Applying fair-treatment and non-discrimination rules to patient marketing

Nobody at Airbnb ever wrote a rule about race. A 2016 field experiment by Edelman, Luca and Svirsky sent otherwise identical booking requests under names signalling different racial backgrounds, and requests from guests with distinctively African American names were accepted roughly 16% less often. The bias sat in thousands of separate host decisions and in a product that showed a face before it showed a booking. Airbnb's answer was structural rather than rhetorical: a non-discrimination policy, more Instant Book inventory, and by 2020 Project Lighthouse, built with Color of Change, to measure a gap the company could not otherwise see.

Hospital marketing fails the same way. No one types "exclude the uninsured" into an ad platform. Exclusion arrives through a seed list, a ZIP radius, a cost-per-acquisition target and an English-only landing pagelanding pageA standalone web page built for a single campaign goal, designed to maximise conversions by removing distractions and focusing visitors on one action.View full definition →. The other lessons in this module ask whether what you say is true and provable. This one asks who hears it, who can act on it, and what your targeting quietly infers about someone's body.

The three rules that touch your marketing

EMTALA: you cannot market your way out of an obligation

EMTALA (the Emergency Medical Treatment and Labor Act, 1986) requires any hospital with an emergency department that accepts Medicare to screen and stabilize anyone who arrives with an emergency, regardless of ability to pay or insurance status. Your messaging cannot contradict that obligation.

Concrete traps:

• An ad implying "insured patients only" or "we serve members" at an ED entrance.
• Signage or web copy suggesting payment is required before emergency screening.
• Outreach that quietly steers uninsured patients away from the ED ("consider urgent care instead if you cannot pay") in a way that reads as deterrence.

The edge case teams miss: EMTALA attaches to the whole campus, so a waiting-room digital screen, a third-party registration kiosk script and the "financial clearance" line in a Google Business profile are all marketing artifacts a surveyor can read. The CMS EMTALA overview is the authoritative free reference.

ACA Section 1557: non-discrimination in how you promote and communicate

Section 1557 of the Affordable Care Act (2010) prohibits discrimination on race, color, national origin, sex, age and disability in health programs receiving federal funds.

Language access. Roughly 25 million US residents, about 8% of the population aged five and over, have limited English proficiency. National-origin protection means they must be able to reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → and use your communications. The 2024 final rule requires a notice of availability of language assistance in English plus the 15 most common languages spoken in the state, a requirement the 2020 rule had stripped out and litigation has repeatedly tested. Check current HHS Office for Civil Rights guidance before you set a translation budget. England codifies the same idea more bluntly: NHS England's Accessible Information Standard, mandatory since August 2016, requires providers to identify, record, flag and meet each patient's communication needs, and it bites hardest on the unglamorous surfaces (appointment letters, reminders, results) that marketing rarely audits.

Disability access. About one in four US adults lives with a disability. WCAG 2.1 level AA is the practical benchmark, and the Justice Department's 2024 Title II web rule adopted exactly that standard for state and local government entities, which includes public hospital districts. An orthopedic microsite a screen reader cannot parse is a marketing failure before it is a legal one.

Sex and the targeting question. The scope of 1557's sex provisions, including gender identity, has moved through successive rulemakings and court challenges. Treat the boundary as unsettled and verify rather than assume.

The subtler exposure is targeting itself. Meta's special ad category restrictions, which strip fine-grained targeting for housing, employment, credit and social issues, do not cover healthcare. The platform will let you build the audience. You are the only guardrailguardrailRules and controls that keep an AI system inside safe, legal and on-brand boundaries, blocking outputs and actions that cross the line.View full definition →.

Financial-assistance transparency: charity care is a marketing obligation

Non-profit hospitals must comply with Internal Revenue Code Section 501(r) to keep tax-exempt status. Among other things, 501(r) requires them to maintain a written Financial Assistance Policy (FAP), widely publicize it, limit charges for FAP-eligible patients, and make reasonable efforts to determine eligibility before extraordinary collection actions.

"Widely publicize" is a marketing verb. The FAP and a plain-language summary belong on the website, in the languages of the significant populations served, and visible in the building. Revocation of exemption is rare; what actually happens is Schedule H scrutiny, a state attorney general asking why eligible patients were never told, and a reporter putting your premium-suite campaign next to your collections docket. The IRS 501(r) requirements page is the primary free source.

Where these rules reshape real campaigns

Service-line promotions

Service lines are the most heavily marketed and highest margin, so fair-treatment risk concentrates there. The mechanism is usually a lookalike audiencelookalike audienceAn audience created by ad platforms to target new prospects who resemble your best existing customers, based on shared traits and behaviors.View full definition →. Seed the model with last year's converted joint-replacement patients and it will faithfully reproduce the payer mix, income band and language of that panel, then spend your budget finding more of the same. Nobody excluded anyone. The optimisation did it.

The second-order version is worse: if you judge the campaign on cost per booked consult, the algorithm learns that low-converting neighborhoods waste money and drifts away from them week after week. Your reporting will call this efficiency.

Ask before launch:

1. What seeded the audience, and what does that seed's demographic profile look like against your service area?
2. Does the creative imply access is conditional on insurance or payment?
3. Can LEP and disabled patients receive and act on this message?

