The promises and challenges of clinical AI in community paediatric medicine.

The widespread adoption of virtual care technologies has quickly reshaped healthcare operations and delivery, particularly in the context of community medicine. In this paper, we use the virtual care landscape as a point of departure to envision the promises and challenges of artificial intelligence (AI) in healthcare. Our analysis is directed towards community care practitioners interested in learning more about how AI can change their practice along with the critical considerations required to integrate AI into their practice. We highlight examples of how AI can enable access to new sources of clinical data while augmenting clinical workflows and healthcare delivery. AI can help optimize how and when care is delivered by community practitioners while also improving practice efficiency, accessibility, and the overall quality of care. Unlike virtual care, however, AI is still missing many of the key enablers required to facilitate adoption into the community care landscape and there are challenges we must consider and resolve for AI to successfully improve healthcare delivery. We discuss several critical considerations, including data governance in the clinic setting, healthcare practitioner education, regulation of AI in healthcare, clinician reimbursement, and access to both technology and the internet.

An integration engineering framework for machine learning in healthcare.

Background and Objectives: Machine 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. Methods: Applied 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. Results: We 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. Conclusion: Clinical 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.

Regulating the Safety of Health-Related Artificial Intelligence.

This article analyzes whether Canada's present approach to regulating health-related artificial intelligence (AI) can address relevant safety-related challenges. Focusing primarily on Health Canada's regulation of medical devices with AI, it examines whether the existing regulatory approach can adequately address general safety concerns, as well as those related to algorithmic bias and challenges posed by the intersections of these concerns with privacy and security interests. It identifies several issues and proposes reforms that aim to ensure that Canadians can access beneficial AI while keeping unsafe products off Canadian markets and motivating safe, effective use of AI products for appropriate purposes and populations.

Legal concerns in health-related artificial intelligence: a scoping review protocol.

BACKGROUND: Medical innovations offer tremendous hope. Yet, similar innovations in governance (law, policy, ethics) are likely necessary if society is to realize medical innovations' fruits and avoid their pitfalls. As innovations in artificial intelligence (AI) advance at a rapid pace, scholars across multiple disciplines are articulating concerns in health-related AI that likely require legal responses to ensure the requisite balance. These scholarly perspectives may provide critical insights into the most pressing challenges that will help shape and advance future regulatory reforms. Yet, to the best of our knowledge, there is no comprehensive summary of the literature examining legal concerns in relation to health-related AI. We thus aim to summarize and map the literature examining legal concerns in health-related AI using a scoping review approach. METHODS: The scoping review framework developed by (J Soc Res Methodol 8:19-32, 2005) and extended by (Implement Sci 5:69, 2010) and the Preferred Reporting Items for Systematic Reviews and Meta-Analysis extension for scoping reviews (PRISMA-ScR) guided our protocol development. In close consultation with trained librarians, we will develop a highly sensitive search for MEDLINE® (OVID) and adapt it for multiple databases designed to comprehensively capture texts in law, medicine, nursing, pharmacy, other healthcare professions (e.g., dentistry, nutrition), public health, computer science, and engineering. English- and French-language records will be included if they examine health-related AI, describe or prioritize a legal concern in health-related AI or propose a solution thereto, and were published in 2012 or later. Eligibility assessment will be conducted independently and in duplicate at all review stages. Coded data will be analyzed along themes and stratified across discipline-specific literatures. DISCUSSION: This first-of-its-kind scoping review will summarize available literature examining, documenting, or prioritizing legal concerns in health-related AI to advance law and policy reform(s). The review may also reveal discipline-specific concerns, priorities, and proposed solutions to the concerns. It will thereby identify priority areas that should be the focus of future reforms and regulatory options available to stakeholders in reform processes. TRIAL REGISTRATION: This protocol was submitted to the Open Science Foundation registration database. See https://osf.io/zav7w .

Assessment of Machine Learning-Based Medical Directives to Expedite Care in Pediatric Emergency Medicine.

Importance: Increased wait times and long lengths of stay in emergency departments (EDs) are associated with poor patient outcomes. Systems to improve ED efficiency would be useful. Specifically, minimizing the time to diagnosis by developing novel workflows that expedite test ordering can help accelerate clinical decision-making. Objective: To explore the use of machine learning-based medical directives (MLMDs) to automate diagnostic testing at triage for patients with common pediatric ED diagnoses. Design, Setting, and Participants: Machine learning models trained on retrospective electronic health record data were evaluated in a decision analytical model study conducted at the ED of the Hospital for Sick Children Toronto, Canada. Data were collected on all patients aged 0 to 18 years presenting to the ED from July 1, 2018, to June 30, 2019 (77 219 total patient visits). Exposure: Machine learning models were trained to predict the need for urinary dipstick testing, electrocardiogram, abdominal ultrasonography, testicular ultrasonography, bilirubin level testing, and forearm radiographs. Main Outcomes and Measures: Models were evaluated using area under the receiver operator curve, true-positive rate, false-positive rate, and positive predictive values. Model decision thresholds were determined to limit the total number of false-positive results and achieve high positive predictive values. The time difference between patient triage completion and test ordering was assessed for each use of MLMD. Error rates were analyzed to assess model bias. In addition, model explainability was determined using Shapley Additive Explanations values. Results: There was a total of 42 238 boys (54.7%) included in model development; mean (SD) age of the children was 5.4 (4.8) years. Models obtained high area under the receiver operator curve (0.89-0.99) and positive predictive values (0.77-0.94) across each of the use cases. The proposed implementation of MLMDs would streamline care for 22.3% of all patient visits and make test results available earlier by 165 minutes (weighted mean) per affected patient. Model explainability for each MLMD demonstrated clinically relevant features having the most influence on model predictions. Models also performed with minimal to no sex bias. Conclusions and Relevance: The findings of this study suggest the potential for clinical automation using MLMDs. When integrated into clinical workflows, MLMDs may have the potential to autonomously order common ED tests early in a patient's visit with explainability provided to patients and clinicians.

Automatically disambiguating medical acronyms with ontology-aware deep learning.

Modern machine learning (ML) technologies have great promise for automating diverse clinical and research workflows; however, training them requires extensive hand-labelled datasets. Disambiguating abbreviations is important for automated clinical note processing; however, broad deployment of ML for this task is restricted by the scarcity and imbalance of labeled training data. In this work we present a method that improves a model's ability to generalize through novel data augmentation techniques that utilizes information from biomedical ontologies in the form of related medical concepts, as well as global context information within the medical note. We train our model on a public dataset (MIMIC III) and test its performance on automatically generated and hand-labelled datasets from different sources (MIMIC III, CASI, i2b2). Together, these techniques boost the accuracy of abbreviation disambiguation by up to 17% on hand-labeled data, without sacrificing performance on a held-out test set from MIMIC III.

Personalized Ventilation to Multiple Patients Using a Single Ventilator: Description and Proof of Concept.

OBJECTIVES: To design and test a ventilator circuit that can be used for ventilation of two or more patients with a single ventilator, while allowing individualization of tidal volume, fractional concentration of oxygen, and positive end-expiratory pressure to each patient, irrespective of the other patient's respiratory system mechanics. DESIGN: Description and proof of concept studies. SETTINGS: Respiratory therapy laboratory. SUBJECTS: Ventilation of mechanical test lungs. INTERVENTIONS: Following a previously advocated design, we used components readily available in our hospital to assemble two "bag-in-a-box" breathing circuits. Each patient circuit consisted of a flexible bag in a rigid container connected via one-way valve to a test lung, along with an inline positive end-expiratory pressure valve, connected to the ventilator's expiratory limb. Compressed gas fills the bags during "patient" exhalation. During inspiration, gas from the ventilator, in pressure control mode, enters the containers and displaces gas from the bags to the test lungs. We varied tidal volume, "respiratory system" compliance, and positive end-expiratory pressure in one lung and observed the effect on the tidal volume of the other. MEASUREMENTS AND MAIN RESULTS: We were able to obtain different tidal volume, dynamic driving pressure, and positive end-expiratory pressure in the two lungs under widely different compliances in both lungs. Complete obstruction, or disconnection at the circuit connection to one test lung, had minimal effect (< 5% on average) on the ventilation to the co-ventilated lung. CONCLUSIONS: A secondary circuit "bag-in-the-box" system enables individualized ventilation of two lungs overcoming many of the concerns of ventilating more than one patient with a single ventilator.

From Clinic to Computer and Back Again: Practical Considerations When Designing and Implementing Machine Learning Solutions for Pediatrics.

Purpose of review: Machine learning (ML), a branch of artificial intelligence, is influencing all fields in medicine, with an abundance of work describing its application to adult practice. ML in pediatrics is distinctly unique with clinical, technical, and ethical nuances limiting the direct translation of ML tools developed for adults to pediatric populations. To our knowledge, no work has yet focused on outlining the unique considerations that need to be taken into account when designing and implementing ML in pediatrics. Recent findings: The nature of varying developmental stages and the prominence of family-centered care lead to vastly different data-generating processes in pediatrics. Data heterogeneity and a lack of high-quality pediatric databases further complicate ML research. In order to address some of these nuances, we provide a common pipeline for clinicians and computer scientists to use as a foundation for structuring ML projects, and a framework for the translation of a developed model into clinical practice in pediatrics. Throughout these pathways, we also highlight ethical and legal considerations that must be taken into account when working with pediatric populations and data. Summary: Here, we describe a comprehensive outline of special considerations required of ML in pediatrics from project ideation to implementation. We hope this review can serve as a high-level guideline for ML scientists and clinicians alike to identify applications in the pediatric setting, generate effective ML solutions, and subsequently deliver them to patients, families, and providers.

Doccurate: A Curation-Based Approach for Clinical Text Visualization.

Before seeing a patient, physicians seek to obtain an overview of the patient's medical history. Text plays a major role in this activity since it represents the bulk of the clinical documentation, but reviewing it quickly becomes onerous when patient charts grow too large. Text visualization methods have been widely explored to manage this large scale through visual summaries that rely on information retrieval algorithms to structure text and make it amenable to visualization. However, the integration with such automated approaches comes with a number of limitations, including significant error rates and the need for healthcare providers to fine-tune algorithms without expert knowledge of their inner mechanics. In addition, several of these approaches obscure or substitute the original clinical text and therefore fail to leverage qualitative and rhetorical flavours of the clinical notes. These drawbacks have limited the adoption of text visualization and other summarization technologies in clinical practice. In this work we present Doccurate, a novel system embodying a curation-based approach for the visualization of large clinical text datasets. Our approach offers automation auditing and customizability to physicians while also preserving and extensively linking to the original text. We discuss findings of a formal qualitative evaluation conducted with 6 domain experts, shedding light onto physicians' information needs, perceived strengths and limitations of automated tools, and the importance of customization while balancing efficiency. We also present use case scenarios to showcase Doccurate's envisioned usage in practice.

A Critical Analysis of Anesthesiology Podcasts: Identifying Determinants of Success.

BACKGROUND: Audio and video podcasts have gained popularity in recent years. Increasingly, podcasts are being used in the field of medicine as a tool to disseminate information. This format has multiple advantages including highly accessible creation tools, low distribution costs, and portability for the user. However, despite its ongoing use in medical education, there are no data describing factors associated with the success or quality of podcasts. OBJECTIVE: The goal of the study was to assess the landscape of anesthesia podcasts in Canada and develop a methodology for evaluating the quality of the podcast. To achieve our objective, we identified the scope of podcasts in anesthesia specifically, constructed an algorithmic model for measuring success, and identified factors linked to both successful podcasts and a peer-review process. METHODS: Independent reviewers performed a systematic search of anesthesia-related podcasts on iTunes Canada. Data and metrics recorded for each podcast included podcast's authorship, number posted, podcast series duration, target audience, topics, and social media presence. Descriptive statistics summarized mined data, and univariate analysis was used to identify factors associated with podcast success and a peer-review process. RESULTS: Twenty-two podcasts related to anesthesia were included in the final analysis. Less than a third (6/22=27%) were still active. The median longevity of the podcasts' series was just 13 months (interquartile range: 1-39 months). Anesthesiologists were the target audience for 77% of podcast series with clinical topics being most commonly addressed. We defined a novel algorithm for measuring success: Podcast Success Index. Factors associated with a high Podcast Success Index included podcasts targeting fellows (Spearman R=0.434; P=.04), inclusion of professional topics (Spearman R=0.456-0.603; P=.01-.03), and the use of Twitter as a means of social media (Spearman R=0.453;P=.03). In addition, more than two-thirds (16/22=73%) of podcasts demonstrated evidence of peer review with podcasts targeting anesthesiologists most strongly associated with peer-reviewed podcasts (Spearman R=0.886; P=.004) CONCLUSIONS: We present the first report on the scope of anesthesia podcasts in Canada. We have developed a novel tool for assessing the success of an anesthesiology podcast series and identified factors linked to this success measure as well as evidence of a peer-review process for a given podcast. To enable advancement in this area of anesthesia e-resources, podcast creators and users should consider factors associated with success when creating podcasts. The lack of these aspects may be associated with the early demise of a podcast series.

A Biomechanical Assessment of Kyphoplasty as a Stand-Alone Treatment in a Human Cadaveric Burst Fracture Model.

