Interoperability and shared standards across agencies

A caseworker in a county child welfare office flags a 9-year-old for chronic absenteeism risk. The school district's attendance system has the data. The child welfare agency has the case file. Both records describe the same child. Neither system knows it.

Why? The school district identifies the student by a state education ID. The welfare agency identifies him by a case management number tied to a different name spelling his guardian used at intake. No shared key exists to link the two records. Multiply this by every school, hospital, court, and housing authority a vulnerable family touches, and you get the defining data problem of the public sector: rich data trapped in incompatible silos.

This lesson covers the identifier and schemaschemaA schema is the formal blueprint that defines how data is structured, named, typed, and related within a database, file, or message.View full definition → standards that solve this, the governance metrics that track interoperability health, and the benchmarks used to judge whether a data-sharing effort is actually working.

Why silos exist by design, not accident

Government systems were built agency by agency, often decades apart, funded by different grant streams with different rules attached.

• FERPA (Family Educational Rights and Privacy Act) restricts school record sharing in the US.
• HIPAA (Health Insurance Portability and Accountability Act) governs health data.
• 42 CFR Part 2 adds extra restrictions on substance use treatment records.

Each law was written to protect one type of sensitive data, without a common technical or legal bridge between systems. The result: legally-mandated silos that predate any conversation about data interoperability.

This matters for fluency. When you hear "our systems can't talk to each other" in a public agency, it's rarely pure incompetence. It's usually a mix of legal constraint, legacy IT, and absent standards.

The core building block: identifiers

You cannot join two datasets about the same person, place, or case without a shared identifier or a reliable matching method.

Deterministic matching joins records on an exact shared key (a Social Security Number, a state student ID, a National Provider Identifier for clinicians). Fast and precise, but only works if the key is populated consistently and correctly, which in practice it often isn't.

Probabilistic matching (also called fuzzy matching or record linkage) scores similarity across multiple fields (name, date of birth, address, phone) when no shared key exists. It produces a match confidence score rather than a certainty. This is the workhorse technique for connecting welfare, health, and education records across agencies that never agreed on a common ID.

A simplified illustration of how a match score gets built:

match_score = (
    0.4 * name_similarity      # e.g., Jaro-Winkler string distance
  + 0.3 * dob_exact_match      # 1 if identical, 0 if not
  + 0.2 * address_similarity
  + 0.1 * phone_match
)
# Threshold: match_score > 0.85 = auto-link
# 0.6-0.85 = flag for human review
# < 0.6 = no link

Agencies like the Actionable Intelligence for Social Policy (AISP) network at the University of Pennsylvania publish open toolkits on building these cross-agency "integrated data systems" responsibly.

Schema standards: speaking the same data language

Even with a matched identifier, two systems need to describe the *content* the same way. This is what schema and interchange standards do.

Key standards professionals should recognize:

• HL7 FHIR (Fast Healthcare Interoperability Resources): the dominant standard for exchanging health data between hospitals, insurers, and public health agencies. Mandated in US federal rules from the Office of the National Coordinator for Health IT (ONC).
• NIEM (National Information Exchange Model): a US federal data-sharing schema used across justice, emergency management, and human services agencies so a "person" or "incident" record means the same thing everywhere it's used.
• Ed-Fi: a data standard widely adopted by US state education agencies to standardize student, course, and attendance records.
• X-Road: Estonia's data exchange layer, arguably the world's most cited example of national interoperability infrastructure, connecting nearly all government e-services since the early 2000s. Referenced by the European Commission's interoperability resources.
• SDMX (Statistical Data and Metadata eXchange): used by Eurostat, the OECD, and national statistics offices to exchange comparable statistical series across countries.

The pattern across all of these: a shared schema turns a data *transfer* into a data *integration*. Without it, agencies exchange files that require manual re-mapping every time, which is expensive, slow, and error-prone.

Governance metrics that track interoperability health

Once a data-sharing pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.View full definition → exists, you need metrics to know if it's actually working. Core ones used in public sector data governancedata governanceData governance is the set of policies, roles, and processes that ensure data is accurate, secure, well-defined, and used responsibly across an organization.View full definition →:

Under the EU's GDPRGDPREU regulation governing how organizations collect, store and use personal data, with fines tied to global revenue for breaches.View full definition → (General Data Protection Regulation), cross-agency sharing must document a lawful basis (Article 6) and, for special category data like health or welfare status, an additional condition under Article 9. Governance dashboards in EU public bodies typically track "legal basis documented" as a compliance metric alongside technical match rates.

A worked example: if a county integrated data system attempts to link 10,000 child welfare case records to a state education database and successfully matches 7,400, the match rate is 74%. If a later audit finds 220 of those matches were incorrect (wrong child), the false positive rate is roughly 3% (220/7,400). Both numbers should be reported together. A high match rate with a hidden high error rate is worse than a modest match rate you can trust.

Benchmarks: what "good" looks like

There's no single global benchmark for interoperability, but reference points help calibrate expectations.

• Estonia's X-Road model is frequently cited (OECD, European Commission) as enabling most government-to-government data checks to happen automatically, without a citizen resubmitting the same information twice, an approach called the "once-only principle."
• In the US, the AISP Integrated Data Systems network reports that mature county-level integrated systems (linking welfare, justice, education, and health data) commonly report probabilistic match rates in the 70 to 90% range, though exact figures vary widely by county and data qualitydata qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.View full definition → (this is a general estimate, not a fixed standard).
• The EU's Interoperable Europe Act (entered into force 2024) sets a policy framework requiring EU public administrations to assess interoperability impact before deploying new systems, an important regulatory signal that this is shifting from best practice to obligation.

When evaluating an agency's interoperability claim, ask three fluency questions:

1. What identifier or matching method underlies the join?
2. What's the documented match rate and error rate, not just "we share data now"?
3. What is the legal basis, and who audits it?

Key Takeaways

• Cross-agency data joins fail primarily due to two gaps: no shared identifier and no shared schema, layered on top of legal restrictions like FERPA, HIPAA, and 42 CFR Part 2 in the US, or GDPR's lawful-basis requirements in the EU.
• Probabilistic record linkage (fuzzy matching on name, DOB, address) is the practical workaround when agencies lack a common ID; it produces a confidence score, not a certainty, and needs human review at the margins.
• Schema standards (HL7 FHIR for health, NIEM for justice/human services, Ed-Fi for education, X-Road and SDMX internationally) let systems agree on what a data field *means*, not just how to transmit it.
• Track interoperability with paired metrics: match rate alongside false positive rate, plus data latency and consent coverage. A high match rate without an error rate reported alongside it should be treated as an incomplete claim.
• Regulatory momentum (the EU's Interoperable Europe Act, US federal FHIR mandates via ONC) is pushing interoperability from voluntary best practice toward compliance obligation, so this is increasingly an audit and governance topic, not just an IT one.