Data Stewardship vs Data Governance: What's the Difference?

Data stewardship and data governance are distinct but complementary functions in research data management. Data governance establishes the policies, standards, and decision rights that determine how data should be handled across an organization or research project. Data stewardship executes those policies through day-to-day management activities such as metadata creation, quality checking, and dataset curation. For bioinformatics researchers, analysts, and life-science professionals, understanding this distinction matters because it clarifies who is accountable for data-related decisions and who is responsible for implementing them. This article defines both concepts, compares their scope and activities, and provides a practical framework for assigning responsibilities in a research project.

Defining Data Governance in Research Contexts

Data governance refers to the framework of policies, standards, roles, and decision rights that guide how data is collected, stored, shared, and used. In a research setting, governance answers questions about what data can be collected, who can access it, how long it must be retained, and what ethical or legal constraints apply. Governance operates at the level of rules and accountability instead of hands-on data handling.

The Genomic Data Sharing Policy from the National Institutes of Health illustrates governance in action for genomic research. This policy establishes expectations for data deposition, access controls, and responsible use of human genomic data. Researchers who receive NIH funding must comply with these governance requirements, which shape how their data management plans are written and executed. The policy functions as a governance instrument because it sets binding rules that individual projects must follow.

Governance structures can exist at multiple levels simultaneously. An individual research project may have its own governance plan that specifies data sharing timelines and access restrictions. An institution may impose additional governance requirements related to data security or intellectual property. National and international bodies may govern data sharing across borders or within specific research communities. The FAIR Guiding Principles published in Scientific Data represent a governance framework adopted by many research funders and repositories. These principles specify that data should be findable, accessible, interoperable, and reusable, providing a common standard against which data management practices can be evaluated.

Governance also encompasses the processes for making decisions when conflicts arise. For example, a governance framework might specify who has authority to approve data sharing requests, how access decisions are documented, and what appeals process exists for denied requests. Without such decision rights, data access becomes ad hoc and inconsistent.

Defining Data Stewardship in Research Contexts

Data stewardship encompasses the operational activities required to manage data according to established policies and standards. Stewards perform the hands-on work of organizing files, creating metadata, validating data quality, preparing datasets for deposition, and maintaining data over time. The term derives from the concept of caring for resources on behalf of others, reflecting the responsibility researchers have to manage data responsibly for the scientific community.

The EMBL-EBI Training program at the European Bioinformatics Institute provides resources that support data stewardship skill development. Training in areas such as data organization, metadata standards, and submission procedures equips researchers and support staff with the practical competencies needed to manage biological data effectively. The existence of such training programs reflects the recognition that stewardship requires specialized knowledge beyond general research skills.

Data stewardship activities vary depending on the data type and research context. For genomic sequence data, stewardship might involve ensuring that files are formatted according to repository requirements, that quality metrics are recorded, and that submission packages include all required metadata. For clinical datasets, stewardship might involve de-identification, version control, and documentation of data transformations. For imaging data, stewardship might involve file format conversion, compression decisions, and annotation of acquisition parameters.

The NCBI Data Resources at the National Center for Biotechnology Information provide infrastructure that supports data stewardship. Repositories such as GenBank, the Sequence Read Archive, and the Database of Genotypes and Phenotypes accept submissions from researchers and impose specific formatting and metadata requirements. Stewards must understand these requirements to deposit data successfully. The repositories themselves also perform stewardship functions by maintaining data integrity, providing persistent identifiers, and ensuring long-term accessibility.

Core Differences Between Stewardship and Governance

The primary distinction between data stewardship and data governance lies in their respective functions. Governance establishes the rules and accountability structures. Stewardship implements those rules through concrete actions. Governance asks what should be done and who should decide. Stewardship asks how the work will be done and who will do it.

Accountability differs between the two roles. Governance assigns decision rights to specific individuals or bodies, such as a data access committee or an institutional review board. These decision-makers are accountable for the policies they create and the approvals they grant. Stewards are accountable for executing tasks correctly and reporting problems that require governance-level decisions. A steward who discovers a data quality issue may be responsible for documenting the issue and escalating it, but the decision about whether to correct the data or exclude it from analysis typically rests with governance authorities.

