RAPID_SOL_25_119_FINAL.pdf

Innovative Solutions Opening (ISO) Rare Disease AI/ML for Precision Integrated

Diagnostics (RAPID) Proactive Health Office (PHO)

Advanced Research Projects Agency for Health

ARPA-H-SOL-25-119

January 10, 2025

Draft RAPID Innovative Solutions Opening, ARPA-H-SOL-25-119

TABLE OF CONTENTS

1.0 INNOVATIVE SOLUTIONS OPENING (ISO) SUMMARY INFORMATION

1.1 ISO PURPOSE

1.2 ISO QUESTIONS AND ANSWERS

2.0 RAPID PROGRAM INFORMATION

2.1 RAPID BACKGROUND

2.2 PROGRAM DESCRIPTION

2.3 NATIONAL HEALTH IMPACT

2.4 RAPID PROGRAM SCOPE

2.5 TECHNICAL APPROACH AND STRUCTURE

2.5.1 TA1: EXISTING RARE DISEASE DATA AGGREGATION

2.5.2 TA2: NOVEL RARE DISEASE DATA COLLECTION

2.5.3 TA3: RARE DISEASE DATA COMMONS

2.5.4 ADDITIONAL CONSIDERATIONS (APPLIES TO TAS 1-3)

2.5.5 PATIENT EXPERIENCE AND ENGAGEMENT

2.5.6 PROGRAM SUPPORT: PATIENT EXPERIENCE (PX) PARTNER NETWORK

2.6 PROGRAM TIMELINE, PHASES & MILESTONES

2.6.1 PHASE I (MONTHS 0-21): FOUNDATION BUILDING

2.6.2 PHASE II (MONTHS 22-54): DEVELOPMENT, DEPLOYMENT, AND VALIDATION

2.6.3 INDEPENDENT VERIFICATION AND VALIDATION (IV&V)

2.6.4 PROGRAM DELIVERABLES

2.6.5 PROGRAM METRICS

2.6.6 TA1 DATA AGGREGATION

2.6.7 TA2 DATA COLLECTION

2.6.8 TA3 DATA COMMONS

2.7 GENERAL REQUIREMENTS

2.7.1 DATA STANDARDS

2.7.2 INTELLECTUAL PROPERTY (IP) AND OPEN SOFTWARE STANDARDS

3.0 ELIGIBILITY INFORMATION

3.1 ELIGIBLE APPLICANTS

3.1.1 PROHIBITION ON PERFORMER PARTICIPATION FROM FEDERALLY FUNDED RESEARCH

AND DEVELOPMENT CENTERS (FFRDCS) AND GOVERNMENT ENTITIES

3.1.2 NON-US ENTITIES

3.1.3 AWARD LIMITATIONS

3.2 SYSTEM FOR AWARD MANAGEMENT (SAM)

4.0 SUBMISSION AND EVALUATION PROCESS

4.1 SOLUTION SUMMARY SUBMISSIONS

4.2 SOLUTION SUMMARY AND PROPOSAL SUBMISSION INFORMATION

4.2.1 GENERAL

4.2.2 SUBMISSION PORTAL

4.3 SOLUTION SUMMARY AND PROPOSAL SUBMISSION DEADLINES

4.4 PROPRIETARY INFORMATION

5.0 REVIEW AND EVALUATION OF FULL PROPOSALS

5.1 CONFORMING PROPOSAL SUBMISSIONS

5.2 PROPOSAL EVALUATION CRITERIA

5.2.1 CRITERION 1: OVERALL SCIENTIFIC AND TECHNICAL MERIT

5.2.2 PROPOSER’S CAPABILITIES AND/OR RELATED EXPERIENCE

5.2.3 BUDGET ANALYSIS

5.3 EVALUATION AND SELECTION PROCESS

5.4 HANDLING OF SELECTION-SENSITIVE INFORMATION

5.5 EVALUATION AND AWARD GENERAL GUIDELINES

6.0 ADMINISTRATIVE AND NATIONAL POLICY REQUIREMENTS

6.1 ORGANIZATIONAL CONFLICTS OF INTEREST (OCI)

6.1.1 AGENCY SUPPLEMENTAL OCI POLICY

6.1.2 GOVERNMENT OCI PROCEDURES

6.2 INTELLECTUAL PROPERTY

6.3 HUMAN SUBJECTS RESEARCH

6.4 ELECTRONIC INVOICING AND PAYMENTS

6.5 GOVERNMENT-FURNISHED PROPERTY/EQUIPMENT

6.6 ASSOCIATE PERFORMER AGREEMENT

APPENDIX A: ASSOCIATE PERFORMER AGREEMENT (APA)

APPENDIX B: DATA OF INTEREST

APPENDIX C: SOLUTION SUMMARY FORMAT & INSTRUCTIONS

APPENDIX D: FULL PROPOSAL FORMAT AND INSTRUCTIONS

1.0 INNOVATIVE SOLUTIONS OPENING (ISO) SUMMARY INFORMATION

Federal Agency: Advanced Research Projects Agency for Health (ARPA-H), Proactive Health Office (PHO)

Program Title: Rare disease AI/ML and Precision Integrated Diagnostics (RAPID)

Announcement Type: Solicitation

ISO Solicitation Number: ARPA-H-SOL-25-119

Dates (all deadlines are 5pm EST):

• Proposers’ Day: January 23, 2025

• Final date to submit ISO Questions: February 4, 2025

• Solution Summary Due Date: February 14, 2025

• Patients’ Day: February 25, 2025

• Full Proposal Due Date: April 11, 2025

Concise Description of the Funding Opportunity: RAPID aims to accelerate rare disease diagnosis by catalyzing large-scale data curation to drive development of the world’s most accurate and comprehensive diagnostic models, enabling the identication of undiagnosed patients at scale.

