Health Belief Model: S-Tier Behavioral Designer’s Guide

Tuberculosis screening trucks rolled through American neighborhoods in the 1950s. The X-rays were free. The line was short. The lung that you might save was your own. And almost nobody showed up.

That bewilderment is the founding moment of the Health Belief Model. Irwin Rosenstock, Godfrey Hochbaum, Stephen Kegels, and Howard Leventhal were public-health researchers staring at the same paradox a thousand product teams stare at today: a benefit that is real, a barrier that is small, and a target audience that walks past anyway. They went looking for what was happening inside the head of the person walking past, and what they found became one of the most-cited behavior-change models in the history of social science.

If you have ever wondered why your evidence-backed feature with the well-designed onboarding flow still has a 9% activation rate, the Health Belief Model is the lens that will make sense of it. The reader is doing belief math in their head whether you design for it or not. Most products are losing that math without knowing the variables.

Speed Run Notes

• HBM says people act on a health threat only when four beliefs line up: it could happen to me, it would be bad, the action would help, and the cost is worth it. Self-Efficacy was added as the fifth in 1988.
• Of those five, “Perceived Barriers” is the single strongest predictor of behavior across 40 years of meta-analyses. The lesson for designers: shrink the barriers before you amplify the threat.
• Cues to Action are the firing pin. A text message at 8:00 AM, a friend’s diagnosis, a yellow CDC banner. Without a cue, the whole apparatus sits idle in working memory.
• Where HBM collapses: it predicts intention more cleanly than behavior, treats decision-making as too rational, and ignores how culture, habit, and emotion route around its tidy boxes.
• Maps onto Octalysis with surgical clarity. Susceptibility and Severity live in Loss & Avoidance; Benefits span Accomplishment and Ownership; Self-Efficacy is built through Creativity; Cues fire from Social Influence.
• The S-Tier design move: stop trying to scare the user into acting. Lower the cost of the first step until the math tips on its own, then drop a cue at the moment of natural arousal.

Author Credibility: Yu-kai Chou

Yu-kai Chou created the Octalysis Framework after studying gamification since 2003 — years before the term entered mainstream vocabulary. As a Human-Systems Architect & Behavioral Designer, his framework has been applied by LEGO, Microsoft, Porsche, Coca-Cola, Salesforce, and MrBeast, impacting over 1.5 Billion Users.

Chou has taught the Octalysis methodology at Harvard, Stanford, Yale, Tesla, Google, BCG, and IDEO.

His work has been cited by Harvard, Stanford, MIT, Forbes, Wall Street Journal, Wired, US Department of Energy, NIST, NSF, NCBI, US Department of Education, ClinicalTrials.gov, and Google Scholar — with 3,700+ more academic publications. Explore his books here.

TEDxLausanne · Gamification to Improve Our World17 min

What Is the Health Belief Model?

The Health Belief Model (HBM) is a value-expectancy theory of health behavior. The compressed version: a person will take a recommended health action when they believe the threat is real and serious, when they believe the action would meaningfully reduce that threat, and when the perceived cost of acting is lower than the perceived cost of not acting. Add a trigger and a sense that they can pull off the action, and the behavior fires.

Rosenstock first articulated the model in print in 1966 in the Milbank Memorial Fund Quarterly, then anchored its definitive 1974 statement in Health Education Monographs. Marshall Becker’s 1974 special issue of that same journal compiled the early empirical work. By 1988, Rosenstock, Strecher, and Becker had grafted Albert Bandura’s Self-Efficacy onto the model as a fifth construct, because the original four kept under-predicting behavior in cases where the person believed every other variable and still did nothing because they did not believe they could pull it off.

The model started inside the U.S. Public Health Service trying to answer one specific failure: people would not show up for free tuberculosis screenings. By the 1970s it had been applied to vaccination, dental care, hypertension, family planning, and seat-belt use. Today it is the unspoken design substrate of cancer-screening reminders, vaccination campaigns, smoking-cessation apps, mental-health-stigma interventions, prescription-adherence nudges, and chronic-disease-management dashboards. It is also, quietly, the framework most consumer-product behavioral designers ignore because it does not have a clean Silicon Valley provenance.

That is a mistake. The Health Belief Model is the cleanest pre-rational decomposition of “why didn’t they act?” in the literature. Once you can name the four or five beliefs the user is silently scoring, you can fix the right one instead of bolting on another notification.

The Five (Originally Four) Constructs, Decoded

Most explanations of HBM list the constructs without explaining the design implications of each. That treats the model like a textbook artifact instead of a working tool. Each construct is a separate lever, and each lever responds to a different intervention.

Perceived Susceptibility: “Could this happen to me?”

Susceptibility is the believed probability of getting the condition. It is the variable that public-health campaigns most often try to move, and the variable they are usually worst at moving. A poster saying “1 in 8 women will develop breast cancer” attempts to lift susceptibility. The reader’s pre-rational response is almost always “but not me.” Optimism bias, identifiable-victim effect, and base-rate neglect all conspire against the statistic. The susceptibility lever moves much more reliably when the threat is personalized through a familiar example. A friend’s diagnosis. A genetic-test result. A symptom you can feel.

This is why every effective health-app onboarding flow that addresses susceptibility starts with the user’s own data, not the population’s. Apple’s Heart Health notifications show your resting heart rate trending against your baseline. The susceptibility lift comes from the personalization, not from the underlying statistic. Designers who try to move susceptibility with epidemiology are fighting the brain’s design.

Perceived Severity: “How bad would it be?”

Severity is the believed badness of the consequences. It includes both physical consequences (pain, disability, death) and social consequences (job loss, stigma, family disruption). Severity is the second-weakest predictor across HBM meta-analyses, which surprises designers who instinctively lead with severity. People know smoking causes lung cancer. People know obesity causes diabetes. People know unsafe sex transmits HIV. Severity is rarely the missing variable.

There is one category where severity is the limiting belief: conditions where the severity has been culturally downplayed. Maternal mortality in low-income countries. Cervical cancer before public-health campaigns. Heart disease in women (where the symptoms differ from men and were dismissed as anxiety for decades). When severity has been culturally suppressed, lifting it does move behavior. Outside those cases, severity-heavy messaging mostly just makes people change the channel.

Perceived Benefits: “What do I gain by acting?”

Benefits is the believed effectiveness of the recommended action. A flu shot prevents flu. A statin lowers cholesterol. A screening catches cancer early. The empirical pattern here is striking: people often correctly believe that the action would help, but believe the help is too small to be worth the cost. This is where benefits and barriers do their math against each other.

The design lesson is that abstract benefits (“reduces your risk by 40%”) almost always lose to concrete benefits (“you can play with your grandchildren without getting winded”). The 40% statistic is a population-level probability that does not feel like ownership of an outcome. The grandchildren image is a personal future the user can see. That is the difference between a benefit framed for Core Drive 8 (Loss & Avoidance) and one framed for Core Drive 2 (Development & Accomplishment) plus Core Drive 4 (Ownership & