Journal of the American Medical Informatics Association最新文献

Objectives: We describe new curriculum materials for engaging secondary school students in exploring the "big data" in the NIH All of Us Research Program's Public Data Browser and the co-design processes used to collaboratively develop the materials. We also describe the methods used to develop and validate assessment items for studying the efficacy of the materials for student learning as well as preliminary findings from these studies.

Materials and methods: Secondary-level biology teachers from across the United States participated in a 2.5-day Co-design Summer Institute. After learning about the All of Us Research Program and its Data Browser, they collaboratively developed learning objectives and initial ideas for learning experiences related to exploring the Data Browser and big data. The Genetic Science Learning Center team at the University of Utah further developed the educators' ideas. Additional teachers and their students participated in classroom pilot studies to validate a 22-item instrument that assesses students' knowledge. Educators completed surveys about the materials and their experiences.

Results: The "Exploring Big Data with the All of Us Data Browser" curriculum module includes 3 data exploration guides that engage students in using the Data Browser, 3 related multimedia pieces, and teacher support materials. Pilot testing showed substantial growth in students' understanding of key big data concepts and research applications.

Discussion and conclusion: Our co-design process provides a model for educator engagement. The new curriculum module serves as a model for introducing secondary students to big data and precision medicine research by exploring diverse real-world datasets.

Objectives: To highlight the use of calibration weighting to improve the precision of estimates obtained from All of Us data and increase the return of value to communities from the All of Us Research Program.

Materials and methods: We used All of Us (2017-2022) data and raking to obtain prevalence estimates in two examples: discrimination in medical settings (N = 41 875) and food insecurity (N = 82 266). Weights were constructed using known population proportions (age, sex, race/ethnicity, region of residence, annual household income, and home ownership) from the 2020 National Health Interview Survey.

Results: About 37% of adults experienced discrimination in a medical setting. About 20% of adults who had not seen a doctor reported being food insecure compared with 14% of adults who regularly saw a doctor.

Conclusions: Calibration using raking is cost-effective and may lead to more precise estimates when analyzing All of Us data.

Objectives: To improve firearm injury encounter classification (new vs follow-up) using machine learning (ML) and compare our ML model to other common approaches.

Materials and methods: This retrospective study used data from the St Louis region-wide hospital-based violence intervention program data repository (2010-2020). We randomly selected 500 patients with a firearm injury diagnosis for inclusion, with 808 total firearm injury encounters split (70/30) for training and testing. We trained a least absolute shrinkage and selection operator (LASSO) regression model with the following predictors: admission type, time between firearm injury visits, number of prior firearm injury emergency department (ED) visits, encounter type (ED or other), and diagnostic codes. Our gold standard for new firearm injury encounter classification was manual chart review. We then used our test data to compare the performance of our ML model to other commonly used approaches (proxy measures of ED visits and time between firearm injury encounters, and diagnostic code encounter type designation [initial vs subsequent or sequela]). Performance metrics included area under the curve (AUC), sensitivity, and specificity with 95% confidence intervals (CIs).

Results: The ML model had excellent discrimination (0.92, 0.88-0.96) with high sensitivity (0.95, 0.90-0.98) and specificity (0.89, 0.81-0.95). AUC was significantly higher than time-based outcomes, sensitivity was slightly (but not significantly) lower than other approaches, and specificity was higher than all other methods.

Discussion: ML successfully delineated new firearm injury encounters, outperforming other approaches in ruling out encounters for follow-up.

Conclusion: ML can be used to identify new firearm injury encounters and may be particularly useful in studies assessing re-injuries.

Objective: This study aims to develop machine learning models that provide both accurate and equitable predictions of 2-year stroke risk for patients with atrial fibrillation across diverse racial groups.

Materials and methods: Our study utilized structured electronic health records (EHR) data from the All of Us Research Program. Machine learning models (LightGBM) were utilized to capture the relations between stroke risks and the predictors used by the widely recognized CHADS2 and CHA2DS2-VASc scores. We mitigated the racial disparity by creating a representative tuning set, customizing tuning criteria, and setting binary thresholds separately for subgroups. We constructed a hold-out test set that not only supports temporal validation but also includes a larger proportion of Black/African Americans for fairness validation.

Results: Compared to the original CHADS2 and CHA2DS2-VASc scores, significant improvements were achieved by modeling their predictors using machine learning models (Area Under the Receiver Operating Characteristic curve from near 0.70 to above 0.80). Furthermore, applying our disparity mitigation strategies can effectively enhance model fairness compared to the conventional cross-validation approach.

Discussion: Modeling CHADS2 and CHA2DS2-VASc risk factors with LightGBM and our disparity mitigation strategies achieved decent discriminative performance and excellent fairness performance. In addition, this approach can provide a complete interpretation of each predictor. These highlight its potential utility in clinical practice.

