# Ethics & privacy extract — AI_READI D4D datasheet
# Source record: data/d4d_concatenated/claudecode_agent/2026-08-06_claude-opus-5-1m-generic-v3-schema2_rep1/AI_READI_d4d.yaml
# Source sha256: e5dc8a12c5584b4b0bc21aba9fc62a65bea1288d9f4042662ebe3e01202bd380
# Generated record: AI-extracted from project documentation (schema 2.0.0, run 2026-08-06_claude-opus-5-1m-generic-v3-schema2_rep1)
# Extract created: 2026-08-07
# Slots included (19): collection_consents, collection_notifications, confidential_elements, consent_revocations, content_warnings, data_protection_impacts, ethical_reviews, human_subject_research, informed_consent, ip_restrictions, is_deidentified, license_and_use_terms, participant_compensation, participant_privacy, prohibited_uses, regulatory_restrictions, retention_limit, sensitive_elements, subpopulations
# Values are verbatim from the record; 'source_caveats' and 'notes'
# fields carry the record's own evidence commentary.

collection_consents:
- consent_details: Written informed consent was provided by all participants before any protocol activity.
    For most participants, informed consent was performed remotely by computer, tablet or smart phone
    through a link in the hardcopy invitation letter or the invitation email that directed them into the
    REDCap database; participants who did not select electronic consent were consented in person at the
    start of the visit, followed by completion of all questionnaires. Consent was obtained in a private
    room by a study team member, with the opportunity to ask questions before signing and the option of
    additional time and a scheduled follow-up discussion.
collection_notifications:
- notification_details: Each individual was aware of the data collection, as this was active data collection
    directly from participants rather than passive collection or secondary use of existing data. Individuals
    in each recruitment pool were mailed a hardcopy invitation letter and sent an invitation email, both
    personalised to direct them into the online REDCap recruitment interface using links, access codes
    and QR codes. Within the REDCap interface, individuals could read an overview of the research programme,
    expectations for participation and answers to Frequently Asked Questions, download the informed consent
    document, request a call back from study staff, complete a screening survey for qualification, and
    enroll by signing the electronic consent. Those without internet access could call research staff
    for alternative methods of enrolment.
confidential_elements:
- confidential_elements_present: false
  confidentiality_details: The healthsheet states that the dataset does not contain data that might be
    considered confidential and that no personally identifiable information is included in the dataset.
consent_revocations:
- revocation_details: Participants were permitted to withdraw consent at any time and cease study participation.
    However, any data that had been shared or used up to that point would stay in the dataset. This is
    clearly communicated in the consent document.
content_warnings:
- content_warnings_present: false
  notes: The healthsheet answers "No" to whether the dataset contains data that, if viewed directly, might
    be offensive, insulting, threatening, or otherwise pose any safety risk such as psychological safety
    and anxiety.
data_protection_impacts:
- impact_details: No data protection impact analysis has been conducted. The v3.0.0 healthsheet answers
    the question "Has an analysis of the potential impact of the dataset and its use on data subjects
    (e.g., a data protection impact analysis) been conducted?" with "No, a data protection impact analysis
    has not been conducted."
ethical_reviews:
- review_details: AI-READI was approved by the Institutional Review Board of the University of Washington
    (approval number STUDY00016228), including reliance agreements with the IRBs of the University of
    Alabama at Birmingham and the University of California, San Diego. Initial approval was received on
    20 December 2022; the approval letter is published in the dataset documentation. Annual renewal reporting
    on the status and progress of the study is required within 90 days of expiration.
  reviewing_organization: University of Washington Institutional Review Board
- notes: Through engagement with the Native Biodata Consortium, the project provides a co-learning opportunity
    for understanding challenges specific to American Indian and Alaska Native communities and for evaluating
    tools such as data usage agreements, contract negotiations, transparent access agreements and forms
    of assured return of benefit and sustainable input, control and monitoring of Tribal data.
  review_details: Ethical, legal and social implications are considered at every stage of the project
    cycle, with guidance from bioethicists; the project reviewed existing documentation approaches such
    as datasheets and healthsheets to identify the most suitable one for biomedical data, and implements
    a robust data dissemination system, a strict licence agreement and data watermarking for both public
    and controlled sets to prevent re-identification attempts.
human_subject_research:
  ethics_review_board: Institutional Review Board of the University of Washington, with reliance agreements
    with the IRBs of the University of Alabama at Birmingham and the University of California, San Diego.
  involves_human_subjects: true
  irb_approval:
  - STUDY00016228, University of Washington Institutional Review Board; initial approval received 20 December
    2022, with the approval letter published at https://docs.aireadi.org/v3/Approval_STUDY00016228_Lee_initial.pdf.