STUDY DESIGN: In vitro biomechanics study. OBJECTIVE: To determine whether kyphoplasty is an adequate stand-alone treatment for restoring biomechanical stability in the spine after experiencing high-energy vertebral burst fractures. SUMMARY OF BACKGROUND DATA: Kyphoplasty in the treatment of high-energy vertebral burst fractures has been shown by previous studies to significantly improve stiffness when used in conjunction with pedicle screw instrumentation. However, it is not known whether kyphoplasty as a stand-alone treatment may be an acceptable method for restoring biomechanical stability of a spinal motion segment post-burst fracture while allowing flexibility of the motion segment through the intervertebral discs. METHODS: Young cadaveric spines (15-50 yr old; 3 males and 1 female; bone mineral density 0.27-0.31 gHA/cm) were divided into motion segments consisting of 3 intact vertebrae separated by 2 intervertebral discs (T11-L1 and L2-L4). Mechanical testing in axial, flexion/extension, lateral bending, and torsion was performed on each specimen in an intact state, after an experimentally simulated burst fracture and postkyphoplasty. Computed tomography was used to confirm the burst fractures and quantify cement placement. RESULTS: Between the intact and burst-fractured states significant decreases in stiffness were seen in all loading modes (63%-69%). Burst fracture increased the average angulation of the vertebral endplates 147% and decreased vertebral body height by an average of 40%. Postkyphoplasty, only small recoveries in stiffness were seen in axial, flexion/extension, and lateral bending (4%-12%), with no improvement in torsional stiffness. Large angular deformations (85%) and height loss (31%) remained postkyphoplasty as compared with the intact state. CONCLUSION: Lack of overall improvement in biomechanical stiffness indicates failure of kyphoplasty to sufficiently restore stability as a stand-alone treatment after high-energy burst fracture. The lack of stability can be explained by an inability to biomechanically repair the compromised intervertebral discs. LEVEL OF EVIDENCE: 3.

Determining the best graft-sealant combination for skull base repair using a soft tissue in vitro porcine model.

BACKGROUND: The purpose of this work was to compare the absolute breaking strength of various soft tissue skull base (SB) repairs in an in vitro porcine model. METHODS: A burst pressure (BP) testing system was designed using an axial loading force to create increasing hydraulic pressure. Defects measuring 0.5 × 0.5 cm were created in fascia lata samples. Defects were repaired using various grafts (pericranium and 2 different dural substitutes, Alloderm(®) and Durasis(®)) measuring 1.0 × 1.0 cm to cover the deficient area. Grafts were further reinforced onto the fascia background with either fibrin glue (Tisseel(®)) or hydrogel sealant (DuraSeal™). Each combination of graft and sealant was conducted 6 times and tested 24 hours after the repair. RESULTS: The mean BP (±standard deviation [SD]) were as follows: DuraSeal™-Alloderm, 12.5 ± 5.8 mmHg; DuraSeal™-Durasis, 21.8 ± 20.7 mmHg; DuraSeal™-pericranium, 44.7 ± 30.1 mmHg; Tisseel-Alloderm, 30.6 ± 26.3 mmHg; Tisseel-Durasis, 15.8 ± 18.6 mmHg; and Tisseel-pericranium, 95.5 ± 86 mmHg. One-way analysis of variance showed that the strongest type of repair was Tisseel-pericranium when adjusting for the others (p < 0.0001). The difference in mean BP of repair with DuraSeal™ vs Tisseel(®) was not statistically significant (p = 0.22). Comparing sealants, the use of Alloderm(®) or Durasis(®) decreased the strength of the repair in comparison to pericranium (p < 0.0001). Bonferroni analysis showed a significant difference between pericranium and Alloderm(®) (p < 0.05) and between pericranium and Durasis(®) (p < 0.05) but not between Alloderm(®) and Durasis(®) (p > 0.05). CONCLUSION: In this model, the strongest type of repair (pressure 6 times higher than normal intracranial pressure) was the combination of Tisseel(®)-pericranium. Our data will help guide surgeons who repair SB defects to choose the best graft and sealant.

Can a semi-automated surface matching and principal axis-based algorithm accurately quantify femoral shaft fracture alignment in six degrees of freedom?

Accurate alignment of femoral shaft fractures treated with intramedullary nailing remains a challenge for orthopaedic surgeons. The aim of this study is to develop and validate a cone-beam CT-based, semi-automated algorithm to quantify the malalignment in six degrees of freedom (6DOF) using a surface matching and principal axes-based approach. Complex comminuted diaphyseal fractures were created in nine cadaveric femora and cone-beam CT images were acquired (27 cases total). Scans were cropped and segmented using intensity-based thresholding, producing superior, inferior and comminution volumes. Cylinders were fit to estimate the long axes of the superior and inferior fragments. The angle and distance between the two cylindrical axes were calculated to determine flexion/extension and varus/valgus angulation and medial/lateral and anterior/posterior translations, respectively. Both surfaces were unwrapped about the cylindrical axes. Three methods of matching the unwrapped surface for determination of periaxial rotation were compared based on minimizing the distance between features. The calculated corrections were compared to the input malalignment conditions. All 6DOF were calculated to within current clinical tolerances for all but two cases. This algorithm yielded accurate quantification of malalignment of femoral shaft fractures for fracture gaps up to 60 mm, based on a single CBCT image of the fractured limb.