AMIA ... Annual Symposium proceedings. AMIA Symposium最新文献

Clinician informatics leadership has been identified as an essential component of addressing the 'implementation to benefits realization gap' that exists for many digital health technologies. Chief Medical Informatics Officers (CMIOs), and Chief Nursing Informatics Officers (CNIOs) are well-positioned to ensure the success of these initiatives. However, while the CMIO role is fairly well-established in Canada, there is limited uptake of CNIO roles in the country. The main objective of this work is to build on the current progress of the CMIO role and explore how the CNIO role can be best positioned for uptake and value across healthcare organizations in Canada. A qualitative study was conducted. Ten clinician leaders in CMIO, CNIO, and related roles in Canada were interviewed about the value of these roles and strategies for supporting the uptake of the role. This study provides the foundation for future initiatives for supporting and showcasing the value of the CNIO in a digitally enabled healthcare organization.

Patient representation learning methods create rich representations of complex data and have potential to further advance the development of computational phenotypes (CP). Currently, these methods are either applied to small predefined concept sets or all available patient data, limiting the potential for novel discovery and reducing the explainability of the resulting representations. We report on an extensive, data-driven characterization of the utility of patient representation learning methods for the purpose of CP development or automatization. We conducted ablation studies to examine the impact of patient representations, built using data from different combinations of data types and sampling windows on rare disease classification. We demonstrated that the data type and sampling window directly impact classification and clustering performance, and these results differ by rare disease group. Our results, although preliminary, exemplify the importance of and need for data-driven characterization in patient representation-based CP development pipelines.

Identifying disease-gene associations is important for understanding molecule mechanisms of diseases, finding diagnostic markers and therapeutic targets. Many computational methods have been proposed to predict disease related genes by integrating different biological databases into heterogeneous networks. However, it remains a challenging task to leverage heterogeneous topological and semantic information from multi-source biological data to enhance disease-gene prediction. In this study, we propose a knowledge graph-based disease-gene prediction system (GenePredict-KG) by modeling semantic relations extracted from various genotypic and phenotypic databases. We first constructed a knowledge graph that comprised 2,292,609 associations between 73,358 entities for 14 types of phenotypic and genotypic relations and 7 entity types. We developed a knowledge graph embedding model to learn low-dimensional representations of entities and relations, and utilized these embeddings to infer new disease-gene interactions. We compared GenePredict-KG with several state-of-the-art models using multiple evaluation metrics. GenePredict-KG achieved high performances [AUROC (the area under receiver operating characteristic) = 0.978, AUPR (the area under precision-recall) = 0.343 and MRR (the mean reciprocal rank) = 0.244], outperforming other state-of-art methods.

Home healthcare (HHC) agencies provide care to more than 3.4 million adults per year. There is value in studying HHC narrative notes to identify patients at risk for deterioration. This study aimed to build machine learning algorithms to identify "concerning" narrative notes of HHC patients and identify emerging themes. Six algorithms were applied to narrative notes (n = 4,000) from a HHC agency to classify notes as either "concerning" or "not concerning." Topic modeling using Latent Dirichlet Allocation bag of words was conducted to identify emerging themes from the concerning notes. Gradient Boosted Trees demonstrated the best performance with a F-score = 0.74 and AUC = 0.96. Emerging themes were related to patient-clinician communication, HHC services provided, gait challenges, mobility concerns, wounds, and caregivers. Most themes have been cited by previous literature as increasing risk for adverse events. In the future, such algorithms can support early identification of patients at risk for deterioration.

This review reports the user experience of symptom checkers, aiming to characterize users studied in the existing literature, identify the aspects of user experience of symptom checkers that have been studied, and offer design suggestions. Our literature search resulted in 31 publications. We found that (1) most symptom checker users are relatively young; (2) eight relevant aspects of user experience have been explored, including motivation, trust, acceptability, satisfaction, accuracy, usability, safety/security, and functionality; (3) future symptom checkers should improve their accuracy, safety, and usability. Although many facets of user experience have been explored, methodological challenges exist and some important aspects of user experience remain understudied. Further research should be conducted to explore users' needs and the context of use. More qualitative and mixed-method studies are needed to understand actual users' experiences in the future.

