The Unspeakable Nature of Death & Dying During Childhood: A Silenced Phenomenon in Pediatric Care

In pediatric settings, the concept of hope is frequently positioned as a fundamental aspect of care and at odds with the possibility and proximity of death. This arguably fosters silence about death and dying in childhood despite evidence indicating the benefits of open communication at the end of life. In this paper, we describe the unspeakable nature of death and dying in childhood, including its conceptual and clinical causes and dimensions, its persistence, and the associated challenges for children and youth facing critical illnesses, their families, and society. We explore how the tension between hope and death can be reframed and apply our analysis to the context of medical assistance in dying for mature minors in Canada. Considering the lack of related literature, this paper offers initial reflections to form a framework for the unspeakable nature of death and dying in childhood and to advance the crucial need for research.

Optimizing warfarin dosing for patients with atrial fibrillation using machine learning

While novel oral anticoagulants are increasingly used to reduce risk of stroke in patients with atrial fibrillation, vitamin K antagonists such as warfarin continue to be used extensively for stroke prevention across the world. While effective in reducing the risk of strokes, the complex pharmacodynamics of warfarin make it difficult to use clinically, with many patients experiencing under- and/or over- anticoagulation. In this study we employed a novel implementation of deep reinforcement learning to provide clinical decision support to optimize time in therapeutic International Normalized Ratio (INR) range. We used a novel semi-Markov decision process formulation of the Batch-Constrained deep Q-learning algorithm to develop a reinforcement learning model to dynamically recommend optimal warfarin dosing to achieve INR of 2.0-3.0 for patients with atrial fibrillation. The model was developed using data from 22,502 patients in the warfarin treated groups of the pivotal randomized clinical trials of edoxaban (ENGAGE AF-TIMI 48), apixaban (ARISTOTLE) and rivaroxaban (ROCKET AF). The model was externally validated on data from 5730 warfarin-treated patients in a fourth trial of dabigatran (RE-LY) using multilevel regression models to estimate the relationship between center-level algorithm consistent dosing, time in therapeutic INR range (TTR), and a composite clinical outcome of stroke, systemic embolism or major hemorrhage. External validation showed a positive association between center-level algorithm-consistent dosing and TTR (R2 = 0.56). Each 10% increase in algorithm-consistent dosing at the center level independently predicted a 6.78% improvement in TTR (95% CI 6.29, 7.28; p < 0.001) and a 11% decrease in the composite clinical outcome (HR 0.89; 95% CI 0.81, 1.00; p = 0.015). These results were comparable to those of a rules-based clinical algorithm used for benchmarking, for which each 10% increase in algorithm-consistent dosing independently predicted a 6.10% increase in TTR (95% CI 5.67, 6.54, p < 0.001) and a 10% decrease in the composite outcome (HR 0.90; 95% CI 0.83, 0.98, p = 0.018). Our findings suggest that a deep reinforcement learning algorithm can optimize time in therapeutic range for patients taking warfarin. A digital clinical decision support system to promote algorithm-consistent warfarin dosing could optimize time in therapeutic range and improve clinical outcomes in atrial fibrillation globally.

Machine Learning Approaches for the Image-Based Identification of Surgical Wound Infections: Scoping Review

BACKGROUND

Surgical site infections (SSIs) occur frequently and impact patients and health care systems. Remote surveillance of surgical wounds is currently limited by the need for manual assessment by clinicians. Machine learning (ML)-based methods have recently been used to address various aspects of the postoperative wound healing process and may be used to improve the scalability and cost-effectiveness of remote surgical wound assessment.

OBJECTIVE

The objective of this review was to provide an overview of the ML methods that have been used to identify surgical wound infections from images.

METHODS

We conducted a scoping review of ML approaches for visual detection of SSIs following the JBI (Joanna Briggs Institute) methodology. Reports of participants in any postoperative context focusing on identification of surgical wound infections were included. Studies that did not address SSI identification, surgical wounds, or did not use image or video data were excluded. We searched MEDLINE, Embase, CINAHL, CENTRAL, Web of Science Core Collection, IEEE Xplore, Compendex, and arXiv for relevant studies in November 2022. The records retrieved were double screened for eligibility. A data extraction tool was used to chart the relevant data, which was described narratively and presented using tables. Employment of TRIPOD (Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis) guidelines was evaluated and PROBAST (Prediction Model Risk of Bias Assessment Tool) was used to assess risk of bias (RoB).