Anticipated Awards: Multiple Other Transaction Agreement awards.

Program Budget: ARPA-H will not publicly release the approved overall Program budget and has not established budget ceilings (or a range) for projects within the Program portfolio. Proposer funding will depend on the proposed statements of work as well as overall Program portfolio composition.

Cost Sharing Requirements: Cost sharing may be requested.

Agency Contact: All inquiries shall be sent to RAPID@arpa-h.gov.

1.1 ISO Purpose

ARPA-H seeks proposals from all eligible entities (see Section 3, Eligibility Information) to accomplish the RAPID program goals as described in this solicitation package.

Ultimately, ARPA-H intends to negotiate multiple OT Agreements with Proposers whose proposals are most advantageous to the Government.

1.2 ISO Questions and Answers

1.2.1 ARPA-H will post Questions and Answers (Q&As) to the RAPID Program website on an on-going basis and will not respond to questions directly. All questions must be submitted in English and must include the name, e-mail address, and telephone number of a point of contact. Proposers submitting questions to individual Government team members (e.g., Program Manager) should not expect a response. ARPA-H will attempt to answer questions in a timely manner;

however, questions submitted after the Q&A due date may not be answered.

Further, duplicative questions may be combined and rephrased to streamline responses.

1.2.3 Proposers’ Day

1. ARPA-H will host a Proposers’ Day in support of the RAPID Program as described in Special Notice ARPA-H-SN-25-124 (see SAM.gov). The purpose of the event is to provide potential Proposers with information on the RAPID program, promote discussion, and encourage team networking.

2. Interested Proposers are not required to attend; materials formally presented at Proposers’ Day will be posted to http://www.sam.gov.

3. ARPA-H will not reimburse potential Proposers for participation at the Proposers’ Day (nor for time and effort related to the submission of Solution Summaries or full proposals).

2.0 RAPID PROGRAM INFORMATION

2.1 RAPID Background

An estimated 30 million Americans have one of 10,000 identied rare diseases, affecting 1 in 10 individuals in the US and more than 350 million people globally. It’s estimated that 50% of individuals with a rare disease remain undiagnosed or misdiagnosed due to a combination of factors including disease heterogeneity, low disease prevalence, lack of awareness among healthcare professionals, and limited availability and use of conrmatory tests. The lengthy diagnostic “odyssey” endured by individuals with a rare disease lasts 6 years on average, but in many cases can extend for decades. The impact of delayed diagnosis on individuals and society is pronounced, often resulting in inappropriate care, irreversible disease progression, and signicantly increased medical costs.

2.2 Program Description

2.2.1 RAPID is a 4.5-year (21-month base period and 33-month option period) program that aims to end or signicantly reduce the prolonged diagnostic odysseys affecting millions of rare disease patients worldwide. RAPID seeks to develop novel diagnostic support tools for population-scale rare disease detection, systematically identifying and diagnosing rare disease patients more efficiently and accurately. This approach aims to overcome current barriers stemming from gaps in clinical assessment. By aggregating and harmonizing fragmented data and employing advanced analytic techniques, RAPID aims to drive the development and deployment of innovative rare disease detection models. These models will be designed to operate both within and outside traditional healthcare systems, enhancing the precision and scope of disease identication.

2.2.2 RAPID also emphasizes equitable access to rare disease diagnostics, focusing on reducing costs, improving accessibility, and enhancing user experiences. The program aims to democratize access to advanced diagnostic tools, ensuring that underserved and marginalized communities benet from these innovations. By prospectively validating models in diverse clinical settings, RAPID seeks to transform the landscape of rare disease diagnosis, ensuring timely and effective care for millions of underserved patients worldwide and demonstrating the efficacy of advanced diagnostic systems.

2.2.3 Ultimately, RAPID aims to achieve measurable improvements in patient outcomes by reducing average diagnostic delays in target diseases by ≥50%, signicantly decreasing the rate of misdiagnoses, and enhancing access to diagnostics for underserved populations. The program also seeks to increase clinical trial enrollment through earlier and more accurate patient identication and facilitate the discovery of actionable biomarkers to support therapeutic development. These outcome-driven targets are intended to alleviate the physical, emotional, and nancial burdens of prolonged diagnostic uncertainty, improve quality of life for patients and families, and accelerate advancements across the rare disease ecosystem.

2.3 National Health Impact

Rare disease patients endure an average diagnostic odyssey of more than 6 years, including 17 medical interventions and 3 misdiagnoses. Delayed diagnosis can result in irreversible disease progression, worsening patient outcomes by increasing risk of preventable death and disability, as well as complicating treatment efforts. An accurate and timely diagnosis can result in better disease management, identication of potential therapeutics, and avoidance of unnecessary treatments and procedures. A rare disease diagnosis can also provide much-needed clarity and access to support networks and resources, improving the mental health and quality of life for patients, their families, and caregivers. More accurate prevalence estimates can also guide research and public health priorities, while faster patient identication supports clinical trial enrollment, accelerating treatment development — crucial for rare diseases lacking approved therapies.

