AI READI Dataset Documentation

Datasheet for Dataset - Human Readable Format

🎯

Motivation

Why was the dataset created?

DescriptionID
Better understand salutogenesis (the pathway from disease to health) in Type 2 Diabetes Mellitus using a hypothesis-agnostic, harmonized, multi-domain dataset designed specifically for AI/ML research. The dataset aims to provide critical insights into how individuals can transition from diabetes toward health resilience through pseudotime manifold analysis of participants at different disease severity stages.
purpose-001
Establish standards, best practices, and guidelines for collection, preparation, and sharing of medical/health data sets targeted for AI/ML applications. This includes guidance from bioethicists on ethical and equitable data collection and management practices, with adherence to FAIR principles (Findable, Accessible, Interoperable, Reusable).
purpose-002
Address the lack of racial and ethnic diversity in T2DM research by creating a dataset that is triple-balanced across race/ethnicity (Asian, Black, Hispanic, White), biological sex (male, female), and diabetes severity (no diabetes, pre-diabetes/lifestyle-controlled, medication-controlled, insulin-controlled). This addresses demographic inequities in many previous epidemiological studies and clinical trials.
purpose-003
Create a flagship biomedical dataset as part of the NIH Bridge2AI program that is ethically sourced, trustworthy, well-defined and publicly available to facilitate the widespread use of AI/ML approaches in diabetes research and beyond.
purpose-004
  • ID
    funder-001
    Description
    NIH Common Fund Bridge2AI Program funded through National Institutes of Health grant 1OT2OD032644-01 (Bridge2AI Salutogenesis Data Generation), administered by NIH Office of the Director. Opportunity Number OTA-21-008. Project dates: September 1, 2022 to August 31, 2025 (later extended to November 30, 2026). Total funding in 2022: $5,026,499 (Direct: $4,569,437, Indirect: $457,062). Additional support from NIH grants P30DK035816 (University of Washington Nutrition and Obesity Research Center), UL1TR003096, and Research to Prevent Blindness.
📊

Composition

What do the instances represent?

  • ID
    instance-001
    Description
    Individual participants aged 40 and older with and without Type 2 Diabetes Mellitus (T2DM). Target enrollment is 4,000 people, triple-balanced by self-reported race/ethnicity (Asian, Black, Hispanic, White), T2DM severity (no diabetes, pre-diabetes/lifestyle-controlled diabetes, diabetes treated with oral medications or non-insulin injections, insulin-controlled diabetes), and biological sex (male, female). Participants must speak, read, and understand English. Exclusion criteria include pregnancy and type 1 diabetes mellitus. Recruitment occurs from patients aged ≥40 years who had medical encounters within each health system site (UAB, UCSD, UW) between 2020 and 2025.
DescriptionID
No Diabetes - Participants without Type 2 Diabetes Mellitus (no ICD-10 codes R73.09 or E11.X in electronic health records)
subpop-001
Pre-diabetes or Lifestyle-Controlled Diabetes - Participants with pre-diabetes or diabetes controlled through lifestyle modifications (ICD-10 code R73.09)
subpop-002
Medication-Controlled Diabetes - Participants with T2DM treated with oral medications or non-insulin injectable medications
subpop-003
Insulin-Controlled Diabetes - Participants with T2DM requiring insulin treatment
subpop-004
Asian Participants - Self-reported race/ethnicity Asian
subpop-005
Black Participants - Self-reported race/ethnicity Black
subpop-006
Hispanic Participants - Self-reported race/ethnicity Hispanic
subpop-007
White Participants - Self-reported race/ethnicity White
subpop-008
Male Participants - Biological sex male
subpop-009
Female Participants - Biological sex female
subpop-010
Access UrlsDescriptionIDName
https://fairhub.io/datasets/2, https://aireadi.org/goals/data-sharing
Publicly accessible dataset distributed through FAIRhub platform (https://fairhub.io/), a Microsoft Azure cloud-based data management platform developed specifically for AI-READI project. Public dataset includes survey data, blood and urine lab results, fitness activity levels, clinical measurements (monofilament, cognitive function testing), retinal images, ECG, blood sugar levels, and environmental variables (home air quality). Agreement with license required to download. First public release (pilot data) made available May 2024. Second release (data collected through July 31, 2024) made available November 2024. Yearly releases planned through project end. All data stored as 'AI-ready' enabling immediate AI/ML research without reformatting.
format-001FAIRhub Public Access Dataset
https://fairhub.io/datasets/2, https://aireadi.org/goals/data-sharing
Controlled access dataset contains all data in publicly accessible set plus more sensitive data available only for scientists whose institutions have completed legal and privacy use agreements with AI-READI research program. Additional data elements include 5-digit zip code, sex, race, ethnicity, genetic sequencing data, past health records, medications, and traffic/accident reports. Requirements for controlled access datasets currently being developed by Data Access Committee. Access procedures available at https://aireadi.org/goals/data-sharing.
format-002FAIRhub Controlled Access Dataset
https://fairhub.io/datasets/2
Retinal imaging data in DICOM (Digital Imaging and Communications in Medicine) standard format. 44 images per participant from six imaging devices. Images from some devices require conversion from proprietary formats (.fda, .sdt) to DICOM before upload to FAIRhub. Includes color fundus photos, optical coherence tomography (OCT), optical coherence tomography angiography (OCTA), infrared imaging, and fundus autofluorescence. Both eyes imaged per participant.
format-003Retinal Imaging Data - DICOM Format
https://fairhub.io/datasets/2
Clinical data mapped to Observational Medical Outcomes Partnership (OMOP) Common Data Model standard format. Includes demographics, vitals, medications (with RxNorm codes), clinical measurements, and laboratory test results. Enables use of OHDSI tool stack for analysis and integration with other OMOP-formatted datasets.
format-004Clinical Data - OMOP Common Data Model
https://fairhub.io/datasets/2
Wearable device data (Garmin VivoSmart 5 fitness tracker) transformed from proprietary .FIT format to mHealth standard. Includes physical activity metrics (steps, heart rate, sleep duration, oxygen saturation) collected over 10-day home monitoring period. CSV format for environmental sensor data (temperature, humidity, nitrogen oxides, VOCs, particulate matter, light intensity).
format-005Wearable Device Data - mHealth Standard
Contact UAB CCTS for biorepository access procedures
Biospecimen samples stored at UAB Center for Clinical and Translational Science biorepository available for future ancillary studies. Includes buffy coats, genomic DNA, EDTA plasma, serum, PAXgene RNA, and peripheral blood mononuclear cells. Finite number of samples available. Procedures for reviewing and prioritizing written requests will be developed before biorepository is complete. Samples shipped on dry ice to approved researchers.
format-006Biorepository Access
🔍