RESULTS

In total, 10 of the 715 unique records screened met the eligibility criteria. In these studies, the clinical contexts and surgical procedures were diverse. All papers developed diagnostic models, though none performed external validation. Both traditional ML and deep learning methods were used to identify SSIs from mostly color images, and the volume of images used ranged from under 50 to thousands. Further, 10 TRIPOD items were reported in at least 4 studies, though 15 items were reported in fewer than 4 studies. PROBAST assessment led to 9 studies being identified as having an overall high RoB, with 1 study having overall unclear RoB.

CONCLUSIONS

Research on the image-based identification of surgical wound infections using ML remains novel, and there is a need for standardized reporting. Limitations related to variability in image capture, model building, and data sources should be addressed in the future.

Maximizing the value of patient and public involvement in the digital health co-design process: A qualitative descriptive study with design leaders and patient-public partners

Digital health interventions have enormous potential to support patients and the public in achieving their health goals. Nonetheless, many digital health interventions are failing to effectively engage patients and the public. One solution that has been proposed is to directly involve patients and the public in the design process of these digital health interventions. Although there is consensus that involving patients and the public in collaborative design is valuable, design teams have little guidance on how to maximize the value of their collaborative design work. The main objective of this study was to understand how the value of patient and public involvement in digital health design can be maximized, from the perspective of design leaders and patient-public partners. Using a qualitative descriptive methodology, we conducted semi-structured interviews with 19 design leaders and 9 patient-public partners. Interviewees agreed that involving patients and the public was valuable, however, they questioned if current collaborative methods were optimized to ensure maximal value. Interviewees suggested that patient and public collaborative design can add value through four different mechanisms: (1) by allowing the design process to be an empowering intervention itself, (2) by ensuring that the digital health intervention will be effectively engaging for users, (3) by ensuring that the digital health intervention will be seamlessly implemented in practice, and (4) by allowing patient-public collaborations extend beyond the initial product design. Overall, interviewees emphasized that although collaborative design has historically focused on improving the digital health product itself, patients and the public have crucial insights on implementation planning as well as how collaborative design can be used as its own empowering intervention. The results of this paper provide clarity about the ways that patient and public collaborative design can be made more valuable. Digital health design teams can use these results to be more intentional about their collaborative design approaches.

A versatile, fast and unbiased method for estimation of gene-by-environment interaction effects on biobank-scale datasets

Identification of gene-by-environment interactions (GxE) is crucial to understand the interplay of environmental effects on complex traits. However, current methods evaluating GxE on biobank-scale datasets have limitations. We introduce MonsterLM, a multiple linear regression method that does not rely on model specification and provides unbiased estimates of variance explained by GxE. We demonstrate robustness of MonsterLM through comprehensive genome-wide simulations using real genetic data from 325,989 individuals. We estimate GxE using waist-to-hip-ratio, smoking, and exercise as the environmental variables on 13 outcomes (N = 297,529-325,989) in the UK Biobank. GxE variance is significant for 8 environment-outcome pairs, ranging from 0.009 - 0.071. The majority of GxE variance involves SNPs without strong marginal or interaction associations. We observe modest improvements in polygenic score prediction when incorporating GxE. Our results imply a significant contribution of GxE to complex trait variance and we show MonsterLM to be well-purposed to handle this with biobank-scale data.

Optimizing Patient Record Linkage in a Master Patient Index Using Machine Learning: Algorithm Development and Validation

BACKGROUND

To provide quality care, modern health care systems must match and link data about the same patient from multiple sources, a function often served by master patient index (MPI) software. Record linkage in the MPI is typically performed manually by health care providers, guided by automated matching algorithms. These matching algorithms must be configured in advance, such as by setting the weights of patient attributes, usually by someone with knowledge of both the matching algorithm and the patient population being served.

OBJECTIVE

We aimed to develop and evaluate a machine learning-based software tool, which automatically configures a patient matching algorithm by learning from pairs of patient records previously linked by humans already present in the database.

METHODS

We built a free and open-source software tool to optimize record linkage algorithm parameters based on historical record linkages. The tool uses Bayesian optimization to identify the set of configuration parameters that lead to optimal matching performance in a given patient population, by learning from prior record linkages by humans. The tool is written assuming only the existence of a minimal HTTP application programming interface (API), and so is agnostic to the choice of MPI software, record linkage algorithm, and patient population. As a proof of concept, we integrated our tool with SantéMPI, an open-source MPI. We validated the tool using several synthetic patient populations in SantéMPI by comparing the performance of the optimized configuration in held-out data to SantéMPI's default matching configuration using sensitivity and specificity.