The temporal scope also differs. Governance tends to operate on longer cycles, with policies reviewed and updated periodically. Stewardship operates continuously as data is created, processed, and shared. A governance framework might be reviewed annually, while stewardship activities occur daily or weekly throughout the research lifecycle.

The Toward Trustworthy Autonomous Data Ecosystems Synthesizing Governance Requirements for Machine to Machine Communication review highlights that data stewardship and governance capabilities are complementary in complex data environments. The review identifies data stewardship, provenance management, explainability, auditability, semantic interoperability, accountability, and lifecycle governance as capabilities that must work together. This finding applies to research data management as well, where governance provides the framework and stewardship provides the execution.

How Stewardship and Governance Complement Each Other

Effective research data management requires both governance and stewardship functioning in concert. Governance without stewardship results in policies that are never implemented. Stewardship without governance results in inconsistent practices that vary by individual preference and cannot be audited.

A practical example illustrates this complementarity. A research consortium studying antimicrobial resistance might adopt a governance framework specifying that all whole-genome sequencing data must be deposited in a public repository within six months of generation. The governance framework also specifies which metadata fields are mandatory and what quality thresholds must be met. Stewards within each participating laboratory then execute these requirements by preparing submission packages, validating sequence quality, and coordinating with repository staff. The governance framework ensures consistency across the consortium, while stewardship ensures that the work actually gets done.

The Physical Activity and Healthy Ageing and Longevity Data Interoperability Reuse and Innovation in a Federated European Context analysis demonstrates how governance and stewardship gaps manifest in practice. The analysis identifies systematic gaps across cataloguing, metadata, semantics, provenance, access governance, and computability layers. These gaps arise when governance frameworks do not specify adequate standards and when stewardship practices do not implement the standards that exist. The analysis emphasizes that governance metadata is rarely machine-actionable, meaning that even when policies exist, they cannot be automatically enforced or discovered.

In bioinformatics, the complementarity is particularly evident in data submission workflows. Governance determines which repositories are acceptable for different data types, what access levels are permitted, and what ethical approvals are required. Stewardship determines how the data is formatted, annotated, and validated before submission. Both functions must operate correctly for data to be successfully shared and reused.

At a Glance: Stewardship vs Governance Comparison

Framework for Assigning Responsibilities in a Research Project

Assigning responsibilities between governance and stewardship requires deliberate planning at the project outset. The following framework provides a structured approach for research teams.

Step 1: Inventory Data Assets and Identify Governance Requirements

Begin by cataloguing the types of data the project will generate or use. For a bioinformatics project, this might include raw sequencing reads, processed alignment files, variant calls, phenotypic metadata, and analysis scripts. For each data type, identify applicable governance requirements from funders, institutions, and regulatory bodies. The Genomic Data Sharing Policy provides an example of funder-level requirements that would apply to NIH-funded genomic research.

Document the following for each data type: collection or generation method, expected volume, sensitivity level, retention requirements, sharing obligations, and applicable standards. This inventory becomes the foundation for both governance planning and stewardship task assignment.

Step 2: Define Governance Roles and Decision Rights

Establish who has authority to make decisions about data policies and exceptions. Common governance roles in research projects include a principal investigator who holds ultimate accountability, a data access committee that reviews sharing requests, and an institutional official who ensures compliance with organizational policies.

Create a decision rights matrix that specifies who can approve data sharing, who can authorize deviations from standard protocols, and who resolves disputes about data interpretation or quality. Document this matrix so that all team members understand the governance structure.

Step 3: Assign Stewardship Responsibilities

Identify individuals who will perform day-to-day data management tasks. In many research groups, this includes bioinformaticians who process and analyze data, laboratory managers who oversee sample tracking, and dedicated data managers who handle submission and curation.

For each stewardship role, define specific responsibilities and performance expectations. A data manager might be responsible for ensuring that all datasets have complete metadata before submission. A bioinformatician might be responsible for documenting analysis pipelines and versioning code. Clear role definitions prevent gaps where tasks fall between responsibilities.

Step 4: Establish Communication and Escalation Pathways

Define how stewards report problems that require governance decisions. For example, if a steward discovers that a dataset contains potentially identifiable information that was not disclosed in the original ethics approval, the steward needs a clear pathway to escalate this issue to the appropriate governance authority.

