JAMIA Open最新文献

Objectives: Absolute risk models estimate an individual's future disease risk over a specified time interval. Applications utilizing server-side risk tooling, the R-based iCARE (R-iCARE), to build, validate, and apply absolute risk models, face limitations in portability and privacy due to their need for circulating user data in remote servers for operation. We overcome this by porting iCARE to the web platform.

Materials and methods: We refactored R-iCARE into a Python package (Py-iCARE) and then compiled it to WebAssembly (Wasm-iCARE)-a portable web module, which operates within the privacy of the user's device.

Results: We showcase the portability and privacy of Wasm-iCARE through 2 applications: for researchers to statistically validate risk models and to deliver them to end-users. Both applications run entirely on the client side, requiring no downloads or installations, and keep user data on-device during risk calculation.

Conclusions: Wasm-iCARE fosters accessible and privacy-preserving risk tools, accelerating their validation and delivery.

Objective: Decision support can improve shared decision-making for breast cancer treatment, but workflow barriers have hindered widespread use of these tools. The goal of this study was to understand the workflow among breast cancer teams of clinicians, patients, and their family caregivers when making treatment decisions and identify design guidelines for informatics tools to better support treatment decision-making.

Materials and methods: We conducted observations of breast cancer clinicians during routine clinical care from February to August 2022. Guided by the work system model, a human factors engineering model that describes the elements of work, we recorded all aspects of clinician workflow using a tablet and smart pencil. Observation notes were transcribed and uploaded into Dedoose. Two researchers inductively coded the observations. We identified themes relevant to the design of decision support that we classified into the 4 components of workflow (ie, flow of information, tasks, tools and technologies, and people).

Results: We conducted 20 observations of breast cancer clinicians (total: 79 hours). We identified 10 themes related to workflow that present challenges and opportunities for decision support design. We identified approximately 48 different decisions discussed during breast cancer visits. These decisions were often interdependent and involved collaboration across the large cancer treatment team. Numerous patient-specific factors (eg, work, hobbies, family situation) were discussed when making treatment decisions as well as complex risk and clinical information. Patients were frequently asked to remember and relay information across the large cancer team.

Discussion and conclusion: Based on these findings, we proposed design guidelines for informatics tools to support the complex workflows involved in breast cancer care. These guidelines should inform the design of informatics solutions to better support breast cancer decision-making and improve patient-centered cancer care.

Importance: Electronic health record textual sources such as medication signeturs (sigs) contain valuable information that is not always available in structured form. Commonly processed through manual annotation, this repetitive and time-consuming task could be fully automated using large language models (LLMs). While most sigs include simple instructions, some include complex patterns.

Objectives: We aimed to compare the performance of GPT-3.5 and GPT-4 with smaller fine-tuned models (ClinicalBERT, BlueBERT) in extracting the average daily dose of 2 immunomodulating medications with frequent complex sigs: hydroxychloroquine, and prednisone.

Methods: Using manually annotated sigs as the gold standard, we compared the performance of these models in 702 hydroxychloroquine and 22 104 prednisone prescriptions.

Results: GPT-4 vastly outperformed all other models for this task at any level of in-context learning. With 100 in-context examples, the model correctly annotates 94% of hydroxychloroquine and 95% of prednisone sigs to within 1 significant digit. Error analysis conducted by 2 additional manual annotators on annotator-model disagreements suggests that the vast majority of disagreements are model errors. Many model errors relate to ambiguous sigs on which there was also frequent annotator disagreement.

Discussion: Paired with minimal manual annotation, GPT-4 achieved excellent performance for language regression of complex medication sigs and vastly outperforms GPT-3.5, ClinicalBERT, and BlueBERT. However, the number of in-context examples needed to reach maximum performance was similar to GPT-3.5.

Conclusion: LLMs show great potential to rapidly extract structured data from sigs in no-code fashion for clinical and research applications.

Objectives: Telehealth or remote care has been widely leveraged to provide health care support and has achieved tremendous developments and positive results, including in low- and middle-income countries (LMICs). Social networking platform, as an easy-to-use tool, has provided users with simplified means to collect data outside of the traditional clinical environment. WeChat, one of the most popular social networking platforms in many countries, has been leveraged to conduct telehealth and hosted a vast amount of patient-generated health data (PGHD), including text, voices, images, and videos. Its characteristics of convenience, promptness, and cross-platform support enrich and simplify health care delivery and communication, addressing some weaknesses of traditional clinical care during the pandemic. This study aims to systematically summarize how WeChat platform has been leveraged to facilitate health care delivery and how it improves the access to health care.