RESULTS

The machine learning-optimized configurations correctly detect over 90% of true record linkages as definite matches in all data sets, with 100% specificity and positive predictive value in all data sets, whereas the baseline detects none. In the largest data set examined, the baseline matching configuration detects possible record linkages with a sensitivity of 90.2% (95% CI 88.4%-92.0%) and specificity of 100%. By comparison, the machine learning-optimized matching configuration attains a sensitivity of 100%, with a decreased specificity of 95.9% (95% CI 95.9%-96.0%). We report significant gains in sensitivity in all data sets examined, at the cost of only marginally decreased specificity. The configuration optimization tool, data, and data set generator have been made freely available.

CONCLUSIONS

Our machine learning software tool can be used to significantly improve the performance of existing record linkage algorithms, without knowledge of the algorithm being used or specific details of the patient population being served.

Making Sense of Theories, Models, and Frameworks in Digital Health Behavior Change Design: Qualitative Descriptive Study

BACKGROUND

Digital health interventions are increasingly being designed to support health behaviors. Although digital health interventions informed by behavioral science theories, models, and frameworks (TMFs) are more likely to be effective than those designed without them, design teams often struggle to use these evidence-informed tools. Until now, little work has been done to clarify the ways in which behavioral science TMFs can add value to digital health design.

OBJECTIVE

The aim of this study was to better understand how digital health design leaders select and use TMFs in design practice. The questions that were addressed included how do design leaders perceive the value of TMFs in digital health design, what considerations do design leaders make when selecting and applying TMFs, and what do design leaders think is needed in the future to advance the utility of TMFs in digital health design?

METHODS

This study used a qualitative description design to understand the experiences and perspectives of digital health design leaders. The participants were identified through purposive and snowball sampling. Semistructured interviews were conducted via Zoom software. Interviews were audio-recorded and transcribed using Otter.ai software. Furthermore, 3 researchers coded a sample of interview transcripts and confirmed the coding strategy. One researcher completed the qualitative analysis using a codebook thematic analysis approach.

RESULTS

Design leaders had mixed opinions on the value of behavioral science TMFs in digital health design. Leaders suggested that TMFs added the most value when viewed as a starting point rather than the final destination for evidence-informed design. Specifically, these tools added value when they acted as a gateway drug to behavioral science, supported health behavior conceptualization, were balanced with expert knowledge and user-centered design principles, were complementary to existing design methods, and supported both individual- and systems-level thinking. Design leaders also felt that there was a considerable nuance in selecting the most value-adding TMFs. Considerations should be made regarding their source, appropriateness, complexity, accessibility, adaptability, evidence base, purpose, influence, audience, fit with team expertise, fit with team culture, and fit with external pressures. Design leaders suggested multiple opportunities to advance the use of TMFs. These included improving TMF reporting, design, and accessibility, as well as improving design teams' capacity to use TMFs appropriately in practice.

CONCLUSIONS

When designing a digital health behavior change intervention, using TMFs can help design teams to systematically integrate behavioral insights. The future of digital health behavior change design demands an easier way for designers to integrate evidence-based TMFs into practice.

Developing a Data and Analytics Platform to Enable a Breast Cancer Learning Health System at a Regional Cancer Center

PURPOSE

This study documents the creation of automated, longitudinal, and prospective data and analytics platform for breast cancer at a regional cancer center. This platform combines principles of data warehousing with natural language processing (NLP) to provide the integrated, timely, meaningful, high-quality, and actionable data required to establish a learning health system.

METHODS

Data from six hospital information systems and one external data source were integrated on a nightly basis by automated extract/transform/load jobs. Free-text clinical documentation was processed using a commercial NLP engine.

RESULTS

The platform contains 141 data elements of 7,019 patients with newly diagnosed breast cancer who received care at our regional cancer center from January 1, 2014, to June 3, 2022. Daily updating of the database takes an average of 56 minutes. Evaluation of the tuning of NLP jobs found overall high performance, with an F1 of 1.0 for 19 variables, with a further 16 variables with an F1 of > 0.95.

CONCLUSION

This study describes how data warehousing combined with NLP can be used to create a prospective data and analytics platform to enable a learning health system. Although upfront time investment required to create the platform was considerable, now that it has been developed, daily data processing is completed automatically in less than an hour.

Nailfold capillaroscopy and deep learning in diabetes

OBJECTIVE

To determine whether nailfold capillary images, acquired using video capillaroscopy, can provide diagnostic information about diabetes and its complications.

