Frequently Asked Questions

  • AD (Alzheimer’s disease)
  • AD-CDEs (AD/ADRD CDEs)
  • ADNI (Alzheimer’s Disease Neuroimaging Initiative)
  • ADRC (Alzheimer’s Disease Research Center)
  • ADRD (AD-Related Dementias)
  • AI/ML (Artificial Intelligence/Machine Learning)
  • CDEs (Common Data Elements)
  • CDM (Common Data Model)
  • CMS (Centers for Medicare & Medicaid Services)
  • CTSA (Clinical and Translational Science Award)
  • FAIR (Findable, Accessible, Interoperable, Reusable)
  • KG (Knowledge Graph)
  • NACC (National Alzheimer’s Coordinating Center)
  • NLP (Natural Language Processing)
  • OHNLP (Open Health NLP)
  • ReCARDO (this U24 project)
  • RRN (ReCARDO Research Network)
  • RWD (Real-World Data)
  • RWE (Real-World Evidence)
  • SC (Steering Committee)
  • UIMA (Unstructured Information Management Application)
  • UDS (Uniform Data Set, used by NACC)
  • WS (Work Stream for this project)

ReCARDO is a national ten-institution U24 research network designed to create, validate, apply, and disseminate Common Data Elements (CDEs) optimized for Alzheimer’s disease and AD-related dementias (AD/ADRD).

It integrates AI/ML/NLP, ontology-driven metadata modeling, and federated data ecosystems to harmonize Real-World Data (RWD) and generate high-quality Real-World Evidence (RWE).

Current CDEs (e.g., NACC, NINDS) were not built to support semantic interoperability, federated analyses, or large-scale RWD harmonization across EHRs, CMS claims, ADRC cohorts, and biobanks.
ReCARDO addresses four critical national challenges:
1. Multidisciplinary team integration (clinicians + informaticians + AI scientists)
2. CDE and RWD lifecycle management through ecosystems
3. Ensuring relevance to NIA-funded studies and ADRCs
4. Achieving transparency, rigor, reproducibility, and explainability

ReCARDO uses a structured governance model:
Steering Committee (SC)
• MPIs + Work Stream Leads
• Coordinates operations and scientific direction
• Oversees pilot awards & annual CDE User Meeting

External Advisory Panel (EAP)
• Appointed by NIA
• Provides scientific and strategic feedback

Administrative Hub & Key Functions (KF1–KF5)
Led at UTHealth Houston, includes operations, training, dissemination, governance, and innovation capacity support.

Seven thematic Work Streams (WS) execute scientific aims:

1. WS1: CDE Development & CDE Knowledge Graph
2. WS2: CDE Extraction from RWD
3. WS3: Data, Metadata & Resource Ecosystems
4. WS4: AI/ML/NLP & Computational Phenotyping
5. WS5: Research Questions, Use Cases & Pilot Studies
6. WS6: Genetics, Imaging, Biomarkers & Big Data Integration
7. WS7: Reproducibility & Evaluation
Each WS is co-led across institutions to ensure cross-site synergy.

Aim 1: Develop AD/ADRD CDEs & CDE Knowledge Graph (CDE-Engine)
• Uses ontology QA, semantic modeling, and metadata standards
• Ensures semantic interoperability and machine-readability
• Applies formal ontology QA methods pioneered by the team

Aim 2: Apply & Extract CDEs from RWD (CDE-Data Ecosystem)
• Uses AI/ML/NLP pipelines
• Federated architecture aligns with privacy-preserving principles
• Supports cross-institution harmonization, temporal phenotyping

Aim 3: Dissemination & Workforce Development
• Web Portal
• Training, adoption pathways, documentation
• Local “data scientist mediators” at ADRCs

Aim 4: Governance, Evaluation & Operational Support
• SC, EAP, KF1-KF5
• Transparency & QA/QC integration
• D3M framework for lifecycle management

D3M is the conceptual backbone of ReCARDO.
It organizes the entire project into two iterative “diamonds”:

1. Diamond 1 — CDE Development:
Discover → Define → Evaluate → Refine
2. Diamond 2 — CDE Application to RWD:
Develop → Deliver → Validate → Disseminate
It ensures continuous feedback from real-world application back into CDE refinement.

CDE-Engine (Aim 1)
• Manages lifecycle of CDE creation
• Includes ontology QA, CDE Knowledge Graph
• Ensures traceable provenance (transparency)
CDE-Data (Aim 2)
• Harmonizes RWD using AI/ML/NLP pipelines
• Supports federated execution
• Includes tools for temporal querying (via TEL), cohort discovery, and faceted search

These ecosystems embody FAIR (Findable, Accessible, Interoperable, Reusable) principles.

Three immediately available major RWD resources:

ReCARDO-U24-Strategy-final-4pm
1. ACT (Adult Changes in Thought) – longitudinal cohort with deep phenotyping
2. OneFlorida+ EHR Network – 26M patients across multiple systems
3. Rush ADC cohorts (ROS/MAP) – biospecimens, imaging, claims linkage, autopsy data
Plus extensive omics, imaging, psychometric, and clinical datasets.

The project integrates NeuropsychToMe, which modernizes psychometric scoring:
• Reduces scoring time from 2 hours → 15 minutes
• Produces discrete, machine-readable variables
• Enables normalization and recalibration for AD/ADRD cohorts

This strengthens CDE development for cognitive and behavioral domains.

Work Streams 2 and 4 deploy:
• OHNLP Toolkits for NLP extraction across 30+ CTSA sites
• Large Language Models (e.g., GatorTron, BioMedGPT)
• RWE-oriented algorithms (e.g., distributed pda, federated learning)
• Temporal Ensemble Logic (TEL) for precise temporal cohort definitions

These tools ensure scalable, reproducible RWD harmonization.

ReCARDO incorporates:
• Expertise from NIH BRIDGE2AI Ethics Core
• Distributed/federated RWD processing (no central pooling required)
• Algorithmic fairness evaluations
• Institutional AI governance integration

Reproducibility is embedded at every step via:
• Ontology QA tools
• Empirical validity metrics (EASE, empirical equipoise, covariate balance)
• Transparent provenance via D3M
• Work Stream 7 dedicated entirely to reproducibility & evaluation

Starting in Year 2, ReCARDO will fund six external pilot awards per year.
• Pre-award: WS5 oversees solicitation & review
• Post-award: WS7 oversees evaluation
• Awardees receive access to training, innovation support, and CDE resources

Opportunities include:
• Applying for pilot awards
• Joining Work Stream activities
• Leveraging CDEs and tools for local ADRC studies
• Participating in workshops, training, and the Annual CDE User Meeting
• Connecting through the Web Portal as resources become available

Ten institutions span ADRCs and CTSAs:

• UTHealth Houston (Administrative Hub)
• Indiana University
• Mayo Clinic
• Rush University
• University of Alabama at Birmingham
• University of Florida
• University of Pennsylvania
• UT San Antonio
• University of Washington
• Vanderbilt University

ReCARDO is grounded in:
• TIES: Teamwork, Innovation, Excellence, Stewardship
• RITE: Reproducible, Implementable, Transparent, Explainable
• FAIR: Findable, Accessible, Interoperable, Reusable