Collection Process

How was the data acquired?

https://fairhub.io/datasets/2
AI-READI
Artificial Intelligence Ready and Equitable Atlas for Diabetes Insights (AI-READI)
The AI-READI is a flagship dataset consisting of multimodal data collected from 4,000 individuals with and without Type 2 Diabetes Mellitus (T2DM), harmonized across 3 data collection sites (Birmingham, Alabama; San Diego, California; Seattle, Washington). The dataset was designed with future AI/Machine Learning studies in mind, including recruitment sampling procedures aimed at achieving approximately equal distribution of participants across diabetes severity (triple-balanced by race/ethnicity, biological sex, and T2DM severity), as well as a multi-domain data acquisition protocol (survey data, physical measurements, clinical data, retinal imaging, wearable device data, environmental sensors, biospecimens) to enable downstream AI/ML analyses that may not be feasible with existing data sources such as claims or electronic health records data. The goal is to better understand salutogenesis (the pathway from disease to health) in T2DM. The study follows FAIR principles and incorporates ethical and equitable data collection and management practices.
https://fairhub.io/datasets/2
en
  • Type 2 Diabetes Mellitus
  • T2DM
  • AI-READI
  • Machine Learning
  • Artificial Intelligence
  • multimodal dataset
  • harmonized data
  • multi-site study
  • salutogenesis
  • FAIR principles
  • retinal imaging
  • continuous glucose monitoring
  • wearable devices
  • biorepository
  • biospecimens
  • triple-balanced sampling
  • health equity
  • Bridge2AI
  • cross-sectional study
  • pseudotime manifold analysis
  • FAIRhub
  • health disparities
DescriptionID
Provide the largest publicly accessible multimodal T2DM dataset with 4,000 participants and over 10 variable domains, enabling AI/ML analyses not feasible with existing sources (e.g., claims or EHR alone). Previous datasets lacked the combination of scale, diversity, and multimodal depth needed for comprehensive AI/ML research.
gap-001
Address demographic underrepresentation in T2DM research by recruiting equal proportions across four race/ethnic groups (Asian, Black, Hispanic, White) and both biological sexes. A meta-epidemiological review of T2DM studies published from 2000 to 2020 indicated that samples lacked racial and ethnic diversity, making AI-READI an important step forward in addressing these demographic inequities.
gap-002
Create a model for future AI-ready medical datasets through comprehensive metadata, standardized data formats (OMOP CDM for clinical data, DICOM for imaging, mHealth standards for wearables), FAIR compliance, and ethical data governance practices that can be replicated for other health conditions.
gap-003
RoleNameORCIDAffiliation
Contributorcreator-001-
Contributorcreator-002-
Contributorcreator-003-
Contributorcreator-004-
Contributorcreator-005-
Contributorcreator-006-
Contributorcreator-007-
Contributorcreator-008-
Contributorcreator-009-
Contributorcreator-010-
Contributorcreator-011-
Contributorcreator-012-
Contributorcreator-013-
DescriptionID
Survey and Questionnaire Data: Demographics, Center for Epidemiological Studies Depression Scale, Problem Areas in Diabetes questionnaire, Diabetes Score (self-care assessment), dietary assessment, ophthalmic survey, smoking/alcohol/vaping/marijuana use history, general health history, Social Determinants of Health surveys (food and job insecurity, educational attainment, health insurance, access to healthcare, housing, neighborhood environment, perceptions of discrimination in medical settings, racial/ethnic discrimination), and current medications with RxNorm codes. Collected via REDCap database.
subset-001
Physical Measurements and Vitals: Height, weight, waist and hip circumference, waist-hip ratio, body mass index, systolic and diastolic blood pressure, heart rate. Measured during single visit lasting 2.5-4 hours.
subset-002
Retinal Imaging Data: Multidevice retinal imaging including Aurora IQ (undilated fundus photos), EIDON widefield truecolor confocal fundus system, Spectralis OCT/OCTA, Maestro2 3D OCT-1, Triton DRI OCT, Cirrus 5000, and Fluorescence Lifetime Imaging Ophthalmoscopy. Both eyes imaged per participant. Images exported in raw formats and converted to DICOM standard. Total of 44 images per participant across multiple devices capturing color fundus photos, optical coherence tomography, optical coherence tomography angiography, infrared imaging, and fundus autofluorescence.
subset-003
Electrocardiogram Data: 12-lead ECG performed using Philips Pagewriter TC30 Cardiograph while participant sits in reclining chair or lies supine. Position recorded (0°, 30°, 60°, or 90° relative to supine). Data exported in XML format.
subset-004
