Health literacy is one of my favorite ideas in all of health research. And almost every time I see someone try to measure it, I want to throw my laptop out the window.
These two feelings are not in conflict. Well, they are, but they come from the same place. The concept of health literacy promises something essential: that people’s capacity to make sense of health, illness and care matters, and that it can grow. The scales used to quantify the concept however, in my view, completely break that promise. They take a rich, relational, empowering idea and degrade it into a test of whether or not patients can follow instructions. Let me try to explain…
The good
In 2012, Kristine Sørensen and colleagues in the European Health Literacy Project reviewed 17 definitions and 12 conceptual models of health literacy (Link to article). They condensed them into one integrated definition:
Health literacy is linked to literacy and entails people’s knowledge, motivation and competences to access, understand, appraise, and apply health information in order to make judgments and take decisions in everyday life concerning healthcare, disease prevention and health promotion to maintain or improve quality of life during the life course.
I LOVE that definition. It is not about reading a leaflet. It is about judgments and decisions in everyday life, not just in the clinic. It spans the life course, and the authors stress that these skills develop over time and are linked to lifelong learning. And it contains no less than four verbs: access, understand, appraise, apply. Appraise means interpreting, filtering, judging and evaluating, which sometimes means concluding that the information does not fit you.
The review calls Don Nutbeam’s model (Link to article) the prototypical one for the broader, public health view. Nutbeam distinguishes functional, interactive and critical health literacy. Functional is the basics: reading and writing well enough to manage everyday situations. Interactive is extracting meaning from communication and applying it to changing circumstances. Critical is analyzing information and using it to exert greater control over life events and situations. Each level supports more autonomy and personal empowerment than the one before.
That third level is what makes me love the concept. Kickbusch and colleagues (Link to article) call health literacy a critical empowerment strategy to increase people’s control over their health. Zarcadoolas, Pleasant and Greer describe a health literate person as someone who can apply health concepts to new situations and take part in ongoing dialogues about health, medicine and scientific knowledge. That treats people as agents who evaluate, challenge and contribute to knowledge, not just receive it.
And the concept was never meant to sit only inside individuals. The review quotes the Institute of Medicine: health literacy emerges from the interaction between the skills of individuals and the demands of social systems. Sørensen and colleagues conclude that low health literacy can be addressed in two ways: by helping people become more resourceful, or by improving the “readability of the system”. I love that phrase. A clinic with confusing letters, fragmented portals and no time for questions is not neutral. It reinforces the obstacles and gate-keeping that makes being a patient so much harder.
The bad
Now look at what we actually measure. I am not alone in being unhappy here. Sørensen and colleagues wrote, already in 2012, that the tools available then did not capture all aspects of the concept, and that very few conceptual models had been empirically validated at all.
The oldest instruments are easy targets. The REALM scores how well people pronounce medical words read aloud. The TOFHLA uses fill-in-the-blank hospital texts. The Newest Vital Sign asks people to do sums with an ice cream label. They test reading and numeracy, and few would claim they capture the full concept. I am more interested in the instruments that were built to do better.
The HLS-EU questionnaire
The HLS-EU questionnaire was built directly on the Sørensen model, and it has become one of the most widely used health literacy instruments in Europe. Its full version (47 items) cover the twelve cells of the matrix: access, understand, appraise and apply, across healthcare, disease prevention and health promotion (Link). Each item asks the same question: on a scale from very easy to very difficult, how easy would you say it is to… These are some of the healthcare items:
- …understand what your doctor says to you?
- …understand your doctor’s or pharmacist’s instruction on how to take a prescribed medicine?
- …judge how information from your doctor applies to you?
- …judge when you may need to get a second opinion from another doctor?
- …use information the doctor gives you to make decisions about your illness?
- …follow instructions from your doctor or pharmacist?
(There is also a short version (16 items), HLS-EU-Q16 (Link).)
The healthcare “apply” dimension is spelled out as the ability to make informed decisions on medical issues and comply with medical advice.
If you count the sources of knowledge among the 47 items, in all of them knowledge comes from somewhere else: the doctor, the pharmacist, the media, family and friends. The patient’s own observations are never a source.
To be fair, the items in domain appraisal are realistic and good. Judging how information from your doctor applies to you is exactly the kind of thing I do every day. But look at where doubt is allowed to go. If you question your doctor, the questionnaire sends you to a second opinion from another doctor. There is no item asking how easy it is to tell your doctor something they didn’t know. None asks when your own experience should override the standard advice. And because the score is built on ease, finding it easy to follow instructions counts as more literate.
eHEALS
The eHealth Literacy Scale, eHEALS, was published by Cameron Norman and Harvey Skinner in 2006 (source). It has eight items, and none of them mentions doctors. That looks more promising. But every item is about health resources on the internet (source): knowing what resources exist and where to find them, evaluating them, telling high quality from low quality resources, and feeling confident using information from the internet to make health decisions.
Knowledge is still something out there, produced by others, that the person retrieves and grades. And “high quality” usually means quality by professional standards. Someone who uses the internet to compare experiences with others who have the same condition, or to log and analyse their own symptoms, is practising eHealth literacy in its richest form. eHEALS has no item for that. And don’t even get me started on the glaring absence of AI literacy… (more on that below)
The ugly
The trouble is not that these scales are badly built. Many are carefully validated. The trouble is what they assume.
So do these instruments assume that the physician always knows more than the patient, and is always right? Not quite, and the difference matters. Neither says the doctor is always right; the HLS-EU explicitly asks about second opinions. The assumption runs deeper: knowledge always flows towards the patient. The doctor, the pharmacist, the media and the internet are sources. The patient is a receiver, sometimes a critical one.