    An annual renewal application reporting the status and progress of the study is required and due within
    90 days of expiration.
  notes: The BMJ Open protocol refers to the same number, STUDY00016228, as a "Clinicaltrials.org approval
    number".
  regulatory_compliance:
  - The FAIRhub study description oversight module records the study as not FDA-regulated for either drug
    or device, with human subject review status "Submitted, approved" and no Data Monitoring Committee.
    Written informed consent is provided by all participants. The study is registered on ClinicalTrials.gov
    as NCT06002048.
  special_populations:
  - Adults aged 40 years or older (maximum 85 years per the FAIRhub eligibility module) with and without
    type 2 diabetes, who are able to provide consent and who speak and read English. Pregnancy, gestational
    diabetes and type 1 diabetes are exclusionary, so pregnant women are not enrolled; children are not
    enrolled.
informed_consent:
- consent_documentation: The approved consent form for the principal project site, the University of Washington,
    is published at https://docs.aireadi.org/v3/AI-READI%20Consent_Form_Standard_Mod17May2023_useforpilot.pdf.
    The other clinical sites had IRB reliance and used the same consent form, with minor institution-specific
    language incorporated depending on individual institutional requirements (for example, the State of
    California requires inclusion of the Bill of Rights in the UCSD consent form). Once signed, the electronic
    consent form is automatically emailed to the participant, who may request an additional copy from
    the study coordinator for the duration of the study.
  consent_obtained: true
  consent_scope: Consent covered participation in the full study protocol, banking of biospecimens for
    future research, and placement of data in an open-source repository. Informed consent to participate
    was required before participation in any part of the protocol, including questionnaires. The dataset
    consent metadata records that the public version of the dataset can only be used for type 2 diabetes
    related research while a private version will allow for more generic use.
  consent_type: Written informed consent, obtained either electronically (e-consent signed in REDCap before
    the visit) or in person at the start of the visit for participants who did not select the electronic
    option.
  notes: Potential participants could read all consent documentation electronically before their visit
    and give consent with an electronic signature without verbal communication with a clinical research
    coordinator, or could decide at that point not to participate or to request additional information.
    No study-specific questionnaires other than the screening questionnaire were available to participants
    until consent was given.
  withdrawal_mechanism: Participants were permitted to withdraw consent at any time and cease study participation;
    however, any data that had been shared or used up to that point would stay in the dataset, and this
    is clearly communicated in the consent document.
ip_restrictions:
  notes: Transcribed from the AI-READI Data License Agreement v1.0. The healthsheet answers the question
    about third-party IP-based restrictions by referring to the license.
  restrictions:
  - Title to the data, and all industrial and intellectual property rights therein, remain solely and
    exclusively with the licensor (University of Washington) and its suppliers; the licensee shall not
    take any action inconsistent with such ownership and any rights not expressly granted are reserved.
  - The licensee shall not transfer, license, sublicense, sell, assign, display, share or otherwise convey
    any portion of the data or any derivative work to any third party other than another licensee bound
    by an agreement on identical terms.
  - Licensees must acknowledge the source and any funder of the data in any publications reporting use
    of the data, following the citation instructions in the dataset documentation.
is_deidentified:
  deidentification_details: 'The FAIRhub dataset_description records deIdentType "NoDeIdentification"
    with deIdentDirect and deIdentHIPAA both true and deIdentDates, deIdentNonarr and deIdentKAnon all
    false: no identifiers were collected, so no active de-identification was necessary, but the team checked
    that no identifiable data per US HIPAA were present in the data. The Nature Metabolism comment describes
    the public set as stripped of Protected Health Information as defined by the HIPAA Privacy Rule via
    the "Safe Harbor" method. The data in the public v3.0.0 release contain no protected health information.