Inferred conditions and what the envelope shows

In 2017 Aetna mailed roughly 12,000 members about HIV medication in envelopes whose windows exposed enough of the letter to reveal the subject. The inferenceinferenceThe moment a trained AI model is put to work: it takes a new input and produces an answer, prediction or generated output.View full definition → was correct, and that was the harm. The class settlement reached about $17 million in 2018, with further state settlements on top.

Digital retargetingretargetingShowing ads to users who have previously visited your site or interacted with your brand, to bring them back and drive conversion.View full definition → is the same envelope. A pixel fired on a bariatric surgery page, an addiction services page or a fertility page will follow that person onto a shared family tablet and a work laptop. The HIPAA and TCPA constraints the sibling lesson sets out are the legal floor here; the fair-treatment layer sits above it. For sensitive service lines, drop behavioral retargeting entirely and buy contextual placement instead. You lose some conversion efficiency and you stop broadcasting a diagnosis.

Charity-care messaging

Treat the FAP like a product you actively market. A "Financial assistance" page linked from the main navigation and from every billing communication, a plain-language summary at roughly a middle-school reading level, translated versions matching the LEP populations you serve, and tone consistency: the page welcoming uninsured patients should not be contradicted by the collections letter.

Targeted outreach to protected populations

Outreach *toward* underserved groups is encouraged and supports community-benefit obligations. NHS England's Core20PLUS5 approach, which names the most deprived 20% of the population plus locally identified groups as the focus for a short list of clinical priorities, is a workable template: narrow targeting with a stated clinical rationale is defensible, and a prostate screening push aimed at men over 50 or sickle-cell outreach in affected communities is not a 1557 problem.

The failure mode is omission. In 2018 NHS England disclosed that an IT fault in the breast screening programme meant large numbers of women in the 68 to 71 age band never received a final invitation; the independent review later revised the affected numbers and disputed the early mortality estimates, but the shape of the harm was fixed by then. No campaign was cancelled. An invitation list simply stopped including people, and nobody was measuring who was missing.

🎬 [VIDEO: "Section 1557 Nondiscrimination Basics" - youtube.com - a short plain-language explainer on the ACA's core non-discrimination provisions for health programs]

A fair-treatment check before spend

The sequencing, sign-off authority and escalation path belong to the pre-launch review lesson. What follows is the equity slice that review should contain:

• [ ] Audience provenance: what seeded the targeting, and how the resulting profile compares with the service area census.
• [ ] Exclusion log: any excluded geography, age band or device tier, with a written rationale.
• [ ] EMTALA tone check: no payment or insurance gating in any emergency-related message, on any campus surface.
• [ ] Language access: notice of availability and translations for the populations you actually serve.
• [ ] Accessibility: landing pages and video tested against WCAG 2.1 AA, not assumed.
• [ ] Sensitive-line handling: retargeting suppressed where the ad would disclose an inferred condition.
• [ ] FAP visibility: charity-care information at least as findable as the promoted service.
• [ ] Claims carry the proof file the substantiation lesson specifies.

Keep an auditable record. One line per campaign is enough to show intent:

campaign: ortho_q1_2026
audience_included: metro_service_area_all_zips
audience_excluded: none
lookalike_seed: none (prior-patient seed rejected: reproduces payer mix)
language_notice: es, zh
accessibility_check: pass wcag21aa (2026-01-14)
reviewer: compliance_jdoe

"We deliberately did not exclude" is a strong position. Silence is not.

Enforcement is real, and it is often a marketing artifact

Regulators and plaintiffs rarely start with your intentions. They start with your *artifacts*: the billboard, the ad-platform audience settings, the buried FAP link, the English-only microsite. Marketing produces most of the evidence in a fair-treatment dispute.

Two channels to know:

• HHS Office for Civil Rights: investigates Section 1557 complaints, with corrective action plans and public settlements as typical outcomes.
• IRS and state attorneys general: scrutinize 501(r) compliance, charity-care publicity and collections conduct.

Reputational cost usually exceeds the legal cost, and it lands on whoever wrote the premium-suite ad, not on whoever authorised the lawsuits against patients.

A note on Europe: there is no EMTALA or 501(r) equivalent, but national medical-advertising rules and GDPRGDPREU regulation governing how organizations collect, store and use personal data, with fines tied to global revenue for breaches.View full definition → both apply, and GDPR treats health data as a special category. Targeting European patients on inferred health status needs explicit consent, which makes the Aetna-style inference problem a consent problem rather than only a dignity problem. The mechanics differ; the principle rhymes.

Key Takeaways

• The audience seed is the discrimination decision. Lookalike models trained on last year's converted patients reproduce last year's payer mix and language. Record what seeded every campaign.
• Optimising on cost per conversion drifts away from poorer neighborhoods without anyone choosing that. Watch the week-over-week geographic distribution of spend, not just the CPACPACost Per Acquisition: the total cost to generate one customer or conversion, computed by dividing total spend by the number of acquisitions.View full definition →.
• Correct inference is the harm. Aetna's HIV mailing cost around $17 million because the envelope was right about the recipient. Sensitive service lines should not use behavioral retargeting.
• "Widely publicize" the FAP is a marketing mandate. Charity-care information must be at least as findable as your highest-margin service line, in the languages you serve.
• Access has a testable standard. WCAG 2.1 AA and a real language-availability notice, verified against current HHS OCR guidance, since 1557's scope keeps moving.
• Omission counts. The people your campaign never reached leave no trace unless you measure for them, which is the whole point of Airbnb building Project Lighthouse.