De-identified wound care data is patient-level clinical information stripped of identifiers under HIPAA’s Safe Harbor or Expert Determination standards, which allows the data to move into competitive intelligence, PMCF submissions, and comparative effectiveness work without individual patient authorization. The catch is that “de-identified” is a specific legal status, not a vague assurance that a dataset is safe to cite. Wound care data in particular carries re-identification risks that claims data doesn’t, because longitudinal wound measurements, photographs, and low-volume care settings can act as quasi-identifiers even after the obvious fields are scrubbed. Before you build a payer dossier, a PMCF report, or a market-share claim on a third-party dataset, you need to know which de-identification standard was applied, when, and at what level of granularity. This matters as much for legal exposure as for data credibility with FDA reviewers, MACs, and IDN value analysis committees.
How Does HIPAA Actually Define De-Identified Data?
As of 2026, HIPAA recognizes exactly two methods for rendering protected health information de-identified, both codified at 45 CFR §164.514. The first, Safe Harbor, requires removal of 18 specific identifier categories: names, geographic subdivisions smaller than a state, all elements of dates directly related to an individual (except year), phone numbers, medical record numbers, biometric identifiers, full-face photographs, and so on. Once those 18 categories are gone and the covered entity has no actual knowledge that remaining information could identify someone, the data is de-identified under Safe Harbor.
The second method, Expert Determination, works differently. A qualified statistician applies accepted analytical methods to certify that the risk of re-identification is “very small,” documents that determination, and the dataset can retain more granular fields, such as specific dates of service, five-digit zip codes, or facility-level detail, than Safe Harbor would allow. Most commercial wound care datasets rely on Expert Determination rather than Safe Harbor, precisely because Safe Harbor’s blanket date and geography restrictions would gut the analytical value of the data. Tracking wound staging progression over time requires dates of service. Comparing CTP/CAMP utilization by region requires geographic granularity finer than “state.” Expert Determination lets a vendor retain that detail while still meeting the legal bar, provided the statistical work is done properly and documented.
It’s worth separating de-identified data from a limited dataset, a distinct category under HIPAA that retains dates and geographic detail but is not treated as de-identified. A limited dataset can only be shared under a signed Data Use Agreement restricting its use to research, public health, or healthcare operations, and it still carries some HIPAA obligations on the recipient’s end. Buyers sometimes conflate the two because both retain more clinical richness than Safe Harbor output. But a limited dataset requires a DUA and contractual restrictions on re-identification attempts; true de-identified data, once certified, does not. If a vendor hands you data under a DUA and calls it “de-identified,” ask which category it actually falls under, because the compliance obligations attached to each are not interchangeable.
Why Is Wound Care Data Harder to De-Identify Than Typical Claims Data?
Standard claims data de-identification assumes a large denominator: millions of encounters spread across thousands of providers, where any single record blends into the noise. Wound care data frequently doesn’t have that luxury. A dataset drawn from a rural HBOT program, a single SNF chain, or a small hospital-based wound center can have so few patients matching a given combination of age, wound etiology, and treatment history that even a fully Safe Harbor-scrubbed record becomes identifiable to anyone who knows the clinical context. This is the quasi-identifier problem: no single field is a direct identifier, but the combination narrows the population to one.
Longitudinal wound measurements compound the issue. A sequence of wound surface area readings, staging changes, and visit cadence over 12 or 20 weeks creates a pattern that is, in effect, a biometric signature for that patient’s care episode. Strip the name and medical record number, and the visit-by-visit trajectory can still be matched back to a specific individual if an adversary has access to even partial outside information, such as a local news item about a limb salvage case or a family member’s account of treatment timing. This is a materially different risk profile than a single claims line with a diagnosis code and a date of service.
Product-level detail adds another layer specific to this industry. CTP and CAMP utilization data, when combined with site-of-care and payer mix, can narrow a record to a single patient in low-volume settings, something high-volume pharmacy or claims data rarely faces because dispensing volumes are typically much higher. If a wound center used only one advanced skin substitute product in a given quarter and treated a handful of Medicare Advantage patients with pressure ulcers, that combination of product, payer, wound type, and site may be uniquely identifying even without a name attached.
Wound photography and staging notes require handling beyond the standard 18-identifier scrub entirely. Full-face images are already on the Safe Harbor list, but limb photography with visible tattoos, scars, or other unique markings falls into a gray area that requires deliberate redaction or exclusion policies, not just removal of EXIF metadata and facial features. Any vendor building a dataset that includes wound imagery should have an explicit protocol for this, and you should ask for it directly rather than assuming standard de-identification software covers it.
Where Do Manufacturers and Providers Actually Use De-Identified Wound Data?
The most consequential use case for manufacturers is post-market clinical follow-up. De-identified real-world use patterns can substitute for, or supplement, prospective registry data when submitting to FDA or to notified bodies under EU MDR-style frameworks. Instead of running a slow, expensive prospective PMCF study from scratch, a manufacturer can pull de-identified utilization and outcome trends across care settings to demonstrate continued safety and performance, provided the underlying data’s provenance and de-identification method can withstand regulatory scrutiny.