Document escalation criteria that specify what types of issues require governance involvement. These might include unexpected data quality problems, requests for data use that fall outside approved purposes, or discoveries that affect data sharing obligations.

Step 5: Document and Review the Framework

Record the governance and stewardship assignments in the project data management plan. Review the framework periodically to ensure it remains appropriate as the project evolves. New data types, new collaborators, or new regulatory requirements may necessitate adjustments to either governance structures or stewardship responsibilities.

Practical Workflow for Implementing Both Functions

A practical workflow integrates governance and stewardship activities throughout the research data lifecycle. The following workflow describes how these functions operate at each stage.

Planning and Design Phase

During project planning, governance activities include developing the data management plan, obtaining ethics approvals, and establishing data sharing agreements with collaborators. Stewardship activities include designing file naming conventions, creating metadata templates, and setting up version control systems.

The data management plan serves as the primary governance document that links policies to implementation. It should specify what data will be collected, how it will be documented, where it will be stored, who can access it, and how it will be shared. The plan also identifies which standards will be followed, such as the FAIR Guiding Principles.

Data Collection and Generation Phase

During data collection, governance requirements continue to apply through protocols that specify approved methods and documentation standards. Stewardship activities include recording provenance information, capturing metadata at the point of collection, and performing initial quality checks.

For sequencing projects, stewardship includes recording instrument settings, sample preparation details, and quality metrics such as read depth and error rates. This information is essential for later interpretation and for meeting repository submission requirements.

Data Processing and Analysis Phase

Governance during analysis includes policies about which software versions are approved, how analysis results are validated, and what documentation is required for reproducibility. Stewardship includes implementing these policies by maintaining analysis environments, documenting parameters, and recording output versions.

The Clinical Outcomes and Bacterial Characteristics of Carbapenem-resistant Acinetobacter baumannii Among Patients From Different Global Regions study demonstrates the importance of standardized data collection and analysis in multicenter research. The study enrolled patients across five global regions and required consistent data abstraction and whole-genome analysis protocols. Governance ensured that all sites followed the same procedures, while stewardship at each site ensured that data was collected accurately and completely.

Data Sharing and Publication Phase

Governance determines when data can be shared, with whom, and under what conditions. This includes compliance with funder policies such as the Genomic Data Sharing Policy and with repository-specific requirements. Stewardship includes preparing data packages, validating submissions, and responding to repository feedback.

The Global epidemiology and clinical outcomes of carbapenem-resistant Pseudomonas aeruginosa and associated carbapenemases POP a prospective cohort study illustrates the value of data sharing in infectious disease research. The study involved 44 hospitals across 10 countries and required coordinated data collection and analysis. Effective governance ensured that data sharing agreements were in place, while stewardship ensured that data from each site met quality standards.

Long-term Preservation and Reuse Phase

Governance specifies retention periods, preservation standards, and conditions for future data use. Stewardship includes maintaining data in accessible formats, refreshing storage media, and updating metadata as needed.

The NCBI Data Resources provide long-term preservation infrastructure for biological data. Researchers who deposit data in NCBI repositories rely on the organization to maintain data integrity and accessibility over time. The repositories also provide persistent identifiers that enable citation and tracking of data reuse.

Records and Measurements for Stewardship and Governance

Tracking the effectiveness of both governance and stewardship requires systematic record-keeping. The following measurements provide evidence that data management functions are operating correctly.

Governance Records

Governance records document decisions, policies, and compliance activities. These include meeting minutes from data access committees, records of data sharing requests and approvals, policy review schedules, and compliance audit results. Maintaining these records provides evidence that governance processes are functioning and enables review of past decisions.

Stewardship Records

Stewardship records document the operational activities performed on data. These include metadata logs, quality check results, version histories, submission records, and issue tracking reports. These records provide evidence that data has been managed according to established standards and enable troubleshooting when problems arise.

Quality Metrics

Quality metrics provide quantitative evidence of data management effectiveness. For genomic data, metrics might include the percentage of samples meeting coverage thresholds, the completeness of metadata records, and the time from data generation to repository submission. For clinical data, metrics might include the rate of data entry errors, the timeliness of data updates, and the completeness of follow-up records.

The Antimicrobial susceptibility assays for Neisseria gonorrhoeae a proof-of-principle population-based retrospective analysis demonstrates the importance of data quality in research that informs clinical decisions. The study analyzed susceptibility data for over 23,000 isolates collected over 12 years. The reliability of the findings depended on consistent data collection and quality control throughout the surveillance program.

