BACKGROUND: Learning and teaching interdisciplinary health data science (HDS) is highly challenging, and despite the growing interest in HDS education, little is known about the learning experiences and preferences of HDS students.
OBJECTIVE: We conducted a systematic review to identify learning preferences and strategies in the HDS discipline.
METHODS: We searched 10 bibliographic databases (PubMed, ACM Digital Library, Web of Science, Cochrane Library, Wiley Online Library, ScienceDirect, SpringerLink, EBSCOhost, ERIC, and IEEE Xplore) from the date of inception until June 2023. We followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines and included primary studies written in English that investigated the learning preferences or strategies of students in HDS-related disciplines, such as bioinformatics, at any academic level. Risk of bias was independently assessed by 2 screeners using the Mixed Methods Appraisal Tool, and we used narrative data synthesis to present the study results.
RESULTS: After abstract screening and full-text reviewing of the 849 papers retrieved from the databases, 8 (0.9%) studies, published between 2009 and 2021, were selected for narrative synthesis. The majority of these papers (7/8, 88%) investigated learning preferences, while only 1 (12%) paper studied learning strategies in HDS courses. The systematic review revealed that most HDS learners prefer visual presentations as their primary learning input. In terms of learning process and organization, they mostly tend to follow logical, linear, and sequential steps. Moreover, they focus more on abstract information, rather than detailed and concrete information. Regarding collaboration, HDS students sometimes prefer teamwork, and sometimes they prefer to work alone.
CONCLUSIONS: The studies\' quality, assessed using the Mixed Methods Appraisal Tool, ranged between 73% and 100%, indicating excellent quality overall. However, the number of studies in this area is small, and the results of all studies are based on self-reported data. Therefore, more research needs to be conducted to provide insight into HDS education. We provide some suggestions, such as using learning analytics and educational data mining methods, for conducting future research to address gaps in the literature. We also discuss implications for HDS educators, and we make recommendations for HDS course design; for example, we recommend including visual materials, such as diagrams and videos, and offering step-by-step instructions for students.

Climate change, local epidemics, future pandemics, and forced displacements pose significant public health threats worldwide. To cope successfully, people and communities are faced with the challenging task of developing resilience to these stressors. Our viewpoint is that the powerful capabilities of modern informatics technologies including artificial intelligence, biomedical and environmental sensors, augmented or virtual reality, data science, and other digital hardware or software, have great potential to promote, sustain, and support resilience in people and communities. However, there is no \"one size fits all\" solution for resilience. Solutions must match the specific effects of the stressor, cultural dimensions, social determinants of health, technology infrastructure, and many other factors.

OBJECTIVE: Collaborate, Analyse, Research and Audit (CARA) project set out to provide an infrastructure to enable Irish general practitioners (GPs) to use their routinely collected patient management software (PMS) data to better understand their patient population, disease management and prescribing through data dashboards. This paper explains the design and development of the CARA infrastructure.
METHODS: The first exemplar dashboard was developed with GPs and focused on antibiotic prescribing to develop and showcase the proposed infrastructure. The data integration process involved extracting, loading and transforming de-identified patient data into data models which connect to the interactive dashboards for GPs to visualise, compare and audit their data.
RESULTS: The architecture of the CARA infrastructure includes two main sections: extract, load and transform process (ELT, de-identified patient data into data models) and a Representational State Transfer Application Programming Interface (REST API) (which provides the security barrier between the data models and their visualisation on the CARA dashboard). CARAconnect was created to facilitate the extraction and de-identification of patient data from the practice database.
CONCLUSIONS: The CARA infrastructure allows seamless connectivity with and compatibility with the main PMS in Irish general practice and provides a reproducible template to access and visualise patient data. CARA includes two dashboards, a practice overview and a topic-specific dashboard (example focused on antibiotic prescribing), which includes an audit tool, filters (within practice) and between-practice comparisons.
CONCLUSIONS: CARA supports evidence-based decision-making by providing GPs with valuable insights through interactive data dashboards to optimise patient care, identify potential areas for improvement and benchmark their performance against other practices.Supplementary file 1. Graphical abstract.