Comparative effectiveness and product ranking work is the second major application. Aggregated, de-identified utilization data across CTPs and CAMPs, segmented by wound type (DFU, VLU, pressure ulcer) and care setting (SNF, home visit, hospital-based wound center), feeds market access dossiers and payer negotiations. A manufacturer arguing for favorable coverage needs to show how its product performs in the real world relative to competitors, not just in a controlled trial population. This is where a platform like Intellicure Analytics’ Wound Care Industry Dashboard becomes relevant: it aggregates de-identified, real-world product utilization and healing pattern data across settings so manufacturers and researchers can benchmark performance without having to construct their own registry from scratch.
IDN value analysis committees and GPOs use the same category of data defensively. When a hospital system’s VAC is deciding whether to add or drop a CTP from formulary, de-identified utilization benchmarks let them compare their own network’s outcomes and cost patterns against national data rather than relying solely on manufacturer-supplied claims. That comparison only holds up if the benchmark dataset’s de-identification and sourcing methodology is transparent enough for the committee to trust the comparison is apples-to-apples. A benchmark built from a handful of unnamed academic centers behaves very differently from one built across diverse site-of-care types, and a VAC evaluating a claim should ask which it’s looking at.
What Mistakes Break De-Identification in Practice?
The most common failure is treating “no patient names” as equivalent to “de-identified.” Removing direct identifiers while leaving dates of service, five-digit zip codes, or rare combinations of age, diagnosis, and facility in place does not satisfy Safe Harbor, and it may not satisfy Expert Determination either if those fields weren’t specifically assessed for re-identification risk. Teams building internal analyses sometimes strip a spreadsheet of the obvious columns and assume they’re compliant, without recognizing that a 91-year-old patient with a stage 4 pressure ulcer treated at a 20-bed rural facility is identifiable by that combination alone, name or no name.
Data linkage attacks are a related and often overlooked risk. A dataset that is properly de-identified in isolation can become re-identifiable when combined with a separately purchased marketing list, a claims file, or even public facility utilization reports. This is exactly the kind of risk auditors and payers scrutinize during compliance reviews: not whether the original dataset was handled correctly, but whether the recipient’s downstream use, including merges with other purchased data, reconstructed identities the original vendor never intended to expose. Manufacturers running competitive intelligence work across multiple vendor datasets should have a documented policy on what merges are permitted and what re-identification risk assessment happens before combining sources.
A third mistake is treating an Expert Determination as a permanent, one-time certification. Re-identification risk is not static. As a dataset grows, ages, or gets merged with new sources, the statistical basis for the original “very small risk” finding can erode. A determination made on a 50,000-patient dataset in 2023 does not automatically extend to the same dataset three years and 300,000 patients later, particularly if new fields, such as more granular product codes or additional care settings, were added along the way. A qualified statistician should periodically re-assess the determination, and any vendor or internal team using aging data should be able to tell you when the last re-assessment happened, not just when the original one occurred.
How Should You Evaluate a De-Identified Wound Care Dataset Before Using It?
Before citing a third-party or vendor dataset in a PMCF report, payer dossier, or competitive claim, run it through a short diligence checklist rather than accepting “de-identified” as a self-certifying label.
- Ask for the specific method. Was the dataset de-identified under Safe Harbor or Expert Determination, and can the vendor produce documentation of that determination, not just a verbal assurance?
- Confirm the date and scope of the determination. Does it cover the current data vintage you’re being given, or only the original build from when the dataset was first assembled? A determination that’s several years old and hasn’t been revisited as the dataset expanded is a red flag.
- Check subgroup cell sizes. If you plan to report on a specific CTP within a single site-of-care type, or a rare wound etiology in a small regional footprint, ask what cell-size suppression rules apply at that level of granularity. A dataset that’s safely de-identified in aggregate can become re-identifying the moment you slice it down to a subgroup of a dozen patients.
- Verify chain of custody. Does the vendor source data directly from EHR systems or wound center partners, or is it reselling third-party claims data that’s already changed hands multiple times? Direct sourcing matters both for re-identification risk, since each additional hop increases linkage exposure, and for credibility with regulatory or payer audiences who will ask where the numbers originated.
- Get the statistician’s credentials, not just their sign-off. An Expert Determination is only as strong as the qualifications and methodology of the person making it. Vague references to “statistical review” without a named, qualified expert and a documented methodology should be treated as unverified.
None of this diligence is optional if the dataset is going to support a regulatory submission or a payer-facing claim. A dossier built on data whose de-identification pedigree can’t survive a challenge doesn’t just create legal exposure, it undermines the credibility of the entire analysis in front of the audience you’re trying to persuade.
Provenance Is the Real Test, Not the Label
De-identification is a compliance floor, not a data-quality guarantee. A dataset can meet the legal definition and still be too thin, too dated, or too opaque in its sourcing to support the claim you want to make. Before citing any dataset in a PMCF report, a payer dossier, or a competitive analysis, confirm the specific method used, the vintage of the determination, and the subgroup granularity at which the data remains safely de-identified. Intellicure Analytics documents this provenance for every dataset behind its Wound Care Industry Dashboard and Comparative Effectiveness Studies, so manufacturers, researchers, and IDN teams can trace exactly how the numbers were built before putting them in front of FDA, a payer, or a value analysis committee. Learn more about our services