Common Failure Patterns in Data Management

Understanding common failure patterns helps research teams identify problems early and implement corrective actions. The following patterns frequently emerge when governance and stewardship functions are not properly aligned.

Policy Implementation Gaps

A common failure occurs when governance policies exist but are not implemented in practice. This can happen when policies are written without input from those who will implement them, when resources are insufficient for implementation, or when compliance is not monitored. The Assessing the policy and institutional framework for multisectoral governance and accountability in emergency preparedness and response in Ethiopia study found that national guidelines were widely recognized but implementation varied across administrative levels. Persistent challenges included role ambiguity, inconsistent coordination, and weak feedback mechanisms. These findings apply to research data management as well, where policies may be well-documented but unevenly applied.

Role Confusion

Another common failure occurs when team members do not understand the distinction between governance and stewardship responsibilities. A researcher might make unilateral decisions about data sharing without consulting governance authorities, or a governance body might become involved in operational details that should be handled by stewards. Clear role definitions and documentation help prevent this confusion.

Metadata Neglect

Insufficient metadata is a persistent problem in research data management. When stewards do not capture adequate metadata at the time of data collection, the data becomes difficult to discover, interpret, and reuse. The Physical Activity and Healthy Ageing and Longevity Data Interoperability Reuse and Innovation in a Federated European Context analysis found that metadata inconsistency undermines comparability and semantic fragmentation prevents federated analytics. These problems arise when governance does not specify adequate metadata standards and stewardship does not implement them consistently.

Quality Check Gaps

Data quality problems often emerge when quality checks are not performed at appropriate points in the data lifecycle. Quality checks should occur at data collection, after processing, before analysis, and before submission. Missing checks at any of these points can allow errors to propagate through the research workflow.

Escalation Failures

Stewards sometimes encounter problems that require governance decisions but fail to escalate them appropriately. This can happen when escalation pathways are unclear, when stewards fear negative consequences for reporting problems, or when governance authorities are unresponsive. Clear escalation criteria and a supportive culture help address this failure pattern.

Limitations and Boundaries of Each Function

Both governance and stewardship have inherent limitations that research teams should recognize.

Governance Limitations

Governance cannot anticipate every situation that will arise during a research project. Policies must be interpreted and applied to specific circumstances, which requires judgment. Governance also depends on accurate information from those implementing policies. If stewards do not report problems or provide complete information, governance decisions will be based on incomplete data.

Governance processes can also become bureaucratic and slow, particularly when multiple approval layers are required. This can delay research activities and create frustration among researchers. Effective governance balances the need for oversight with the need for research efficiency.

Stewardship Limitations

Stewardship depends on the skills and knowledge of the individuals performing the work. Inadequate training can lead to inconsistent metadata, poor quality checks, and submission errors. The Providing Research Data Management RDM Services in Libraries Preparedness Roles Challenges and Training for RDM Practice survey found that librarians recognized the importance of research data management but expressed concerns about lack of bandwidth and capacity. The study concluded that institutional commitment to resources and training is crucial for growing data management services. This finding applies to research settings generally, where stewardship capacity is often limited.

Stewardship also depends on the quality of the tools and infrastructure available. Poorly designed data management systems, inadequate storage, and incompatible software can undermine even the most diligent stewardship efforts.

Safety and Regulatory Context for Research Data

Research data management operates within a complex regulatory environment that varies by data type, research domain, and jurisdiction. Understanding this context is essential for both governance and stewardship functions.

Human Subjects Data

Research involving human subjects data is subject to ethical and legal requirements that govern consent, privacy, and data sharing. Governance must ensure that data collection and sharing comply with these requirements. Stewardship must implement technical controls such as de-identification, access restrictions, and secure storage.

The Genomic Data Sharing Policy includes specific provisions for human genomic data, including requirements for informed consent, data use limitations, and oversight of data access. Researchers working with human genomic data must understand these requirements and ensure that both governance and stewardship functions support compliance.

Pathogen and Antimicrobial Resistance Data

Research on pathogens and antimicrobial resistance raises additional governance considerations related to biosecurity and public health. Data on antimicrobial resistance patterns can inform clinical decisions and public health policy, making timely and accurate data sharing particularly important.