Materials and methods: Utilizing Levesque's health care accessibility model, the study explores WeChat's impact across 5 domains: Approachability, Acceptability, Availability and accommodation, Affordability, and Appropriateness.

Results: The findings highlight WeChat's diverse functionalities, ranging from telehealth consultations and remote patient monitoring to seamless PGHD exchange. WeChat's integration with health tracking apps, support for telehealth consultations, and survey capabilities contribute significantly to disease management during the pandemic.

Discussion and conclusion: The practices and implications from WeChat may provide experiences to utilize social networking platforms to facilitate health care delivery. The utilization of WeChat PGHD opens avenues for shared decision-making, prompting the need for further research to establish reporting guidelines and policies addressing privacy and ethical concerns associated with social networking platforms in health research.

Objective: To address database interoperability challenges to improve collaboration among disparate organizations.

Materials and methods: We developed a lightweight system to allow broad but well-controlled data sharing while preserving local data protection policies. We used 2 NIST-developed technologies-Next-generation Database Access Control (NDAC) and the Data Block Matrix (DBM)-to create a proof-of-concept system called the Secure Federated Data Sharing System (SFDS). NDAC controls access to database resources down to the field level based on attributes assigned to users. The DBM manages and shares authoritative user-attribute assignments across a federation of organizations, implemented using a modified open-source permissioned blockchain, to manage and share authoritative user-attribute assignments across a federation of organizations. We used synthetic data to demonstrate a clinical research data-sharing use case using the SFDS.

Results: We demonstrated, through consent, the onboarding of previously unknown users into NDAC via assignments to their DBM-validated attributes, allowing those users policy-preserving access to local database resources. The SFDS main system components-NDAC and DBM-also showed excellent performance metrics.

Discussion: The SFDS provides a generic data-sharing infrastructure that effectively and securely achieves data-sharing objectives. It is completely transparent to the otherwise normal business operations of participating organizations. It requires no changes to database management systems or existing methods of authenticating and authorizing local user access to local resources.

Conclusion: This efficiency, flexibility of deployment, and granularity of control make this new infrastructure solution practical for meeting the data-sharing and protection objectives of the clinical research community.

Objective: High-risk pregnancy (HRP) conditions such as gestational diabetes mellitus (GDM), hypertension (HTN), and peripartum depression (PPD) affect maternal and neonatal health. Patient engagement is critical for effective HRP management (HRPM). While digital technologies and analytics hold promise, emerging research indicates limited and suboptimal support offered by the highly prevalent pregnancy digital solutions within the commercial marketplace. In this article, we describe our efforts to develop a portfolio of digital products leveraging advances in social computing, data science, and digital health.

Methods: We describe three studies that leverage core methods from Digilego digital health development framework to (1) conduct large-scale social media analysis (n = 55 301 posts) to understand population-level patterns in women's needs, (2) architect a digital repository to enable women curate HRP related information, and (3) develop a digital platform to support PPD prevention. We applied a combination of qualitative coding, machine learning, theory-mapping, and programmatic implementation of theory-linked digital features. Further, we conducted preliminary testing of the resulting products for acceptance with sample of pregnant women for GDM/HTN information management (n = 10) and PPD prevention (n = 30).

Results: Scalable social computing models using deep learning classifiers with reasonable accuracy have allowed us to capture and examine psychosociobehavioral drivers associated with HRPM. Our work resulted in two digital health solutions, MyPregnancyChart and MomMind are developed. Initial evaluation of both tools indicates positive acceptance from potential end users. Further evaluation with MomMind revealed statistically significant improvements (P < .05) in PPD recognition and knowledge on how to seek PPD information.

Discussion: Digilego framework provides an integrative methodological lens to gain micro-macro perspective on women's needs, theory integration, engagement optimization, as well as subsequent feature and content engineering, which can be organized into core and specialized digital pathways for women engagement in disease management.

Conclusion: Future works should focus on implementation and testing of digital solutions that facilitate women to capture, aggregate, preserve, and utilize, otherwise siloed, prenatal information artifacts for enhanced self-management of their high-risk conditions, ultimately leading to improved health outcomes.

Objective: To automate scientific claim verification using PubMed abstracts.