Cognitive Function Testing: Montreal Cognitive Assessment (MoCA) administered electronically on iPad using MoCA Duo Application. Total possible score is 30, with higher numbers representing better performance. Generates total score, section subscores, Memory Index Score, and task completion times in CSV format.
subset-005
Peripheral Neuropathy Assessment: Monofilament test assessing touch perception on both feet using 10g filament in three locations per foot (10 times per location). Participants respond yes/no whether they feel the filament with eyes closed and shoes/socks removed.
subset-006
Visual Function Testing: Photopic and mesopic visual acuity measured with Electronic Visual Acuity tester (right and left eyes tested separately). Photopic and mesopic letter contrast sensitivity measured with Mars chart. Mesopic conditions achieved by viewing through 2.0 neutral density filter. Autorefraction data (spherical and cylindrical components expressed as diopter with cylindrical axis) obtained using Topcon KR 800 autorefractor for each eye.
subset-007
Clinical Laboratory Tests: Comprehensive blood and urine testing including HbA1c, complete blood count, EDTA plasma tests (NT-proBNP, troponin-T, C-peptide, insulin), serum tests (CRP-HS, total cholesterol, triglycerides, HDL/LDL cholesterol, glucose, blood urea nitrogen, creatinine, electrolytes, calcium, protein, albumin, globulin, bilirubin, alkaline phosphatase, AST, ALT), and urine tests (creatinine, albumin). Testing performed by CLIA-certified local laboratories (CBC on fresh samples) and University of Washington Nutrition and Obesity Research Center (NORC) testing facility (frozen samples). Results provided in CSV format.
subset-008
Continuous Glucose Monitoring: Dexcom G6 Continuous Glucose Monitor worn for 10 days at home, measuring blood glucose (mg/dL) every 5 minutes. Participants masked to results during monitoring period. Data exported in CSV format. Participants receive Dexcom report with average blood glucose levels (hourly, daily, overall) and Glucose Management Indicator after 10-day period.
subset-009
Physical Activity Monitoring: Garmin VivoSmart 5 fitness tracker worn for 10 days at home, collecting number of steps, heart rate, sleep duration (circadian and diurnal rhythm), and oxygen saturation. Device chosen for privacy (no video/audio, no GPS, not synced with participant devices). Data exported in FIT file format and converted to mHealth standard. Participants indicate which wrist worn and dominant hand.
subset-010
Home Environmental Monitoring: Custom-designed environmental sensor placed in home for 10 days, collecting ambient temperature, relative humidity, nitrogen oxides (NO and NO2), volatile organic compounds, particulate matter (PM1.0, PM2.5, PM4, PM10), and multi-spectral light intensity measurements (11 channels). Device designed with privacy as primary concern (no video/audio, no GPS). Data exported in CSV format. Participants indicate sensor placement location in home.
subset-011
Biorepository Specimens: Blood derivatives including buffy coats (for DNA extractions and future genomics), genomic DNA, EDTA plasma (for future proteomics and metabolomics), serum (for future proteomics and metabolomics), PAXgene vacutainers (for future RNA isolations and gene expression studies), and peripheral blood mononuclear cells (for future immunological studies and generation of induced pluripotent stem cells). Biospecimens stored at UAB Center for Clinical and Translational Science (CCTS) biorepository at -70°C. Samples batch-shipped from UCSD and UW to UAB on dry ice approximately quarterly. Standardized operating procedures ensure consistent handling.
subset-012
DescriptionID
Triple-Balanced Sampling Design: Purposive recruitment to achieve approximately equal distribution across three dimensions: (1) four race/ethnic groups (Asian, Black, Hispanic, White), (2) four categories of T2DM severity (no diabetes, pre-diabetes/lifestyle-controlled diabetes, diabetes treated with oral medications or non-insulin injections, insulin-controlled diabetes), and (3) two biological sexes (male, female). Building balanced data sets is critical for development of unbiased machine learning models. Rather than targeting demographic distribution of US population, study recruits equal proportions across all groups.
sampling-001
Multi-Site Geographic Sampling: Recruitment across three geographically diverse data collection sites (Birmingham, Alabama; San Diego, California; Seattle, Washington) to capture regional diversity. All study groups recruited from each site, but proportion of each group varies depending on demographic prevalence in each site's geographical area.
sampling-002
Wave-Based Recruitment: Participants recruited in waves to facilitate efficient sampling and allow coordinators to respond to interested persons with expediency. Composition and size of each wave influenced by observed participation characteristics of race/ethnicity, gender, severity of T2DM, and site according to health records. As recruitment progresses, composition monitored for diversity and inclusion and adjusted by under- and oversampling groups as needed.