RESEARCH DESIGN AND METHODS

Nailfold video capillaroscopy was performed in 120 adult patients with and without type 1 or type 2 diabetes, and with and without cardiovascular disease. Nailfold images were analyzed using convolutional neural networks, a deep learning technique. Cross-validation was used to develop and test the ability of models to predict five5 prespecified states (diabetes, high glycosylated hemoglobin, cardiovascular event, retinopathy, albuminuria, and hypertension). The performance of each model for a particular state was assessed by estimating areas under the receiver operating characteristics curves (AUROC) and precision recall curves (AUPR).

RESULTS

A total of 5236 nailfold images were acquired from 120 participants (mean 44 images per participant) and were all available for analysis. Models were able to accurately identify the presence of diabetes, with AUROC 0.84 (95% confidence interval [CI] 0.76, 0.91) and AUPR 0.84 (95% CI 0.78, 0.93), respectively. Models were also able to predict a history of cardiovascular events in patients with diabetes, with AUROC 0.65 (95% CI 0.51, 0.78) and AUPR 0.72 (95% CI 0.62, 0.88) respectively.

CONCLUSIONS

This proof-of-concept study demonstrates the potential of machine learning for identifying people with microvascular capillary changes from diabetes based on nailfold images, and for possibly identifying those most likely to have diabetes-related complications.

Re-examining medical assistance in dying for mature minors in Canada: Reflections for health leaders

In Canada, Medical Assistance in Dying (MAiD) is legal for many Canadians based on several criteria, though minors who are deemed sufficiently capable to make medical decisions (ie, mature minors) remain ineligible. In this article, we provide insight into recent philosophical and legal evidence related to MAiD for mature minors. We begin by providing an overview of literature pertaining to MAiD for mature minors in particular (including evidence from Belgium and the Netherlands), followed by a discussion on the lessons that can be learnt from Canada's MAiD implementation process (in general) and other forms of paediatric end-of-life care. As a whole, we aim to highlight some key takeaway messages for health leaders to consider as deliberations on MAiD for mature minors continue.

Machine learning for detecting centre-level irregularities in randomized controlled trials: A pilot study

Centralized statistical monitoring is sometimes employed as an alternative to onsite monitoring for randomized control trials. Current central monitoring methods have limitations, in that they are relatively resource intensive and do not necessarily generalize to studies where an irregularity pattern has not been observed before. Machine learning has been effective in detecting irregularities in industries such as finance and manufacturing, but to date none have been applied to clinical trials. We conducted a pilot study for the use of machine learning to identify center-level irregularities in data from multicenter clinical trials. We employed unsupervised machine learning methods, which do not rely on labelled data, and therefore allow for the automated discovery of previously unseen irregularity patterns while maintaining flexibility when applied to new data with different structures. This pilot study employs unsupervised machine learning to compute distance matrices between centres, which we used to produce centre-level continuous features. We then used a one-class support vector machine to learn the underlying distribution of each data set to identify data that was substantially different from these distributions. We evaluated our approach against current automatable centralized monitoring methods on two trials with known irregularities. While current approaches performed well on one trial (AUROC 0.752 for monitoring vs. 0.584 for machine learning), our techniques performed substantially better on the other (AUROC 0.140 for monitoring vs 0.728 for machine learning). The results of this pilot study suggest both the feasibility and the potential value of a machine learning-based approach to irregularity detection in RCTs.

Optimizing Health Coaching for Patients With Type 2 Diabetes Using Machine Learning: Model Development and Validation Study

Background

Health coaching is an emerging intervention that has been shown to improve clinical and patient-relevant outcomes for type 2 diabetes. Advances in artificial intelligence may provide an avenue for developing a more personalized, adaptive, and cost-effective approach to diabetes health coaching.

Objective

We aim to apply Q-learning, a widely used reinforcement learning algorithm, to a diabetes health-coaching data set to develop a model for recommending an optimal coaching intervention at each decision point that is tailored to a patient’s accumulated history.

Methods

In this pilot study, we fit a two-stage reinforcement learning model on 177 patients from the intervention arm of a community-based randomized controlled trial conducted in Canada. The policy produced by the reinforcement learning model can recommend a coaching intervention at each decision point that is tailored to a patient’s accumulated history and is expected to maximize the composite clinical outcome of hemoglobin A1c reduction and quality of life improvement (normalized to [ ​0, 1 ​], with a higher score being better). Our data, models, and source code are publicly available.

Results

Among the 177 patients, the coaching intervention recommended by our policy mirrored the observed diabetes health coach’s interventions in 17.5% (n=31) of the patients in stage 1 and 14.1% (n=25) of the patients in stage 2. Where there was agreement in both stages, the average cumulative composite outcome (0.839, 95% CI 0.460-1.220) was better than those for whom the optimal policy agreed with the diabetes health coach in only one stage (0.791, 95% CI 0.747-0.836) or differed in both stages (0.755, 95% CI 0.728-0.781). Additionally, the average cumulative composite outcome predicted for the policy’s recommendations was significantly better than that of the observed diabetes health coach’s recommendations (tn-1=10.040; P<.001).

