Your health system just spent nine months negotiating a business associate agreement with a new CDP vendor. Legal signed off. The CISO ticked the security review. The marketing team launched a campaign the same week the contract went live. You assume you are covered. You are not. The BAA is a contract about one vendor. It does nothing about the seven other systems that touch the same patient data before it reaches that vendor, or the ad platform you sync audiences to on Friday afternoon.
The BAA is a contract, not a compliance program
A BAA defines what the vendor may do with PHI and binds it to specific Security Rule obligations: permitted disclosures, safeguards, breach reporting, subcontractor flow-down. It covers the entity that signed it. It does not extend to Meta, Google Ads, The Trade Desk, your SMS gateway, your agency partner, or the analytics tag somebody added last quarter. It does not authorize anything HIPAA does not already allow. Retargeting and lookalike modeling usually need the ad platform to receive PHI directly, and almost none will sign a BAA for that. No paperwork retroactively legalizes a disclosure that was impermissible from the start.
Worse, vendors often limit BAAs to specific products, configurations, or fields. Some prohibit the exact identifiers your team plans to use. Some exclude AI add-ons or ad integrations from scope. A BAA used outside its scope is paperwork, not protection. Your compliance officer knows this. Your marketing operations lead probably does not.
Where the real risk actually lives
The website is half the exposure. Health systems increasingly pipe EHR data into CDPs and journey tools: appointment history, service lines, discharge dates, sometimes diagnosis codes. That data is PHI the moment it lands. The platform vendor is a business associate and its BAA has to cover the modules and features you actually use, not just the ones the salesperson mentioned.
Then watch the exits. An audience synced from a CDP to an ad platform is a patient list handed to a company that will not sign a BAA. A journey triggered by a diagnosis can reveal it in a subject line or a text preview. The HIPAA minimum necessary rule applies: send each platform the fields the use case needs, not the whole feed.
The risk you do not know about is bigger than the risk you do. A 2024 federal court ruling vacated the part of federal tracking guidance that covered IP addresses on public pages about conditions or providers. The rest of the rule stands. Patient portals, apps, and pages where people book care or enter symptoms remain in scope. Class actions under state wiretap and medical confidentiality laws are the bigger financial exposure, and class periods reach to when the tracking started. A tag installed in 2022 is still a liability in 2026.
What marketers actually need to do
Map the stack in both directions. On the collection side, inventory every tag and SDK on every web and app property, including the ones nobody remembers adding, and note what each sends from which pages. On the data side, trace every feed out of the EHR: which fields go into which platforms and where those platforms send data next. Ad audiences, enrichment vendors, SMS gateways, and agency exports all count.
For each destination, check the BAA against its actual scope, then check whether the use itself is permitted under HIPAA. Server-side tagging changes where data is collected, not what gets sent onward. Forwarding PHI to a vendor without a BAA through a different pipe is the same disclosure. A vendor promise to strip PHI after receipt does not cure it. Federal guidance says so directly.
Pull legal and compliance in early, not at final sign-off. A written, good-faith interpretation holds up far better than an unwritten assumption. Marketers waiting for one universal compliant answer will keep waiting. The answer is organizational, and somebody has to decide it on purpose.
Sources
- Why HIPAA compliance in martech goes beyond BAAs -- MarTech, October 9 2026