sampling-003
DescriptionID
Electronic Health Record Screening and Mail/Email Recruitment: Source population identified by screening electronic health records at each health system site (UAB, UCSD, UW) for patients aged ≥40 years with medical encounters between 2020 and 2025. Patients with T2DM and pre-diabetes identified by ICD-10 diagnosis codes R73.09 and E11.X. Contact pools mailed hardcopy invitation letters and sent invitation emails, both personalized with links, access codes, and QR codes directing to REDCap recruitment interface. Enrollment began July 18, 2023 and continues until November 30, 2026.
collection-001
Online Informed Consent and Questionnaire Completion: Most participants provide informed consent remotely via computer, tablet, or smartphone through REDCap database. Once in REDCap interface, individuals read research overview, expectations, and FAQs, download informed consent document, request call back from study staff, complete screening survey for qualification, and enroll by signing electronic consent. Once enrolled, access to study questionnaires provided. Alternative in-person consent available for those without internet access.
collection-002
Single-Visit In-Person Data Collection: Protocol performed as single visit lasting 2.5-4 hours at one of three data collection sites (Birmingham, San Diego, Seattle). All research personnel underwent training by data manager at UAB using detailed Manual of Procedures and completed certification process including mandatory institutional compliance training for human subject research. Coordinators required to successfully administer all protocol elements to at least three volunteer practice subjects meeting MOP standards before enrolling and testing participants. Preliminary pilot enrollment occurred July 18 - November 30, 2023. Formal data collection began December 1, 2023.
collection-003
Home-Based 10-Day Monitoring: At end of in-person visit, participants sent home with three data monitoring devices: (1) Dexcom G6 Continuous Glucose Monitor, (2) Garmin VivoSmart 5 fitness tracker, and (3) custom-designed environmental sensor. Coordinators provide instructions for continuous use for 10 days. Participants return devices to study staff by overnight mail (materials and fees provided) or in-person after 10 days. Devices chosen with participant privacy as primary concern (no video/audio, no GPS, not synced with participant devices).
collection-004
DescriptionID
Standardized Clinical Assessment Protocol: Vitals measured using automatic oscillometric medically approved blood pressure device (two measurements 2 minutes apart). ECG performed using Philips Pagewriter TC30 Cardiograph with participant position recorded. Monofilament test performed on both feet. MoCA cognitive assessment administered electronically on iPad. Visual acuity and contrast sensitivity assessed under photopic and mesopic conditions using Electronic Visual Acuity tester and Mars chart. Autorefraction obtained using Topcon KR 800 autorefractor.
acquisition-001
Multidevice Retinal Imaging Protocol: Both eyes imaged on each participant using six retinal imaging devices capturing 44 total images per participant. All images collected under dilated conditions except Optomed. Devices include Aurora IQ (4 images), EIDON widefield system (12 images), Spectralis OCT/OCTA (6 images), Maestro2 3D OCT-1 (6 images), Triton DRI OCT (6 images), Cirrus 5000 (8 images), and Fluorescence Lifetime Imaging Ophthalmoscopy (2 images). Images exported in raw format and some converted to DICOM standard (.fda files to DICOM, .sdt files to DICOM).
acquisition-002
Biospecimen Collection and Processing: Blood (53 mL total) collected in multiple vacutainer types: EDTA vacutainers (whole blood for CBC, plasma for lab tests and biobanking, buffy coats for DNA extraction), serum separator vacutainers (serum for lab tests and biobanking), CPT mononuclear cell preparation tubes (peripheral blood mononuclear cells for immunological studies and iPSC generation), and PAXgene RNA vacutainers (stabilized unisolated RNA for gene expression studies). Urine collected for lab testing. Local processing using standardized operating procedures. Samples for NORC lab testing stored locally at -70°C and batch-shipped to University of Washington on dry ice quarterly. Biobanking samples shipped to UAB CCTS for central biorepository.
acquisition-003
REDCap Data Management and Cloud Upload: REDCap used for collecting patient-reported questionnaire data and clinical visit data (medications, vitals, visual acuity, contrast sensitivity, monofilament, CBC results). Device data exported in raw formats: ECG (.xml), MoCA (.csv with scores and completion times), retinal imaging (various formats converted to DICOM), environmental sensors (.csv), CGMs (.csv), fitness trackers (.FIT converted to mHealth standard), NORC lab tests (.csv). All data mapped to applicable standards (OMOP CDM for clinical data, DICOM for imaging, mHealth for wearables). Data uploaded at regular intervals to AI-READI-specific FAIRhub platform (Microsoft Azure cloud-based). Data managers at each site responsible for quality control before upload.