The Antibiotic Prescribing Patterns for Urinary Tract Infections and Pneumonia by Prescriber Type and Specialty in Nursing Home Care 2016-2018 study demonstrates how surveillance data can inform antimicrobial stewardship efforts. The study identified differences in prescribing patterns by prescriber type and nursing home specialization, information that can guide targeted interventions. Research that generates such data must ensure that governance frameworks support appropriate data sharing while protecting any sensitive information.

Data Security Requirements

Research data may be subject to security requirements imposed by institutions, funders, or regulations. Governance must specify security standards and access controls. Stewardship must implement these controls through practices such as encryption, access logging, and secure data transfer.

The Post-Quantum Cryptography Migration Readiness in Global Capability Centres A Governance Framework for Enterprise Transition analysis highlights that governance readiness, instead of technical capability, often constitutes the primary barrier to implementing security measures. The study found that algorithm selection is a governance decision with measurable operational consequences. This principle applies to research data security, where governance must make informed decisions about security standards and stewards must implement them effectively.

Professional Escalation Criteria

Research teams should establish clear criteria for escalating data management issues to governance authorities. The following situations warrant escalation.

Data Quality Issues Affecting Interpretation

When data quality problems could affect research conclusions, stewards should escalate the issue to governance authorities. This includes unexpected error rates, missing data that cannot be recovered, or inconsistencies between data sources. Governance must decide whether the data can be used, whether additional collection is needed, or whether the analysis plan must be modified.

Unauthorized Data Access or Use

Any suspected unauthorized access to or use of research data requires immediate escalation. This includes access by individuals without approval, use of data beyond approved purposes, or security breaches. Governance authorities must determine the appropriate response, which may include notification of institutional officials, affected participants, or regulatory bodies.

Conflicts Between Policies

When applicable policies conflict, stewards should escalate the issue instead of making unilateral decisions. For example, a funder policy might require open data sharing while an institutional policy restricts sharing of certain data types. Governance authorities must resolve such conflicts and document their decisions.

Requests for Data Use Outside Approved Purposes

When researchers or external parties request data use that falls outside the originally approved purposes, the request must be escalated to governance authorities. This ensures that data use remains consistent with consent requirements and ethical approvals.

Discovery of Previously Unknown Sensitive Information

If stewards discover that data contains sensitive information that was not previously identified, such as potentially identifiable details or unexpected genetic findings, this must be escalated immediately. Governance authorities must determine how to handle the information and whether notifications are required.

Frequently Asked Questions

What is the main difference between data stewardship and data governance?

Data governance establishes the policies, standards, and decision rights that determine how data should be managed. Data stewardship implements those policies through day-to-day activities such as metadata creation, quality checking, and repository submission. Governance answers what rules apply and who decides, while stewardship answers how the work gets done and who performs it.

Who should be responsible for data governance in a research project?

The principal investigator typically holds ultimate accountability for data governance, but governance responsibilities are often shared across multiple roles. A data access committee may review sharing requests, an institutional official may ensure compliance with organizational policies, and funders may impose governance requirements. The specific structure depends on the project scope, data sensitivity, and applicable regulations.

What skills does a data steward need in bioinformatics?

A data steward in bioinformatics needs practical skills in data organization, metadata standards, quality assessment, and repository submission. Familiarity with domain-specific formats and standards is essential. Training resources such as those provided by EMBL-EBI Training can help develop these skills. Stewards also need communication skills to document issues and escalate problems to governance authorities.

How do the FAIR principles relate to governance and stewardship?

The FAIR Guiding Principles provide a governance framework that specifies desired characteristics for research data: findable, accessible, interoperable, and reusable. Governance adopts these principles as standards, while stewardship implements them through specific practices such as assigning persistent identifiers, creating rich metadata, and using standard data formats.

Can one person perform both governance and stewardship functions?

In small research projects, one person may perform both functions, but this creates risks. The separation of duties allows governance decisions to be reviewed independently from implementation. When the same person both sets policies and implements them, there is less opportunity for oversight and fewer checks on errors or biases. Even in small projects, it is advisable to document decisions and seek external review where possible.

What records should a research project maintain for data governance?