    Information related to the sex and race/ethnicity of the participants, as well as medications used,
    has also been removed, in the case of sex and race/ethnicity to prevent stigmatization of findings.
    More sensitive variables (5-digit zip code, sex, race, ethnicity, genetic sequencing data, past health
    records, medications, and traffic and accident reports) are retained in a controlled-access dataset
    available only to scientists whose institutions have completed legal and privacy use agreements with
    the AI-READI research programme. Raw data are not shared because they may accidentally include personal
    health information or personally identifiable information, for example in free-text fields.'
  identifiable_elements_present: false
  identifiers_removed: Protected Health Information as defined by the HIPAA Privacy Rule (Safe Harbor
    method); Sex; Race and ethnicity; Medication information.
  method: HIPAA Safe Harbor verification; no direct identifiers collected
  notes: The healthsheet questions asking whether measures were taken to avoid re-identification and whether
    there was any pre-processing for de-identification of the patients were left unanswered in the v3.0.0
    record.
license_and_use_terms:
  contact_person: Aaron Lee, MD, central contact for the study, contact@aireadi.org; general contact form
    at https://aireadi.org/contact.
  data_use_permission: disease_specific_research
  license_terms: Version 3.0.0 is released under a custom license, "AI-READI custom license v2.0" (https://doi.org/10.5281/zenodo.17555036),
    specifically tailored to enable reuse of the AI-READI dataset (and other clinical datasets) for commercial
    or research purposes while placing strong requirements around data usage, security and secondary sharing
    to protect study participants, especially when data are reused for artificial intelligence and machine
    learning applications. The corresponding v1.0 license text present in the source bundle (University
    of Washington Data License Agreement, Zenodo 10.5281/zenodo.10642459) grants a non-exclusive, non-transferable
    license to download, reproduce and use the data and create derivative works for research and commercial
    purposes; forbids conveying any portion of the data or derivative works to third parties other than
    another identically bound licensee; permits distribution of models trained on the data provided they
    do not contain the data and provided reasonable efforts are made to minimise memorisation or reconstruction;
    forbids clinical treatment decisions based on the data; forbids re-identification or contact of data
    subjects; requires compliance with NIH Genomic Data Sharing Policy security and privacy standards;
    requires acknowledgement of the source and funder in publications; provides the data "as is" without
    warranty; limits the licensor's aggregate liability to one US dollar; requires indemnification of
    the licensor; and terminates automatically on any breach, after which the licensee must delete all
    copies.
  notes: 'Access to the dataset requires several steps: logging in through a verified ID system; agreeing
    to use the data only for type 2 diabetes related research; and agreeing to the license terms. The
    FAIRhub record labels the license "Health Data License" and the dataset accessType is "PublicDownloadSelfAttestationRequired".
    The license additionally requires acknowledgement of the source and any funder of the data in publications
    reporting use of the data, a condition that corresponds to the "publication_required" permission category
    as well as the recorded disease-specific restriction; only one value could be recorded in data_use_permission.
    The BMJ Open protocol describes an additional controlled-access set containing all publicly available
    data plus more sensitive data, available only to scientists whose institutions have completed legal
    and privacy use agreements with the AI-READI research programme, with requirements set by a Data Access
    Committee.'
participant_compensation:
- compensation_amount: USD 200
  compensation_provided: true
  compensation_rationale: Compensation for the burden of the study visit, which lasts 2.5-4 hours, plus
    a subsequent 10-day at-home monitoring period and return of the study devices.
  compensation_type: Monetary stipend, plus reimbursement of reasonable travel costs
  notes: The healthsheet states that study subjects received a compensation of $200 for the study visit,
    funded through the grant. The UW IRB protocol form states that the $200 stipend is provided when the
    final data are received from participants and the study devices are returned, typically at least two
    weeks after the study visit, that the payment is not prorated, and that the study covers reasonable
    costs for parking, public transit or rideshare fees related to the study visit; the same type and
    amount of compensation applies to participants asked to return in year 4. Costs for transport exceeding
    $25 may not be reimbursed. Study staff effort was supported by NIH award OT2OD032644 according to
    percentage of effort, with salaries aligned to the funding guidelines at each site.