Overly restricted and poorly designed eligibility criteria reduce the generalizability of the results from clinical trials. We conducted a study to identify and quantify the impacts of study traits extracted from eligibility criteria on the age of study populations in Alzheimer's Disease (AD) clinical trials. Using machine learning methods and SHapley Additive exPlanation (SHAP) values, we identified 30 and 34 study traits that excluded older patients from AD trials in our 2 generated target populations respectively. We also found that study traits had different magnitudes of impacts on the age distributions of the generated study populations across racial-ethnic groups. To our best knowledge, this was the first study that quantified the impact of eligibility criteria on the age of AD trial participants. Our research is a first step in addressing the overly restrictive eligibility criteria in AD clinical trials.

Unplanned 30-day cancer readmissions are an important outcome of cancer hospitalization and can significantly raise mortality rates and costs for both the patient and the hospital. This paper aimed to develop a predictive model using machine learning and electronic health records to predict unplanned 30-day cancer readmissions and further develop it as a clinical decision support system. The three-stage study design followed the 2022 AMIA Artificial Intelligence Evaluation Showcase. In the first stage, the technical performance of the model was determined (81% of AUROC) and contributing factors were identified. In the second stage, the technical feasibility and workflow considerations of using such a predictive model were explored through semi-structured interviews. In the third stage, a decision tree analysis and a cost estimation showed that the model can reduce unplanned readmissions significantly if timely action is taken and that preventing a single readmission may significantly reduce costs.

Assessments of Life-space Mobility (LSM) evaluate the locations of movement and their frequency over a period of time to understand mobility patterns. Advancements in and miniaturization of GPS sensors in mobile devices like smartwatches could facilitate objective and high-resolution assessment of life-space mobility. The purpose of this study was to compare self-reported measures to GPS-based LSM extracted from 27 participants (44.4% female, aged 65+ years) who wore a smartwatch for 1-2 weeks at two different site locations (Connecticut and Florida). GPS features (e.g., excursion size/span) were compared to self-reported LSM with and without an indicator for needing assistance. Although correlations between self-reported measures and GPS-based LSM were positive, none were statistically significant. The correlations improved slightly when needing assistance was included, but statistical significance was achieved only for excursion size (r=0.40, P=0.04). The poor correlations between GPS-based and self-reported indicators suggest that they capture different dimensions of life-space mobility.

A significant proportion of clinical physiologic monitoring alarms are false. This often leads to alarm fatigue in clinical personnel, inevitably compromising patient safety. To combat this issue, researchers have attempted to build Machine Learning (ML) models capable of accurately adjudicating Vital Sign (VS) alerts raised at the bedside of hemodynamically monitored patients as real or artifact. Previous studies have utilized supervised ML techniques that require substantial amounts of hand-labeled data. However, manually harvesting such data can be costly, time-consuming, and mundane, and is a key factor limiting the widespread adoption of ML in healthcare (HC). Instead, we explore the use of multiple, individually imperfect heuristics to automatically assign probabilistic labels to unlabeled training data using weak supervision. Our weakly supervised models perform competitively with traditional supervised techniques and require less involvement from domain experts, demonstrating their use as efficient and practical alternatives to supervised learning in HC applications of ML.

Our objective was to detect common barriers to post-acute care (B2PAC) among hospitalized older adults using natural language processing (NLP) of clinical notes from patients discharged home when a clinical decision support system recommended post-acute care. We annotated B2PAC sentences from discharge planning notes and developed an NLP classifier to identify the highest-value B2PAC class (negative patient preferences). Thirteen machine learning models were compared with Amazon's AutoGluon deep learning model. The study included 594 acute care notes from 100 patient encounters (1156 sentences contained 11 B2PAC) in a large academic health system. The most frequent and modifiable B2PAC class was negative patient preferences (18.3%). The best supervised model was Extreme Gradient Boosting (F1: 0.859), but the deep learning model performed better (F1: 0.916). Alerting clinicians of negative patient preferences early in the hospitalization can prompt interventions such as patient education to ensure patients receive the right level of care and avoid negative outcomes.