LiSep LSTM: A Machine Learning Algorithm for Early Detection of Septic Shock.

Sepsis is a major health concern with global estimates of 31.5 million cases per year. Case fatality rates are still unacceptably high, and early detection and treatment is vital since it significantly reduces mortality rates for this condition. Appropriately designed automated detection tools have the potential to reduce the morbidity and mortality of sepsis by providing early and accurate identification of patients who are at risk of developing sepsis. In this paper, we present "LiSep LSTM"; a Long Short-Term Memory neural network designed for early identification of septic shock. LSTM networks are typically well-suited for detecting long-term dependencies in time series data. LiSep LSTM was developed using the machine learning framework Keras with a Google TensorFlow back end. The model was trained with data from the Medical Information Mart for Intensive Care database which contains vital signs, laboratory data, and journal entries from approximately 59,000 ICU patients. We show that LiSep LSTM can outperform a less complex model, using the same features and targets, with an AUROC 0.8306 (95% confidence interval: 0.8236, 0.8376) and median offsets between prediction and septic shock onset up to 40 hours (interquartile range, 20 to 135 hours). Moreover, we discuss how our classifier performs at specific offsets before septic shock onset, and compare it with five state-of-the-art machine learning algorithms for early detection of sepsis.

A clinical decision support tool for improving adherence to guidelines on anticoagulant therapy in patients with atrial fibrillation at risk of stroke: A cluster-randomized trial in a Swedish primary care setting (the CDS-AF study).

BACKGROUND: Atrial fibrillation (AF) is associated with substantial morbidity, in particular stroke. Despite good evidence for the reduction of stroke risk with anticoagulant therapy, there remains significant undertreatment. The main aim of the current study was to investigate whether a clinical decision support tool (CDS) for stroke prevention integrated in the electronic health record could improve adherence to guidelines for stroke prevention in patients with AF. METHODS AND FINDINGS: We conducted a cluster-randomized trial where all 43 primary care clinics in the county of Östergötland, Sweden (population 444,347), were randomized to be part of the CDS intervention or to serve as controls. The CDS produced an alert for physicians responsible for patients with AF and at increased risk for thromboembolism (according to the CHA2DS2-VASc algorithm) without anticoagulant therapy. The primary endpoint was adherence to guidelines after 1 year. After randomization, there were 22 and 21 primary care clinics in the CDS and control groups, respectively. There were no significant differences in baseline adherence to guidelines regarding anticoagulant therapy between the 2 groups (CDS group 70.3% [5,186/7,370; 95% CI 62.9%-77.7%], control group 70.0% [4,187/6,009; 95% CI 60.4%-79.6%], p = 0.83). After 12 months, analysis with linear regression with adjustment for primary care clinic size and adherence to guidelines at baseline revealed a significant increase in guideline adherence in the CDS (73.0%, 95% CI 64.6%-81.4%) versus the control group (71.2%, 95% CI 60.8%-81.6%, p = 0.013, with a treatment effect estimate of 0.016 [95% CI 0.003-0.028]; number of patients with AF included in the final analysis 8,292 and 6,508 in the CDS and control group, respectively). Over the study period, there was no difference in the incidence of stroke, transient ischemic attack, or systemic thromboembolism in the CDS group versus the control group (49 [95% CI 43-55] per 1,000 patients with AF in the CDS group compared to 47 [95% CI 39-55] per 1,000 patients with AF in the control group, p = 0.64). Regarding safety, the CDS group had a lower incidence of significant bleeding, with events in 12 (95% CI 9-15) per 1,000 patients with AF compared to 16 (95% CI 12-20) per 1,000 patients with AF in the control group (p = 0.04). Limitations of the study design include that the analysis was carried out in a catchment area with a high baseline adherence rate, and issues regarding reproducibility to other regions. CONCLUSIONS: The present study demonstrates that a CDS can increase guideline adherence for anticoagulant therapy in patients with AF. Even though the observed difference was small, this is the first randomized study to our knowledge indicating beneficial effects with a CDS in patients with AF. TRIAL REGISTRATION: ClinicalTrials.gov NCT02635685.

Clinical decision support for stroke prevention in atrial fibrillation (CDS-AF): Rationale and design of a cluster randomized trial in the primary care setting.

BACKGROUND: Atrial fibrillation (AF) is associated with substantial morbidity, in particular stroke. Despite good evidence for the reduction of stroke risk with anticoagulant therapy, there remains a significant undertreatment. The main aim of the current study is to investigate whether a clinical decision support tool for stroke prevention (CDS) integrated in the electronic health record can improve adherence to guidelines for stroke prevention in patients with AF. METHODS: We will conduct a cluster randomized trial where 43 primary care clinics in the county of Östergötland, Sweden (population 444,347), will be randomized to be part of the CDS intervention or serve as controls. The CDS will alert responsible physicians of patients with AF and increased risk for thromboembolism according to the CHA2DS2VASc (Congestive heart failure, Hypertension, Age ≥ 74 years, Diabetes mellitus, previous Stroke/TIA/thromboembolism, Vascular disease, Age 65-74 years, Sex category (i.e. female sex)) algorithm without anticoagulant therapy. The primary end point will be adherence to guidelines after 1 year. CONCLUSION: The present study will investigate whether a clinical decision support system integrated in an electronic health record can increase adherence to guidelines regarding anticoagulant therapy in patients with AF.

