AHRQ SI for Healthcare Improvement Initiative

Contents

• Challenge Title 
• Tagline 
• Overview  
• A. Clinical Evidence Synthesis Transformation Challenge 
  • Detailed Description 
  • Judging 
  • How to Enter
  • Additional Information Regarding Registration
  • Additional Information Regarding Submissions
  • Resources
  • Frequently Asked Questions
• B. SI Tech Sprint for Healthcare Research and Implementation Quality and Safety  
  • Detailed Description 
  • Judging 
  • How to Enter
  • Additional Information Regarding Registration
  • Additional Information Regarding Submissions
  • Resources
  • Frequently Asked Questions
• Legal Authority 
• Rules

Challenge Title

AHRQ Super Intelligence for Healthcare Improvement Initiative

Tagline

The Agency for Healthcare Research and Quality (AHRQ) seeks an individual, team, or entity to build and successfully demonstrate an automated means of producing high-quality, reproducible systematic evidence reviews and other evidence products in shorter timelines, enabling substantially increased throughput to expand and improve AHRQ's in-house production capacity. In addition, AHRQ desires the same or another individual, team, or entity to build practical, trustworthy, and reliable super intelligence (SI) applications that use AHRQ data to improve healthcare research, implementation quality, and safety.

Return to Contents

Overview

A. Clinical Evidence Synthesis Transformation Challenge

The rapid expansion of new medical knowledge, therapeutics and technologies, combined with rising healthcare spending that often fails to produce better patient outcomes, has created an unprecedented demand for systematic evidence analysis. Systematic reviews identify, evaluate, and combine published medical studies and related sources to answer consequential questions, for example, whether a drug, procedure, diagnostic test, preventive service, or model of care should be used, for which patients, and under what circumstances. Such reviews help patients, physicians and other healthcare providers, healthcare systems, payers, policymakers, and researchers understand what is known about a topic, what remains uncertain, and where additional research would be most valuable.

AHRQ currently supports a full range of evidence products ranging from comprehensive systematic clinical evidence reviews to rapid evidence products, technical briefs and evidence maps, methods, reports, medical test reviews, and living or updated reviews. These products are scientifically valuable. However, conventional evidence synthesis employs labor-intensive searches of journal articles, trial registries, regulatory records, supplemental files, prior reviews, and structured databases, source screening, data extraction, critical appraisal, statistical analysis, report preparation, review, and updating for each review. Throughput is limited and completed projects are difficult to maintain and update, often becoming obsolete soon after they are finalized.

This $5.85 million challenge seeks to transform evidence analysis by providing breakthrough ways to help answer the preceding questions by automating the laborious, expensive, low-throughput systematic review process using new technologies such as SI to increase productivity, reduce costs, and improve quality, standardization, and reproducibility. It will strengthen the scientific basis for decisions about optimal and appropriate care by transforming how this evidence is analyzed and produced.

Recent advances in super intelligence, information retrieval, natural-language processing, machine learning, software agents, knowledge representation, automated statistical analysis, and human-computer collaboration create an opportunity to redesign the entire evidence-analysis process. The idea behind the challenge is not merely to append an SI tool to each currently existing manual step, but to make the underlying evidence base and the full range of derived products continuously updateable, queryable, auditable, and reusable, so that all products are generated from a common evidence foundation.

The challenge seeks a working system and operating model that can support the range of clinical evidence products AHRQ produces today and can also create new products yet to be conceived that become useful or feasible only when advanced technology changes the cost, speed, and scale of evidence synthesis. Rather than requesting a static report, decision-makers could potentially interrogate a living evidence base, receive alerts when conclusions change, or obtain a specific analysis within hours.

On the other hand, AHRQ intends this challenge to be outcome-based and technology-neutral within the preceding objectives. Super intelligence will, by necessity, likely become a central component, but participants may use any combination of models, retrieval systems, knowledge graphs, ontologies, deterministic software, statistical packages, automated agents, human-in-the-loop methods, and other technologies that accomplish their goals. Therefore, a fully autonomous system is not required. However, a competitive solution should automate routine work aggressively, make uncertainty and exceptions visible, and reserve human effort for consequential judgments. All solutions must maintain or enhance scientific rigor and produce unbiased, evidence-focused results that adhere to fit-for-purpose principles.

The challenge will have two phases and a proposed prize pool of up to $5.85 million. In Phase 1, participants will submit an end-to-end architecture, prototype or proof of capability, operating plan, quantified performance claims, and pathway for secure AHRQ operation. Up to five finalists will receive $100,000 each for a total of up to $500,000 for Phase 1.

In Phase 2, finalists will build and operate working systems to solve common, prospective, or hidden benchmark questions including a comprehensive review, rapid-response product, living evidence update, AHRQ-operated run, innovation product, and stress or red-team testing. AHRQ intends to select one Phase 2 Grand Prize winner who can receive $5.35 million, the largest SI grand prize to date. Individuals, teams, and entities, including existing AHRQ Evidence-based Practice Centers, other evidence synthesis centers, academic institutions, healthcare organizations, and established government contractors, are encouraged to apply.

B. SI Tech Sprint for Healthcare Research and Implementation Quality and Safety

The $2.1 million SI Tech Sprint will foster the AHRQ SI-able data ecosystem. AHRQ is seeking high-value applications that make material, traceable use of AHRQ data resources through AHRQ-produced competition data services, i.e., application programming interfaces (APIs) and Model Context Protocol (MCP). Solutions should help users conduct valid research or translate evidence and data into implementation, quality, and safety actions while preserving methodological validity, privacy, security, provenance, transparency, and appropriate human oversight. The SI Tech Sprint includes two tracks. The SI for Healthcare Research track will support applications that help researchers formulate questions, identify appropriate AHRQ resources, conduct methodologically valid analyses, reproduce results, and communicate findings, uncertainty, and limitations. The SI for Implementation Quality and Safety track will support applications that help healthcare organizations, policymakers, quality and safety leaders, and other authorized users translate evidence and data into implementation planning, benchmarking, monitoring, prioritization, workflow redesign, and evaluation. Agent2Agent (A2A) interoperability may be used where it materially improves a multi-agent or cross-organization workflow, but A2A is not required.

Participants in both SI Tech Sprint tracks must use at least one AHRQ resource (Medical Expenditure Panel Survey – MEPS; Healthcare Cost and Utilization Project – HCUP, excluding controlled microdata; Consumer Assessment of Healthcare Providers and Systems – CAHPS; Surveys on Patient Safety Culture – SOPS; Network of Patient Safety Databases – NPSD; or other AHRQ datasets) in a material and auditable manner. The competition judges will determine which aspects of the specialized expertise required to use these resources correctly can be safely incorporated into software and which should remain visible to users, require expert review, or cause an application to decline an unsupported request.

The proposed prize pool is $2.1 million. Each track will offer first-, second-, and third-place awards of $400,000, $240,000, and $160,000, respectively. Two additional $250,000 awards will recognize trustworthy SI and data use and effective use of AHRQ interfaces and agent interoperability. The competition will remain application-oriented and capable of recognizing multiple successful tools rather than selecting a single enterprise system.

Return to Contents

A. Clinical Evidence Synthesis Transformation Challenge

Challenge Structure

This challenge has two phases. Up to five participants selected to receive a Phase 1 finalist prize may participate in Phase 2. Phase 2 will require development, live operation, and independent evaluation of a working system. AHRQ intends to select one Grand Prize winner at the conclusion of Phase 2.

Phase 1: Solution Design and Proof of Capability

Participants will describe an end-to-end technical and operating architecture, demonstrate a prototype or existing capability, propose a plan for supporting current and new AHRQ evidence products, disclose expected human labor and third-party dependencies, and present a credible path to secure operation by AHRQ personnel. Phase 1 will emphasize transformative potential, technical feasibility, scientific integrity, and the ability to move beyond incremental automation of legacy workflows.

Phase 2: Build, Transfer, and Independent Benchmark Evaluation

Finalists will develop and operate their systems against common AHRQ benchmark work packages, including prospective or hidden topics designed to reduce the risk of benchmark memorization or training-data leakage. Finalists will submit complete evidence products and reproducibility packages, support AHRQ-operated test runs, and demonstrate at least one new technology-enabled evidence product. AHRQ will measure quality, speed, human labor, throughput, reproducibility, maintainability, security, and the degree to which AHRQ staff can operate the solution in house.