2.4 RAPID Program Scope

2.4.1 The RAPID program takes an integrated approach to systematically identify undiagnosed rare disease patients at scale, leveraging novel training data, advanced AI/ML techniques, and real-world validation. The program will be conducted over two consecutive phases. At present, ARPA-H is soliciting innovative solutions for Phase I and Phase II that address the following Technical Areas (TAs): TA1, TA2, and TA3, introduced below. Proposers may apply to, and potentially be selected for, multiple TAs; however, each TA must be addressed by a separate solution summary and proposal. If submitting multiple proposals, clearly indicate this at the top of each submission. A subsequent solicitation for TA4 Effective Deployment and Real-World Validation will be issued separately at a later date. TA4 is anticipated to be an open solicitation, and subsequent performers will only participate in Phase II. A Government Independent Verication and Validation (IV&V) partner will be selected separately and will provide continuous support and oversight throughout the duration of the program. The goals of each TA are outlined below:

1. Massive-Scale Rare Disease Dataset (TA1): Develop the largest curated dataset of longitudinal rare disease patient health data by integrating information from a fragmented data landscape. This extensive resource, spanning thousands of rare and ultra-rare diseases, is purpose-built to accelerate the development and validation of advanced detection algorithms and diagnostic tools. Sustainable systems will ensure the continuous aggregation and curation of patient data, driving transformative advancements in rare disease detection and diagnosis.

2. Novel Diagnostic Indicators and Population-Scale Discovery (TA2):

Leverage cost-effective, widely accessible, and remotely deployable digital tools to collect data directly from patients, creating a comprehensive, multimodal dataset of rare disease patient health information. This dataset is designed to drive the discovery of novel diagnostic indicators and enable scalable detection of rare diseases in previously undiagnosed individuals, with an emphasis on reaching underrepresented and underserved populations.

3. Sustainable Platform for AI Diagnostic Development (TA3): Establish an enduring Rare Disease Data Commons (RDDC) with a benchmark dataset, enabling the accelerated development and empirical evaluation of rare disease diagnostic models. This dynamic, AI/ML-optimized platform will support interoperable data collection, sharing, analysis, and evaluation among key stakeholders, driving advancements in diagnostic decision support systems for rare diseases.

4. Model Benchmarking (IV&V): Following the initial phases of data collection and platform development, the RAPID Program will initiate a benchmarking effort to empirically evaluate rare disease detection models. The ARPA-H IV&V partner will facilitate the benchmarking process and collaborate with performers to integrate benchmarking tools, methodologies, and evaluation results into the TA3-developed platform.

These results will provide critical insights to guide the selection of TA4 performers for Phase II.

2.4.2 In year two of the program, a Phase II solicitation will be released to solicit proposals for TA4: Effective Deployment and Real World Validation. TA4 performers will leverage data and infrastructure from TA1-3 to develop, deploy, and validate novel rare disease detection systems in both clinical and direct-to-patient settings, with real-world performance assessed through prospective validation. Data collection and processing for TA1-3 will begin in Phase I and continue through Phase II (refer to Figure 2 below for details).

Figure 1: RAPID Technical Areas (TAs) Overview. Data collected in TA1 and TA2 will be utilized by TA3 for further processing and normalization for use in the Rare Disease Data Commons and the Rare Disease Benchmark Dataset.

This data will then be leveraged for model benchmarking and be available to TA4 performers for the training, testing, and validation of models in Phase II.

2.4.3 Proposers are encouraged to read the detailed description of the program’s requirements for each TA below before responding to this solicitation.

2.5 Technical Approach and Structure

2.5.1 TA1: Massive Scale Rare Disease Dataset

1. TA1 will aggregate longitudinal health data of rare disease patients from diverse sources, including but not limited to Electronic Health Records (EHRs). TA1 data will enable the creation of the Rare Disease Benchmark Dataset a comprehensive data source optimized for developing and testing algorithms used in rare disease detection and diagnosis.

2. TA1 solutions must address the unique challenges of aggregating rare disease data, including fragmented data across institutions and organizations, misdiagnosis, underreporting, and limitations of medical coding systems like the International Classication of Diseases, Tenth Edition (ICD-10). Proposals should detail scalable strategies for integrating rare disease data across disparate sources, ensuring comprehensive coverage, and overcoming issues related to low prevalence and incomplete records.

3. Submissions must present a clear technological and strategic approach, with a focus on broad accessibility, patient engagement, and long-term sustainability. A detailed plan for overcoming data fragmentation, enabling scientic impact, and managing consent and Data Use Agreements (DUAs) is required.

4. TA1 proposals should address the following:

a. Disease Selection

1) A clear methodology for selecting rare diseases with a justication based on factors such as prevalence, etiology, availability of Human Phenotype Ontology (HPO) characterizations, and potential impact on under-diagnosed or under-characterized conditions.

2) An approach for collecting control data, including a justication for why the selected data is suitable as a control cohort and any additional data processing tasks required to achieve a suitable control dataset.

b. Data Format and Data Provider Selection

1) A comprehensive strategy for selecting and integrating diverse data sources and types, with the goal of maximizing the volume of usable data while navigating potential limitations imposed by data oversight requirements. This approach should detail methods for identifying, aggregating, and linking data from a wide spectrum of providers, including but not limited to healthcare organizations, insurers, diagnostic laboratories, patient advocacy groups, and research institutions.