Towards mHealth Systems for Support of Psychotherapeutic Practice: A Qualitative Study of Researcher-Clinician Collaboration in System Design and Evaluation.

We examined clinicians' and researchers' experiences from participation in collaborative research on the introduction of Internet and mobile information systems (mHealth systems) in psychotherapeutic routines. The study used grounded theory methodology and was set in a collaboration that aimed to develop and evaluate mHealth support of psychotherapy provided to young people. Soundness of the central objects developed in the design phase (the collaboration contract, the trial protocol, and the system technology) was a necessary foundation for successful collaborative mHealth research; neglect of unanticipated organizational influences during the trial phase was a factor in collaboration failure. The experiences gained in this study can be used in settings where collaborative research on mHealth systems in mental health is planned.

Studying distributed cognition of simulation-based team training with DiCoT.

Health care organizations employ simulation-based team training (SBTT) to improve skill, communication and coordination in a broad range of critical care contexts. Quantitative approaches, such as team performance measurements, are predominantly used to measure SBTTs effectiveness. However, a practical evaluation method that examines how this approach supports cognition and teamwork is missing. We have applied Distributed Cognition for Teamwork (DiCoT), a method for analysing cognition and collaboration aspects of work settings, with the purpose of assessing the methodology's usefulness for evaluating SBTTs. In a case study, we observed and analysed four Emergo Train System® simulation exercises where medical professionals trained emergency response routines. The study suggests that DiCoT is an applicable and learnable tool for determining key distributed cognition attributes of SBTTs that are of importance for the simulation validity of training environments. Moreover, we discuss and exemplify how DiCoT supports design of SBTTs with a focus on transfer and validity characteristics. Practitioner Summary: In this study, we have evaluated a method to assess simulation-based team training environments from a cognitive ergonomics perspective. Using a case study, we analysed Distributed Cognition for Teamwork (DiCoT) by applying it to the Emergo Train System®. We conclude that DiCoT is useful for SBTT evaluation and simulator (re)design.

The Nurse Watch: Design and Evaluation of a Smart Watch Application with Vital Sign Monitoring and Checklist Reminders.

Computerized wearable devices such as smart watches will become valuable nursing tools. This paper describes a smart-watch system developed in close collaboration with a team of nurses working in a Swedish ICU. The smart-watch system provides real-time vital-sign monitoring, threshold alarms, and to-do reminders. Additionally, a Kanban board, visualized on a multitouch screen provides an overview of completed and upcoming tasks. We describe an approach to implement automated checklist systems with smart watches and discuss aspects of importance when implementing such memory and attention support. The paper is finalized with an in-development formative evaluation of the system.

Cognitive design of a digital desk for the emergency room setting.

Digital desk technology has a still mainly unexplored potential to support the everyday work of collaborating clinicians. This paper presents ER Desk - a digital desk that was designed to specifically support a team of healthcare professionals working in an emergency room setting. The underlying design requirements were elicited in a comprehensive distributed cognition study of paper-based practices in an emergency room of a middle-sized Swedish hospital. We present the user interface and visualization requirements for digital desks for small clinical emergency room teams. Moreover, we discuss key design issues more generally with a focus on supporting team awareness, cognition, and collaborative routines of healthcare personnel working in clinical environments such as emergency rooms and intensive care units.

Cell phone-supported cognitive behavioural therapy for anxiety disorders: a protocol for effectiveness studies in frontline settings.

BACKGROUND: Reviews of randomized controlled trials (RCTs) of cognitive behavioural therapy (CBT) for anxiety disorders have reported large pre- to post-treatment within-group effect sizes on measures of anxiety when supplied in therapist consultations and in technology-supported settings. However, the stringent experimental control of RCTs results in a lack of external validity, which limits the generalizability of findings to real-world frontline clinical practice. We set out to examine the specification of a protocol for study of the effectiveness of cell phone-supported CBT for in situ management of anxiety disorders. METHODS AND DESIGN: Nominal group methods were used for requirements analysis and protocol design. Making a distinction between different forms of technology-supported therapy, examination of therapists' role, and implementing trials in existing organizational and community contexts were found to be the central requirements in the protocol. DISCUSSION: The resulting protocol (NCT01205191 at clinicaltrials.gov) for use in frontline clinical practice in which effectiveness, adherence, and the role of the therapists are analyzed, provides evidence for what are truly valuable cell phone-supported CBT treatments and guidance for the broader introduction of CBT in health services.

Mobile phone computing for in-situ cognitive behavioral therapy.