participant_privacy:
- anonymization_method: No identifiers were collected, so no active de-identification was necessary; the
    team verified that no data identifiable per US HIPAA were present. The public release is stripped
    of Protected Health Information via the HIPAA Safe Harbor method and additionally of sex, race/ethnicity
    and medication information.
  data_linkage: Within the released dataset, records are linked across datatypes only by the participant's
    study ID number, which names the participant folder in each device directory; no direct identifiers
    are released alongside the data.
  privacy_techniques: HIPAA Safe Harbor removal and verification of Protected Health Information; removal
    of sex and race/ethnicity from the public set to prevent stigmatization of findings; removal of medication
    information from the public set; a two-tier release model in which more sensitive variables are available
    only in a controlled-access dataset requiring completed legal and privacy use agreements; data watermarking
    for both the public and controlled sets; a robust data dissemination system with a strict licence
    agreement, verified-ID login and self-attestation before download; selection of at-home monitoring
    devices that do not capture video or audio, are not equipped with GPS location monitoring and are
    not synced with participant-owned devices.
  reidentification_risk: A theoretical risk of future re-identification is acknowledged. AI algorithms
    can infer individual patient characteristics and could facilitate individual-level identification;
    to prevent re-identification attempts the project implements a robust data dissemination system, a
    strict licence agreement and data watermarking for both public and controlled sets. The license further
    prohibits any attempt to identify or contact data subjects or to establish their membership in a particular
    group.
prohibited_uses:
- notes: Transcribed from the University of Washington AI-READI Data License Agreement v1.0 (Zenodo record
    10.5281/zenodo.10642459), which is the only license text present in the source bundle. Version 3.0.0
    of the dataset is distributed under "AI-READI custom license v2.0" (https://doi.org/10.5281/zenodo.17555036),
    whose text is not present in the bundle, so it cannot be established from the bundle that this clause
    carries forward unchanged into v2.0.
  prohibition_reason: Licensees shall not make clinical treatment decisions based on the data, as it is
    intended solely as a research resource (AI-READI Data License Agreement v1.0, section 3(i)).
- notes: Transcribed from the University of Washington AI-READI Data License Agreement v1.0 (Zenodo record
    10.5281/zenodo.10642459), which is the only license text present in the source bundle. Version 3.0.0
    of the dataset is distributed under "AI-READI custom license v2.0" (https://doi.org/10.5281/zenodo.17555036),
    whose text is not present in the bundle, so it cannot be established from the bundle that this clause
    carries forward unchanged into v2.0.
  prohibition_reason: Licensees shall not use or attempt to use the data, alone or in concert with other
    information, to compromise or otherwise infringe the confidentiality of information on an individual
    data subject or their right to privacy, to identify or contact any individual data subject or group
    of data subjects, to extract or extrapolate identifying information, to establish a particular data
    subject's membership in a particular group of persons, or otherwise to cause harm or injury to any
    data subject (section 3(ii)).
- notes: Transcribed from the University of Washington AI-READI Data License Agreement v1.0 (Zenodo record
    10.5281/zenodo.10642459), which is the only license text present in the source bundle. Version 3.0.0
    of the dataset is distributed under "AI-READI custom license v2.0" (https://doi.org/10.5281/zenodo.17555036),
    whose text is not present in the bundle, so it cannot be established from the bundle that this clause
    carries forward unchanged into v2.0.
  prohibition_reason: Licensees shall not transfer, license, sublicense, sell, assign, display, share
    or otherwise convey any portion of the data or any derivative work to any third party other than another
    licensee bound by an agreement with the licensor on identical terms (section 2).