Key Dates

Phase 1 submission period begins: Monday, Jan. 11, 2027, 5 p.m. ET

Phase 1 submission period ends: Friday, March 12, 2027, 5 p.m. ET

Phase 1 finalist prizes announced: Friday, April 30, 2027

Phase 2 begins: Monday, May 3, 2027

Phase 2 final system submissions due: Friday, Nov. 5, 2027, 5 p.m. ET

Live and AHRQ-operated evaluation period: Monday, Nov. 8, 2027, through Friday, Dec. 10, 2027

Grand Prize winner announced: Friday, Dec. 17, 2027

Prizes

Total Cash Prizes

$5,850,000

Cash Prize Breakdown

AHRQ has indicated its intent to award up to $5.85 million in total cash prizes from AHRQ funds, subject to the terms of this challenge announcement.

1. For Phase 1, AHRQ intends to select up to five finalist submissions to receive up to $100,000 each, for a total Phase 1 prize pool of up to $500,000.
2. For Phase 2, AHRQ intends to select one Grand Prize winner to receive up to $5.35 million.

AHRQ may select fewer than five Phase 1 finalists, may decline to make a Phase 2 award if no solution meets the minimum quality and reproducibility thresholds, and may modify prize amounts consistent with applicable law and the published rules. AHRQ, a component of the United States Department of Health and Human Services (HHS), is responsible for paying prizes from its funds.

Use of Cash Prizes

Cash prizes, not grants or contracts, may be awarded under this challenge announcement. HHS does not limit how winners may use prize money awarded to them. Phase 1 finalists may use prize money toward Phase 2 development, but they are not required to do so. Selection as a winner does not guarantee a future grant, cooperative agreement, procurement contract, or other federal award.

Detailed Description

Problem and Opportunity

The amount, complexity, and rate of production of health and healthcare research continue to grow. Evidence relevant to a single decision may be distributed across several journal articles, trial registries, regulatory records, supplemental files, prior reviews, structured databases, and other sources. Conventional review processes often recreate searches, eligibility decisions, extraction forms, analytic datasets, and narrative summaries that could be reused. Much of the work remains serial, document-centered, and difficult to update.

AHRQ seeks to move from a model in which each evidence product is largely a separate project toward a model in which reusable evidence objects, transparent analyses, and continuously maintained evidence structures support many products and users. The desired transformation should allow a smaller number of AHRQ personnel to oversee a much larger portfolio, shorten the time between a decision need and an evidence response, and increase the share of products that AHRQ can initiate, operate, update, and publish with its own staff and infrastructure.

The challenge is intended to attract both established evidence-synthesis organizations and nontraditional participants, including individual technologists, SI laboratories, software companies, data-science teams, statistical-methods groups, and interdisciplinary collaborations. AHRQ is seeking a solution that is scientifically credible enough for consequential healthcare decisions and technologically ambitious enough to reinvent the operating model of clinical evidence synthesis and analysis.

Challenge Objective

The objective is to create and demonstrate a technology-enabled evidence production system that can:

1. Support the full range of AHRQ evidence-synthesis products and analytic functions, including comprehensive reviews, rapid products, technical briefs and evidence maps, living updates, medical test and diagnostic reviews, methods products, and products addressing benefits, harms, implementation, patient safety, health systems, and comparative effectiveness.
2. Reduce calendar time and human labor dramatically while preserving or improving the completeness, correctness, relevance, and clarity of the resulting evidence.
3. Increase throughput by an order of magnitude or more, measured against a common benchmark and inclusive of all human and computational resources used.
4. Enable AHRQ personnel to configure, run, inspect, correct, update, and export evidence products without routine dependence on an external evidence-review team.
5. Produce outputs that are highly reproducible, with source-level provenance, versioned data and models, executable analysis, and an audit trail sufficient for independent verification.
6. Create at least one new evidence product or service that becomes feasible or useful because advanced technology changes the speed, scale, cost, or structure of evidence synthesis.
7. Operate within a credible federal implementation pathway, including accessibility, security, privacy, intellectual-property, data-rights, interoperability, and maintainability requirements.

Current and Potential Evidence Products

Participants should not assume that AHRQ's current report categories must remain separate or document-based. Phase 1 submissions should explain which existing products the proposed system will support, which distinctions among products may become less useful, and what new outputs should be created. At minimum, the proposed architecture should be capable of producing the evidence and analytic components needed for the following current product families:

• Comprehensive systematic reviews and comparative effectiveness reviews.
• Rapid evidence products and rapid-response reviews.
• Technical briefs, scoping products, and evidence maps.
• Knowledge gap analyses.
• Living reviews, surveillance products, and targeted updates.
• Medical-test and diagnostic evidence reviews, including linked evidence when direct outcome evidence is incomplete.
• Methods reports, evidence synthesis guidance, and analyses of research gaps or future research needs.
• Evidence products that address implementation, contextual factors, patient safety, health system interventions, benefits, harms, and applicability.

Each Phase 1 participant must also propose at least one technology-enabled product that is not merely a faster version of a conventional report. Examples include, but are not limited to:

• A continuously updated, queryable evidence service that can generate a decision-specific answer and its supporting analysis on demand.
• A machine-readable evidence package or evidence graph linking populations, interventions, comparators, outcomes, study designs, judgments, extracted values, source passages, and executable analyses.
• An evidence change alert that identifies new studies, estimates whether they are likely to change a conclusion, and triggers a targeted update.
• An uncertainty, disagreement, or contradiction map that makes evidentiary conflicts and their likely causes visible to decision-makers.
• A research-gap and value-of-information product that prioritizes questions whose resolution could materially improve decisions.
• An interactive product that permits users to explore results by population, setting, outcome, analytic assumption, or decision threshold while preserving appropriate uncertainty and warnings.
• A product that combines synthesized published evidence with a prespecified, reproducible analysis of structured federal or real-world data, where lawful and appropriate.

Required Functional Capabilities

A competitive solution should address the evidence-production lifecycle as an integrated system. AHRQ does not require that every function be performed by one model or one software product. The participant may assemble modular technologies, provided the combined solution is coherent, auditable, legally usable, and maintainable. Capabilities should include, as applicable:

1. Question formulation and refinement, including analytic frameworks, key questions, population, intervention, comparator, outcomes, timing, and setting (PICOTS) or analogous structures, outcome definitions, subgroup hypotheses, and explicit estimands.
2. Systematic retrieval from multiple bibliographic, registry, regulatory, gray literature, and structured data sources; deduplication; citation network expansion; and documentation sufficient to reproduce the search.
3. Automated or SI-assisted eligibility screening with calibrated confidence, transparent reasons, disagreement handling, and high recall safeguards.
4. Structured extraction of study characteristics, interventions, comparators, outcomes, denominators, time points, effect measures, statistical uncertainty, and other critical fields, with every extracted item linked to the exact supporting source location.
5. Recognition and handling of randomized trials, observational studies, diagnostic-accuracy studies, qualitative and mixed-methods research, implementation studies, prior systematic reviews, and other designs relevant to AHRQ products.
6. Risk-of-bias, quality, applicability, and certainty or strength-of-evidence assessment, with clear separation between machine-generated recommendations and human judgments and with escalation of ambiguous or consequential cases.
7. Deterministic and executable statistical analysis, including effect-size calculation, meta-analysis, heterogeneity assessment, sensitivity analysis, subgroup or meta-regression analysis where justified, publication-bias assessment when appropriate, and support for more advanced synthesis methods when required.
8. Narrative synthesis and report generation that distinguish evidence from inference, represent uncertainty accurately, avoid unsupported causal language, and trace all material claims to evidence or analysis.
9. Continuous surveillance, change detection, version control, updating, and comparison of conclusions over time.
10. Export of human-readable and machine-readable products, data, code, application programming interfaces (APIs), structured evidence tables, excluded study lists with reasons, model and prompt records, and audit logs.
11. Human exception management, including abstention, confidence thresholds, materiality ranking, targeted quality-control sampling, and interfaces that direct reviewers to the cases most likely to affect conclusions.
12. Administrative, security, and governance functions including role-based access, logging, source and license tracking, model-version control, reproducible environments, and the ability to prevent unauthorized retention or use of federal information.