acquisition-004
DescriptionID
Data Format Standardization and Transformation: Raw data from imaging devices transformed from proprietary formats to standard formats prior to upload to FAIRhub. Retinal images exported in raw formats (.fda, .sdt) and converted to DICOM standard. Fitness tracker data exported in .FIT format and converted to mHealth standard. All clinical data mapped to Observational Medical Outcomes Partnership (OMOP) Common Data Model. Medications coded using RxNorm codes.
preproc-001
Quality Control and Data Validation: Data managers at each data collection site (UAB, UCSD, UW) responsible for quality control of all data before uploading to FAIRhub. All coordinators underwent certification process requiring successful administration of all protocol elements to at least three volunteer practice subjects meeting Manual of Procedures standards. Preliminary pilot enrollment (July 18 - November 30, 2023) conducted to secure sufficient familiarity for all coordinators before formal data collection began December 1, 2023.
preproc-002
De-identification and Privacy Protection: Data de-identified before release. Controlled access dataset contains 5-digit zip code, sex, race, ethnicity, genetic sequencing data, past health records, medications, and traffic/accident reports. Public dataset excludes these sensitive elements. Home monitoring devices specifically chosen with privacy as primary concern (no video/audio capture, no GPS location monitoring, not synced with participant-owned devices). Retinal imaging technicians trained to detect life-threatening or vision-threatening conditions for immediate participant referral.
preproc-003
AI-Readiness Preparation: All data stored and shared using FAIRhub as 'AI-ready' which enables immediate AI/ML research without reformatting and preprocessing. Data harmonized across three data collection sites. Comprehensive metadata provided. FAIR principles (Findable, Accessible, Interoperable, Reusable) implemented throughout data lifecycle. Detailed dataset documentation provided at https://docs.aireadi.org/.
preproc-004
DescriptionID
Multi-Site Data Harmonization: Data collected from three geographically diverse sites (Birmingham, San Diego, Seattle) harmonized to ensure consistency. Standardized operating procedures implemented for biospecimen handling across sites. All coordinators trained using same Manual of Procedures. Certification process ensures coordinators meet consistent standards. Central laboratories used for most clinical testing (University of Washington NORC for plasma, serum, whole blood, urine; local CLIA-certified labs for CBC only).
cleaning-001
Biospecimen Batch Processing: Centralized processing strategies implemented for consistency. Samples for NORC lab testing stored locally at -70°C and batch-shipped to University of Washington quarterly on dry ice. Plasma, serum, and buffy coats from UCSD and UW batch-shipped to UAB CCTS for integration into central biobank. Genomic DNA extractions from stored buffy coats performed at UAB CCTS. Peripheral blood mononuclear cell isolation centralized at UAB CCTS (CPT tubes pre-centrifuged locally, then overnight delivery to UAB). PAXgene RNA vacutainers stable for 72 hours allowing overnight shipment from UCSD/UW to UAB CCTS.
cleaning-002
DescriptionID
Comprehensive Metadata and Documentation: Detailed dataset documentation provided using Healthsheet template at https://docs.aireadi.org/. Documentation includes overall dataset description, details for each data domain (clinical context, data acquisition, relevant variables, data processing details, file formats, data standards, metadata, example outputs). One version of documentation associated with each dataset version. AI-readiness section describes FAIR principles implementation and ethical practices. Additional resources section provides related materials.
labeling-001
Community Advisory Board Input: Community Advisory Board of 11 persons from three sites including diversity in race and ethnicity as represented in AI-READI sample contributes to protocol development. Ensures community perspectives integrated into study design and implementation.
labeling-002
DescriptionIDMaintainer DetailsName
Multi-institutional consortium managing dataset maintenance including three data collection sites (University of Alabama at Birmingham, University of California San Diego, University of Washington) and supporting institutions. Data managers at each site oversee data quality control before upload to FAIRhub platform. Central coordination through University of Washington (contact PI Aaron Lee).