- notes: Transcribed from the University of Washington AI-READI Data License Agreement v1.0 (Zenodo record
    10.5281/zenodo.10642459), which is the only license text present in the source bundle. Version 3.0.0
    of the dataset is distributed under "AI-READI custom license v2.0" (https://doi.org/10.5281/zenodo.17555036),
    whose text is not present in the bundle, so it cannot be established from the bundle that this clause
    carries forward unchanged into v2.0.
  prohibition_reason: Licensees shall not disseminate models developed from the data without first undertaking
    all reasonable efforts to minimise the likelihood that data can be memorized, derived, reconstructed
    or reconstituted through the use or construction of such models (section 2).
regulatory_restrictions:
  confidentiality_level: restricted
  hipaa_compliant: compliant
  notes: The public dataset is stripped of Protected Health Information as defined by the HIPAA Privacy
    Rule using the Safe Harbor method, and the project verified that no data identifiable per US HIPAA
    were present. The healthsheet answers the question about export controls and other regulatory restrictions
    by referring to the license. Requirements for the controlled-access dataset were being developed by
    a Data Access Committee at the time of the BMJ Open protocol publication.
  regulatory_restrictions:
  - The licensee agrees to comply with all data security and privacy standards established by the US National
    Institutes of Health under its Genomic Data Sharing (GDS) Policy from time to time, including the
    NIH Security Best Practices for Controlled-Access Data Subject to the NIH GDS Policy.
  - The licensee agrees to comply with all applicable laws, regulations and restrictions relating to the
    distribution and use of the data and of models developed from it.
retention_limit:
  retention_details: There are no limits on the retention of the data associated with the instances; participants
    were not told that their data would be retained only for a fixed period and then deleted.
sensitive_elements:
- sensitive_elements_present: true
  sensitivity_details: Race and ethnic origin (self-reported). Not included in the public dataset; available
    only in the controlled-access dataset.
- sensitive_elements_present: true
  sensitivity_details: Biological sex. Removed from the public dataset to prevent stigmatization of findings;
    available only in the controlled-access dataset.
- sensitive_elements_present: true
  sensitivity_details: Location data in the form of 5-digit zip code. Held under controlled access only.
- sensitive_elements_present: true
  sensitivity_details: Genetic sequencing data. Held under controlled access only; genomic DNA is extracted
    from banked buffy coats for future studies such as whole genome sequencing.
- sensitive_elements_present: true
  sensitivity_details: Past health records and current medications (recorded with RxNorm codes). Held
    under controlled access only.
- sensitive_elements_present: true
  sensitivity_details: Motor vehicle traffic and accident reports, including driving record and accident
    reports obtained for the previous three years. Held under controlled access only.
subpopulations:
- distribution: 'Aggregate counts documented in the v3.0.0 README recommended-split table for all 2,280
    participants: Hispanic 380, Asian 545, Black 519, White 836. Within the recommended split the training
    set contains Hispanic 204, Asian 369, Black 343, White 660, while the validation and test sets each
    contain 88 participants in each of the four groups.'
  identification: Self-reported race/ethnicity in four groups (Asian, Black, Hispanic, White) is collected
    at screening but is withheld from the public dataset and released only under controlled access, so
    the public release does not identify this demographic sub-population at the instance level.
  notes: The Nature Metabolism comment states that information related to sex and race/ethnicity is stripped
    from the public set to prevent stigmatization of findings.
  subpopulation_elements_present: false
- distribution: 'Aggregate counts for all 2,280 participants: male 951, female 1,329. Training set male
    599, female 977; validation set male 176, female 176; test set male 176, female 176.'
  identification: Biological sex (male, female) is collected at screening but is withheld from the public
    dataset and released only under controlled access.
  subpopulation_elements_present: false
- distribution: 'Aggregate counts for all 2,280 participants: no diabetes 776, lifestyle 560, oral 686,
    insulin 258. Training set 600/384/487/105; validation set 88/88/109/67; test set 88/88/90/86.'
  identification: 'Diabetes status group in four categories used to balance the cohort: no diabetes; pre-diabetes
    or type 2 diabetes controlled by lifestyle adjustments; type 2 diabetes controlled by oral or non-insulin
    injectable medications; and insulin-controlled type 2 diabetes.'
  subpopulation_elements_present: true
- distribution: 'Mean age in years for the v3.0.0 cohort: overall 60.8 +/- 11.3; training set 61.1 +/-
    11.3; validation set 60.3 +/- 10.8; test set 60.5 +/- 11.4.'
  identification: Age of participants (inclusion criterion 40 years or older; the FAIRhub eligibility
    module records a maximum age of 85 years).
  subpopulation_elements_present: true
