Policy analysis · access & communication · evidence checked

Health literacy as health-system infrastructure

Health literacy is normally measured as something a patient has. Framed that way, the remedy is education. Framed as a system property, the remedy is design — and the obligation moves to whoever writes the denial letter, the consent form and the discharge instructions.

What the reframing changes

The four verbs in the Australian formulation — find, understand, evaluate, use — are not attributes of a reader in isolation. Each one is jointly produced by a person and a system. A patient cannot find what is not published, cannot understand what is written at a reading level above the population median, cannot evaluate a claim without a stated source, and cannot use an instruction that omits what to do when the instruction fails.

That is why the framing matters for policy rather than for patient education. If comprehension is a system output, it can be specified, measured and required. If it is a patient attribute, it can only be lamented.

Where it bites in the United States

Three surfaces do most of the damage, and none of them is a clinical encounter.

The denial letter. A coverage denial is a legal document that triggers a time-limited appeal right. Its comprehensibility determines whether the right is real. The KPSGILL model payer standard reaches duration, continuity, a named accountable clinician and automation disclosure; it does not currently reach readability, and the Australian framework is a reasonable prompt to ask whether it should.

The consent form. Consent documents have lengthened in response to liability rather than comprehension, and length is negatively correlated with being read. A signature on an unread document satisfies a record requirement and not an ethical one.

Machine-generated patient communication. AI-drafted patient messages are now produced at volume. They inherit the fluency of the model, which is not the same as the comprehensibility of the message, and they are not currently tested for either. This is the point at which health literacy stops being a communications topic and becomes part of the AI governance readiness question.

Six questions, none of them yet a KPSGILL position

  • Should health literacy be reported as a quality metric, and measured at the system or the encounter level?
  • Should informed-consent design incorporate comprehension testing rather than readability scoring?
  • Should payer communications carry a readability requirement enforceable by the regulator that already reviews them?
  • Should an adverse determination carry a plain-language statement of what was denied, why, and what to do next?
  • Should AI-generated patient communications be literacy-tested before deployment, and by whom?
  • Should communication competence be part of professional licensure, and if so, assessed how?

Why these remain questions. Readability standards have a documented failure mode: they are satisfiable by mechanical simplification that leaves comprehension unchanged while generating compliance paperwork. Before KPSGILL recommends any of the above, the evidence that a stated standard changes what patients actually understand — rather than what documents score — needs assessment. That work is not published on this site, so no position is taken.

Sources