MCBK 2022 Lightning Round Abstracts

the

The Fast Evidence Interoperability Resources (FEvIR) Platform is a hub for data exchange for scientific evidence and guidance that uses a standard (Fast Healthcare Interoperability Resources [FHIR]) extended for Evidence-Based Medicine related knowledge. Converter tools between any system and the FEvIR Platform will enable interoperable data exchange to mobilize CBKs.
A COVID-19 Knowledge Accelerator (COKA) Common Metadata Framework Working Group created 129 data element structure definitions to specify metadata for 12 of these categories (all but preservation). Each data element is specified with an element name, a datatype, whether the element is required, whether the element can contain multiple values, and whether the element is a container that includes data elements.
Two crosswalks matching data elements across dataset schemas were Facing a series of compound spreadsheets and recognizing the inefficiency for scaling to support many more dataset schemas, we invented a SchemaElement StructureDefinition to provide a common reusable structure for representation of a data element structure definition and mapping it to a data element structure in a different dataset schema.
Data conversion tools on the FEvIR Platform (ClinicalTrials.gov-to-FEvIR Converter and RIS-to-FEvIR Converter so far) are being used to inform the development of the Common Metadata Framework and to show how the Common Metadata Framework will facilitate the efficiency of creating additional data conversion tools to support interoperable data exchange between systems using different dataset schemas.
completeness, the deployment of good modeling practices is inevitable.
Contrary to the current praxis of focusing on the processable knowledge representation (KR), we have to start with the representation of the epistemological knowledge of the involved domains and its correct formalization to transform the outcome into the processable KR. That way, interoperability advances from data exchange through information sharing, process management, knowledge management to interdisciplinary knowledge space management, thereby also acknowledging individual skills and experiences. In other words, we advance from data exchange to knowledge sharing. Our solution is an architecture-centric, system-theoretical, ontology-based, and policy-driven model and framework   The relationship between an investigated intervention and the outcome of interest is judged by P-value and/or 95% Confidence Interval (CI). If "P-value greater than 0.05," and/or "the lower CI smaller than 1 and the upper CI greater than 1," it indicates to "has no difference," The "60-30-10" challenge is one of EPIC proportions within healthcare, and continues to worsen, with wastage, process inefficiencies and non-optimal patient care and safety practices.
Learning Health Systems (LHS) have the potential to utilize biomedical health data in real time, through rapid and continuous cycles of data interrogation, implementing insights to practice, feedback, and practice change. However, there is a lack of an appropriately skilled interdisciplinary, informatics workforce that are able to leverage computable biomedical knowledge to design innovative solutions. Therefore, there is a need to develop tailored professional development training to foster skilled interdisciplinary learning communities in the healthcare workforce in Australia. In order to evaluate the utility of program we undertook a mixed methods evaluation consisting of pre and post-surveys with ratings scales for usefulness, engagement, value, and applicability for various aspects of the course. Participants also completed identical measures of self-efficacy pre and post, with a scale that was mapped to specific skills and tasks that should have been achievable following each of the topics covered. Post-course participants were invited to participate in a semi-structured interview process and surveys to elaborate on survey questions and dive deeper into themes around utility, future applicability, barriers, recommendations, and identity in digital health.
Results: From the evaluation, it was evident that participants found the teaching model engaging, useful, valuable and applicable to their work and LHS projects. In the self-efficacy component, we observed a significant increase in perceive confidence for all topics, when comparing pre and post-course ratings. Overall, it was evident that the program gave participants a framework to organize their knowledge; a common understanding and shared language to converse with other disciplines; changed the way they perceived their role and the possibilities of data and technologies; and finally, LHS provided a toolkit to operate from. For the LHS Academy Fellows, there was also a notable increase in the number of fellows perceiving themselves as leaders at the end of their foundational coursework.

Conclusion:
We present examples of LHS-specific education programs as means of educating the health workforce to adopt the LHS model into standard practice. However, it is evident that such a movement will require a global and coordinated effort with significant training for the workforce to be able to understand and utilize data effectively, to improve practice. "transient assemblies of functions, information items, or components spread over information infrastructures and the Internet, a condition that sets them strongly apart from physical objects" (due to their malleability and interactivity). 1,2 To make CBK artifacts stable and usable, our packages add content alongside CBK, including identifiers, pointers, documentation, tests, and metadata. 3 Here, we share the results of several implementation choices.