maintainer-001University of Washington - Contact PI site, data collection, biospecimen processing (NORC lab), University of Alabama at Birmingham - Data collection, central biorepository (CCTS), University of California San Diego - Data collection, retinal imaging expertise, FAIRhub platform team - Cloud infrastructure, FAIR compliance, data standards, Data Access Committee - Controlled access procedures (in development)AI-READI Consortium
Committee responsible for developing and overseeing requirements for controlled access dataset. Manages legal and privacy use agreements with institutions requesting access to sensitive data elements. Procedures currently in development. Contact information and access requirements available at https://aireadi.org/goals/data-sharing.
maintainer-002Develops controlled access requirements, Reviews institutional data use agreements, Oversees privacy and legal compliance, Manages biorepository access proceduresData Access Committee
DescriptionIDNameSensitive Elements PresentSensitivity Details
Dataset contains protected health information requiring de-identification before release. Controlled access dataset includes 5-digit zip code, sex, race, ethnicity, genetic sequencing data, past health records, medications, and traffic/accident reports. Public dataset excludes these sensitive elements. All data de-identified before release with separate controlled access and public access tiers.
sensitive-001Protected Health Information and DemographicsTrue5-digit zip code (controlled access only), Sex, race, ethnicity (controlled access only), Genetic sequencing data (controlled access only, biorepository), Past health records (controlled access only), Medications with RxNorm codes (controlled access only), Traffic and accident reports (controlled access only), Complete electronic health record data used for screening, Biospecimens enabling future genetic studies
Comprehensive clinical laboratory test results including HbA1c, complete blood count, biomarkers for heart failure, diabetes, kidney function, liver function, and overall health status. Laboratory test results provided to participants by HIPAA-compliant encrypted email with information on normative values. Some test results returned during visit (heart rate, blood pressure, visual acuity).
sensitive-002Clinical and Laboratory DataTrueHbA1c and other diabetes biomarkers, Complete blood count results, Cardiovascular biomarkers (troponin, NT-proBNP), Kidney function markers (creatinine, BUN, urine albumin), Liver function tests, Lipid panels and metabolic markers, Continuous glucose monitoring data (10-day period)
Comprehensive retinal imaging from multiple devices (44 images per participant) potentially revealing health conditions. Retinal imaging technicians trained to detect life-threatening or vision-threatening conditions (retinal detachment, tumor, optic disc edema) for immediate participant referral. All retinal images collected under dilated conditions (except Optomed). Images converted to DICOM standard with embedded metadata.
sensitive-003Retinal Imaging and Medical ImagesTrue44 retinal images per participant from six devices, Potential detection of diabetic retinopathy, cardiovascular disease, neurological conditions, Immediate referral protocol for emergency conditions, DICOM format with metadata
Ten-day home monitoring data from environmental sensors, continuous glucose monitors, and fitness trackers. Devices specifically chosen with privacy as primary concern (no video/audio, no GPS location monitoring, not synced with participant devices). Includes physical activity patterns, sleep patterns, home air quality, and continuous glucose levels.
sensitive-004Home Monitoring and Behavioral DataTruePhysical activity patterns and sleep duration, Continuous glucose measurements (every 5 minutes for 10 days), Home air quality and environmental conditions, Privacy-preserving device design (no video/audio/GPS), Participants masked to glucose monitoring results during collection
ID
retention-001
Name
Long-term Preservation via FAIRhub and NIH Repositories
Description
Dataset preserved long-term through FAIRhub platform (Microsoft Azure cloud-based) and NIH-approved repositories. All dataset versions preserved with persistent identifiers for reproducibility. Biorepository samples stored at -70°C at UAB Center for Clinical and Translational Science with institutional support. No planned deletion or retention limits. Project funded through November 30, 2026 with data releases continuing through that date. Long-term maintenance supported through institutional commitments and NIH data sharing policies.
Retention Details
  • Indefinite preservation through FAIRhub platform
  • Biospecimens stored at -70°C at UAB CCTS biorepository
  • All dataset versions preserved with persistent identifiers
  • No planned deletion or data expiration
  • NIH data sharing policy compliance
  • Institutional support for long-term maintenance
🚀