Cognitive behavioral therapy (CBT) for psychological disorders is becoming increasingly popular on the Internet. However, when using this workstation approach, components such as training and learning relaxation skills, problem solving, exposure exercises, and sleep management guidance must be done in the domestic environment. This paper describes design concepts for providing spatially explicit CBT with mobile phones. We reviewed and analyzed a set of treatment manuals to distinguish elements of CBT that can be improved and supported using mobile phone applications. The key advantage of mobile computing support in CBT is that multimedia can be applied to record, scale, and label anxiety-provoking situations where the need arises, which helps the CBT clients formulate and convey their thoughts and feelings to relatives and friends, as well as to therapists at subsequent treatment sessions.

Ubiquitous computing to support co-located clinical teams: using the semiotics of physical objects in system design.

OBJECTIVES: Co-located teams often use material objects to communicate messages in collaboration. Modern desktop computing systems with abstract graphical user interface (GUIs) fail to support this material dimension of inter-personal communication. The aim of this study is to investigate how tangible user interfaces can be used in computer systems to better support collaborative routines among co-located clinical teams. METHODS: The semiotics of physical objects used in team collaboration was analyzed from data collected during 1 month of observations at an emergency room. The resulting set of communication patterns was used as a framework when designing an experimental system. Following the principles of augmented reality, physical objects were mapped into a physical user interface with the goal of maintaining the symbolic value of those objects. RESULTS: NOSTOS is an experimental ubiquitous computing environment that takes advantage of interaction devices integrated into the traditional clinical environment, including digital pens, walk-up displays, and a digital desk. The design uses familiar workplace tools to function as user interfaces to the computer in order to exploit established cognitive and collaborative routines. CONCLUSION: Paper-based tangible user interfaces and digital desks are promising technologies for co-located clinical teams. A key issue that needs to be solved before employing such solutions in practice is associated with limited feedback from the passive paper interfaces.

Information infrastructure for inter-organizational mental health services: an actor network theory analysis of psychiatric rehabilitation.

In the supply of mental health services to communities, data and information are managed not only by clinical organizations, but also by welfare state agencies and charities. The aim of this study is to use methods of analysis from actor network theory to identify organizational interventions necessary for the development of an information infrastructure for inter-organizational mental health services. Data was collected in a project aimed at developing an information system that supports inter-organizational psychiatric rehabilitation in a Swedish municipality. Three organizational interventions were identified: an integrated service policy defined by the national government, a common legal framework allowing sharing of high-level client data, and commissioned support for local inter-agency workspaces. It is concluded that organizational interventions must be regarded when configuring an information infrastructure for mental health services. Organizational interventions should also routinely be addressed in systems design methods to be used in inter-organizational settings.

Distributed user interfaces for clinical ubiquitous computing applications.

OBJECTIVES: Ubiquitous computing with multiple interaction devices requires new interface models that support user-specific modifications to applications and facilitate the fast development of active workspaces. METHODS: We have developed NOSTOS, a computer-augmented work environment for clinical personnel to explore new user interface paradigms for ubiquitous computing. NOSTOS uses several devices such as digital pens, an active desk, and walk-up displays that allow the system to track documents and activities in the workplace. RESULTS: We present the distributed user interface (DUI) model that allows standalone applications to distribute their user interface components to several devices dynamically at run-time. This mechanism permit clinicians to develop their own user interfaces and forms to clinical information systems to match their specific needs. We discuss the underlying technical concepts of DUIs and show how service discovery, component distribution, events and layout management are dealt with in the NOSTOS system. CONCLUSION: Our results suggest that DUIs--and similar network-based user interfaces--will be a prerequisite of future mobile user interfaces and essential to develop clinical multi-device environments.

Design requirements for ubiquitous computing environments for healthcare professionals.

Ubiquitous computing environments can support clinical administrative routines in new ways. The aim of such computing approaches is to enhance routine physical work, thus it is important to identify specific design requirements. We studied healthcare professionals in an emergency room and developed the computer-augmented environment NOSTOS to support teamwork in that setting. NOSTOS uses digital pens and paper-based media as the primary input interface for data capture and as a means of controlling the system. NOSTOS also includes a digital desk, walk-up displays, and sensor technology that allow the system to track documents and activities in the workplace. We propose a set of requirements and discuss the value of tangible user interfaces for healthcare personnel. Our results suggest that the key requirements are flexibility in terms of system usage and seamless integration between digital and physical components. We also discuss how ubiquitous computing approaches like NOSTOS can be beneficial in the medical workplace.

NOSTOS: a paper-based ubiquitous computing healthcare environment to support data capture and collaboration.

In this paper, we present a new approach to clinical workplace computerization that departs from the window-based user interface paradigm. NOSTOS is an experimental computer-augmented work environment designed to support data capture and teamwork in an emergency room. NOSTOS combines multiple technologies, such as digital pens, walk-up displays, headsets, a smart desk, and sensors to enhance an existing paper-based practice with computer power. The physical interfaces allow clinicians to retain mobile paper-based collaborative routines and still benefit from computer technology. The requirements for the system were elicited from situated workplace studies. We discuss the advantages and disadvantages of augmenting a paper-based clinical work environment.