Required Scientific and Reproducibility Characteristics

The challenge will not reward speed or volume at the expense of scientific validity. A system that produces many outputs but cannot show where its evidence came from, how its calculations were performed, or whether its conclusions can be reproduced will not be competitive. The following characteristics are central:

• Source grounding. Every material factual or quantitative statement must be linked to a verifiable source passage, structured record, or executable analysis. Fabricated, nonresolvable, or materially mismatched citations are unacceptable.
• Explicit uncertainty. The system must represent uncertainty, missing information, disagreement, and limits of applicability rather than concealing them behind fluent prose.
• Deterministic computation. Quantitative synthesis and transformations should be executed by inspectable code or validated statistical software rather than generated as unaudited free text.
• Reproducible workflow. An independent AHRQ user must be able to rerun the workflow from the preserved inputs, versions, configurations, and code and obtain the same or materially equivalent results.
• Versioned provenance. The system must record sources, search dates, data versions, model versions, prompts or instructions where relevant, human interventions, decision history, and output lineage.
• Appropriate human oversight. The system should minimize routine manual repetition but identify cases that require expert judgment. Participants should explain why each human checkpoint is necessary and how it affects quality, time, and cost.
• Robustness and generalizability. Performance must extend beyond a single disease, publication type, or benchmark. The system should identify when it is operating outside its validated domain.
• Integrity of evaluation. Participants must disclose all human work, subcontracted work, benchmark-specific tuning, prior exposure to benchmark topics, and use of reference products. Undisclosed manual production or data leakage may be grounds for disqualification.

Illustrative Performance Outcomes

AHRQ will publish a Phase 2 Benchmark Manual before Phase 2 begins. The Manual will define work units, reference methods, scoring tolerances, material errors, permitted resources, logging requirements, and any minimum thresholds. The following proposed targets communicate the intended scale of transformation; final thresholds may be adjusted before Phase 2 to preserve feasibility and fairness.

For timing purposes, AHRQ-controlled peer review, public comment, clearance, and publication intervals will be reported separately from the participant-controlled evidence production interval. Participants may not improve apparent speed by narrowing the question, omitting sources or outcomes, replacing required judgments with generic text, or shifting work to undisclosed human reviewers.

Phase 1 Submission Requirements

Phase 1 submissions should present a credible, testable plan rather than a general statement that SI can improve evidence synthesis. A submission should include:

1. A description of the proposed system, its current maturity, and the specific transformation it would produce in AHRQ evidence synthesis.
2. An end-to-end architecture showing data sources, retrieval, screening, extraction, appraisal, analysis, synthesis, report generation, updating, human review, provenance, and export.
3. A product matrix identifying which current AHRQ products the system supports and at least one proposed new product, including the decision need it serves and why that product is feasible only or primarily because of the proposed technology.
4. A prototype demonstration, benchmark result, or other concrete evidence of capability. Demonstrations should disclose the topic, data, human assistance, system versions, costs, and evaluation method.
5. A Phase 2 work plan and schedule, including milestones, personnel, computing resources, third-party dependencies, risk management, and contingency plans.
6. A plan to measure and report calendar time, total human labor, computational use, cost, accuracy, reliability, reproducibility, and throughput.
7. A plan for AHRQ in-house operation, including training, workflow configuration, interfaces for nonprogrammer users, administrative controls, maintenance, and how AHRQ will avoid routine reliance on the participant for substantive review work.
8. A security, privacy, accessibility, data-rights, and intellectual-property plan, including disclosure of all third-party models, software, databases, licenses, and restrictions relevant to federal use, including: for each component, disclose limits on government access to, inspection, retention, reuse, or disclosure of underlying data, full text, source code, models, model weights, and documentation. The plan must identify limits on government reproduction, modification, publication, or sharing of outputs, extracted data, evidence products, and evaluation results, and identify restrictions on operation by AHRQ or its authorized contractors, including hosting, security testing, transfer, and continued use after the challenge. The plan must also disclose provider retention or use of government inputs or outputs for model training, and identify applicable permissions, licenses, subscriptions, fees, attribution obligations, use or user restrictions, and expiration or termination conditions. Finally, the plan must explain how AHRQ can lawfully perform the evaluation, operation, reproducibility, and publication activities specified in this challenge, consistent with third-party licenses. Identify any and all restrictions that could potentially prevent those activities, and how such restrictions will be addressed. Disclosure of a restriction, without a detailed plan explaining how it will be resolved or addressed to allow AHRQ to lawfully perform the affected activities, does not satisfy the applicable requirement.

Phase 2 Benchmark Work Packages

AHRQ anticipates evaluating finalists on common work packages that test reconstruction of defensible evidence products and operation under prospective conditions. The final design will be described in the Phase 2 Benchmark Manual. AHRQ currently anticipates the following components:

1. Comprehensive review. A prospective or hidden clinical or health-services question that requires an analytic framework, systematic search, study selection, structured extraction, critical appraisal, quantitative or structured narrative synthesis, strength-of-evidence judgments, and a complete draft report.
2. Rapid response. A second question announced on a compressed schedule, requiring a decision-ready rapid evidence product with transparent streamlining decisions and explicit limitations.
3. Living update. An update to an existing evidence base after release of a controlled set of new studies, requiring identification of new evidence, targeted extraction and analysis, comparison with the prior conclusion, and an explanation of whether and why the conclusion changed.
4. AHRQ-operated run. Trained AHRQ personnel will use the finalist system in a clean or controlled environment to configure, run, inspect, correct, update, and export one or more products. Participant personnel may provide technical support but may not silently perform the evidence review.
5. Innovation product. Each finalist will demonstrate the new product proposed in Phase 1 or another product approved by AHRQ. Evaluation will emphasize decision usefulness, scientific defensibility, reuse of structured evidence, and impact on evidence service.
6. Stress and red-team tests. AHRQ may introduce difficult abstracts, contradictory sources, duplicated publications, retractions, incomplete reporting, changed outcome definitions, tables embedded in images or supplements, prompt injection content, citation traps, or out-of-domain tasks to evaluate robustness and appropriate abstention.

Reference Standard and Independent Evaluation

AHRQ may establish an independent reference process using qualified evidence reviewers, information specialists, clinicians, statisticians, methodologists, and subject-matter experts who are not affiliated with the finalists. Reference products may be created in parallel, adjudicated after submission, or assembled from validated components. The reference process is intended to evaluate material correctness, not to force every finalist to reproduce one stylistic or methodological choice when multiple scientifically defensible choices exist.

Evaluation may include citation-level and field-level accuracy, search recall, eligibility agreement, errors in intervention arm assignment, denominators, outcomes and time points, risk-of-bias agreement, statistical replication, conclusion concordance, uncertainty calibration, robustness to repeated runs, and whether errors would change a decision or the overall conclusion. AHRQ may make evaluation datasets, scoring scripts, and de-identified adjudication results public after the challenge when legally and scientifically appropriate.

AHRQ In-House Operating Model and Transition

An important purpose of the challenge is to increase the percentage of evidence products that AHRQ can perform in house that are scientifically sound and based on unbiased evidence. For this challenge, in-house operation means that trained AHRQ personnel can formulate or import a review question, configure the workflow, launch and monitor searches, inspect and adjudicate exceptions, rerun analyses, update evidence, generate outputs, and export the complete reproducibility package without routine performance of those tasks by the winner or another evidence-review contractor.

The winning system should support a practical division of labor in which technology performs high-volume, repeatable tasks, AHRQ staff make policy and scientific judgments, and outside experts are used selectively for domain consultation, independent peer review, specialized methods, or system maintenance. The winner must provide training, documentation, an administrative guide, a reproducible deployment package, and a transition plan sufficient for AHRQ to operate the core workflow. AHRQ may test the system with users who did not participate in its development.

AHRQ-Provided Resources

To support Phase 2, AHRQ may make common resources available to all finalists on substantially equivalent terms. These resources may include selected completed and ongoing evidence products, structured data from AHRQ-supported repositories, benchmark corpora, protocols, study reports, access to approved bibliographic or full text sources, application programming interfaces, computing resources, a secure evaluation environment, reusable analytic components, technical support, and access to AHRQ program and methods personnel. Availability will depend on law, licensing, appropriations, security, privacy, and programmatic considerations.