Materials and methods: We developed CliVER, an end-to-end scientific Claim VERification system that leverages retrieval-augmented techniques to automatically retrieve relevant clinical trial abstracts, extract pertinent sentences, and use the PICO framework to support or refute a scientific claim. We also created an ensemble of three state-of-the-art deep learning models to classify rationale of support, refute, and neutral. We then constructed CoVERt, a new COVID VERification dataset comprising 15 PICO-encoded drug claims accompanied by 96 manually selected and labeled clinical trial abstracts that either support or refute each claim. We used CoVERt and SciFact (a public scientific claim verification dataset) to assess CliVER's performance in predicting labels. Finally, we compared CliVER to clinicians in the verification of 19 claims from 6 disease domains, using 189 648 PubMed abstracts extracted from January 2010 to October 2021.

Results: In the evaluation of label prediction accuracy on CoVERt, CliVER achieved a notable F1 score of 0.92, highlighting the efficacy of the retrieval-augmented models. The ensemble model outperforms each individual state-of-the-art model by an absolute increase from 3% to 11% in the F1 score. Moreover, when compared with four clinicians, CliVER achieved a precision of 79.0% for abstract retrieval, 67.4% for sentence selection, and 63.2% for label prediction, respectively.

Conclusion: CliVER demonstrates its early potential to automate scientific claim verification using retrieval-augmented strategies to harness the wealth of clinical trial abstracts in PubMed. Future studies are warranted to further test its clinical utility.

Introduction: Cloud-based solutions are a modern-day necessity for data intense computing. This case report describes in detail the development and implementation of Amazon Web Services (AWS) at Emory-a secure, reliable, and scalable platform to store and analyze identifiable research data from the Centers for Medicare and Medicaid Services (CMS).

Materials and methods: Interdisciplinary teams from CMS, MBL Technologies, and Emory University collaborated to ensure compliance with CMS policy that consolidates laws, regulations, and other drivers of information security and privacy.

Results: A dedicated team of individuals ensured successful transition from a physical storage server to a cloud-based environment. This included implementing access controls, vulnerability scanning, and audit logs that are reviewed regularly with a remediation plan. User adaptation required specific training to overcome the challenges of cloud computing.

Conclusion: Challenges created opportunities for lessons learned through the creation of an end-product accepted by CMS and shared across disciplines university-wide.

Objectives: Partially observed confounder data pose a major challenge in statistical analyses aimed to inform causal inference using electronic health records (EHRs). While analytic approaches such as imputation are available, assumptions on underlying missingness patterns and mechanisms must be verified. We aimed to develop a toolkit to streamline missing data diagnostics to guide choice of analytic approaches based on meeting necessary assumptions.

Materials and methods: We developed the smdi (structural missing data investigations) R package based on results of a previous simulation study which considered structural assumptions of common missing data mechanisms in EHR.

Results: smdi enables users to run principled missing data investigations on partially observed confounders and implement functions to visualize, describe, and infer potential missingness patterns and mechanisms based on observed data.

Conclusions: The smdi R package is freely available on CRAN and can provide valuable insights into underlying missingness patterns and mechanisms and thereby help improve the robustness of real-world evidence studies.

Objectives: To provide a real-world example on how and to what extent Health Level Seven Fast Healthcare Interoperability Resources (FHIR) implements the Findable, Accessible, Interoperable, and Reusable (FAIR) guiding principles for scientific data. Additionally, presents a list of FAIR implementation choices for supporting future FAIR implementations that use FHIR.

Materials and methods: A case study was conducted on the Medical Information Mart for Intensive Care-IV Emergency Department (MIMIC-ED) dataset, a deidentified clinical dataset converted into FHIR. The FAIRness of this dataset was assessed using a set of common FAIR assessment indicators.

Results: The FHIR distribution of MIMIC-ED, comprising an implementation guide and demo data, was more FAIR compared to the non-FHIR distribution. The FAIRness score increased from 60 to 82 out of 95 points, a relative improvement of 37%. The most notable improvements were observed in interoperability, with a score increase from 5 to 19 out of 19 points, and reusability, with a score increase from 8 to 14 out of 24 points. A total of 14 FAIR implementation choices were identified.

Discussion: Our work examined how and to what extent the FHIR standard contributes to FAIR data. Challenges arose from interpreting the FAIR assessment indicators. This study stands out for providing a real-world example of a dataset that was made more FAIR using FHIR.

Conclusion: To the best of our knowledge, this is the first study that formally assessed the conformance of a FHIR dataset to the FAIR principles. FHIR improved the accessibility, interoperability, and reusability of MIMIC-ED. Future research should focus on implementing FHIR in research data infrastructures.