Methods for causal inference from observational data are common in human disease epidemiology and social sciences but are used relatively little in plant pathology. We draw upon an extensive data set of the incidence of hop plants with powdery mildew (Podosphaera macularis) collected from yards in Oregon during 2014 to 2017 and associated metadata on grower cultural practices, cultivar susceptibility to powdery mildew, and pesticide application records to understand variation in and causes of growers\' fungicide use and associated costs. An instrumental causal forest model identified growers\' spring pruning thoroughness, cultivar susceptibility to two of the dominant pathogenic races of P. macularis, network centrality of a yards during May-June and June-July time transitions, and the initial strain of the fungus were important variables determining the number of pesticide active constituents applied by growers and the associated costs they incurred in response to powdery mildew. Exposure-response function models fit after covariate weighting indicated both the number of pesticide active constituents applied and their associated costs scaled linearly with the seasonal mean incidence of plants with powdery mildew. While the causes of pesticide use intensity are multifaceted, biological and production factors collectively influence the incidence of powdery mildew, which has a direct exposure-response relationship on the number of pesticide active constituents that growers apply and their costs. Our analyses point to several potential strategies for reducing pesticide use and costs for management of powdery mildew on hop. We also highlight the utility of these methods for causal inference in observational studies.

UNASSIGNED: Insufficient physical activity (PA) has long been a global health issue, and a number of studies have explored correlates of PA to identify the mechanisms underlying inactive lifestyles. In the literature, dozens of correlates have been identified at different (eg, individual, environmental) levels, but there is little or no direct evidence for the mutual associations of these correlates. This study analysed 44 variables identified as theoretically and empirically relevant for PA to clarify the factors directly and indirectly associated with PA.
UNASSIGNED: A cross-sectional survey dataset of 19 005 Japanese-speaking adults (mean age=53.50 years, SD=17.40; 9706 women) was analysed. The data encompassed demographic and anthropometric variables; self-reported PA levels; perceived social support and environments (eg, awareness of urban facilities for PA); psychological traits and health-behaviour characteristics (eg, personality, motivation, self-efficacy, decisional balance, process of change strategies); and technology use (eg, mobile health apps).
UNASSIGNED: Network analyses were performed to select meaningful associations (partial correlations) among variables, which identified nine variables directly positively associated with PA: job/employment status, self-efficacy, perceived social support, intrinsic motivation, stage of change, counter conditioning, self-reevaluation, environment and technology use. Indirect associations (two-step neighbourhood) were identified for 40 (out of 44) variables, implying that most of the known PA-correlates are associated with PA-at least indirectly.
UNASSIGNED: These identified associations echo the importance of the multilevel perspective in understanding how people maintain (in)active lifestyles. Interventions for PA could have mixed-level targets, including intraindividual characteristics, social support and physical and digital environments.

Metadata describe and provide context for other data, playing a pivotal role in enabling findability, accessibility, interoperability, and reusability (FAIR) data principles. By providing comprehensive and machine-readable descriptions of digital resources, metadata empower both machines and human users to seamlessly discover, access, integrate, and reuse data or content across diverse platforms and applications. However, the limited accessibility and machine-interpretability of existing metadata for population health data hinder effective data discovery and reuse.
To address these challenges, we propose a comprehensive framework using standardized formats, vocabularies, and protocols to render population health data machine-readable, significantly enhancing their FAIRness and enabling seamless discovery, access, and integration across diverse platforms and research applications.
The framework implements a 3-stage approach. The first stage is Data Documentation Initiative (DDI) integration, which involves leveraging the DDI Codebook metadata and documentation of detailed information for data and associated assets, while ensuring transparency and comprehensiveness. The second stage is Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) standardization. In this stage, the data are harmonized and standardized into the OMOP CDM, facilitating unified analysis across heterogeneous data sets. The third stage involves the integration of Schema.org and JavaScript Object Notation for Linked Data (JSON-LD), in which machine-readable metadata are generated using Schema.org entities and embedded within the data using JSON-LD, boosting discoverability and comprehension for both machines and human users. We demonstrated the implementation of these 3 stages using the Integrated Disease Surveillance and Response (IDSR) data from Malawi and Kenya.
The implementation of our framework significantly enhanced the FAIRness of population health data, resulting in improved discoverability through seamless integration with platforms such as Google Dataset Search. The adoption of standardized formats and protocols streamlined data accessibility and integration across various research environments, fostering collaboration and knowledge sharing. Additionally, the use of machine-interpretable metadata empowered researchers to efficiently reuse data for targeted analyses and insights, thereby maximizing the overall value of population health resources. The JSON-LD codes are accessible via a GitHub repository and the HTML code integrated with JSON-LD is available on the Implementation Network for Sharing Population Information from Research Entities website.
The adoption of machine-readable metadata standards is essential for ensuring the FAIRness of population health data. By embracing these standards, organizations can enhance diverse resource visibility, accessibility, and utility, leading to a broader impact, particularly in low- and middle-income countries. Machine-readable metadata can accelerate research, improve health care decision-making, and ultimately promote better health outcomes for populations worldwide.