It is not that the patient is assumed to know less. It is that the patient knowing something the physician doesn’t is not a possibility the instruments can register at all. Where doubt is allowed, it is routed back into the professional system.
The Sørensen review itself names the pattern. The “medical” models of health literacy, it notes, all concern a person’s capacity to function in the role of a patient within the healthcare system. One model lists compliance with medication as a dimension of health literacy. The American Medical Association defines it as the skills needed to function in the healthcare environment.
They locate the problem in the individual. A low score becomes a property of the patient, “limited health literacy”, rather than a sign that the information, the encounter or the system failed to meet them. The readability of the system, which the review names as half the solution, is nowhere in the number.
They mistake critical literacy for non-compliance. Let me give an example from my own life with Parkinson’s. I was taking an extended-release medication, prescribed once a day in the morning. I noticed the effect was strong in the morning and wore off later in the day. What if I took lower doses, spread across the day? At my next appointment, my neurologist said that according to the manufacturer, once a day really should be enough. But if I wanted to, I could try.
I did. After the usual adjustment period, the effect was markedly more even across the whole day. I went back and told him. Later I heard from other patients who see him that he had started prescribing the medication the same way to them.
That is access, understand, appraise and apply in their fullest sense. I questioned the manufacturer’s information and the default instructions, ran a careful experiment on myself, and acted on the result. Yet on an item about how easy it is to follow my doctor’s instructions, or a definition that includes complying with medical advice, it looks like deviation. The most health literate thing I did would hardly register.
And look at what happened next: the knowledge travelled from me to my clinician, and on to patients I have never met. In the integrated model’s figure, health information flows into the person. There is no arrow for knowledge flowing out. No scale I know of measures it either.
They make experiential knowledge invisible. The review does mention experience, but only as an antecedent: prior experience with illness and the healthcare system is something that feeds health literacy. It is never part of what health literacy is. Someone who has lived with a long-term condition for decades may know things about their body, their treatment and their daily life that no clinician knows. You can be an expert in your own condition and still score as “inadequate” because a word list tripped you up.
They are context-free. Paasche-Orlow and Wolf, one of the 17 definitions in the review, insist that health literacy must always be examined in the context of the specific tasks to be accomplished. Yet most instruments produce a single, general score. The same person can be highly literate about their own condition and lost in an unfamiliar specialty, or confident at home and silenced in a rushed consultation. One number flattens all of that.
Put together, the measures turn an emancipatory concept into a screening tool for whom we need to simplify things for. That is useful for some purposes. But it is not health literacy as the concept describes it.
There is also a risk in what happens next. Goodhart’s law (Link), in its best-known paraphrase, says that when a measure becomes a target, it ceases to be a good measure. Health literacy scores are increasingly used to monitor populations and evaluate interventions. If a health system sets out to raise its HLS-EU scores, the quickest route is to make patients find it easier to understand and follow what professionals tell them. The scores would rise. The critical health literacy the concept was meant to foster could quietly fall. We would be optimising for good receivers.
What measuring health literacy well might look like
I don’t think the answer is to stop measuring. Sørensen and colleagues argued for a concept validation approach: build the scales from the concept, and let the scales test the concept in return. I agree. I just want us to take the concept at its word. A few starting points:
- Measure both sides of the encounter. If health literacy is relational, then the clarity of the system, the time given and the willingness of professionals to listen belong in the measure too.
- Count appraisal and critique as competence. Asking a good question, spotting a weak claim, or adapting advice to your life should raise a score, not lower it.
- Make room for experiential knowledge. Ask what people know about their own condition and how they use it, not only what they can decode from others.
- Accept that it is situated. Measure in context, for specific conditions and settings, rather than chasing one universal number.
- Design the measures with patients. The people being measured are well placed to say what being health literate actually requires in their lives.
The concept keeps moving even while the measures stand still. In a December 2025 commentary for the National Academy of Medicine, Hugo Campos and Liz Salmi propose critical AI health literacy (source). They define it as the ability to use AI tools strategically to analyse what shapes health, including power structures, to evaluate AI outputs for bias and institutional alignment, and to act on individual and collective health goals. They build on Abel and Benkert’s definition of critical health literacy as reflection turned into action, and they are clear that writing good prompts is only the bare minimum. What strikes me most is one of their case examples. Liz Salmi learned through an online patient community that the guidelines for her condition had changed, used AI to condense the evidence into a one-page summary, and sent it to her doctor, who read it and changed her treatment. Knowledge flowed from patient to clinician, just as it did in my extended-release story. For Campos and Salmi, the real promise of AI lies not in supplying answers but in sharpening patients’ ability to question them. Try capturing that with an eHEALS item about telling high-quality from low-quality health resources on the internet.
Closing
What we measure ends up defining what we mean. As long as our instruments tells healthcare and researchers that a health literate patient is a patient who reads well and follows instructions, “health literacy” will come to mean exactly that, regardless of what our definitions say.
I want to keep loving this concept. For that, I need the measures to catch up with it. The measures need to see people as knowers in our own right, not just as compliant and slightly critical readers. Healthcare also needs to be held accountable for the readability and usability of its information, websites, and patient portals. Both need to recognize that the patient who helped improve a prescribing practice was not a non-compliant patient. Instead, they were demonstrating health literacy in its truest form.
Until then, I’ll keep cheering for the idea and refuse to accept that a low score is the patient’s fault. The responsibility belongs to the healthcare systems that write the letters, build the portals, and to the researchers who design the questionnaires.