2) Data should comprise diverse types, forming a comprehensive longitudinal medical record. Proposals should prioritize aggregating data in a format suitable for export from a health system, simulating a diagnostic support model using data from multiple providers. [Note:

While structuring and normalization of data will be accomplished as part of TA3, aggregating both raw and standardized data is preferred when possible].

c. Data Accounting

Detailed methodologies for estimating and validating the content of unstructured data sources and assessing the completeness of individual patient records.

d. Data Curation and Annotation

1) A robust plan for expert curation of rare disease data, including approach to specialized disease coding (e.g., ORPHACODE, OMIM, ICD-11, ICD-10) and metadata annotation.

2) Strategy for accurately annotating patient data to identify pre- and post-diagnosis elements. Teams should develop approaches or tools capable of scaling the expert annotation process, ensuring high-quality data ready for further processing in TA3.

3) If developing novel tools for expert annotation, describe capabilities such as scalability, collaborative interfaces, machine learning assistance, and quality control methods to ensure consistent, high-quality curation across diverse diseases and data sources.

4) If developing or utilizing tools to enable expert annotation, describe strategies for integrating annotation tools into the broader data management workow and ensuring their scalability as the program grows.

e. Data Usability and Access

1) Strategies to maximize the usability of aggregated data for use in the Rare Disease Benchmark Dataset, including approaches for de-identication, such as obtaining consent for TA3 to de-identify the data, or using synthetic data generation. Proposals should outline anticipated distributions of data access and acceptable use of aggregated data.

2) All data collected under TA1 must be available for processing and integration by TA3 performers as part of the development of the Rare Disease Data Commons. TA1 performers will be expected to collaborate with TA2 and TA3 performers to ensure data handoff for further processing in TA3.

f. Technical Infrastructure

1) A comprehensive technical infrastructure plan that supports scalable rare disease data aggregation. This plan should include strategies for managing data from varied sources, ensuring accessibility across data types.

2) Robust data processing workows, including secure transfer methodologies to interface with TA2 and TA3 performers, and an outline of measures for ensuring data integrity, security, and privacy compliance - e.g. Health Insurance Portability and Accountability Act (HIPAA), General Data Protection Regulation (GDPR).

3) Explanation of how technical decisions support the program’s strategic goals for data access, usability, and ownership rights throughout the program lifecycle.

4) Strategies for obtaining, tracking, and updating patient consent, including methods for implementing ne-grained access controls based on user roles and data sensitivity.

g. Teaming

Approaches to interfacing with external teams and subject matter experts, as well as ways to build collaboration among diverse stakeholders which may include health systems, patient advocacy groups, or patient registries.

h. Scalability and Sustainability

1) Describe how the proposed architecture will be highly scalable and adaptable for continuous growth and adaptation of rare disease data, including how the architecture accommodates external changes, such as evolving healthcare data systems, while addressing potential scalability limitations.

2) Describe approach for incorporating new data and newly diagnosed patients to enable sustainable scalability of operations beyond initial program metrics.

3) Detail plans for growing disease coverage based on early successes and emerging partnerships. Address scaling of operations and engagement strategies appropriate to the track (e.g. healthcare systems). Demonstrate how feedback from initial efforts will inform selection of new disease targets.

i. Ethics

Include an ethical framework that prioritizes patient trust and welfare. This framework should clearly delineate data ownership rights, ensure open access without imposing unfair nancial barriers, and align with the program’s core mission of improving patients' lives and advancing rare disease research and diagnosis.

2.5.2 TA2: Novel Diagnostic Indicators and Population-Scale Discovery

1. TA2 aims to develop a novel, large-scale, multi-modal dataset collected directly from rare disease patients through a user-friendly digital system.

Solutions should be patient-centric including a highly usable interface for collecting novel and diverse data types, such as but not limited to, patient-reported outcomes, photos, video, audio, or other inputs that can be utilized in the training and validation of rare disease diagnostic models.

The dataset should enable the development of scalable models that support the diagnosis of patients who are underrepresented in EHRs or have limited access to the healthcare system, such as low-income households, racial/ethnic minorities, and rural residents. The data collection systems should also have the capability to obtain patient consent for collecting their EHR data. Submissions must include a comprehensive plan for collecting the appropriate consent and Data Use Agreements (DUAs) to enable broad access to aggregated data both during and after program completion.

2. TA2 proposals should address the following:

a. Targeted Population Selection

1) Describe and justify a methodology for selecting targeted rare diseases to reach the required metrics, emphasizing under-diagnosed conditions with identiable multi-modal diagnostic signals or incomplete characterizations for advancing scientic understanding, demonstrating how the chosen diseases will advance diagnostics, improve patient outcomes, and potentially impact underserved populations.

2) Outline a methodology for selecting “control” populations and collecting relevant data, with justication for the intended selection methods to create a control group, such as age-matched populations.

b. Outreach Plans

Outline partnerships and strategies for patient identication and outreach, detailing plans to engage rare disease communities and ensure representation of historically underserved groups (e.g., low-income households, racial/ethnic minorities, rural residents) with the objective of improving rare disease diagnosis among underrepresented groups while maintaining the privacy of potentially vulnerable populations.

c. Data Denitions

Dene a comprehensive data dictionary for the collection effort, including a core set of data elements required from all participants and any additional proposed elements. Justify each data point based on its relevance to diagnostic modalities, phenotyping potential, and early diagnosis capabilities. Outline the collection methodology and address mitigations for potential biases, especially for supplementary data requiring specialized tools.

d. Data Usability and Access

1) TA2 performers will be expected to collect data from both rare disease patients and patients without rare disease to serve as control data. Of the rare disease respondents, performers should obtain consent from >50% to collect EHR data for use in the Rare Disease Benchmark Dataset.