Conclusions: Our research presents a practical example of addressing clinical challenges through the All of Us Research Program data. The disparity mitigation framework we proposed is adaptable across various models and data modalities, demonstrating broad potential in clinical informatics.

Background: Acute hepatic porphyria (AHP) is a group of rare but treatable conditions associated with diagnostic delays of 15 years on average. The advent of electronic health records (EHR) data and machine learning (ML) may improve the timely recognition of rare diseases like AHP. However, prediction models can be difficult to train given the limited case numbers, unstructured EHR data, and selection biases intrinsic to healthcare delivery. We sought to train and characterize models for identifying patients with AHP.

Methods: This diagnostic study used structured and notes-based EHR data from 2 centers at the University of California, UCSF (2012-2022) and UCLA (2019-2022). The data were split into 2 cohorts (referral and diagnosis) and used to develop models that predict (1) who will be referred for testing of acute porphyria, among those who presented with abdominal pain (a cardinal symptom of AHP), and (2) who will test positive, among those referred. The referral cohort consisted of 747 patients referred for testing and 99 849 contemporaneous patients who were not. The diagnosis cohort consisted of 72 confirmed AHP cases and 347 patients who tested negative. The case cohort was 81% female and 6-75 years old at the time of diagnosis. Candidate models used a range of architectures. Feature selection was semi-automated and incorporated publicly available data from knowledge graphs. Our primary outcome was the F-score on an outcome-stratified test set.

Results: The best center-specific referral models achieved an F-score of 86%-91%. The best diagnosis model achieved an F-score of 92%. To further test our model, we contacted 372 current patients who lack an AHP diagnosis but were predicted by our models as potentially having it (≥10% probability of referral, ≥50% of testing positive). However, we were only able to recruit 10 of these patients for biochemical testing, all of whom were negative. Nonetheless, post hoc evaluations suggested that these models could identify 71% of cases earlier than their diagnosis date, saving 1.2 years.

Conclusions: ML can reduce diagnostic delays in AHP and other rare diseases. Robust recruitment strategies and multicenter coordination will be needed to validate these models before they can be deployed.

Objective: Artificial intelligence (AI) models trained using medical images for clinical tasks often exhibit bias in the form of subgroup performance disparities. However, since not all sources of bias in real-world medical imaging data are easily identifiable, it is challenging to comprehensively assess their impacts. In this article, we introduce an analysis framework for systematically and objectively investigating the impact of biases in medical images on AI models.

Materials and methods: Our framework utilizes synthetic neuroimages with known disease effects and sources of bias. We evaluated the impact of bias effects and the efficacy of 3 bias mitigation strategies in counterfactual data scenarios on a convolutional neural network (CNN) classifier.

Results: The analysis revealed that training a CNN model on the datasets containing bias effects resulted in expected subgroup performance disparities. Moreover, reweighing was the most successful bias mitigation strategy for this setup. Finally, we demonstrated that explainable AI methods can aid in investigating the manifestation of bias in the model using this framework.

Discussion: The value of this framework is showcased in our findings on the impact of bias scenarios and efficacy of bias mitigation in a deep learning model pipeline. This systematic analysis can be easily expanded to conduct further controlled in silico trials in other investigations of bias in medical imaging AI.

Conclusion: Our novel methodology for objectively studying bias in medical imaging AI can help support the development of clinical decision-support tools that are robust and responsible.

Objective: To evaluate the use of patient portal messaging to recruit individuals historically underrepresented in biomedical research (UBR) to the All of Us Research Program (AoURP) at a single recruitment site.

Materials and methods: Patient portal-based recruitment was implemented at Columbia University Irving Medical Center. Patient engagement was assessed using patient's electronic health record (EHR) at four recruitment stages: Consenting to be contacted, opening messages, responding to messages, and showing interest in participating. Demographic and socioeconomic data were also collected from patient's EHR and univariate logistic regression analyses were conducted to assess patient engagement.

Results: Between October 2022 and November 2023, a total of 59 592 patients received patient portal messages inviting them to join the AoURP. Among them, 24 445 (41.0%) opened the message, 8983 (15.1%) responded, and 3765 (6.3%) showed interest in joining the program. Though we were unable to link enrollment data with EHR data, we estimate about 2% of patients contacted ultimately enrolled in the AoURP. Patients from underrepresented race and ethnicity communities had lower odds of consenting to be contacted and opening messages, but higher odds of showing interest after responding.

Discussion: Patient portal messaging provided both patients and recruitment staff with a more efficient approach to outreach, but patterns of engagement varied across UBR groups.

Conclusion: Patient portal-based recruitment enables researchers to contact a substantial number of participants from diverse communities. However, more effort is needed to improve engagement from underrepresented racial and ethnic groups at the early stages of the recruitment process.