Conclusions

Applying reinforcement learning to diabetes health coaching could allow for both the automation of health coaching and an improvement in health outcomes produced by this type of intervention.

Integrating Behavioral Science and Design Thinking to Develop Mobile Health Interventions: Systematic Scoping Review

Background

Mobile health (mHealth) interventions are increasingly being designed to facilitate health-related behavior change. Integrating insights from behavioral science and design science can help support the development of more effective mHealth interventions. Behavioral Design (BD) and Design Thinking (DT) have emerged as best practice approaches in their respective fields. Until now, little work has been done to examine how BD and DT can be integrated throughout the mHealth design process.

Objective

The aim of this scoping review was to map the evidence on how insights from BD and DT can be integrated to guide the design of mHealth interventions. The following questions were addressed: (1) what are the main characteristics of studies that integrate BD and DT during the mHealth design process? (2) what theories, models, and frameworks do design teams use during the mHealth design process? (3) what methods do design teams use to integrate BD and DT during the mHealth design process? and (4) what are key design challenges, implementation considerations, and future directions for integrating BD and DT during mHealth design?

Methods

This review followed the Joanna Briggs Institute reviewer manual and PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) checklist. Studies were identified from MEDLINE, PsycINFO, Embase, CINAHL, and JMIR by using search terms related to mHealth, BD, and DT. Included studies had to clearly describe their mHealth design process and how behavior change theories, models, frameworks, or techniques were incorporated. Two independent reviewers screened the studies for inclusion and completed the data extraction. A descriptive analysis was conducted.

Results

A total of 75 papers met the inclusion criteria. All studies were published between 2012 and 2021. Studies integrated BD and DT in notable ways, which can be referred to as “Behavioral Design Thinking.” Five steps were followed in Behavioral Design Thinking: (1) empathize with users and their behavior change needs, (2) define user and behavior change requirements, (3) ideate user-centered features and behavior change content, (4) prototype a user-centered solution that supports behavior change, and (5) test the solution against users’ needs and for its behavior change potential. The key challenges experienced during mHealth design included meaningfully engaging patient and public partners in the design process, translating evidence-based behavior change techniques into actual mHealth features, and planning for how to integrate the mHealth intervention into existing clinical systems.

Conclusions

Best practices from BD and DT can be integrated throughout the mHealth design process to ensure that mHealth interventions are purposefully developed to effectively engage users. Although this scoping review clarified how insights from BD and DT can be integrated during mHealth design, future research is needed to identify the most effective design approaches.

Continuous non-invasive remote automated blood pressure monitoring with novel wearable technology: A Validation Study

Background

Wearable continuous monitoring biosensor technologies have the potential to transform postoperative care with early detection of impending clinical deterioration.

Objective

Our aim was to validate the accuracy of Cloud DX Vitaliti continuous vital signs monitor (CVSM) continuous noninvasive blood pressure (cNIBP) measurements in postsurgical patients. A secondary aim was to examine user acceptance of the Vitaliti CVSM with respect to comfort, ease of application, sustainability of positioning, and aesthetics.

Methods

Included participants were ≥18 years old and recovering from surgery in a cardiac intensive care unit (ICU). We targeted a maximum recruitment of 80 participants for verification and acceptance testing. We also oversampled to minimize the effect of unforeseen interruptions and other challenges to the study. Validation procedures were according to the International Standards Organization (ISO) 81060-2:2018 standards for wearable, cuffless blood pressure (BP) measuring devices. Baseline BP was determined from the gold-standard ICU arterial catheter. The Vitaliti CVSM was calibrated against the reference arterial catheter. In static (seated in bed) and supine positions, 3 cNIBP measurements, each 30 seconds, were taken for each patient with the Vitaliti CVSM and an invasive arterial catheter. At the conclusion of each test session, captured cNIBP measurements were extracted using MediCollector BEDSIDE data extraction software, and Vitaliti CVSM measurements were extracted to a secure laptop through a cable connection. The errors of these determinations were calculated. Participants were interviewed about device acceptability.