Outline a consent management strategy on the collected data that includes both the immediate data collection and future research and development uses. This strategy should dene which data can be used in a publicly released Rare Disease Benchmark Dataset and which data should be reserved for controlled research purposes.

Explain how the consent strategy will have the exibility to support potential future research.

2) All data collected under TA2 must be available for processing and integration by TA3 performers as part of the development of the Rare Disease Data Commons. TA2 performers will be expected to collaborate with TA1 and TA3 to ensure data handoff for further processing in TA3.

e. Data Collection Interface

1) Describe a patient-facing digital interface for data collection and patient interaction, with a primary focus on accessibility (i.e., cross-platform compatibility) and usability, including anticipated timelines for development.

Detail the interface's design choices, explaining how they impact other aspects of the proposal, such as collected data quality, user engagement, and usability. This interface will serve as the central platform for executing data collection efforts.

2) Include details on interface adaptability to diverse user needs, including multi-language support, cultural adaptability, and intuitive design for both self-use and assisted data collection.

3) Discuss data collection capabilities, detailing potential technological limitations and their impact on data quality, and explain how these factors inuence the overall data gathering strategy.

f. Data Validation

Propose a comprehensive data validation strategy encompassing both real-time and post-collection processes. Detail automated methods for immediate data validation during collection, ensuring adherence to quality standards and content correctness. Include plans for analytical validation of collected data to maintain consistency across patients and uphold overall data delity.

g. Adaptability

1) Demonstrate an adaptable interface capable of evolving throughout the collection efforts, modifying data collection requirements as needed.

2) Dene the ability to push updates, deploy multiple versions where necessary.

3) Outline how the system will incorporate new populations and ndings, ensuring it remains agile and responsive to evolving research needs and discoveries throughout the project lifecycle.

4) Dene a mechanism for re-contacting patients to gather new or updated information.

h. Security and Privacy

Describe robust security measures and privacy protocols for the collection, storage, and transfer of Protected Health Information (PHI), ensuring compliance with all relevant data protection regulations and industry standards (e.g., HIPAA, GDPR, NIST SP

800-53, ISO/IEC 27001, HITRUST).

i. Technical Infrastructure Detail a comprehensive technical infrastructure capable of hosting and deploying the digital interface, managing patient data transfer and staging, handling user accounts and authentication, securely storing data, and facilitating transfers to authorized entities for analysis and processing.

j. Scalability and Sustainability

1) Proposals should prioritize extensible and sustainable systems over singular efforts, detailing long-term scalability and growth strategies including international expansion, leveraging the ubiquity of digital devices to facilitate global reach.

2) Detail plans for growing disease coverage based on early successes and emerging partnerships. Address scaling of operations and engagement strategies appropriate to the track (e.g. patient advocacy organizations). Demonstrate how feedback from initial efforts will inform selection of new disease targets.

2.5.3 TA3: Sustainable Platform for AI Diagnostic Development

1. The TA3 performer will establish a Rare Disease Data Commons (RDDC), an AI/ML-optimized data platform designed to accelerate the development and evaluation of rare disease detection algorithms. A primary output of TA3 is the novel Rare Disease Benchmark Dataset to catalyze the research and development of diagnostic models by providing essential data and a common data source for comparing model performance.

2. The TA3 performer will work closely with the data collection initiatives (TA1 and TA2) to ingest, host, and process health and medical data, transforming it into the Rare Disease Benchmark Dataset, a research-grade resource optimized for machine learning. The performer will leverage advanced technologies for data structuring and normalization, while developing a user-friendly interface for accessing the Rare Disease Benchmark Dataset, supported by a sustainable and accessible protocol that manages all relevant DUAs. In collaboration with TA1 and TA2, TA3 will also oversee consent processes, ensure data de-identication, and coordinate data ingestion schedules.

3. TA3 proposals should address the following:

a. Data Infrastructure

1) Propose a technical infrastructure capable of securely handling patient data packets from diverse providers, accommodating multiple data formats and various data access methods, including federated access.

2) Describe approach to pre-processing data and secure, efficient solutions to storing processed data.

3) Address how the infrastructure will adapt to different interfacing systems (e.g., local vs. federated data sources).

b. Data Processing

1) Describe an advanced pipeline for structuring and normalizing heterogeneous medical data into standardized formats (e.g., HL7 FHIR, HPO, VCF/GA4GH), including immediate capabilities, performance metrics, and strategies for ongoing adaptation to evolving standards.

2) Include plans for ensuring data interoperability, such harmonizing through use of a Common Data Elements (CDE) dictionary.

3) Describe how you will enhance the Extract, Transform, Load (ETL) pipeline with external knowledge bases and graphical data elements to support downstream knowledge graph development.

4) Describe how you will establish a HIPAA-compliant de-identication plan, in collaboration with TA1 and TA2, leveraging existing services where applicable, and incorporating informed consent and privacy waivers as needed.