Governance records should document policies, decisions, and compliance activities. This includes data management plans, ethics approvals, data sharing agreements, access committee decisions, and audit results. These records provide evidence that governance processes are functioning and enable review of past decisions.

How should data quality issues be escalated in a research project?

Data quality issues should be escalated when they could affect research conclusions, when they involve unauthorized access or use, or when they require decisions beyond the steward's authority. Escalation pathways should be documented in the data management plan. The appropriate governance authority, such as the principal investigator or a data access committee, should decide how to address the issue.

What are common signs that data governance is not working effectively?

Signs of ineffective governance include policies that are not implemented in practice, inconsistent decisions about data access, unclear accountability for data-related problems, and frequent conflicts that require ad hoc resolution. The Assessing the policy and institutional framework for multisectoral governance and accountability in emergency preparedness and response in Ethiopia study identified role ambiguity, inconsistent coordination, and weak feedback mechanisms as common governance challenges. These signs indicate that governance structures need review and revision.

Related Bioinformatics Guides

• Data Sharing and Privacy in Genomic Research
• Predicting AMR from Genomic Data
• Docker and Containerization in Reproducible Research
• The 1000 Genomes Project: Computational Insights
• The Human Genome Project: Computational Triumphs

References and Further Reading

• EMBL-EBI Training. European Bioinformatics Institute.
• NCBI Data Resources. National Center for Biotechnology Information.
• Genomic Data Sharing Policy. National Institutes of Health.
• The FAIR Guiding Principles. Scientific Data.
• Global epidemiology and clinical outcomes of carbapenem-resistant Pseudomonas aeruginosa and associated carbapenemases (POP): a prospective cohort study.. The Lancet. Microbe, 2023.
• Clinical Outcomes and Bacterial Characteristics of Carbapenem-resistant Acinetobacter baumannii Among Patients From Different Global Regions.. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America, 2024.
• Identifying heterogeneity of treatment effect for antibiotic duration in bloodstream infection: an exploratory post-hoc analysis of the BALANCE randomised clinical trial.. EClinicalMedicine, 2025.
• Prevalence of non-communicable diseases among household contacts of people with tuberculosis: A systematic review and individual participant data meta-analysis.. Tropical medicine & international health : TM & IH, 2024.
• Antimicrobial susceptibility assays for Neisseria gonorrhoeae: a proof-of-principle population-based retrospective analysis.. The Lancet. Microbe, 2023.
• Antibiotic Prescribing Patterns for Urinary Tract Infections and Pneumonia by Prescriber Type and Specialty in Nursing Home Care, 2016-2018.. Journal of the American Medical Directors Association, 2024.
• The budget impact of procalcitonin-guided antibiotic stewardship compared to standard of care for patients with suspected sepsis admitted to the intensive care unit in Belgium.. PloS one, 2023.
• Inappropriate Pediatric Orthopaedic Emergency Department Transfers: A Burden on the Health Care System.. Journal of pediatric orthopedics, 2024.
• Toward Trustworthy Autonomous Data Ecosystems Synthesizing Governance Requirements for Machine to Machine Communication. 2026.
• Assessing the policy and institutional framework for multisectoral governance and accountability in emergency preparedness and response in Ethiopia.. 2026.
• Mapping National Governance of AI for Health: Protocol for a Global Scoping Review.. 2026.
• Physical Activity and Healthy Ageing & Longevity Data: Interoperability, Reuse, and Innovation in a Federated European Context. 2026.
• Post-Quantum Cryptography Migration Readiness in Global Capability Centres: A Governance Framework for Enterprise Transition. 2026.
• Data Management Roles for Librarians. 2016.
• Academic Library and Research Data management Roles: The Case of Norwegian Libraries. 2015.
• Master Data Management Roles - Their Part in Data Quality Implementation. MIT International Conference on Information Quality, 2005.
• Providing Research Data Management (RDM) Services in Libraries: Preparedness, Roles, Challenges, and Training for RDM Practice. Data and Information Management, 2019.
• Transdisciplinary Citizen Science Connects Caribbean Hope Spots of Colombia to Improve Coral Reefs Governance. Palgrave Studies in Democracy Innovation and Entrepreneurship for Growth, 2021.
• Advancing citizen-centered public services in Kazakhstan: legal, institutional, and digital governance perspectives. Frontiers in Political Science, 2025.

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.