JAMIA Open最新文献

Objective: Use of secure messaging through the patient portal has increased in recent years. We compile evidence of how patient-initiated secure messaging has been measured, factors associated with use, and effects on individual and organization-level outcomes.

Materials and methods: We conducted a scoping review of articles published through October 2023 by systematically searching PubMed, Embase, and CINAHL. The search identified 2574 articles; 220 were selected for full text review and 78 met eligibility criteria. Factors and outcomes associated with messaging were organized according to the System Engineering Initiative for Patient Safety (SEIPS) 2.0 framework.

Results: Of 78 included studies, 70 (90%) specified the measurement approach: measuring any messaging use versus none (binary measure) (27/70; 39%), intensity of use (34/70; 49%), or both (9/70; 13%). Studies predominantly examined patient (vs clinician) characteristics (48/78; 62%), findings that patients of female sex, White race, higher socioeconomic status, and greater comorbidity were more likely to message and with greater intensity. Factors in other domains of the SEIPS framework such as tasks (7/78; 9%), tools/technology (5/78; 6%), organizational (7/78; 9%), and environmental (11/78; 14%) factors were examined less frequently, with mixed findings. Outcomes of secure messaging (23/78; 30%) were generally favorable with respect to clinical outcomes (10/23; 43%), efficiency (5/23; 22%), and care experience (5/23; 22%) and mixed with respect to health services use.

Discussion: Patient-initiated messaging use has been variably measured, with notable gaps in our understanding of the role of organization-level factors.

Conclusion: Evidence is needed to inform approaches implemented by healthcare systems to manage the growing volume of patient-initiated messages.

Objectives: Artificial Intelligence (AI) is transforming healthcare by improving diagnostics, treatment recommendations, and resource allocation. However, its implementation also raises ethical concerns, particularly regarding biases in AI algorithms trained on inequitable data, which may reinforce health disparities. This article introduces the AI COmmunity-based Ethical Dialogue and DEcision-making (CODE) framework to embed ethical deliberation into AI development, focusing on Electronic Health Records (EHRs).

Materials and methods: We propose the AI CODE framework as a structured approach to addressing ethical challenges in AI-driven healthcare and ensuring its implementation supports health equity.

Results: The framework outlines 5 steps to advance health equity: (1) Contextual diversity and priority: Ensuring inclusive datasets and that AI reflects the community needs; (2) Sharing ethical propositions: Structured discussions on privacy, bias, and fairness; (3) Dialogic decision-making: Collaboratively with stakeholders to develop AI solutions; (4) Integrating ethical solutions: Applying solutions into AI design to enhance fairness; and (5) Evaluating effectiveness: Continuously monitoring AI to address emerging biases.

Discussion: We examine the framework's role in mitigating AI biases through structured community engagement and its relevance within evolving healthcare policies. While the framework promotes ethical AI integration in healthcare, it also faces challenges in implementation.

Conclusion: The framework provides practical guidance to ensure AI systems are ethical, community-driven, and aligned with health equity goals.

Objective: To address biased representation in Electronic Health Records (EHRs) across subpopulations (SPs), which leads to predictive models underperforming for underrepresented groups, we propose a framework to enhance equitable predictive performance.

Materials and methods: We developed a framework using generative adversarial networks (GANs) to create SP-specific synthetic data, which augments the original training datasets. Subsequently, we employed an ensemble approach, training distinct prediction models tailored to each SP.

Results: The proposed framework was evaluated on two datasets derived from the MIMIC database, achieving a performance improvement in Receiver Operating Characteristics Area Under Curve (ROCAUC) ranging from 8% to 31% for underrepresented SPs.

Discussion: The results indicate that targeted synthetic data augmentation and SP-specific model training significantly mitigate the performance disparities observed in conventional predictive models trained on imbalanced EHR data.

Conclusion: Our novel GAN-based framework, combined with an ensemble prediction approach, effectively enhances predictive equity across SPs. The code and ensemble models developed in this study are publicly available, supporting further research and practical adoption of equitable predictive analytics in healthcare.

Objectives: Chronic obstructive pulmonary disease (COPD) is the third leading cause of death worldwide, with up to 70% of cases remaining undiagnosed. This paper proposes a COPD screening tool based on time-frequency representation features of self-recorded respiratory sounds.

Materials and methods: Respiratory sound samples (breath and cough sounds) were extracted from COPD and asymptomatic non-COPD volunteers using a large, scientific-purpose database. We analyzed 39 time-frequency representation features of breath and cough sounds, combined with age, sex, and smoking status, using Autoencoder neural networks and random forest (RF) algorithms. We compared the performance of different breath and cough RF models built to detect COPD: one based exclusively on sound features, one based exclusively on sociodemographic characteristics, and one based on sound features and sociodemographic characteristics.