Uses

What (other) tasks could the dataset be used for?

DescriptionID
Enable multi-domain AI/ML analyses for T2DM by collecting data across survey, clinical, imaging, wearable device, environmental, and biospecimen domains that may not be feasible with existing data sources such as claims or electronic health records data alone. The multimodal nature of the data supports complex machine learning model development and training.
task-001
Support the development of unbiased machine learning models through balanced data collection across demographic groups and diabetes severity levels. The triple-balanced design is critical for preventing algorithmic bias in AI/ML applications and ensuring health equity.
task-002
Study disease trajectories and salutogenesis pathways in T2DM through cross-sectional analysis of participants at different disease stages, enabling pseudotime manifold analysis to predict disease progression and paths to health resilience. Data collection is specifically designed to permit downstream pseudotime manifold analysis by collecting data from participants with differing disease severity.
task-003
Recruit and train a diverse biomedical AI/ML workforce by increasing access to and quality of AI/ML research. The project includes skills and workforce development components to train personnel in AI-ready dataset generation and analysis.
task-004
DescriptionID
AI/ML Model Training and Development for T2DM Research: Primary intended use is training and development of artificial intelligence and machine learning models for Type 2 Diabetes Mellitus research. AI-ready datasets with full metadata, standardized formats, and FAIR compliance enable immediate use in AI/ML pipelines without reformatting and preprocessing. Multimodal data supports complex machine learning model development across multiple health domains.
use-001
Salutogenesis and Disease Trajectory Analysis: Study of salutogenesis pathways (transitions from disease to health) and disease trajectories in T2DM. Cross-sectional data from participants at different disease severity stages enables pseudotime manifold analysis to predict disease progression and paths to health resilience. Dataset designed specifically to permit downstream pseudotime manifold analysis.
use-002
Health Equity and Disparities Research: Research into health disparities and development of equitable AI/ML models. Triple-balanced sampling design (race/ethnicity, biological sex, T2DM severity) enables analysis of differences across demographic groups and supports development of unbiased machine learning models that perform equally well across all populations.
use-003
Biospecimen-Based Ancillary Studies: Biorepository specimens (buffy coats, genomic DNA, plasma, serum, PAXgene RNA, peripheral blood mononuclear cells) available to researchers for future ancillary studies including genomics (whole genome sequencing), proteomics, metabolomics, gene expression studies, immunological studies, and generation of induced pluripotent stem cells. Procedures for reviewing and prioritizing written requests will be developed before biorepository is complete.
use-004
DescriptionID
Clinical Decision-Making Without Validation: Dataset consists of research data from volunteer participants and should not be used for clinical decision-making or patient care without appropriate validation, regulatory oversight, and approval. Results from research assessments provided to participants for informational purposes only, not for clinical diagnosis or treatment.
discouraged-001
Generalization Beyond English-Speaking Populations: Study requires participants to speak, read, and understand English, which may limit generalizability to non-English-speaking populations. AI/ML models trained on this dataset may not perform as well in populations with different language backgrounds or cultural contexts.
discouraged-002
ID
license-001
Name
Creative Commons Attribution-NonCommercial 4.0 International
Description
AI-READI dataset distributed under Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license for the publicly accessible dataset. Controlled access dataset requires additional legal and privacy use agreements between requesting institution and AI-READI research program. Public dataset available upon agreement with license defining permitted uses. Controlled access requirements currently being developed by Data Access Committee. Citation required: Owsley C, Matthies DS, McGwin G, et al. Cross-sectional design and protocol for Artificial Intelligence Ready and Equitable Atlas for Diabetes Insights (AI-READI). BMJ Open 2025;15:e097449. DOI: 10.1136/bmjopen-2024-097449. Biospecimen use requires separate agreements. Commercial use may require additional licensing. All mandatory institutional compliance training for human subjects research required for data users working with identifiable or sensitive data.
License Terms
  • CC BY-NC 4.0 for publicly accessible dataset
  • Controlled access requires institutional data use agreement
  • Attribution required to AI-READI Consortium and publication DOI
  • Citation required - BMJ Open 2025;15:e097449
  • NonCommercial use only without additional licensing
  • Biorepository access requires separate procedures
  • Compliance with ethical guidelines and FAIR principles
  • No clinical use without appropriate validation and approval
  • FAIRhub platform terms of service apply
📤