Participants may use their own data, models, platforms, and licensed sources, but must disclose them and demonstrate that AHRQ can lawfully evaluate and, if selected, use the resulting system. A solution that depends on undisclosed data, inaccessible full text, nontransferable licenses, or services that cannot be used in a federal environment may receive a lower score or be deemed infeasible.

Required Phase 2 Deliverables

Each finalist must provide the deliverables specified in the Phase 2 Benchmark Manual. At minimum, AHRQ anticipates requiring:

1. A working system accessible in an AHRQ-approved environment for the evaluation period.
2. All required evidence products in both human-readable and machine-readable formats.
3. Search records, citation sets, screening decisions, excluded study lists with reasons, extracted data, critical appraisal judgments, evidence tables, analytic datasets, executable code, outputs, and complete provenance and audit logs.
4. System architecture, installation and configuration instructions, software bill of materials, model and data documentation, third-party dependencies, proposed licensing terms, security documentation, accessibility documentation, and known limitations.
5. A quantitative performance report covering accuracy, material errors, time, human labor, compute, cost, throughput, repeated run stability, and comparison with the participant's Phase 1 claims.
6. Training materials and completion of an AHRQ staff training and transfer exercise.
7. A maintenance, update, and transition plan addressing model changes, source changes, monitoring, error correction, vendor continuity, and operation for at least the first 12 months after the challenge if AHRQ elects to continue using the system.
8. A live demonstration and response to technical, scientific, security, and user-experience questions from the evaluation panel.

Technology, Data, and Evaluation Integrity Rules

1. Participants must disclose all models, software, databases, APIs, full-text sources, contractors, and material third-party services used in preparing or operating the submission, including versions and relevant license restrictions.
2. Participants must maintain contemporaneous logs of human labor, manual interventions, system calls, model and configuration changes, and benchmark-specific tuning. AHRQ may audit those records. Undisclosed human production, fabricated telemetry, or material omission may result in disqualification.
3. Participants may not use or disclose nonpublic challenge data except as authorized by AHRQ. Participants may not use nonpublic benchmark materials to train or improve a generally available model or service unless AHRQ expressly authorizes that use in writing.
4. Participants must use literature, data, and full text lawfully. AHRQ does not authorize infringement, circumvention of access controls, or use of content beyond the scope of a license or exception provided by law.
5. AHRQ may test submissions for data leakage, memorization, unsupported claims, prompt injection, cybersecurity vulnerabilities, repeated-run instability, and performance outside the demonstrated domain. Participants must cooperate with reasonable evaluation and red-team procedures.
6. Benchmark inputs and products must not contain protected health information (PHI), personally identifiable information (PII), or directly identifiable establishment information. AHRQ will not provide or authorize access to such research data through this challenge. Other nonpublic materials, such as unreleased benchmark questions, reference answers, or licensed content, remain subject to applicable data-use agreements, licenses, and security requirements. Controlled data that meet this restriction require prior written AHRQ approval before use. These restrictions concern research and benchmark data, not contact information required for registration or prize administration.
7. AHRQ may publish submitted evidence products, methods, performance results, adjudication summaries, and other challenge materials consistent with the licenses granted, applicable law, protection of confidential commercial information, and the challenge rules.

Return to Contents

Judging

HHS Award Approving Official

The HHS Award Approving Official will be the Director of AHRQ. AHRQ is responsible for final decisions regarding prizes paid with AHRQ funds.

Basis Upon Which a Winner Will Be Selected

All submissions will be prescreened for eligibility, completeness, and compliance. Eligible submissions will be reviewed by an evaluation panel composed of technical experts with relevant expertise in evidence synthesis, information science, clinical research, statistics, super intelligence, software engineering, human factors, cybersecurity, accessibility, and federal implementation. Panel members will individually evaluate submissions using the criteria below and may participate in consensus discussions, demonstrations, interviews, and technical validation.

A judging panel composed of federal employees from AHRQ and potentially other agencies will consider the evaluation panel's assessments and recommend finalists and the Grand Prize winner, subject to final approval by the award approving official. AHRQ may use independent contractors or other nonvoting experts to administer benchmarks or provide technical analysis, subject to conflict-of-interest and confidentiality requirements.

Phase 1 Judging Criteria: Solution Design and Proof of Capability

1. Transformative potential – 15 points. To what extent could the proposed solution materially change AHRQ's evidence-production speed, scale, cost, and operating model rather than provide incremental assistance to an otherwise unchanged manual workflow?
2. Scientific and technical merit – 20 points. Is the architecture technically credible and scientifically appropriate? Does it address retrieval, screening, extraction, appraisal, analysis, synthesis, uncertainty, and updating in a coherent way?
3. Quality, reproducibility, and provenance – 15 points. Does the proposal provide a credible approach to source grounding, executable analysis, versioning, audit trails, error detection, abstention, and independent reproduction?
4. Speed, labor reduction, and throughput – 15 points. Are the claimed gains substantial, measurable, inclusive of all human work and computer resources, and supported by prototype data or a persuasive validation plan?
5. AHRQ in-house operation and transition – 15 points. Can AHRQ personnel operate the system after limited training? Does the proposal reduce routine dependence on outside review labor and address deployment, usability, maintenance, and vendor lock-in?
6. Breadth and product innovation – 10 points. Can the solution support multiple current AHRQ product types and study designs? Is the proposed new product genuinely enabled by technology and useful to decision-makers?
7. Implementation feasibility, security, accessibility, and data rights – 5 points. Is there a credible federal implementation pathway and lawful access to needed content and technology?
8. Participant capability and execution plan – 5 points. Does the participant have, or have a credible plan to obtain, the interdisciplinary capability required to complete Phase 2?

Phase 2 Judging Criteria: Demonstrated Performance

1. Scientific validity and product quality – 30 points. How complete, correct, appropriately interpreted, and decision-useful are the submitted products? Do they meet the minimum thresholds and avoid material unsupported claims or errors?
2. Reproducibility, transparency, and auditability – 20 points. Can AHRQ reproduce the evidence base, critical data, analyses, and outputs? Are claims and judgments traceable and are versions and human interventions preserved?
3. Calendar time, human labor, cost, and throughput – 20 points. How much faster and less labor-intensive is the system, and what sustainable throughput does it enable under comparable scope and quality?
4. AHRQ-operated usability and transition – 15 points. How successfully can AHRQ staff operate, inspect, correct, update, and export products without routine reliance on finalist staff?
5. Breadth and new-product value – 10 points. Does the system generalize across products, topics, sources, and designs, and does the innovation product create meaningful new decision capability?
6. Security, maintainability, interoperability, and total operating burden – 5 points. Can the solution be deployed and sustained in a federal environment with acceptable risk, cost, and dependency?

Minimum Conditions for a Phase 2 Award

Regardless of total score, AHRQ does not intend to award the Grand Prize to a submission that fails a material quality or reproducibility threshold specified in the Phase 2 Benchmark Manual. Proposed threshold conditions include:

• No fabricated or nonresolvable citations and no pattern of materially mismatched source attribution.
• No material quantitative error that changes the direction, magnitude, certainty, or practical interpretation of a principal finding without being detected and corrected by the submitted workflow.
• Successful independent reproduction of principal analyses and outputs in an AHRQ-controlled environment.
• Complete disclosure of human labor, systems, data sources, versions, and third-party dependencies.
• Successful AHRQ-operated demonstration of core workflow functions after the specified training period.
• A legally and technically credible pathway for government use of the winning solution.

Return to Contents

How to Enter

To enter this challenge, participants must:

1. Submit a signed HHS Challenge Participant Registration Form as a PDF by the Phase 1 deadline to SIHealthCareImprovement@ahrq.hhs.gov (see https://www.ahrq.gov/si/evidence-synthesis-challenge/register.html).
2. Submit all Phase 1 materials to SIHealthCareImprovement@ahrq.hhs.gov by the Phase 1 deadline.
3. If selected as a finalist, execute any Phase 2 participation, confidentiality, data-use, security, and evaluation-integrity acknowledgments required by AHRQ and submit all Phase 2 materials through the designated mechanism by the applicable deadlines.