Objectives: The goal of this case report is to detail experiences and challenges experienced in the training of Primary Care residents in secondary analysis using All of Us Researcher Workbench. At our large, urban safety net hospital, Primary Care/Internal Medicine residents in their third year undergo a research intensive block, the Research Practicum, where they work as a team to conduct secondary data analysis on a dataset with faculty facilitation. In 2023, this research block focused on use of the All of Us Researcher Workbench for secondary data analysis.

Materials and methods: Two groups of 5 residents underwent training to access the All of Us Researcher Workbench, and each group explored available data with a faculty facilitator and generated original research questions. Two blocks of residents successfully completed their research blocks and created original presentations on "social isolation and A1C" levels and "medical discrimination and diabetes management."

Results: Departmental faculty were satisfied with the depth of learning and data exploration. In focus groups, some residents noted that for those without interest in performing research, the activity felt extraneous to their career goals, while others were glad for the opportunity to publish. In both blocks, residents highlighted dissatisfaction with the degree to which the All of Us Researcher Workbench was representative of patients they encounter in a large safety net hospital.

Discussion: Using the All of Us Researcher Workbench provided residents with an opportunity to explore novel questions in a massive data source. Many residents however noted that because the population described in the All of Us Researcher Workbench appeared to be more highly educated and less racially diverse than patients they encounter in their practice, research may be hard to generalize in a community health context. Additionally, given that the data required knowledge of 1 of 2 code-based data analysis languages (R or Python) and work within an idiosyncratic coding environment, residents were heavily reliant on a faculty facilitator to assist with analysis.

Conclusion: Using the All of Us Researcher Workbench for research training allowed residents to explore novel questions and gain first-hand exposure to opportunities and challenges in secondary data analysis.

Objective: Our objective is to develop and validate TrajVis, an interactive tool that assists clinicians in using artificial intelligence (AI) models to leverage patients' longitudinal electronic medical records (EMRs) for personalized precision management of chronic disease progression.

Materials and methods: We first perform requirement analysis with clinicians and data scientists to determine the visual analytics tasks of the TrajVis system as well as its design and functionalities. A graph AI model for chronic kidney disease (CKD) trajectory inference named DisEase PrOgression Trajectory (DEPOT) is used for system development and demonstration. TrajVis is implemented as a full-stack web application with synthetic EMR data derived from the Atrium Health Wake Forest Baptist Translational Data Warehouse and the Indiana Network for Patient Care research database. A case study with a nephrologist and a user experience survey of clinicians and data scientists are conducted to evaluate the TrajVis system.

Results: The TrajVis clinical information system is composed of 4 panels: the Patient View for demographic and clinical information, the Trajectory View to visualize the DEPOT-derived CKD trajectories in latent space, the Clinical Indicator View to elucidate longitudinal patterns of clinical features and interpret DEPOT predictions, and the Analysis View to demonstrate personal CKD progression trajectories. System evaluations suggest that TrajVis supports clinicians in summarizing clinical data, identifying individualized risk predictors, and visualizing patient disease progression trajectories, overcoming the barriers of AI implementation in healthcare.

Discussion: The TrajVis system provides a novel visualization solution which is complimentary to other risk estimators such as the Kidney Failure Risk Equations.

Conclusion: TrajVis bridges the gap between the fast-growing AI/ML modeling and the clinical use of such models for personalized and precision management of chronic diseases.

Objectives: ABO blood types have widespread clinical use and robust associations with disease. The purpose of this study is to evaluate the portability and suitability of tag single-nucleotide polymorphisms (tSNPs) used to determine ABO alleles and blood types across diverse populations in published literature.

Materials and methods: Bibliographic databases were searched for studies using tSNPs to determine ABO alleles. We calculated linkage between tSNPs and functional variants across inferred continental ancestry groups from 1000 Genomes. We compared r2 across ancestry and assessed real-world consequences by comparing tSNP-derived blood types to serology in a diverse population from the All of Us Research Program.

Results: Linkage between functional variants and O allele tSNPs was significantly lower in African (median r2 = 0.443) compared to East Asian (r2 = 0.946, P = 1.1 × 10-5) and European (r2 = 0.869, P = .023) populations. In All of Us, discordance between tSNP-derived blood types and serology was high across all SNPs in African ancestry individuals and linkage was strongly correlated with discordance across all ancestries (ρ = -0.90, P = 3.08 × 10-23).

Discussion: Many studies determine ABO blood types using tSNPs. However, tSNPs with low linkage disequilibrium promote misinference of ABO blood types, particularly in diverse populations. We observe common use of inappropriate tSNPs to determine ABO blood type, particularly for O alleles and with some tSNPs mistyping up to 58% of individuals.

Conclusion: Our results highlight the lack of transferability of tSNPs across ancestries and potential exacerbation of disparities in genomic research for underrepresented populations. This is especially relevant as more diverse cohorts are made publicly available.