The Challenge of Healthcare Asset Data
Healthcare asset registers are built over decades. A hospital that opened in 1990 has equipment records created by dozens of different biomedical engineers, facilities managers, and procurement teams — each with their own conventions for recording manufacturer names, model numbers, and locations.
The result is a dataset where "Philips Healthcare", "Philips Medical Systems", "PHILIPS", and "Philips" all appear as separate manufacturers. Where "GE CARESCAPE B650" and "B650 Monitor" and "Carescape B650" all refer to the same device. Where "Ward 4B" is the same location as "Cardiology Ward" — because the ward was renamed in 2018 and nobody updated the older records.
This is not a failure of process. It is the natural result of a dataset that grows organically over time, without a shared vocabulary or a normalisation layer to enforce consistency.
What Healthcare Asset Normalisation Covers
A complete normalisation pass for a healthcare asset register addresses six dimensions of data quality.
Manufacturer name standardisation resolves the dozens of variants under which a single manufacturer appears. "Philips Healthcare", "Philips Medical Systems", "PHILIPS", and "Philips" all map to the canonical form "Philips Healthcare". This matters because CMMS platforms like Nuvolo, Accruent, and IBM Maximo match assets against their manufacturer libraries, and a single-character difference creates a duplicate.
Model name normalisation expands abbreviations and shorthand to canonical model names. "B650" becomes "CARESCAPE B650". "S5" becomes "CARESCAPE Patient Monitor S5". This is where device type classification and FDA UDI cross-referencing are most valuable — both databases contain the authoritative model name for every registered medical device.
device type classification assigns every asset a Global Medical Device Nomenclature code and definition. Device type classification is the healthcare equivalent of a DCIM device type — it provides a universal vocabulary that every CMMS, every biomedical engineering team, and every regulatory body understands. A ventilator is not a "breathing machine" or a "vent" or a "resp support unit" — it is device type 34938 "Ventilator, intensive care".
FDA UDI cross-referencing uses the device's UDI barcode to retrieve the authoritative manufacturer name, device description, and model number from the FDA's Global Unique Device Identification Database (GUDID). For medical devices sold in the US market, GUDID is the single most reliable source of ground truth for equipment identification.
Location normalisation converts free-text location strings into structured fields: site, building, floor, department, room, and bay. This must account for ward renaming (a common source of phantom duplicates), building reconfigurations, and the difference between physical location and cost centre assignment. A piece of equipment may be physically in "Ward 4B" but charged to "Cardiology" — both fields need to be captured correctly.
Quality scoring assigns each record a confidence score based on field completeness and source reliability. Healthcare quality scoring weights clinical criticality: a ventilator with a missing serial number is a higher-priority gap than a desk fan with the same issue. The scoring model reflects the regulatory and patient safety implications of incomplete biomedical equipment records.
Why It Matters Before CMMS Import
Every major healthcare CMMS — Nuvolo, Accruent, IBM Maximo, ServiceMax — has an import process that looks straightforward and is not. The import will reject rows where the manufacturer name does not match a known vendor. It will create duplicate device types if the same model appears under two different names. It will fail on location strings that do not correspond to existing site and room objects.
Teams that skip normalisation spend weeks in post-import cleanup. Teams that normalise first — using device type classification, UDI cross-referencing, and structured location parsing — typically complete a clean import in a single session.
The downstream benefits extend beyond the import. A normalised asset register enables accurate maintenance scheduling (you cannot schedule preventive maintenance on a device type you cannot identify), reliable compliance reporting (HTM 01-05, MHRA, CQC inspections all require accurate equipment records), and meaningful benchmarking against peer institutions.
The Struktive Healthcare Workflow
Struktive processes healthcare asset registers through four stages: ingest, classify, enrich, and export.
Ingest accepts the register in any format — Excel, CSV, or direct export from an existing CMMS. The ingest stage identifies the relevant columns (manufacturer, model, serial, location, asset ID) regardless of column naming conventions.
Classify assigns each record a device type code and definition, using a combination of manufacturer-model matching, description parsing, and UDI lookup where barcodes are present in the data.
Enrich cross-references each record against GUDID to confirm or correct manufacturer and model fields, and parses location strings into structured site-building-floor-department-room fields.
Export produces a CMMS-ready dataset with a quality score for each record, a gap report highlighting records that require manual review, and a summary of the normalisation changes applied.
The output is a dataset that imports cleanly into Nuvolo, Accruent, Maximo, or any other CMMS — without the weeks of post-import cleanup that typically follow an unnormalised import.