5) Propose processes to validate data accuracy and completeness, identifying and addressing gaps or inconsistencies.

c. Rare Disease Benchmark Dataset Access

1) Describe a Rare Disease Benchmark Dataset access interface and access protocol, including clear, ethical guidelines for data access, considering varied data use agreements and consent levels. A thoughtful protocol for managing data access should balance openness with necessary protections, ensuring that the Rare Disease Benchmark Dataset serves as a democratized resource in rare disease research.

2) Outline a comprehensive plan for managing researcher access to the platform that ensures robust data privacy and security while facilitating collaborative research. The RDDC should implement a granular access control system, granting researchers access only to the specic data and tools necessary for their approved projects, and should include provisions for creating multiple versions of the Rare Disease Benchmark Dataset, including a publicly accessible version.

3) Include a data governance framework that tracks data provenance/lineage, enables element-level access controls, and maintains secure linkages between identiable and de-identied data while keeping them separated in storage and access.

Note: Once de-identied, Rare Disease Benchmark Dataset distribution may include a controlled public access protocol, such as a contractual DUA and/or mandatory data privacy training, to uphold trust. Alternative methods of controlled public access could follow a broad IRB access protocol, such as the Clinical Data Science IRB (CDS-IRB). The Rare Disease Benchmark Dataset must be available for public use in perpetuity. Performers may not implement exclusionary or restrictive use requirements (e.g.

processing data using proprietary methods).

d. Scalability and Sustainability

1) Outline how the infrastructure will address the data processing pipeline’s sustainability in adapting to changing data standards and regulations, including ability to handle growing data volumes and expanding international data sources.

2) Scalability is crucial for the platform to accommodate the growing volume and variety of rare disease data. Describe the architectural design and infrastructure choices that will enable the platform to scale horizontally and vertically as data and user demands increase to ensure that the platform can handle increasing workloads without compromising performance or availability. Further, ARPA- H may negotiate use of the RDDC for broader health ecosystem data.

3) To ensure the long-term success and impact of the platform, proposals should include detailed plans for continued maintenance and updating of the system. This may involve the establishment of dedicated support and development teams, regular security audits and penetration testing, and the implementation of robust backup and disaster recovery procedures. Proposers should also outline their strategies for maintaining data privacy and security, including compliance with relevant regulations such as HIPAA and GDPR, and the implementation of industry-standard security controls such as encryption, access logging, and intrusion detection.

e. Financial Sustainability

1) Provide a sustainable nancial model that ensures long-term viability without compromising data accessibility or the project's core mission of serving the rare disease community – including steps that would encumber data access because of intermediary data processing. This may include the exploration of non-rare disease applications, such as extending the platform's capabilities to support the diagnosis and management of more common diseases.

Proposals should also consider the potential for partnerships with academic institutions, healthcare organizations, and industry stakeholders to secure funding, expertise, and resources for the continued development and maintenance of the platform.

2) Outline a sustainable funding model that doesn't compromise data quality or create biased access with incentives that hamper the quality and accessibility of the Rare Disease Benchmark Dataset. Financial sustainability should be tied to technical and strategic plans, including Rare Disease Benchmark Dataset update cadences, access protocols, and hosting infrastructure.

f. Benchmarking Framework

1) Include a comprehensive framework for evaluating and benchmarking external models; this framework will be developed in partnership with the IV&V partner and should support a wide range of evaluation techniques, bias scores, noise-injection for robustness, performance benchmarking, and expert review, allowing for a thorough assessment of model quality and validity.

2) Privacy considerations should be considered in the external model evaluation architecture and the method for returning results. The model evaluation capabilities should also be able to support model comparisons, such as model leaderboards, for varying specic or general tasks.

g. Ethics

1) Explain methods for actively involving the rare disease community in shaping the Rare Disease Benchmark Dataset’s evolution, from governance to feature development.

2) Describe safeguards and policies that promote responsible data use, protecting patient privacy while fostering open science.

3) Include a plan for establishing an Institutional Review Board (IRB) to oversee TA3 work and ensure compliance with informed consent procedures established by TA1 and TA2. Note: TA1and TA2 performers are responsible for obtaining data, and therefore for obtaining informed consent.

2.5.4 Additional Considerations (Applies to TAs 1-3)

1. Government Data Rights

Data generated under this program, including de-identied and processed patient data, must be made available to the Government in accordance with the applicable Data Use Agreement (DUA). Unlimited Government Rights are preferred to maximize program impact, though not required. Data must be accessible to performer(s) of Technical Area 3 (TA3) and shared in compliance with all relevant legal, ethical, and contractual requirements, including patient consent agreements, and applicable regulations. Data sharing practices should be designed to align with the program's objectives, ensuring broad usability while upholding ethical and regulatory standards.

2. Optional Rights Negotiation Performers in TA1 and TA2 will have the opportunity to negotiate additional Government rights as part of the contracting process. These optional negotiations will specically address the integration of data collected under the RAPID program into the ARPA-H data commons, fostering agency-wide impact and facilitating future data reuse.

3. Deidentied Data De-identied patient data collected or generated under this program must be accessible to other program performers and available as a public resource under controlled access, such as through restricted databases or access granted upon committee approval. Data must comply with applicable de-identication standards, such as HIPAA’s Safe Harbor or Expert Determination methods, ensuring no information could reidentify patients. If additional consent requirements are specied in a DUA, these must also be met.

4. Identiable Data Identiable data, such as patient photographs, must only be shared with authorized entities under applicable DUAs and explicit patient consent agreements. Such sharing must adhere to all relevant legal, ethical, and contractual guidelines.