Results

The validation analysis included data for 20 patients. The average times from calibration to first measurement in the static position and to first measurement in the supine position were 133.85 seconds (2 minutes 14 seconds) and 535.15 seconds (8 minutes 55 seconds), respectively. The overall mean errors of determination for the static position were –0.621 (SD 4.640) mm Hg for systolic blood pressure (SBP) and 0.457 (SD 1.675) mm Hg for diastolic blood pressure (DBP). Errors of determination were slightly higher for the supine position, at 2.722 (SD 5.207) mm Hg for SBP and 2.650 (SD 3.221) mm Hg for DBP. The majority rated the Vitaliti CVSM as comfortable. This study was limited to evaluation of the device during a very short validation period after calibration (ie, that commenced within 2 minutes after calibration and lasted for a short duration of time).

Conclusions

We found that the Cloud DX’s Vitaliti CVSM demonstrated cNIBP measurement in compliance with ISO 81060-2:2018 standards in the context of evaluation that commenced within 2 minutes of device calibration; this device was also well-received by patients in a postsurgical ICU setting. Future studies will examine the accuracy of the Vitaliti CVSM in ambulatory contexts, with attention to assessment over a longer duration and the impact of excessive patient motion on data artifacts and signal quality.

Trial Registration

ClinicalTrials.gov NCT03493867; https://clinicaltrials.gov/ct2/show/NCT03493867

Post-discharge after surgery Virtual Care with Remote Automated Monitoring-1 (PVC-RAM-1) technology versus standard care: randomised controlled trial

OBJECTIVE

To determine if virtual care with remote automated monitoring (RAM) technology versus standard care increases days alive at home among adults discharged after non-elective surgery during the covid-19 pandemic.

DESIGN

Multicentre randomised controlled trial.

SETTING

8 acute care hospitals in Canada.

PARTICIPANTS

905 adults (≥40 years) who resided in areas with mobile phone coverage and were to be discharged from hospital after non-elective surgery were randomised either to virtual care and RAM (n=451) or to standard care (n=454). 903 participants (99.8%) completed the 31 day follow-up.

INTERVENTION

Participants in the experimental group received a tablet computer and RAM technology that measured blood pressure, heart rate, respiratory rate, oxygen saturation, temperature, and body weight. For 30 days the participants took daily biophysical measurements and photographs of their wound and interacted with nurses virtually. Participants in the standard care group received post-hospital discharge management according to the centre's usual care. Patients, healthcare providers, and data collectors were aware of patients' group allocations. Outcome adjudicators were blinded to group allocation.

MAIN OUTCOME MEASURES

The primary outcome was days alive at home during 31 days of follow-up. The 12 secondary outcomes included acute hospital care, detection and correction of drug errors, and pain at 7, 15, and 30 days after randomisation.

RESULTS

All 905 participants (mean age 63.1 years) were analysed in the groups to which they were randomised. Days alive at home during 31 days of follow-up were 29.7 in the virtual care group and 29.5 in the standard care group: relative risk 1.01 (95% confidence interval 0.99 to 1.02); absolute difference 0.2% (95% confidence interval -0.5% to 0.9%). 99 participants (22.0%) in the virtual care group and 124 (27.3%) in the standard care group required acute hospital care: relative risk 0.80 (0.64 to 1.01); absolute difference 5.3% (-0.3% to 10.9%). More participants in the virtual care group than standard care group had a drug error detected (134 (29.7%) v 25 (5.5%); absolute difference 24.2%, 19.5% to 28.9%) and a drug error corrected (absolute difference 24.4%, 19.9% to 28.9%). Fewer participants in the virtual care group than standard care group reported pain at 7, 15, and 30 days after randomisation: absolute differences 13.9% (7.4% to 20.4%), 11.9% (5.1% to 18.7%), and 9.6% (2.9% to 16.3%), respectively. Beneficial effects proved substantially larger in centres with a higher rate of care escalation.

CONCLUSION

Virtual care with RAM shows promise in improving outcomes important to patients and to optimal health system function.

TRIAL REGISTRATION

ClinicalTrials.gov NCT04344665.

Opening the black box: the promise and limitations of explainable machine learning in cardiology

Many clinicians remain wary of machine learning due to long-standing concerns about "black box" models. "Black box" is shorthand for models that are sufficiently complex that they are not straightforwardly interpretable to humans. Lack of interpretability in predictive models can undermine trust in those models, especially in health care where so many decisions are literally life and death. There has recently been an explosion of research in the field of explainable machine learning aimed at addressing these concerns. The promise of explainable machine learning is considerable, but it is important for cardiologists who may encounter these techniques in clinical decision support tools or novel research papers to have a critical understanding of both their strengths and their limitations. This paper reviews key concepts and techniques in the field of explainable machine learning as they apply to cardiology. Key concepts reviewed include interpretability versus explainability and global versus local explanations. Techniques demonstrated include permutation importance, surrogate decision trees, local interpretable model-agnostic explanations, and partial dependence plots. We discuss several limitations with explainability techniques, focusing on the how the nature of explanations as approximations may omit important information about how black box models work and why they make certain predictions. We conclude by proposing a rule of thumb about when it is appropriate to use black box models with explanations, rather than interpretable models.

