Framework documentation

The RIQA Assurance Framework

A four-phase, modular methodology for independent analytical verification and Independent Result Reconstruction (IRR) across biomedical, clinical, and healthcare analytics domains. Every engagement is pre-specified, version-tracked, and independently reproducible and analytically verifiable under defined conditions.

Four-phase pipeline

How every engagement proceeds

The same four phases apply across all three domains. Domain-specific verification procedures are applied within each phase.

Phase 01

Data Provenance & Transformation Review

Evaluates the lineage and transformation pathway from source data to reported outputs. Covers normalization procedures, batch correction documentation, gating strategy definitions, endpoint derivation files, censoring rule implementations, SAP alignment, ETL logic, look-back window specifications, and crosswalk file versioning.

→ provenance-trace.pdf · transformation-log.json

Phase 02

Statistical & Methodological Assessment

Evaluates alignment between the declared analytical methodology and the structure of the underlying data. Each analytical component is assigned to an IRR class within the RIQA taxonomy. An IRR Methodology Declaration is issued specifying the verification standard to be applied in Phase 03.

→ irr-declaration.json · methodology-assessment.pdf

Phase 03

Independent Result Reconstruction (IRR)

Results are reconstructed from source data and documented analytical procedures. For exact and near-deterministic methods, numerical reconstruction is performed. A sensitivity analysis is separately performed to evaluate conclusion stability under reasonable alternate assumptions.

→ irr-findings.csv · sensitivity-analysis.json

Phase 04

Structured QA Reporting

Generates structured QA reports including findings registers, integrity scoring summaries, and machine-readable provenance artifacts. SHA-256 hashes of all input files are recorded in the audit trail. Every finding traces to a specific catalog entry.

→ riqa-assurance-report.pdf · audit-trail.json · findings-register.csv

Provenance verification flow

End-to-end analytical pipeline

From submission intake to structured QA report — every step is documented, version-controlled, and reproducible.

Severity framework

Finding classification

All findings are classified using a four-tier severity framework. Deductions are subtractive from a 100-point base per dimension.

Material

−25 pts

Reported endpoint does not reproduce from submitted data. Revision required.

Moderate

−9 pts

Methodological concern not changing direction of effect. Disclosure recommended.

Minor

−4 pts

Reporting or documentation gap with no effect on the result.

Informational

0 pts

Best-practice recommendation. No defect identified.

Integrity scoring

Four dimensions, one weighted score

Every engagement produces an overall integrity score and four per-dimension scores on a 0–100 scale.

30%

Independent Result Reconstruction (IRR)

Whether RIQA could reproduce the reported quantitative claims from the raw data.

25%

Statistical methodology

Soundness of test choice, test scale, and multiple-comparisons handling.

25%

Conclusion-to-result alignment

Whether the manuscript's conclusions match what the data support.

20%

Data provenance & transformation

Completeness and traceability of inputs, reagents, and normalization.

Score range	Classification	Interpretation
95–100	Verified	Reproducibility fully demonstrated. Informational notes only.
85–94	Verified with notes	Reproducibility holds; minor or moderate items warrant attention.
70–84	Methodological concerns	Direction of conclusions holds, but specific issues should be addressed.
< 70	Material reproducibility concerns	One or more results cannot be independently reproduced; revision required.

Reconstruction taxonomy

How RIQA classifies analytical methods

Before reconstruction begins, each analytical component is assigned to a class that defines the applicable verification standard.

Class	Method examples	Verification standard
Exact	t-test, chi-square, ANOVA, Kaplan-Meier, ΔΔCt reconstruction	Full numerical agreement within rounding tolerance.
Near-deterministic	Cox PH, logistic regression, ANCOVA, log-rank	Point estimates within defined tolerance; significance and direction must match.
Software-tolerant	MMRM, mixed models, GEE, GLMM	Direction, significance, and order verified; numerical differences documented.
Structural verification	Multiple imputation, Bayesian MCMC, adaptive designs	Correct methodological implementation verified; seed and version documentation required.
Architectural verification	ML pipelines, CMS-HCC models, federated systems	Logic consistency, population construction, and specification-to-implementation alignment.

Anchor standards

Standards alignment

RIQA's provenance requirements are anchored to established community standards. RIQA goes beyond checklist compliance to provide independent quantitative reconstruction.

MIQE 2.0 — Minimum Information for qPCR Experiments

Bustin et al. 2025, Clin Chem 71:634. RIQA-qPCR Livak v1.1 provenance catalog is structured along MIQE 2.0 sections with severity assignments consistent with the essential/desirable distinction.

MIFlowCyt — Minimum Information about a Flow Cytometry Experiment

Lee et al. 2008, Cytometry A 73A:926. RIQA-Flow Herzenberg v1.1 provenance layer is anchored to MIFlowCyt, covering panel, gating tree, compensation, FMO controls, and viability declarations.

ICH E9(R1) — Estimands and Sensitivity Analysis

The RIQA sensitivity analysis framework for clinical trials aligns with ICH E9(R1) estimand principles, evaluating conclusion stability under alternate analytical assumptions.

FAIR Principles — Findable, Accessible, Interoperable, Reusable

RIQA QA outputs include machine-readable JSON artifacts structured for downstream integration, consistent with FAIR data principles for research provenance infrastructure.

Research-use and scope of findings

RIQA provides analytical assurance for research and quality-review purposes. RIQA does not function as a regulatory authority, certifying body, or legal compliance organization. Findings are reproducibility and methodological consistency statements derived from submitted data and declared methodology — not determinations of scientific truth, biological validity, or regulatory compliance.

Download the RIQA White Paper

Full framework documentation · RIQA-WP-001 · 2026 · Open Access CC BY-NC 4.0

View demonstration studies →