5. Management Plan Project teams must include a Principal Investigator (PI), or co-PIs, identied for the management team, as well as one designated primary point of contact (POC) who will serve as the Project Manager. Management plans should provide good faith estimates of metrics and target values that are appropriate and specic to proposed use-cases as well as detailed plans for meeting stated milestones on the prescribed program schedule (see Section 2.6 Program Timeline, Phases & Milestones). Technical risks along with mitigation strategies or alternative approaches should be included where appropriate.

6. Institutional Review Board

Projects must include a draft Institutional Review Board (IRB) protocol or, at minimum, a plan for having an IRB protocol approved by month 6 of performing in the RAPID program for handling sensitive patient data.

Program plans should address risk mitigation and compliance efforts related to handling such data.

2.5.5 Patient Experience and Engagement

1. Proposals for each TA should outline clear strategies for engaging patients and incorporating their perspectives through methods such as surveys, interviews, or focus groups, ensuring their input informs the development of key deliverables.

2. Continuous engagement and feedback loops should be described to ensure patient insights are consistently integrated throughout the project lifecycle, resulting in outcomes that align with patient needs and expectations.

3. Performers are expected to collaborate with Patient Experience Partners, described below, who will provide valuable guidance and resources to enhance engagement strategies. These partnerships will play a critical role in ensuring patient voices are effectively and consistently integrated throughout the program.

2.5.6 Program Support: Patient Experience (PX) Partner Network

1. Performers will be required to partner with PX Partners throughout the period of performance. The PX component of the RAPID program aims to ensure that patient perspectives, feedback, and lived experiences are integrated throughout the program. This component, solicited and funded through the IV&V RAPID Partner, will fund patient advocacy groups and organizations, which will serve as essential partners in fostering patient-centric design and participation, ensuring patient input is continuously incorporated into the development and deployment of RAPID tools and technologies. By leveraging their deep connections with patient communities, these organizations will help foster trust and drive patient participation, ultimately enhancing the success of the program.

2. PX Partners are not eligible for ARPA-H awards under this solicitation and will be selected and funded independently. Entities must choose between serving as a PX Partner and performing under any TA, performance in one area will prevent performance in the other. The RAPID IV&V partner will manage the PX Partner Network.

3. PX Partners will support and collaborate with RAPID program performers to facilitate patient-centered design and achieve the following:

a. Objectives:

1) Amplify Patient Voices: Ensure patient feedback, perspectives, and experiences are integrated into program decisions and technology development.

2) Improve Data Relevance: Align data collection, aggregation, and diagnostic models with the real-world conditions, symptoms, and challenges faced by rare disease patients.

3) Promote Engagement: Actively engage populations underrepresented in rare disease research and historically marginalized in healthcare, including racial and ethnic minorities and rural communities, to ensure their perspectives are reected in data collection and analysis.

4) Build Patient Trust: Foster transparency and trust between the program and patients through active, ongoing engagement, championing responsible and ethical program practices.

b. Specications and deliverables:

1) Awareness and Outreach: Partner with patient advocacy groups to raise awareness about the RAPID program within patient communities, increasing program visibility and engagement.

2) Incentivize Data Participation: Engage and motivate patients and patient advocacy groups to contribute data for TA1 and TA2, ensuring a diverse and representative dataset through patient outreach efforts.

3) Usability Feedback: Facilitate patient participation in user experience research (UXR), ensuring that patient perspectives are incorporated into the design and validation of user interfaces, and that tools are accessible and user-friendly.

4) Trust Building: Develop strategies that build patient trust in the program’s new technologies and platforms, focusing on transparency, privacy, and the benets of participation.

5) Patient Resources: Assist performers with the development of patient-facing resources such as tutorials, FAQs, and educational materials to help patients understand and effectively use the tools and platforms developed in RAPID, maximizing their accessibility, utility, and benet.

Organizations interested in becoming a PX Partner are invited to join an informational event to be held after Proposers’ Day. For details about the event and instructions on applying to the PX Partner Network, please visit the RAPID program page: https://arpa-h.gov/research-and-funding/programs/rapid

2.6 Program Timeline, Phases & Milestones

RAPID is a 4.5 year (54-month), two-phase program with a 21-month Phase I period and an option for a 33-month Phase II period. Multiple awards are anticipated, and it is expected that fewer performers may be funded to participate in Phase II. For TA1, TA2, and TA3 the Phase II options may be exercised, at the Government’s sole discretion, based on technical progress measured against the metrics (Table 1) and milestones dened in the Innovative Solutions Opening, and based on funding availability. Each Phase has key technical milestone goals. A description of the Phases, which are aligned with critical program milestones are described below:

2.6.1 Phase I (months 0-21): Foundation Building

At the start of Phase I, TA1 and TA2 performers will immediately commence data aggregation and collection efforts. TA2 performers will also design, test, and deploy scalable approaches for collecting multimodal data directly from patients.

By the end of Phase I, TA1 and TA2 performers should have submitted data that has been validated by IV&V, satises all metrics, achieves all milestones, and is ready to support benchmarking and algorithm development for TA4 performers.

Throughout the process, ARPA-H will be conducting periodic data assessments (to be dened after performer selection according to program needs), and performers should be ready to submit data Quality Assurance (QA) and progress reports. Initially, TA3 performers will prioritize the development of platform functionalities for processing data, as well as creating, maintaining, and hosting the Rare Disease Benchmark Dataset. The metrics dene high-level goals, and the milestones outline a roadmap of features expected to be fully operational.

