Family medicine · Aug 17, 2026

Health beliefs, why knowing what to do is not enough

Research on health beliefs is 70 years old and has an uncomfortable conclusion for anyone whose work is to inform. Here is what it shows about what moves someone to care for themselves, and about the limits of fear and of explanation.

“I know I should.” It is the sentence I hear most often in my office, and it comes from very different people. From a man who has carried the unopened box of his blood pressure medication in his backpack for three weeks. From a woman who has booked and cancelled her mammogram twice. From a nineteen-year-old who knows the name of what he feels, knows where to look for help, and keeps putting it off. None of them lacks information. They all know exactly what medicine recommends, and even so nothing moves.

For much of my training I treated that distance as a failure of communication. If the person did not do it, it was because I had not explained it well. I spent years explaining better, and it helped very little. What changed my consultations was not learning to explain, it was finding out that an entire field of research is devoted to that distance, and that it began with a question much like my own.

The research began by trying to understand who was not getting the X-ray

Edward Green, Elaine Murphy and Kristina Gryboski write, in a 2021 chapter of the Wiley Encyclopedia of Health Psychology, that the model appeared in the 1950s among behavioral scientists working for the United States Public Health Service. Their problem was concrete. Chest X-ray screening for tuberculosis was available and underused, because many people who were ill did not recognize their own symptoms and did not seek care for what they took to be nothing more than a cough.

Irwin Rosenstock, at the University of Michigan, reviewed that group's first studies in 1974. The clearest of them is Godfrey Hochbaum's, who from 1952 onward interviewed more than 1,200 adults in three cities to understand what made someone go for the test. Among people who held two beliefs at once, that they could have tuberculosis without feeling anything and that finding it early would bring some benefit, 82% had had at least one voluntary chest X-ray in the period studied. Among those who held neither belief, 21%.

What a person asks themselves before deciding

The Health Belief Model organizes that decision into questions people ask themselves without noticing they are asking. Could this happen to me? If it does, how much would it disrupt my life? Does what they are proposing actually work? And what will it cost me, in money, in time, in travel, in embarrassment, in fear of what the test might show? Over time the model gained two more pieces. One is the cue, whatever makes the decision leave the page on a particular day, a campaign on television, an abnormal result, illness in someone close. The other is the person's confidence in being able to do it, self-efficacy, added to the model in 1988.

None of those questions is about information. All of them are about belief, and that is why explanation on its own rarely suffices.

When they went to measure it, two of the four beliefs did not hold up

Christopher Carpenter, at Western Illinois University, gathered in 2010 the studies that measured those beliefs first and behavior later, rather than both in the same interview, precisely so as not to confuse the belief with the justification we give after we have already acted. Eighteen studies remained, with 2,702 people in all. The result takes apart the most used piece of the model. Perceived severity correlated with behavior at 0.15. Perceived risk came in at 0.05, close to zero. The two that held the model up were perceived benefit, at 0.27, and perceived barrier, at 0.30.

Carpenter concludes that, given the weakness of two of the four predictors, the version of the model in which the four beliefs act directly on behavior should no longer be used. Translated into the consulting room, what predicts whether someone will care for themselves is not how threatened that person feels. It is how much they believe the proposed course works, and how much they feel able to pay its price.

Michael Cummings, Alan Jette and Rosenstock himself had already found a clue to this in 1978, when they tested how to measure each of those beliefs in 85 graduate students in public health. Benefit and barrier turned out to be linked by a negative correlation of 0.655, strong enough for the authors to raise the possibility that they are not two separate beliefs but the two ends of a single ruler. The more someone sees that a thing is worth it, the lower its price appears.

The people who were most frightened were the ones who did not get tested

Rosenstock did not hide that limit. In the same 1974 paper he records that Hochbaum identified 16 people who were intensely afraid of tuberculosis, and that not one of them had had a single voluntary chest X-ray in the previous eight years. In a genetic screening program for Tay-Sachs disease that began in 1971 in the Baltimore and Washington area, he described the same pattern and wrote that perceived severity there had reached a level so high that it had become dysfunctional.