Return to Contents

Additional Information Regarding Registration

Registration is required to be eligible for a prize. A participant must register as:

1. An individual, acting on behalf of that individual;
2. A team, acting on behalf of a group of people competing together but not on behalf of an established legal entity, i.e., any organization such as a for-profit or not-for-profit corporation, partnership, cooperative, college or university, charitable organization, tribal entity, that is formally created and legally recognized under applicable law. Each team must designate an eligible team leader responsible for registration, submission, communications, and receipt of prize payment; or
3. An entity, acting on behalf of a legally established organization, institution, or corporation.
   The registration category determines who HHS pays. Before registering, participants may wish to consider legal, tax, intellectual-property, employment, and conflict-of-interest consequences. The HHS Challenge Participant Registration Form will be available under "Resources."  HHS may grant or deny requests to update registration information in its sole discretion.

Return to Contents

Additional Information Regarding Submissions

Phase 1 Submission Requirements

Phase 1 narrative submissions must be limited to 20 single-spaced pages, excluding a cover page, table of contents, references, architecture diagrams, letters of commitment, and a software or model inventory. Use a font no smaller than 11-point Arial and margins no smaller than one inch. AHRQ may specify file formats and size limits in the official submission instructions.

Phase 1 submissions should include:

1. A two-page executive summary, included within the 20-page narrative limit, stating the problem, proposed solution, current maturity, expected speed and throughput gains, and why the approach is disruptive rather than incremental.
2. The technical, scientific, and operating architecture described under Phase 1 submission requirements above.
3. A matrix mapping the solution to current AHRQ evidence-product types and the proposed new product or service.
4. A description of prototype evidence and a link to or file containing a demonstration video of no more than 15 minutes. AHRQ may also request access to a live prototype or sandbox.
5. Quantified performance claims and the basis for those claims, including all human labor and computing resources used.
6. A Phase 2 work plan, staffing plan, transition plan, security and accessibility approach, risk register, and intellectual-property and data-rights disclosure.

Phase 2 Submission Requirements

Phase 2 finalists must comply with the Phase 2 Benchmark Manual and submit the required working system, evidence products, machine-readable data, code, logs, performance report, transition package, and documentation. The Phase 2 narrative report must be limited to 30 single-spaced pages, excluding participant-generated evidence products, appendices, reproducibility packages, source code, machine-readable files, security materials, user manuals, and other artifacts requested by AHRQ.

Finalists must identify any change from Phase 1 architecture, claims, team, data sources, models, or dependencies. Finalists must not remove unfavorable benchmark results, selectively report repeated runs, or substitute unreported human work. AHRQ may require a signed certification concerning completeness and accuracy of the Phase 2 submission.

Return to Contents

Resources

• AHRQ Center for Evidence and Practice Improvement: https://www.ahrq.gov/cpi/centers/cepi/index.html. Describes AHRQ evidence synthesis and the Evidence-based Practice Center (EPC) Program.
• AHRQ Effective Health Care Program: https://effectivehealthcare.ahrq.gov/. Provides examples of evidence reviews, rapid products, protocols, technical briefs, and related materials.
• EPC Research Methods and Tools: https://effectivehealthcare.ahrq.gov/research-methods. Provides AHRQ evidence-synthesis methods guides, tools, education resources, and methods research.
• HHS Challenge Participant Registration Form: https://www.ahrq.gov/si/evidence-synthesis-challenge/register.html. Registration form and instructions.
• Phase 2 Benchmark Manual: https://www.ahrq.gov/si/evidence-synthesis-challenge/benchmark-manual.html. Will define benchmark tasks, metrics, minimum thresholds, permitted resources, reporting requirements, and evaluation procedures.
• Point of Contact/Questions: Dr. Gil Alterovitz, Chief Data Officer and Chief Super Intelligence Officer, SIHealthCareImprovement@ahrq.hhs.gov

Return to Contents

Frequently Asked Questions

1. Must a submission use generative SI?

No. Super intelligence should be central to the transformation, but AHRQ is not prescribing a model type. Solutions may combine generative models, conventional machine learning, information retrieval, rules, ontologies, knowledge graphs, statistical software, and other technologies.

2. Must the system be fully autonomous?

No. AHRQ expects that consequential judgments will sometimes require humans. Competitive solutions should automate routine work, identify uncertainty, abstain when appropriate, and direct limited human effort to exceptions and decisions that could materially affect conclusions.

3. May a single individual win?

Yes. An eligible individual may register, compete, and receive the Grand Prize. The individual must nevertheless meet the same scientific, technical, security, licensing, transition, and deliverable requirements as any other participant.

4. Must a participant be an AHRQ Evidence-based Practice Center or an established evidence-synthesis organization?

No. AHRQ encourages participation by nontraditional entrants and interdisciplinary teams interested in improving the quality and throughput of evidence reviews. Existing evidence-synthesis experience may be valuable but is not an eligibility requirement.

5. Is open-source software required?

No. Open-source, proprietary, and mixed architectures are eligible. The participant must disclose dependencies and provide sufficient government rights, exportability, documentation, and continuity for AHRQ to evaluate and use the solution as required by the challenge.

6. May participants use existing commercial models, databases, or platforms?

Yes, provided their use is lawful and fully disclosed and AHRQ can evaluate the resulting solution. Participants should identify ongoing costs, restrictions, data-retention practices, model-change risks, and whether the component can be used in a federal environment.

7. How will AHRQ prevent a system from appearing fast because humans performed undisclosed work?

AHRQ will require labor logs, workflow telemetry, system logs, model and version records, disclosure of all subcontracted activity, common benchmark tasks, and AHRQ-operated runs. AHRQ may audit records and conduct live or controlled tests.

8. How will AHRQ judge quality when reasonable reviewers may disagree?

AHRQ intends to focus on material correctness, transparent reasoning, reproducible evidence and analysis, and defensible handling of uncertainty. Reference adjudicators may recognize multiple acceptable methods or judgments. The benchmark will distinguish minor stylistic differences from errors likely to change a conclusion or decision.

9. Does the challenge require the winner to reproduce current report formats exactly?

No. The system must be capable of supplying the evidence and analytic components needed for current AHRQ products, but AHRQ expressly seeks new formats and services. Finalists should preserve necessary scientific content while explaining what can be simplified, merged, automated, made interactive, or replaced by a better technology-enabled product.

10. What does increased in-house production mean?

It means that AHRQ staff can perform the recurring evidence workflow using the system, including configuration, monitoring, adjudication of exceptions, analysis, updating, and export, rather than commissioning a separate outside review team for each product. Specialized consultation, independent peer review, infrastructure maintenance, and occasional technical support may remain external.

11. Will AHRQ provide copyrighted full text, proprietary databases, or computing resources?

AHRQ may provide common resources to Phase 2 finalists when lawful and feasible, but participants should not assume access to any particular source or service unless it is identified in the Phase 2 Benchmark Manual. Participants are responsible for proposing a lawful and sustainable source-access strategy.

12. Does winning guarantee a follow-on federal contract or adoption?

No. A prize award does not guarantee procurement, continued access, deployment, or any other federal award. AHRQ may use the challenge results to inform future program, research, acquisition, or implementation decisions consistent with applicable law.

Return to Contents

B. SI Tech Sprint for Healthcare Research and Implementation Quality and Safety

SI Tech Sprint Structure

This is a single-phase SI Tech Sprint with two tracks:

• Track A: SI for Healthcare Research – applications that help researchers and analysts formulate questions, identify appropriate AHRQ resources, construct valid analyses, use AHRQ MCP APIs, reproduce results, and communicate findings.
• Track B: SI for Implementation Quality and Safety – applications that help healthcare organizations, policymakers, quality and safety leaders, and other authorized users translate evidence and data into implementation planning, benchmarking, monitoring, prioritization, workflow redesign, and evaluation.

Key Dates

Public launch webinar: Tuesday, Dec. 1, 2026, 3 p.m. ET.

Submission period begins: Monday, March 1, 2027, 5 p.m. ET

Submission period ends: Thursday, Sept. 9, 2027, 5 p.m. ET

Winners announced (tentative): Tuesday, Oct. 19, 2027, time to be announced

Technical onboarding sessions and office hours: Dates to be announced

Prizes

Total Cash Prizes

$2,100,000

Cash Prize Breakdown

AHRQ Prize Funds

AHRQ may award up to a total cash prize pool of $2.1 million to the SI Tech Sprint winners from its fiscal year 2026 funds.