Distribution

How will the dataset be distributed?

CC BY-NC 4.0
🔄

Maintenance

How will the dataset be maintained?

ID
updates-001
Name
Yearly Dataset Releases and Ongoing Enrollment
Description
Dataset released approximately yearly with progressive accumulation of participants. Pilot data from July 18 - November 30, 2023 released May 2024. Data collected through July 31, 2024 released November 2024. Enrollment continues until November 30, 2026 with target of 4,000 total participants. After one year of enrollment, sites on schedule for reaching final sample of 4,000 within three years of data collection. As enrollment is ongoing, pilot data release and periodic updates may not have achieved balanced distribution across groups yet. Final dataset expected late 2026 or early 2027. Long-term preservation plans through FAIRhub platform and NIH data repositories.
Frequency
Approximately yearly releases through November 2026
Update Details
  • Pilot data release (May 2024) - Data from July 18 - November 30, 2023
  • Version 1.0 release (November 2024) - Data collected through July 31, 2024
  • Ongoing enrollment until November 30, 2026
  • Target enrollment 4,000 participants
  • Final dataset release expected late 2026 or early 2027
  • Each release includes updated documentation at https://docs.aireadi.org/
👥

Human Subjects

Does the dataset relate to people?

ID
hsr-001
Name
AI-READI Human Subjects Research
Description
Cross-sectional study of 4,000 human participants aged 40 and older with and without Type 2 Diabetes Mellitus. Approved by Institutional Review Board (IRB) of University of Washington (approval number STUDY00016228) including reliance agreements with IRBs of University of Alabama at Birmingham and University of California San Diego. Written informed consent provided by all participants (electronic consent via REDCap or in-person consent for those without internet access). Community Advisory Board of 11 persons from three sites contributes to protocol development. Study incorporates ethical and equitable data collection and management practices with guidance from bioethicists. Recruitment designed to address historical underrepresentation of racial and ethnic minorities in T2DM research.
Involves Human Subjects
True
Ethics Review Board
  • University of Washington Institutional Review Board (approval STUDY00016228)
  • University of Alabama at Birmingham IRB (reliance agreement)
  • University of California San Diego IRB (reliance agreement)
  • Community Advisory Board (11 members across three sites)
  • Bioethics guidance team (Kadija Ferryman, Camille Nebeker)
Generated on 2026-04-21 15:34:02 using Bridge2AI Data Sheets Schema