Results: Models including breathing features outperformed models exclusively based on sociodemographic characteristics. Specifically, the model combining sociodemographic characteristics and breathing features achieved an area under the curve (AUC), accuracy, sensitivity, and specificity of 0.901, 0.836, 0.871, and 0.761, respectively, in the test set, representing a substantial increase in AUC when compared to the model based exclusively on sociodemographic characteristics (0.901 vs 0.818).

Discussion: Our results suggest that a lightweight collection of the time-frequency representation features of self-recorded beathing sounds could effectively improve the predictive performance of COPD screening or case-finding questionnaires.

Conclusion: COPD screening through self-recorded breathing sounds could be easily integrated as a low-cost first step in case-finding programs, potentially contributing to mitigate COPD underdiagnosis.

Objectives: Measurement-based care (MBC) improves clinical decision-making but remains underutilized in the Veterans Health Administration due to barriers such as provider awareness, time constraints, and user-experience limitations. This study describes the development of the War Related Illness and Injury Study Center Veteran Dashboard (VetDash), a patient-level clinical dashboard designed to integrate the VA's Collect, Share, Act model into the provider workflow.

Materials and methods: VetDash was developed using R Shiny, utilizing data from the WRIISC Clinical Intake Packet Database. It integrates patient-reported health data and military history into a dashboard hosted on a Linux-based Shiny Server within the VA's intranet.

Results: VetDash includes four modules: Patient Characteristics, Patient Health Symptoms, Patient Exposures, and Patient Self-Report Measures. Providers can visualize patient-reported symptoms, military exposures, and self-report measures, and compare patients to provider-defined cohorts.

Discussion and conclusion: VetDash facilitates MBC integration into the clinical workflow, potentially improving personalized patient care. Future research should evaluate its impact on clinical decisions, outcomes, and explore further enhancements.

Objectives: School violence risk prevention in the United States relies on manual assessments that are time-consuming and subjective. We developed a machine learning algorithm named Automated RIsk Assessment (ARIA), using natural language processing (NLP) to find linguistic patterns in standardized interview questions that can predict risk of aggression. Our goal was to evaluate the incremental change in performance with the addition of each question to simulate situations where interviews cannot be completed.

Materials and methods: Students were interviewed with 2 14-question risk assessments, the Brief Rating of Aggression by Children and Adolescents (BRACHA) and the School Safety Scale (SSS), that encouraged open-ended answers to the interview questions. The reference standard was defined as the subject's likeliness to display aggression in the future as determined by a forensic psychiatrist. Feature sets were extracted to represent the addition of 1 question at a time in a typical interview, up to and including the 28 total main questions along with other sub-questions that arose. The ARIA NLP pipeline tokenized each feature set, then extracted n-gram features (n ≤ 5) that captured contextual and semantic information. The features were evaluated using an L2-regularized logistic regression classifier and L2-regularized support vector machine (L2-SVM) classifier.

Results: Between May 1, 2015 and February 6, 2021, 412 assessment interviews were conducted. When compared to clinical judgement, ARIA performed with an area under the Receiver Operating Characteristic curve of 0.9 after 10 BRACHA questions, suggesting that it remains powerful even with truncated interviews. The full BRACHA had similar performance to the BRACHA + SSS assessment.

Discussion and conclusion: ARIA could use incomplete risk assessment interviews to provide modest recommendations even if interview completion is not possible. This could help to reduce the burden for the social worker or school counselor who may be using ARIA in less-than-ideal conditions.

Objective: To compare clinical summaries generated from simulated patient primary care electronic health records (EHRs) by GPT-4, to summaries generated by clinicians on multiple domains of quality including utility, concision, accuracy, and bias.

Materials and methods: Seven primary care physicians generated 70 simulated patient EHR notes, each representing 10 patient contacts with the practice over at least 2 years. Each record was summarized by a different clinician and by GPT-4. artificial intelligence (AI)- and clinician-authored summaries were rated blind by clinicians according to 8 domains of quality and an overall rating.

Results: The median time taken for a clinician to read through and assimilate the information in the EHRs before summarizing, was 7 minutes. Clinicians rated clinician-authored summaries higher than AI-authored summaries overall (7.39 vs 7.00 out of 10; P = .02), but with greater variability in clinician-authored summary ratings. AI and clinician-authored summaries had similar accuracy and AI-authored summaries were less likely to omit important information and more likely to use patient-friendly language.

Discussion: Although AI-authored summaries were rated slightly lower overall compared with clinician-authored summaries, they demonstrated similar accuracy and greater consistency. This demonstrates potential applications for generating summaries in primary care, particularly given the substantial time taken for clinicians to undertake this work.

Conclusion: The results suggest the feasibility, utility and acceptability of using AI-authored summaries to integrate into EHRs to support clinicians in primary care. AI summarization tools have the potential to improve healthcare productivity, including by enabling clinicians to spend more time on direct patient care.