LESSON 1-Choose identifier systems intentionally
To find CBK artifacts on the WWW, new persistent unique identifiers (PUIDs) must be minted and assigned to each one. We appreciate the Handle System. 4,5 However, we have instead implemented Archival Resource Keys (ARKs), an identifier system born at the California Digital Library and sustained by the ARK Alliance. 6,7 ARKs offer local control over identifier shoulders and subnamespaces. 8 We use ARK subnamespaces to reflect our evolving ontology of CBK artifact parts. 9 In this way, ARKs enable a standardized pointer scheme for the subcomponents of our packaged CBK artifacts.

LESSON 2-Leverage resolver-registries to increase access
To access CBK artifacts with persistent identifiers on the WWW, we have explored how to take advantage of the Name-to-Thing ID resolver (N2T.net). 9 To increase access, identifier resolver technology like N2T.net provides lookup and HTTP redirection to various CBK artifact servers and repositories.

LESSON 3-Use service descriptions to achieve interoperability
To interoperate with CBK artifacts via common webservices, we incorporate machine-readable service descriptions into our packages.
Our initial approach uses service descriptions conforming to the OpenAPI 3x standard for RESTful webservices. 10,11 Using these service descriptions and some homegrown technology for activating CBK artifacts, 12   The event will feature several clinical scenarios, potentially: antimicrobial prescribing, urgent referrals for suspected cancer, diabetes medication, or rare disease diagnosis.
The plan is to have two in-person collaborathons, in November 2022 and February 2023. Both events will be multidisciplinary, but with some tasks that are primarily clinical and some that are primarily technical. The primarily clinical tasks at the first event are to decompose selected NICE narrative into useful tagged components and to note any common principles for content decomposition across specialty topics. The primarily technical tasks are to represent a selected NICE narrative recommendation or structured pathway in a fully specified logical model using CQL or BPM+. At the second events the planned tasks are to (1) implement a CDS Hook or lookup for the tagged fragments identified in the first meeting in a simulated EPR or app and (2) implement a prepared logical model (CQL or BPM+) in a simulated EPR or app.
Before the first collaborathon, we will develop online tutorials on FHIR, CQL, and BPM+ for participants and clarify the clinical use cases. Before the second collaborathon, we need to build the requisite infrastructure: FHIR servers, CQL engine, BPM+ engine, and synthetic data.
We are recruiting clinicians to participate along with invited vendor teams. We look forward to sharing our learning with the global MCBK community.  The proposed ERKD model consists of three sub-models: a knowledge-driven model, a big data-driven model, and a knowledgeand-data fusion model for integrating the results generated by knowledge-driven and data-driven models. In the knowledge-driven model, domain knowledge is represented using belief rules, which can represent medical uncertainties by introducing new parameters including rule weight, antecedent attribute weight and belief degrees in consequents; and the inference mechanism is implemented using the evidential reasoning (ER) approach, which can reason with the parameters for uncertainty representation to produce a distributed result with combined degree of belief in each possible consequent. In the big datadriven model, real-world evidence is mined from real-world big data using a generalized Bayesian method and the evidence is profiled using a belief distribution format with evidence weight, reliability, independence degree, and degree of belief in each medical outcome contained in the evidence; and the ER rule with the updated and improved ER algorithm is used to do evidence combination. In the knowledgeand-data fusion model, the ER approach is used to integrated the results generated by the data-driven and knowledge-driven models. The proposed ERKD model takes advantages of domain knowledge and real-world big data, and the results generated by the knowledgedriven and data-driven sub-models are treated with equal importance in the knowledge-and-data fusion sub-model. In addition, uncertainties in medical domain knowledge can be well represented and inferred in the ERKD model, and the big data-driven model is not a black-box tool here, and the real-world evidence and the ER rulebased evidence combination process in the big data-driven model are explainable to clinicians. Next, we plan to apply the ERKD model to CKD complications and adverse outcomes prediction. We believe that CBK has the greatest potential to improve health outcomes in low-resource primary care settings, and yet, there has been very little study of such efforts. Our experience suggests that a straightforward clinical decision support system implemented broadly in an outpatient primary care network can more than double clinician adherence to guideline recommendations, which ultimately ensures better outcomes for patients.  Yet, more needs to be done to understand CBK knowledge repository policies, standards, and practices for promoting and using trustworthy CBK artifacts. We surveyed the policies and procedures that CBK repositories in the United States currently employ to convey trust.
Our findings found general trends in governing principles and provided a basis for suggested "desiderata" with which repositories may use to convey trust in CBK.

