Cross-Agency Entity Resolution Across Federal and State Filings

Valuable regulatory and operational questions almost always span at least two filing regimes. A healthcare analyst wants to compare a hospital's federal Medicare cost report against its audited state financial disclosure and clinical quality scores. A logistics underwriter wants to link an interstate carrier's USDOT census record to roadside inspection citations and hazardous materials incidents. A retirement plan researcher needs to follow an employer plan across annual Form 5500 filings and pension insurance records even when the sponsor restructures.

Each agency assigns identifiers around its own statutory mandate rather than a shared national entity graph:

• Acute-care hospitals (hospitals.meruai.co): CMS tracks Medicare-certified providers by six-character CMS Certification Number (CCN) and organizational NPIs, whereas California's Department of Health Care Access and Information (HCAI) keys facility financial and utilization reports to state facility IDs with different campus-consolidation boundaries and fiscal-year cutoffs.
• Skilled nursing facilities (helpelders.meruai.co): CMS Payroll-Based Journal staffing telemetry, Care Compare quality ratings, and state health department citation rosters differ on how ownership changes, DBA renamings, and license transfers split historical continuity.
• Retirement plans & depositories (erisa.meruai.co, banks.meruai.co): Department of Labor Form 5500 filings key plans by sponsor EIN plus three-digit Plan Number, while banking regulators separate Federal Reserve RSSD IDs, FDIC certificate numbers, and routing identifiers across holding companies and charters.

Joining these tables on normalized legal names or street strings produces silent false merges whenever multi-campus systems share corporate headquarters addresses or regional subsidiaries share similar trade names.

Linking federal CMS Healthcare Cost Report Information System (HCRIS) filings to California HCAI annual financial disclosures illustrates why identifier matching is only half of entity resolution. Even after a hospital's six-character CCN is matched to its California HCAI facility ID, three structural differences can make naive side-by-side ratios misleading:

1. Consolidated license versus single-campus reporting: A health system may file one consolidated Medicare cost report under a primary CCN covering two acute campuses while submitting separate campus-level disclosures to the state, or vice versa. Our crosswalk tags cardinality (1:1, 1:N, N:1) before computing per-bed or margin comparisons.
2. Fiscal period alignment: Hospitals report on their own fiscal year ends (June 30, September 30, December 31), and ownership transfers create stub periods shorter than 365 days. Cross-agency comparisons must align overlapping fiscal intervals and annualize or flag short-period filings explicitly.
3. Accounting basis and line-item definitions: Medicare allowable cost schedules serve reimbursement settlement, whereas state financial disclosures track GAAP-aligned operating revenue, charity care, and payer-mix categories. Both series are valuable precisely because they measure different statutory concepts, provided the interface names each schedule rather than blending them into an unlabeled number.

We apply the exact same structural checks inside private enterprise engagements when integrating CRM, ERP, billing, and case-management systems that define customers, matters, or locations at different levels of granularity.

A current owner is credited or blamed for inspections recorded under a prior operator.

The entity table stores only the latest ownership attribute and joins five years of historical events without effective-date bounds.

What to do: Model ownership, name, and parent-system affiliations as temporal intervals (valid_from, valid_to) so historical events remain attached to the operator of record on the event date.

System-wide financial totals double-count multi-campus hospitals or bank branches.

A 1:N crosswalk join replicates the parent consolidated filing onto every child location row before aggregation.

What to do: Enforce cardinality-aware joins: aggregate only at the grain of the reporting filing, and display parent-consolidated context on child profiles with a clear consolidation badge.

Address standardization merges two distinct suites, clinics, or carrier terminals on the same industrial parcel.

Geocoding collapses suite numbers or co-located licensees onto one rooftop coordinate and uses location as a primary merge key.

What to do: Treat physical proximity as a candidate-generation signal only, never sufficient to merge two distinct statutory licenses without secondary key agreement.