His conclusion is the one I carry into the consulting room. The recommended behavior appears at an optimal balance among motive, vulnerability, severity and the ratio between benefit and cost, and when that set sits either very low or very high the person is not to be expected to do what is recommended. Too little fear does not move anyone, and too much fear paralyzes.

In diabetes, believing the treatment works weighs more

John Harvey and Vanessa Lawson, at the endocrinology and diabetes center of Cardiff University, in Wales, reviewed in 2009 twenty years of this literature applied to self-care in diabetes. They open by recalling a meta-analysis of 30 studies of patient education in chronic disease, in which improving knowledge on its own was rarely enough to improve adherence to treatment. Some people with a good level of knowledge adhere poorly, and the review says so plainly.

The classic model's numbers there are modest as well. The review records that, in the meta-analysis by Harrison and colleagues, the four main dimensions accounted for less than 10% of the variation in measured behavior. What appears consistently is a different belief. The perception that treatment is effective is the one most related to outcomes, including glycated hemoglobin.

And there is a finding that runs against what intuition suggests. High perceived threat, the sense that the disease is serious and that complications are coming, was associated with missing diabetes appointments, and with worse self-care among adolescents. Harvey and Lawson sum it up by saying that believing diabetes is serious works as a motivator when the person also believes the treatment works, and not necessarily when what dominates is the management of emotion, sometimes through avoidance or denial.

Belief is not a personality trait, which is why it moves

Rita Orji, Julita Vassileva and Regan Mandryk, at the University of Saskatchewan in Canada, followed the eating behavior of 576 people in 2012 to test how far the original model reaches. On average the classic determinants explain somewhere around 20% of the variation in healthy behavior, which leaves 80% unexplained. When the group added beliefs the model had not been looking at, among them how much the person cares about the matter and how much they consider future consequences, predictive capacity rose from 40% to 71%. In every model they tested, the strongest determinant was self-efficacy.

There is also a sign that these beliefs respond to a small amount of well-done work. A team from STIKes Mitra Husada Medan delivered four 60-minute sessions in 2025, one a week, to 50 adolescents aged 12 to 16 in a rural village in northern Sumatra, Indonesia, working through exactly these questions applied to mental health. The group that took part finished with a mean score of 87.6 on access to care, against 76.5 in the group that did not, with an effect size of 1.06. The share of adolescents in the low anxiety band rose from 72% to 92%, and the 3 who had started in the high anxiety band finished outside it.

It is a small study, from a single community, with self-reported measures and no long follow-up, and the authors list those limitations before any reader can point them out. It does not prove that four conversations solve anything. It shows the direction, which is to work on the belief and not only on the leaflet.

Today I ask before I explain

I stopped opening a conversation about a treatment with an explanation of the treatment. I start by asking what the person thinks it will do for them, and I listen to the end, including when the answer is that it will do nothing. Then I ask what gets in the way, and I press a little, because the first answer tends to be polite and the second tends to be true. It costs money, it makes me sleepy, it makes me gain weight, the clinic opens while I am at work, my husband thinks I am overreacting, I am afraid of what the test will show. A barrier that has been named becomes a problem two people can work on, and a barrier left unnamed goes on deciding by itself.

I also stopped betting on fear. Frightening someone more is the easiest tool a doctor has at hand and one of the least effective, and the research has been saying so since the 1970s. What has proved most useful in my office is far less dramatic, which is to show clearly and without promises what that treatment usually does, and to help the person believe they can do it.

If you are putting off something you know you need to do, do not try to solve the whole postponement today. Answer two questions out loud, on your own, honestly. What do I think this will bring me? And what does it cost me? A concrete reason will very likely surface, and a concrete reason has an address. Take that reason to your next appointment, because it is the thread by which the knot begins to loosen. None of this replaces the individual assessment of whoever cares for you.

Dr. Marcelo Gobbo Jr.
Family and Community Physician (CRM-MG 74.511 · RQE 69038), with a master’s degree in Psychology from the Federal University of Uberlândia and a Laboratory Associate appointment in the Department of Psychiatry at Yale School of Medicine, based at the Program for Recovery and Community Health. He cares for neurodiverse families and people living with severe mental illness, in person and remotely.