暂无摘要。

BACKGROUND: The Veterans Affairs Surgical Quality Improvement Program (VASQIP) trains surgical quality nurses (SQNs) at each Veterans Affairs (VA) hospital to extract or verify 187 variables from the medical record for all cardiac surgical cases. For ten preoperative laboratory values, VASQIP has a semiautomated (SA) system in which local lab values are automatically extracted, verified by SQNs, and lab values recorded at other VA facilities are manually extracted. The objective of this study was to develop and validate a method to automate the extraction of these ten preoperative laboratory values and compare results with the current SA method.
METHODS: We developed methods to extract ten preoperative laboratory values and measurement dates from the VA Corporate Data Warehouse using Logical Observation Identifiers Names and Codes. Automated (A) versus SA information extraction was compared in terms of agreement, conformance to data definitions, proximity to surgery, and missingness.
RESULTS: For surgeries with both A and SA lab values, the intraclass correlation coefficients for the ten variables ranged from 0.90 to 0.98. For several variables, the A method resulted in much lower rates of missing data (e.g., 2.4% versus 22.5% missing data for high-density lipoprotein) and eliminated out-of-date-range entries.
CONCLUSIONS: Although SQN-extracted data are widely considered the gold standard within National Surgical Quality Improvement Programs, there may be advantages to fully automating extraction of lab values, including high congruence with SA SQN-extracted or verified values and lower rates of missingness and out-of-date-range data.

BACKGROUND: The impact of missing data on individual continuous glucose monitoring (CGM) data is unknown but can influence clinical decision-making for patients.
OBJECTIVE: We aimed to investigate the consequences of data loss on glucose metrics in individual patient recordings from continuous glucose monitors and assess its implications on clinical decision-making.
METHODS: The CGM data were collected from patients with type 1 and 2 diabetes using the FreeStyle Libre sensor (Abbott Diabetes Care). We selected 7-28 days of 24 hours of continuous data without any missing values from each individual patient. To mimic real-world data loss, missing data ranging from 5% to 50% were introduced into the data set. From this modified data set, clinical metrics including time below range (TBR), TBR level 2 (TBR2), and other common glucose metrics were calculated in the data sets with and that without data loss. Recordings in which glucose metrics deviated relevantly due to data loss, as determined by clinical experts, were defined as expert panel boundary error (εEPB). These errors were expressed as a percentage of the total number of recordings. The errors for the recordings with glucose management indicator <53 mmol/mol were investigated.
RESULTS: A total of 84 patients contributed to 798 recordings over 28 days. With 5%-50% data loss for 7-28 days recordings, the εEPB varied from 0 out of 798 (0.0%) to 147 out of 736 (20.0%) for TBR and 0 out of 612 (0.0%) to 22 out of 408 (5.4%) recordings for TBR2. In the case of 14-day recordings, TBR and TBR2 episodes completely disappeared due to 30% data loss in 2 out of 786 (0.3%) and 32 out of 522 (6.1%) of the cases, respectively. However, the initial values of the disappeared TBR and TBR2 were relatively small (<0.1%). In the recordings with glucose management indicator <53 mmol/mol the εEPB was 9.6% for 14 days with 30% data loss.
CONCLUSIONS: With a maximum of 30% data loss in 14-day CGM recordings, there is minimal impact of missing data on the clinical interpretation of various glucose metrics.
BACKGROUND: ClinicalTrials.gov NCT05584293; https://clinicaltrials.gov/study/NCT05584293.

BACKGROUND: Despite the increasing availability of electronic healthcare record (EHR) data and wide availability of plug-and-play machine learning (ML) Application Programming Interfaces, the adoption of data-driven decision-making within routine hospital workflows thus far, has remained limited. Through the lens of deriving clusters of diagnoses by age, this study investigated the type of ML analysis that can be performed using EHR data and how results could be communicated to lay stakeholders.
METHODS: Observational EHR data from a tertiary paediatric hospital, containing 61 522 unique patients and 3315 unique ICD-10 diagnosis codes was used, after preprocessing. K-means clustering was applied to identify age distributions of patient diagnoses. The final model was selected using quantitative metrics and expert assessment of the clinical validity of the clusters. Additionally, uncertainty over preprocessing decisions was analysed.
RESULTS: Four age clusters of diseases were identified, broadly aligning to ages between: 0 and 1; 1 and 5; 5 and 13; 13 and 18. Diagnoses, within the clusters, aligned to existing knowledge regarding the propensity of presentation at different ages, and sequential clusters presented known disease progressions. The results validated similar methodologies within the literature. The impact of uncertainty induced by preprocessing decisions was large at the individual diagnoses but not at a population level. Strategies for mitigating, or communicating, this uncertainty were successfully demonstrated.
CONCLUSIONS: Unsupervised ML applied to EHR data identifies clinically relevant age distributions of diagnoses which can augment existing decision making. However, biases within healthcare datasets dramatically impact results if not appropriately mitigated or communicated.