysis of multimodal data that can then be displayed in several ways including simple trends, cross histograms, correlations, and spectral analysis charts. The software is intended for research so it stores the raw signals acquired from bedside monitors for subsequent reprocessing, thus providing the means of building a data repository for testing novel analytical methods [20].
Others have focused on multimodal data collection linked with clinical annotation. In London, Gade and his colleagues reported the development of the Improved Monitoring for Brain Dysfunction in Intensive Care and Surgery (IBIS) data library that contained continuous EEG signals, multimodal evoked potential recordings, and ECG [21]. The system captured trend data from patient monitors, laboratory data, and some clinical annotations. In 2001, Kropyvnytskyy and colleagues [22] reported a similar system in Boston initially developed for the MIT-BIH Arrhythmia Database (and now used for publicly available databases on the National Institutes of Health-sponsored PhysioNet Web site).
Sorani and colleagues from San Francisco General Hospital [23] also developed a system that captured over 20 physiological variables (plus date, time, and annotated clinical information) from Viridia bedside monitors (Philips), Licox tissue oxygen monitors (Integra Neurosciences, Plainsboro, NJ), and Draeger ventilators (Luebeck, Germany). Data was collected automatically at 1-min intervals and was output into text files. Monitoring data was integrated by special custom-developed middleware. Goldstein and colleagues from Oregon Health Sciences University developed a physiologic data acquisition system that could capture and archive parametric data (such as blood pressure and heart rate) along with the underlying waveforms to assess dynamic changes in the patient’s physiologic state [24]. The system consisted of a laptop computer, a standard PCMCIA serial card (Socket Communications, Newark, CA), RS232 serial interface cables, and custom software. The system acquired analog data from devices incorporating antialiasing filters along with analog-to-digital conversion. Parametric data was sampled at a rate of 0.98 Hz and continuous wave data either at 500 Hz (ECG) or 125 Hz (pressures, arterial saturation, and respiration). Communications with the monitoring devices were managed by special software and the collected signal data were sent to a patient data server and workstation where the files were archived. Standard analytical software packages, such as MATLAB™, facilitated advanced mathematical analyses including time and frequency domain methods and linear and nonlinear signal metrics. Annotation of important clinical events, such as changes in a patient’s condition or timing of drug administration, was limited. In 2006, the same group reported the next generation of their system that added event markers and clinical annotation [25].
Moody and colleagues from Massachusetts General Hospital initially reported on their initial efforts in developing the MIMIC (Multiparameter Intelligent Monitoring for Intensive Care) database [26]. An updated version, MIMIC II, was reported in 2002 [27]. Each record consisted of four continuously monitored waveforms (two leads of ECG, arterial blood pressure, and pulmonary artery pressure) sampled at 125 Hz, along with other basic parameters (heart rate, oxygen saturation, cardiac output) collected every minute. The waveforms and parameters were originally acquired from Philips bedside patient monitors using their RDE software tool. Using a customized archiving agent, the waveform and parameter data were later stored onto storage drives and converted from Philips’ proprietary format into an open-source format (WFDB), thereby making it accessible to others for research. Various tools were used to analyze data. The 1-min parameter data were processed using wavelet analysis to identify potentially relevant clinical events. Matching waveform records to clinical data was based on unique identifiers such as medical record numbers, dates of admission, and patient names. A text search engine was created to allow users to query the database for key words and patterns of interest. In 2011, this MIMIC II database was made public, available for research [28].
While MIMIC II represents a major achievement, because physiological data and clinical annotations are collected separately, the two datasets are poorly synchronized. Also, physiological data and clinical annotations have different time granularity,
making it difficult to retrospectively determine the timing of a clinical