Machine Learning-Based Early Warning Systems for Clinical Deterioration: Systematic Scoping Review

Background

Timely identification of patients at a high risk of clinical deterioration is key to prioritizing care, allocating resources effectively, and preventing adverse outcomes. Vital signs–based, aggregate-weighted early warning systems are commonly used to predict the risk of outcomes related to cardiorespiratory instability and sepsis, which are strong predictors of poor outcomes and mortality. Machine learning models, which can incorporate trends and capture relationships among parameters that aggregate-weighted models cannot, have recently been showing promising results.

Objective

This study aimed to identify, summarize, and evaluate the available research, current state of utility, and challenges with machine learning–based early warning systems using vital signs to predict the risk of physiological deterioration in acutely ill patients, across acute and ambulatory care settings.

Methods

PubMed, CINAHL, Cochrane Library, Web of Science, Embase, and Google Scholar were searched for peer-reviewed, original studies with keywords related to “vital signs,” “clinical deterioration,” and “machine learning.” Included studies used patient vital signs along with demographics and described a machine learning model for predicting an outcome in acute and ambulatory care settings. Data were extracted following PRISMA, TRIPOD, and Cochrane Collaboration guidelines.

Results

We identified 24 peer-reviewed studies from 417 articles for inclusion; 23 studies were retrospective, while 1 was prospective in nature. Care settings included general wards, intensive care units, emergency departments, step-down units, medical assessment units, postanesthetic wards, and home care. Machine learning models including logistic regression, tree-based methods, kernel-based methods, and neural networks were most commonly used to predict the risk of deterioration. The area under the curve for models ranged from 0.57 to 0.97.

Conclusions

In studies that compared performance, reported results suggest that machine learning–based early warning systems can achieve greater accuracy than aggregate-weighted early warning systems but several areas for further research were identified. While these models have the potential to provide clinical decision support, there is a need for standardized outcome measures to allow for rigorous evaluation of performance across models. Further research needs to address the interpretability of model outputs by clinicians, clinical efficacy of these systems through prospective study design, and their potential impact in different clinical settings.

Extracting Clinical Features From Dictated Ambulatory Consult Notes Using a Commercially Available Natural Language Processing Tool: Pilot, Retrospective, Cross-Sectional Validation Study

Background

The increasing adoption of electronic health records (EHRs) in clinical practice holds the promise of improving care and advancing research by serving as a rich source of data, but most EHRs allow clinicians to enter data in a text format without much structure. Natural language processing (NLP) may reduce reliance on manual abstraction of these text data by extracting clinical features directly from unstructured clinical digital text data and converting them into structured data.

Objective

This study aimed to assess the performance of a commercially available NLP tool for extracting clinical features from free-text consult notes.

Methods

We conducted a pilot, retrospective, cross-sectional study of the accuracy of NLP from dictated consult notes from our tuberculosis clinic with manual chart abstraction as the reference standard. Consult notes for 130 patients were extracted and processed using NLP. We extracted 15 clinical features from these consult notes and grouped them a priori into categories of simple, moderate, and complex for analysis.

Results

For the primary outcome of overall accuracy, NLP performed best for features classified as simple, achieving an overall accuracy of 96% (95% CI 94.3-97.6). Performance was slightly lower for features of moderate clinical and linguistic complexity at 93% (95% CI 91.1-94.4), and lowest for complex features at 91% (95% CI 87.3-93.1).

Conclusions

The findings of this study support the use of NLP for extracting clinical features from dictated consult notes in the setting of a tuberculosis clinic. Further research is needed to fully establish the validity of NLP for this and other purposes.

Introduction to CAUSES: Description of Weather and Climate Models and Their Near-Surface Temperature Errors in 5 day Hindcasts Near the Southern Great Plains

We introduce the Clouds Above the United States and Errors at the Surface (CAUSES) project with its aim of better understanding the physical processes leading to warm screen temperature biases over the American Midwest in many numerical models. In this first of four companion papers, 11 different models, from nine institutes, perform a series of 5 day hindcasts, each initialized from reanalyses. After describing the common experimental protocol and detailing each model configuration, a gridded temperature data set is derived from observations and used to show that all the models have a warm bias over parts of the Midwest. Additionally, a strong diurnal cycle in the screen temperature bias is found in most models. In some models the bias is largest around midday, while in others it is largest during the night. At the Department of Energy Atmospheric Radiation Measurement Southern Great Plains (SGP) site, the model biases are shown to extend several kilometers into the atmosphere. Finally, to provide context for the companion papers, in which observations from the SGP site are used to evaluate the different processes contributing to errors there, it is shown that there are numerous locations across the Midwest where the diurnal cycle of the error is highly correlated with the diurnal cycle of the error at SGP. This suggests that conclusions drawn from detailed evaluation of models using instruments located at SGP will be representative of errors that are prevalent over a larger spatial scale.