Once baseline capabilities are established, TA3 performers will shift their efforts toward enabling the benchmarking of machine learning models.

2.6.2 Phase II (months 22-54): Development, Deployment, and Validation

1. During Phase II, TA1 and TA2 performers will build upon their initial efforts, with an expanded focus on scaling to address ultra-rare conditions and incorporating international data sources. Data QA and progress reports will continue to be conducted similarly to Phase I, along with similar IV&V analyses. TA3 performers will prioritize the continuous enhancement of core platform capabilities while advancing functionalities for real-time data exchange and collaboration with data providers, deploying models in clinical settings, and facilitating external model evaluation.

2. It is expected that TA4, solicited at a later date, will be incorporated during Phase II. TA4 model development will involve an iterative process with continuous cycles of model development and testing. At yearly intervals, model performance will be evaluated against baseline metrics, and improvements will be expected in subsequent intervals based on these assessments. Additionally, ARPA-H will conduct spot checks on model performance, and IV&V will analyze model architecture.

3. During the second half of Phase II, TA1 and TA2 performers will continue to scale their systems and nalize efforts to achieve the most challenging metrics and milestones. Their success will depend on the initial systems and groundwork established in the previous phase. TA3 performers will focus on continuous development of RDDC features to meet yearly metrics, while also deploying new functionalities for researcher access and clinical deployment. ARPA-H and IV&V analyses will proceed in a manner similar to the prior phases.

4. The ultimate objective of Phase II is to transition from controlled clinical validation to real-world deployment in its later stages. During the deployment period, performers will iteratively rene their methods to align with real-world conditions, leveraging techniques validated by the IV&V team to enhance their models. Performers will be evaluated based on their performance and outcomes in real-world applications.

2.6.3 Independent Verication and Validation (IV&V)

1. IV&V will be conducted by an ARPA-H identied entity with the necessary capabilities to evaluate performance during the program. Performers are expected to work closely with the IV&V team throughout the program’s duration. In particular, evaluations will take place before scheduled Phase transitions to ensure validation and real-world performance of developed technologies in Phases I & II.

2. For TA1 and TA2, IV&V will focus on data quality and completeness, ensuring that the underlying data matches the high-level descriptive metrics. For TA3, IV&V will validate that the features function as described and that security standards are met. IV&V will also manage the Patient Experience Partners program support.

3. As noted above, the IV&V entity will also provide PX Partners for Performers.

2.6.4 Program Deliverables

RAPID anticipates several deliverables from the cumulative outputs of each TA.

TA1 will produce the raw data to be used to develop an expertly curated, novel Rare Disease Benchmark Dataset, and ideally a sustainable system for identifying and aggregating rare disease data. TA2 will produce both a large-scale, diverse, multi-modal dataset representing several rare diseases and a control population alongside a highly usable digital system for collecting the data. TA3 will produce a rare disease data commons that has the functionality to ingest various formats of health data and medical records to produce an AI-optimized dataset provided through a controlled access interface.

Figure 2: RAPID Program Timeline, Key Milestones and Deliverables across Phase I & II: TA1-TA4. Timeframes are relative to the start of the program. Note dotted red lines depict program phase transitions or key milestones, while gray-dotted lines depict end of scal-year.

2.6.5 Program Metrics

1. Specic metrics will be used to track progress as performers complete tasks for each individual TA. Performers will be required to report progress against these metrics as needed to the Program Manager, and to the designated IV&V team(s). These metrics will be used to evaluate success and will inform the Government team as they decide which performers will continue in the Program.

2. Table 1 denes the aggregation metrics by subcategory for the yearly target minimums. Proposers are also encouraged to include in their proposal any additional quantitative metrics and milestones, beyond those listed in Table 1, that they plan to use for tracking progress toward the nal goals of each TA. If mutually agreed upon during negotiations, these additional metrics will be incorporated into the OT agreement alongside the established evaluation criteria. The ability to meet target metrics on time will be a major factor for advancing to later milestones of the RAPID Program. Clarifying details on specic metrics are provided in Table 2 and Table 3.

Table 1: RAPID Program Metrics across TA1. Timeframes are relative to the start of the program. Select metrics are explained in more detail in the following section.

TA1 AGGREGATION METRICS SUB-CATEGORY

TARGET

(Minimums)

Year 1 Year 2 Year 3 Year 4 Year 5 Grand Totals

Totals Total Disease Count ≥400 ≥400 ≥525 ≥620 ≥755

≥2,700 diseases

Total Sample Size ≥22,000 ≥21,500 ≥20,100 ≥11,200 ≥5,750 ≥80,550 samples

Higher prevalence aggregation (likely in the range ~10-50/100k)

Disease Count ≥200 ≥100 ≥75 ≥20 ≥5 ≥400 diseases

Sample Size min. ≥100 ≥200 ≥250 ≥500 ≥1,000 ≥73,750 samples

Lower prevalence aggregation (likely in the range ~0.1-9/100k)

Disease Count ≥200 ≥300 ≥450 ≥600 ≥750 ≥2,300 diseases

Sample Size min. ≥10 ≥5 ≥3 ≥2 ≥1 ≥6,800 samples

Number of international sources -- -- ≥1 ≥2 ≥3 ≥5

Control Data Control data must be…