HHS may make fewer awards or no award in a category if no submission meets the applicable minimum standard or other requirements in this announcement.

A participant may receive no more than one SI Tech Sprint prize, including cross-track awards. This limit does not apply to prizes in the Clinical Evidence Synthesis Transformation Challenge.

Use of Cash Prizes

Cash prizes (not grants or contracts) may be awarded under this SI Tech Sprint announcement. HHS does not limit how winners may use prize money awarded to them.

Detailed Description

AHRQ develops and supports nationally significant data resources describing healthcare use, spending, insurance, access, health status, hospital and facility encounters, patient experience, safety culture, patient safety events, healthcare quality, and disparities. Resources contemplated for this SI Tech Sprint include the Medical Expenditure Panel Survey (MEPS); public aggregate products from the Healthcare Cost and Utilization Project (HCUP), including its national databases; the Consumer Assessment of Healthcare Providers and Systems (CAHPS); the Surveys on Patient Safety Culture (SOPS); the Network of Patient Safety Databases (NPSD).

These resources can require specialized expertise to use correctly, including selecting the correct unit of analysis, applying survey or discharge weights, accounting for complex sample designs, interpreting benchmark populations, recognizing the limits of voluntary reporting systems, complying with disclosure restrictions, and avoiding unsupported causal or patient-specific conclusions. The SI Tech Sprint is intended to identify which aspects of that expertise can be safely encapsulated in software and which must remain visible to the user, require qualified review, or cause an application to decline a requested analysis.

The overall goal is to accelerate practical, trustworthy use of AHRQ data from research through implementation quality and safety. The SI Tech Sprint seeks usable applications and reusable design knowledge rather than a one-time demonstration. It will also help AHRQ understand how well competition-facing MCP interfaces support real-world workflows, where safe defaults can reduce complexity, where warnings or expert review are needed, and where interoperability or documentation should improve.

Track A: SI for Healthcare Research

Track A seeks applications that help researchers and analysts formulate questions, identify appropriate AHRQ resources, construct valid analyses, use AHRQ Model Context Protocol (MCP) APIs, integrate authorized external MCP APIs when appropriate, reproduce results, and communicate findings. Illustrative uses include guided cohort or population definition, survey-weighted descriptive analysis, hospital or facility utilization analysis, evidence generation, hypothesis development, quality measurement, and methods that help users understand uncertainty and limitations. Submissions should not be rewarded for producing a simple answer when the data or method cannot support that answer.

Track B: SI for Implementation Quality and Safety

Track B seeks applications that help healthcare organizations, policymakers, quality and safety leaders, and other authorized users translate evidence and data into implementation planning, benchmarking, monitoring, prioritization, workflow redesign, and evaluation. Illustrative uses include identifying quality or safety improvement opportunities, selecting relevant evidence or measures, comparing performance with appropriate benchmarks, estimating resource implications, and monitoring implementation.

Data, MCP, A2A, and Trustworthy Use

A qualifying submission must make material, traceable use of at least one AHRQ competition resource and explain why the selected resource is appropriate, how it was accessed, which version was used, which methodological requirements apply, and how the application prevents or communicates misuse. Where AHRQ provides an MCP interface for a competition resource, participants should use that interface for SI-agent access unless an approved technical reason requires another approach.

Authorized external MCP APIs, A2A agents, underlying APIs, and SI services may supplement, contextualize, validate, or operationalize AHRQ data, but they may not replace meaningful use of an AHRQ resource. Participants must disclose each external source, service, server, or agent; what information is exchanged; its role in the application; and relevant risks. A2A is encouraged when it adds material value, such as coordinating specialized research, quality, safety, or partner-service agents, but submissions will not receive credit for unnecessary agent complexity.

Applications must support trustworthy use. Depending on the operation, an application should safely automate a validated task, guide the user through material assumptions or confirmations, restrict a task to appropriately qualified experts, or refuse a prohibited or unsupported request. Material outputs should preserve provenance, assumptions, uncertainty, limitations, and reproducibility information in an analysis receipt or equivalent manifest.

Scope and Safety Boundaries

In the SI Tech Sprint, AHRQ does not authorize the use or submission of protected health information (PHI), personally identifiable information (PII), or directly identifiable establishment information as research data and will not provide access to such information through this SI Tech Sprint. Participants must use public-use, synthetic, or appropriately de-identified competition data that satisfy this restriction. Submission or release of controlled HCUP microdata are not authorized by this announcement; permitted HCUP resources are public aggregate products that meet applicable disclosure rules. Other controlled AHRQ resources remain subject to their applicable data-use agreements and licenses. Such resources may not be transmitted to an external service, MCP server, or A2A agent unless AHRQ expressly approves that service and transfer in writing. Public demonstrations must use synthetic or disclosure-reviewed aggregate research data and must not contain prohibited information.

The SI Tech Sprint does not authorize clinical deployment, connection to a production AHRQ environment, regulatory clearance or certification, or use of an application as patient-specific clinical decision support or a medical device unless a separate, expressly authorized scope is established. SI Tech Sprint recognition does not constitute AHRQ endorsement, certification, regulatory approval, or authorization for clinical use.

Qualifying finalist and award-winning applications are expected to be documented in a public app gallery. The gallery will describe intended users, AHRQ resources used, methods, validated use cases, safeguards, limitations, reproducibility materials, and review status.

SI Tech Sprint-Specific Technical and Data-Use Rules

• Each qualifying submission must make material, traceable use of at least one AHRQ competition resource. A submission that uses only AHRQ documentation or branding without using AHRQ data or measures in a material way will not qualify.
• External MCP APIs, A2A agents, underlying APIs, and SI services may be used only when their use is lawful, authorized, disclosed, and consistent with the terms of all data sources. Controlled AHRQ data may not be transmitted to an external service, MCP server, or A2A agent unless AHRQ has expressly approved that service and transfer in writing.
• Participants must not use SI Tech Sprint resources to re-identify persons or establishments, support prohibited linkages, release prohibited information or disclose controlled HCUP cells with 10 or fewer observations. Public demonstrations must use synthetic or disclosure-reviewed aggregate data that does not contain PHI, PII, or directly identifiable establishment information.
• Submissions must preserve provenance and auditability; communicate material assumptions, uncertainty, and limitations; and enforce required privacy, security, data-use, and human-approval controls. A warning may not be used as a substitute for blocking a prohibited operation.
• No application may connect to a production AHRQ environment without separate authorization.

Return to Contents

Judging

HHS Award Approving Official

The HHS Award Approving Official will be the AHRQ Director. AHRQ is responsible for final decisions regarding prizes paid with AHRQ funds.

Basis Upon Which a Winner Will Be Selected

Before numerical scoring, each submission must pass the approved eligibility, data-use, re-identification, disclosure, restricted-data transmission, security, accessibility, reproducibility, and prohibited-claim gates. A submission that fails a mandatory gate will not advance based on its numerical score.

Judging Criteria

Total: 100 points. Mandatory eligibility and compliance gates apply before numerical scoring.

How to Enter

To enter this SI Tech Sprint, participants must:

• register as an individual, team, or entity using one signed AHRQ SI Tech Sprint Participant Registration Form per submission through https://www.ahrq.gov/si/tech-sprint/register.html; and
• submit the SI Tech Sprint entry and all required materials through https://www.ahrq.gov/si/tech-sprint/submit.html by the submission deadline.

Return to Contents

Additional Information Regarding Registration

Registration is required to be eligible to win a cash prize. When you register using the "AHRQ SI Tech Sprint Participant Registration Form," you will need to decide whether you are registering as:

1. an individual (on behalf of yourself);
2. a team (on behalf of a group of people competing together, but not on behalf of an established legal entity). If registering as a team, each participating team is required to designate an eligible team leader who will register and submit on behalf of the team members. The team leader is responsible for all communications with HHS; or
3. an established legal entity (on behalf of a legally established organization, institution, or corporation). Your decision to register as an individual, team, or entity on the “AHRQ SI Tech Sprint Participant Registration Form” will determine who HHS pays any prize money from this SI Tech Sprint. Please see “Prize Payment Rules” in the “Rules” section. Before registering, you may wish to consider any potential tax consequences of registering as an individual, team, or entity.