How much do cancer specialists earn? A comparison of physician fees and remuneration in oncology and radiology in high-income countries

The main driver of higher spending on health care in the US is believed to be substantially higher fees paid to US physicians in comparison with other countries. We aim to compare physician incomes in radiology and oncology considering differences in relation to fees paid, physician capacity and volume of services provided in five countries: the United States, Canada, Australia, France and the United Kingdom. The fee for a consultation with a specialist in oncology varies threefold across countries, and more than fourfold for chemotherapy. There is also a three to fourfold variation in fees for ultrasound and CT scans. Physician earnings in the US are greater than in other countries in both oncology and radiology, more than three times higher than in the UK; Canadian oncologists and radiologists earn considerably more than their European counterparts. Although challenging, benchmarking earnings and fees for similar health care activities across countries, and understanding the factors that explain any differences, can provide valuable insights for policy makers trying to enhance efficiency and quality in service delivery, especially in the face of rising care costs.

Looking Beyond Income and Education: Socioeconomic Status Gradients Among Future High-Cost Users of Health Care

INTRODUCTION

Healthcare spending occurs disproportionately among a very small portion of the population. Research on these high-cost users (HCUs) of health care has been overwhelmingly cross-sectional in nature and limited to the few sociodemographic and clinical characteristics available in health administrative databases. This study is the first to bridge this knowledge gap by applying a population health lens to HCUs. We investigate associations between a broad range of SES characteristics and future HCUs.

METHODS

A cohort of adults from two cycles of large, nationally representative health surveys conducted in 2003 and 2005 was linked to population-based health administrative databases from a universal healthcare plan for Ontario, Canada. Comprehensive person-centered estimates of annual healthcare spending were calculated for the subsequent 5 years following interview. Baseline HCUs (top 5%) were excluded and healthcare spending for non-HCUs was analyzed. Adjusted for predisposition and need factors, the odds of future HCU status (over 5 years) were estimated according to various individual, household, and neighborhood SES factors. Analyses were conducted in 2014.

RESULTS

Low income (personal and household); less than post-secondary education; and living in high-dependency neighborhoods greatly increased the odds of future HCUs. After adjustment, future HCU status was most strongly associated with food insecurity, personal income, and non-homeownership. Living in highly deprived or low ethnic concentration neighborhoods also increased the odds of becoming an HCU.

CONCLUSIONS

Findings suggest that addressing social determinants of health, such as food and housing security, may be important components of interventions aiming to improve health outcomes and reduce costs.

Public payments to physicians in Ontario adjusted for overhead costs

We used data collected in the 2010 National Physician Survey and public payment data published in the Institute for Clinical and Evaluative Sciences report Payments to Ontario Physicians from Ministry of Health and Long-Term Care Sources 1992/93 to 2009/10 to estimate 2009/2010 net physician income from public payments for Ontario physicians by specialty. Incorporating overhead substantially affects estimates of physician income and changes relative position. For example, ophthalmologists were ranked second when only public payments were considered but eighth when overhead was included. Conversely, hospital-based specialties such as anaesthesia, radiation oncology and emergency medicine rank significantly higher after overhead is included.

Vegetations in infective endocarditis. Clinical relevance and diagnosis by cross sectional echocardiography

Cross sectional echocardiography identified vegetations in 45 of 62 (73%) patients who had clinical evidence of active infective endocarditis. The sensitivity of this technique in diagnosing vegetations in infective endocarditis was 93% and the specificity 89%. The predictive value of a positive test was 96% and that of a negative test 80%. Vegetations were detected with a similar frequency on the aortic and mitral valves. The incidence of valvar incompetence, congestive heart failure, and the need for surgical intervention was similar in the patients with and without vegetations. Embolism occurred in 47% of those patients with vegetations and in 12% of those without. The mortality rate was 27% in those with vegetations, and no patient without vegetations died. Thus cross sectional echocardiography is accurate in diagnosing vegetations in patients with infective endocarditis, and this finding identifies patients at high risk of embolic complications and death.