Objectives: To evaluate clinical decision support (CDS) of varying complexities and care settings represented using Health Information Technology (HIT) standards-Clinical Quality Language (CQL) for representing clinical logic and Fast Healthcare Interoperability Resources (FHIR) for health information exchange.

Materials and methods: This Institutional Review Board-approved, retrospective study was performed at an academic medical center (January 1, 2023-December 31, 2023). Recommendations extracted from patient-centered outcomes guidelines were translated into standardized syntax (SNOMED CT) and representations (CQL, FHIR). Clinical decision support Hooks applications were developed for: CDS1-provides education for emergency department (ED) patients with venous thromboembolism; CDS2-recommends CT pulmonary angiogram in ED patients with suspected pulmonary embolism (PE) and uses FHIR Questionnaire resources for representing interactive content; CDS3-recommends mammography/breast magnetic resonance imaging surveillance in outpatients with breast cancer history. We randomly selected 50 ED patients with suspected PE and 50 outpatients undergoing breast imaging surveillance. We compared outcomes of false-positive alerts and the accuracy of CDS1, the more complex CDS2, and CDS3 for outpatients.

Results: Clinical decision support Hooks applications used CQL logic for trigger expressions and logic files and provided recommendations to ED and outpatient providers. CDS1 had a false-positive alert and accuracy of 11.1% and 98%, respectively, not significantly different from CDS2 (0.0% false-positive alerts, P = .33 and 96% accuracy, P = .56) or from CDS3 (0.0% false-positive alerts, P = .15 and 100% accuracy, P = .31).

Discussion: Health Information Technology standards can represent recommendations of varying complexities in various care settings.

Conclusion: The potential to represent CDS using standardized syntax and formats can help facilitate the dissemination of CDS-consumable artifacts.

Objective: With extended life expectancy, the number of people in need of care has been growing. To optimally support them, it is important to know the patterns and conditions of their daily life that influence the need for support, and thus, the classification of the care need. In this study, we aim to utilize a large corpus consisting of care benefits applications to do an explorative analysis of factors affecting care need to support the tedious work of experts gathering reliable criteria for a care need assessment.

Materials and methods: We compare state-of-the-art methods from explainable artificial intelligence (XAI) as means to extract such patterns from over 72 000 German care benefits applications. We train transformer models to predict assessment results as decided by a Medical Service Unit from accompanying text notes. To understand the key factors for care need assessment and its constituent modules (such as mobility and self-therapy), we apply feature attribution methods to extract the key phrases for each prediction. These local explanations are then aggregated into global insights to derive key phrases for different modules and severity of care need over the dataset.

Results: Our experiments show that transformers-based models perform slightly better than traditional bag-of-words baselines in predicting care need. We find that the bag-of-words baseline also provides useful care-relevant phrases, whereas phrases obtained through transformer explanations better balance rare and common phrases, such as diagnoses mentioned only once, and are better in assigning the correct assessment module.

Discussion: Even though XAI results can become unwieldy, they let us get an understanding of thousands of documents with no extra annotations other than existing assessment outcomes.

Conclusion: This work provides a systematic application and comparison of both traditional and state-of-the-art deep learning based XAI approaches to extract insights from a large corpus of text. Both traditional and deep learning approaches provide useful phrases, and we recommend using both to explore and understand large text corpora better. We will make our code available at https://github.com/oguzserbetci/explainer.

Objective: To compare Observational Medical Outcomes Partnership (OMOP) Logical Observation Identifiers Names and Codes (LOINC) and Veterans Aging Cohort Study (VACS) methods for extracting laboratory chemistry data from Veterans Health Administration (VA) electronic health records (EHR).

Materials and methods: Laboratory chemistry test results for 16 laboratory tests commonly assess in Veterans in VACS HIV (N = 143 830) followed in the VA 2015-2019 were extracted from the EHR and compared using 2 different data extraction approaches.

Results: The LOINC approach captured laboratory results from all 130 VA stations for all 16 labs. The VACS approach captured laboratory results from 128 to130 stations. Both approaches yielded laboratory results for a patient on a given date for 97% or more of the observations for 10 of the tests, 94%-97% for 5 of the tests, and 89% for 1 test. The percentage of exact matches on the value of the test result exceeded 99% for 10 of the laboratory tests and 92% for all other laboratory tests.

Discussion: Both approaches resulted in extraction of similar amounts of data in terms of individual patients, VA stations and total observations for all 16 tests. Both approaches yielded high agreement on test results in terms of identical values and correlation of test results for all tests.

Conclusion: The high level of agreement between VACS and LOINC approaches in this study demonstrate the favorable use of the LOINC approach for extracting laboratory results for most tests due to the ease and efficiency of this approach without compromising validity.