The “AHRQ SI Tech Sprint Participant Registration Form” is available under the “Resources” section for this SI Tech Sprint.

Return to Contents

Additional Information Regarding Submissions

Submission Requirements

Each submission is required to provide enough information and materials for administrative screening, technical validation, reproducibility review, and judging. At minimum, the submission must include:

• REQUIREMENT: Submission title; participant/team name and affiliations; selected track; intended users; and the problem or decision the application is designed to support.
• REQUIREMENT: The AHRQ data resource(s) and version(s) used; why each resource is appropriate; how the resource was accessed; and the methodological requirements and limitations that apply.
• REQUIREMENT: A description of AHRQ MCP resources/tools used, any external MCP APIs or other APIs, any SI services, and any A2A agents or delegations; what information is exchanged; and the role of each external service in the application.
• REQUIREMENT: A method summary and an explanation of the application’s safe defaults, warnings, user confirmations, expert-review triggers, prohibited uses, and refusals.
• REQUIREMENT: Reproducibility and provenance materials sufficient for technical reviewers to reproduce material results, including code, configuration, documentation, test data, or executable packaging as applicable, together with an analysis receipt or equivalent manifest for material outputs.
• REQUIREMENT: Known limitations; data-use, privacy, and security controls; accessibility status; and the proposed implementation or adoption path.

Required Evaluation Content

To maintain consistency with the Judging section, each submission must address the following evaluation areas using the same criteria that judges will apply:

REQUIREMENT – Appropriate and material use of AHRQ data (20 points): Resource selection, correct unit and population, weighting or design, measure specifications, versioning, limitations, and auditability.

REQUIREMENT – Trustworthy SI and data safeguards (25 points): Safe defaults, transparency, uncertainty, warnings, expert-review triggers, refusals, privacy, disclosure controls, and model limitations.

REQUIREMENT – Track-specific user value and impact (20 points): Research utility for Track A or implementation quality and safety utility for Track B; relevance of the problem and decision.

REQUIREMENT – Usability and accessibility (15 points): Human-centered design, clarity for intended users, novice safety, accessibility, and understandable explanations.

REQUIREMENT – Technical quality and reproducibility (10 points): Reliability, testability, security, software quality, reproducibility, provenance, and maintainability.

REQUIREMENT – Innovation, scalability, MCP API, and agent interoperability (10 points): Novel capability, feasible adoption path, reusable components, cross-agency value, and useful insight into MCP API and A2A improvements.

Mandatory Compliance Gates

REQUIREMENT: Before numerical scoring, each submission must pass the approved eligibility, data-use, re-identification, disclosure, restricted-data transmission, security, accessibility, reproducibility, and prohibited-claim gates. Participants must provide the information and materials needed for HHS and its technical reviewers to evaluate those gates.

Finalists may be required to provide additional software-assurance and reproducibility materials, including software dependency information, secrets-scan results, a software bill of materials, or other materials specified in the finalist instructions.

Return to Contents

Resources

• AHRQ SI Tech Sprint Participant Registration Form: https://www.ahrq.gov/si/tech-sprint/register.html
• AHRQ SI Tech Sprint landing page: https://www.ahrq.gov/si/tech-sprint/index.html 
• Point of Contact/Questions: Dr. Gil Alterovitz, Chief Data Officer and Chief Super Intelligence Officer, SIHealthCareImprovement@ahrq.hhs.gov

Return to Contents

Frequently Asked Questions

Must my submission use AHRQ data?

Yes. A qualifying submission must make material, traceable use of at least one AHRQ competition resource. AHRQ documentation or branding alone is not sufficient.

Is MCP required?

Where AHRQ provides an MCP interface for a competition resource, participants should use that interface for SI-agent access unless an approved technical reason requires another approach. The final technical guide will identify available competition interfaces.

Is A2A interoperability required?

No. A2A is encouraged only when it materially improves an appropriate multi-agent or cross-organization workflow. A single-agent or deterministic architecture may be more appropriate for some use cases.

Can I use external data, MCP APIs, agents, or SI services?

Yes, when their use is lawful, authorized, disclosed, and consistent with all applicable data terms. AHRQ data must remain central and auditable in every qualifying submission, and controlled AHRQ data may not be sent to an unapproved external service or agent.

Can I use controlled HCUP data?

This announcement does not authorize controlled HCUP microdata. Participants may use HCUP products that comply with applicable disclosure rules and the "Scope and Safety Boundaries."

Does selection as a finalist or winner mean AHRQ endorses or certifies the application?

No. SI Tech Sprint recognition does not constitute AHRQ endorsement, certification, regulatory approval, or authorization for clinical use.

Return to Contents

Legal Authority

Challenges are authorized by the America Creating Opportunities to Meaningfully Promote Excellence in Technology, Education, and Science (COMPETES) Reauthorization Act of 2010, as amended (15 United States Code (U.S.C.) § 3719).

AHRQ, a component of the United States Department of Health and Human Services, is dedicated to health services research. As set forth in 42 U.S.C. § 299, AHRQ's mission includes enhancing the quality, appropriateness, and effectiveness of health services and access to such services through scientific research and the promotion of improvements in clinical and health-system practices. Consistent with AHRQ's statutory mandates, this challenge seeks to improve healthcare quality, safety, outcomes, affordability, access, and value by transforming the speed, scale, reproducibility, and usability of evidence synthesis and analysis.

Return to Contents

Rules

Eligibility Rules

1. To be eligible to win a prize under these challenges, a participant, whether an individual, team, or entity:
   1. Shall have registered to participate under the rules and instructions promulgated by HHS and published in this announcement, including the "Prize Payment Rules" and "How to Enter" sections.
   2. Shall have complied with all requirements set forth in this announcement.
   3. In the case of a private entity, shall be incorporated in and maintain a primary place of business in the United States; in the case of an individual or team participant, shall be a citizen or permanent resident of the United States.
   4. Shall not be a federal entity or a federal employee acting within the scope of federal employment.
   5. Shall not be an employee of HHS or any HHS component acting in a personal capacity.
   6. If employed by a federal agency or entity other than HHS, shall consult an agency ethics official to determine whether federal ethics rules limit or prohibit acceptance of a prize.
   7. Shall not be a judge of these challenges, another party involved with their design, production, execution, or administration, or an immediate family member of such a party (i.e., spouse, parent, stepparent, child, or stepchild).
   8. If an individual or team member, shall be 18 years of age or older at the time of submission.
2. A participant will not be deemed ineligible solely because the participant used federal facilities or consulted with federal employees during the challenge, provided the facilities and employees were made available to all participants on substantially equivalent terms.
3. A single eligible individual may register, compete, and win. A participant is not required to form a corporation, obtain an academic affiliation, or join a team, although the participant remains responsible for satisfying all requirements and providing all required capabilities.

Participation Rules

1. Participants may not use federal funds from a grant award or cooperative agreement to develop challenge submissions or to fund activities in support of challenge submissions.
2. Federal contractors may not use federal contract funds to develop challenge submissions or to fund activities in support of challenge submissions. Such costs are unallowable and are not allocable to government contracts.
3. By participating, each participant agrees to assume all risks and waive claims against the federal government and its related entities, except in the case of willful misconduct, for injury, death, damage, or loss of property, revenue, or profits arising from participation, whether direct, indirect, or consequential and whether arising through negligence or otherwise.
4. Based on the subject matter and anticipated activities, no participant is required to obtain liability insurance, demonstrate financial responsibility, or indemnify the federal government against third-party claims as a condition of participation, unless AHRQ publishes a modification stating otherwise.
5. Each participant (whether an individual, team, or entity) warrants that they are the sole author or owner of, or have sufficient rights to use, modify, evaluate, and license, all copyrightable works, software, models, data, documentation, and other materials included in the submission, that the works are wholly original with the participant (or are an improved version of an existing work that the participant has sufficient rights to use and improve), and that the submission does not knowingly infringe any copyright or any other third-party rights of which the participant is aware.
6. Each participant shall comply with all applicable federal, state, and local laws, regulations, and policies, including requirements concerning privacy, security, accessibility, records, research integrity, copyright, data use, and export control.
7. Participation constitutes full and unconditional agreement to the published rules. Winning is contingent upon fulfilling all requirements and completing winner-verification and payment documents by the deadline stated in the notification.
8. HHS may disqualify a submission that, in HHS’ judgment, is inconsistent with its mission, is ineffective or harmful, contains material misrepresentation, fails required testing, compromises evaluation integrity, or violates law or the challenge rules, or fails to deliver rights, source code, model artifacts, documentation, interfaces, deployment materials, and escrow arrangements described in the Intellectual Property and Government Use provision, or otherwise agreed to by authorized AHRQ representatives.