MCBK's Trust and Policy Working Group (TPWG) conducted an
online survey, identifying a convenience sample of 24 knowledge repositories based on expert knowledge and review of the field. The survey was designed to inquire about policies and procedures for conveying trust in CBK. We asked 91 questions (60 structured and 31 unstructured) about the organizations themselves as well as questions that were aligned to the "TRUST" principles for data repositories. We vetted the questions through multiple rounds with TFWG members as well as with non-members during the 2021 MCBK Annual Conference. We conducted a quantitative analysis of data from respondents who had completed at least 40% of the survey questions to generate summary frequencies of the answers, and organized governance practices into three categories: "common," "somewhat common," or "uncommon." Thirteen of the 24 CBK representatives (54%) sufficiently completed the survey. All 13 indicated to different degrees their adherence to policies that conveyed aspects of TRUST. Policies that were more commonly addressed were policies related to: Transparency, which was conveyed by having policies pertaining to provenance, credentialed contributors, and the provision of metadata; Responsibility, repositories reportedly provided knowledge in machine-readable formats, included implementation guidelines, and adhered to standards; and Technology, which included functions that enabled end-users to verify, search, and filter for knowledge products. Less common were practices that addressed: User Focused procedures that enabled consumers to know about user licensing requirements or query the use of knowledge artifacts; and Sustainability, less than a majority post described their sustainability plans. Of note, it was "uncommon" for patients to reportedly play a role in repositories' governance or decision-making processes. Based on our findings, we developed 29 desiderata, organized by the TRUST principles, that we believe will promote TRUST among the CBK repositories and catalyze maturation in the CBK ecosystem.
This, to our knowledge, is the first such survey to inquire about knowledge repository governance practices to promote trust in CBK.
We intend to field another survey with an increased response rate and that reports any changes in CBK governance policies, particularly policies pertaining to patient involvement.
The criticality of care processes as computable artifacts This lightning session will introduce BPM+, brief the status of the community, provide a tour of the products that have been developed. It will explore the utility of BPM+ and familiarize attendees with the impacts computable processes can make in guideline adherence, accelerating institutional adoption, care quality, and quality measurement. In our recorded talk, we will introduce learning materials, recommend reading and videos, and present challenges to librarians, knowledge managers, and information professional to support health professionals, researchers, and patients during the current and future evolution of CBK ecosystems to improve human health worldwide. In our live thematic group discussion, we will invite sharing of health prob- Library) from their inception to March 2022. The key phrases included "knee osteoarthritis" or "KOA" and "predict* tool," "predict* model,"

Mobilizing knowledge conclusions with CDS Hooks
"algorithm" or "nomogram," and "knee pain," "knee complaint*." The Prediction model Risk of Bias Assessment Tool (PROBAST) was used to assess the quality of the included studies. Delivery of the Right Decision Service platform as a national set of tools which:

Results
• Convert validated evidence, guidance and best practice into a diverse range of computerized knowledge formats. This includes actionable knowledge summaries, interactive visual pathways, and structured rules (algorithms). Algorithms can be bespoke or written in standards-based decision support languages-GDL or CQL. • Store structured and semi-structured computerized knowledge in a single national repository. • Deploy, reuse and customize these computerized knowledge entities within web and mobile applications, and integration with electronic health record systems.
The Right Decision platform has delivered over 20 national and local decision support solutions using these tools.
Work strand two: processes.
• A pipeline of clinical safety and quality management processes which provide assurance around safe and effective use of the decision support tools in practice. • An evaluation framework based on a logic model, setting out short-, medium-and long-term outcomes and measures. • An implementation and spread framework being developed to facilitate widespread adoption of the Right Decision Service and the decision support solutions it delivers. This will provide: • Guidance on roles, process and system change, and implementation methodologies. • A service model for local and national implementation of decision support, including service goals, components, roles, governance and accountability requirements.
Work strand three: people.
• A workforce development program for knowledge, information and data staff, as key to engineering decision support solutions and facilitating implementation. • A competency framework and program of learning opportunities to build confidence across the health and social care workforce in designing, interpreting, evaluating and implementing decision support solutions.