Intellectual Property and Government Use

1. License grant; scope of use. Each participant grants the United States Government (acting through HHS and AHRQ) an irrevocable, paid-up, royalty-free, nonexclusive, worldwide license — with the right to sublicense as provided in paragraph (6) — to reproduce, publish, post, link to, share, display, and evaluate the submission and challenge deliverables. Each participant further grants the United States Government an irrevocable, paid-up, royalty-free, nonexclusive, worldwide license to use, practice, operate, host, reproduce, and prepare and use derivative works of the submitted solution, and to have any of the foregoing done for or on behalf of the Government by its contractors, for any government purpose, including evaluation, deployment, operation, maintenance, security patching, modification, enhancement, integration, retraining, and adaptation of the solution. The license granted in this paragraph is not limited to the purposes enumerated in this announcement. To the extent a participant holds patents with claims that read on the submitted solution, the participant grants the United States Government a license of the same scope under those patents (including any continuations, divisionals, reissues, and foreign counterparts) to make, have made, use, and practice the solution for government purposes.
2. Retained rights; deliverables for the Phase 2 winner. Participants retain all other intellectual-property rights in their submissions, as applicable. Transfer of ownership to the Government is not required. However, as a condition of the award, the Clinical Evidence Synthesis Transformation Challenge Phase 2 winner must deliver to AHRQ sufficient rights, source code, build and configuration files, documentation, application programming and other interfaces, model artifacts and weights (including any fine-tuned artifacts), and deployment materials to enable AHRQ, and its contractors acting on its behalf, to test, operate, reproduce, modify, maintain, and prepare derivative works of the solution, and to have those things done on its behalf, without legal or technical obstacles that defeat the purposes of the challenge. The winner must also place the deliverables described in this paragraph, and any credentials or materials necessary to rebuild and redeploy the solution, into a source-code escrow or equivalent arrangement designated by AHRQ and must grant AHRQ a step-in right to access and use those materials to continue operating and maintaining the solution in the event the winner ceases to support it, becomes insolvent, or otherwise fails to sustain the solution.
3. Third-party and dependent components; participant facilitation. A Clinical Evidence Synthesis Transformation Challenge Phase 2 submission must identify all components, whether proprietary, open-source, or service-based, for which AHRQ would need a separate license, subscription, cloud service, model-access agreement, database license, application programming interface, or continuing vendor support in order to operate, maintain, or modify the solution, and must disclose for each the licensor, the applicable license or terms (including any copyleft or source-disclosure obligations), the expected terms and costs, and any use, field-of-use, or redistribution restrictions. The participant warrants that this disclosure is accurate and complete. At AHRQ's request, and to the extent within the participant's ability, the participant must reasonably facilitate AHRQ's acquisition of the rights identified under this paragraph, including by making introductions to the relevant licensors, providing existing license documentation, and using commercially reasonable efforts to assist AHRQ in obtaining those rights on terms no less favorable than those available to the participant; nothing in this paragraph obligates AHRQ to acquire any such component or obligates the participant to acquire rights on AHRQ's behalf. AHRQ may consider lock-in, continuity, source-disclosure obligations, and total cost of operation in judging feasibility and maintainability. 
4. Participants are not granted any license to use HHS names, seals, logos, or trademarks and may not imply endorsement by HHS or the federal government. Any authorized use is subject to prior written approval and applicable HHS public-affairs requirements.
5. Warranties. By participating, each participant warrants that: (a) it has the right to grant the licenses set forth above and there are no legal obstacles to doing so; (b) the participant is the owner of, or is otherwise authorized to license, all intellectual property in the submission other than the third-party components disclosed under paragraph (3); (c) each individual and entity that contributed to the submission, including employees, subcontractors, and co-authors, has assigned or licensed to the participant rights sufficient to support the licenses granted above; (d) to the participant's knowledge after reasonable inquiry, the submission and the submitted solution, when used as contemplated by this announcement, do not infringe or misappropriate the intellectual property or other proprietary rights of any third party; and (e) the disclosures made under paragraphs (3) and (8) are accurate and complete. To the extent permitted by law, the participant will indemnify and hold harmless the United States Government against third-party claims that the submission or submitted solution, as delivered, infringes or misappropriates that third party's intellectual property, except to the extent the claim arises from the Government's own modifications or from components the Government elected to substitute. Nothing in these challenge rules requires any participant to waive claims against the Government arising out of the Government's unauthorized use or disclosure of the participant's intellectual property, trade secrets, or confidential business information. The warranties in this paragraph survive the conclusion of the challenge.
6. Public release and downstream licensing. The license in paragraph (1) includes the right of the Government to reproduce, distribute, and make the submission and challenge deliverables available to the public, and to grant downstream licenses to the public, including release under an open-source license, for any government purpose. The Government will exercise this right consistent with the confidentiality protections in paragraph (9).
7. Grantee; sharing within Government; contractors. All licenses granted under these challenge rules run to the United States Government and may be exercised by any federal agency for government purposes, and by contractors acting for or on behalf of the Government. Notwithstanding any reference to a specific component or agency, the licenses are not personal to HHS or AHRQ and may be shared among federal agencies; the licenses may be assigned to a successor federal entity that assumes the relevant function.
8. Data provenance and training data. A Clinical Evidence Synthesis Transformation Challenge Phase 2 submission must disclose the provenance of all non-governmental data used to develop, train, fine-tune, validate, or operate the submitted solution, including the source of each dataset, the license or authorization under which it was obtained, and any use, redistribution, field-of-use, privacy, or consent restrictions applicable to that data or to models derived from it. The participant warrants that all such data was lawfully obtained and that its use in developing and operating the solution, and the Government's contemplated use of the solution, are consistent with the applicable data rights and restrictions. The participant must identify any data whose restrictions would limit the Government's ability to operate, reproduce, modify, or publicly release the solution, and any personally identifiable or protected health information incorporated into or required by the solution.
9. Confidential and proprietary information. A participant that includes trade secrets or confidential commercial or financial information in a submission must clearly and conspicuously mark those specific portions at the time of submission in the manner specified in this announcement. The Government will handle properly marked material in accordance with applicable law, including the Freedom of Information Act and its exemptions, and will not exercise the publication, public-distribution, or public-release rights in paragraphs (1) and (6) as to properly marked material without the participant's prior written consent, except as required by law. Failure to mark material as required may result in its treatment as publicly releasable. Overbroad or blanket confidentiality markings that do not reasonably identify specific protectable material need not be honored. Nothing in this paragraph limits the licenses granted for government purposes, and nothing in these challenge rules requires a participant to waive claims against the Government arising out of the Government's unauthorized use or disclosure of the participant's intellectual property, trade secrets, or confidential business information.

Prize Payment Rules

1. HHS will use the participant's registration as an individual, team, or entity to determine who receives prize payment.
   1. For a winning individual, HHS will pay the prize to that individual.
   2. For a winning team, HHS will pay the prize to the eligible team leader. Division of prize money among team members is at the team leader's discretion, and HHS will not arbitrate or resolve disputes among team members.
   3. For a winning entity, HHS will pay the prize to the entity.
2. Prizes will be paid by electronic funds transfer and may be subject to federal income tax. AHRQ will comply with Internal Revenue Service withholding and reporting requirements where applicable.
3. Participating entities are encouraged, but not required, to request and obtain a Unique Entity ID (UEI) through SAM.gov because doing so may expedite prize payment.
4. Prizes offered by AHRQ will be paid directly by AHRQ.

Other Rules

HHS reserves the right, in its sole discretion, to cancel, suspend, modify, or terminate the challenge or any phase; change the schedule or evaluation procedures; request clarification or additional demonstrations; select fewer finalists; or decline to award any prize if no submission is deemed worthy or if funds are unavailable. Any material modification will be published through the official challenge channel.

Return to Contents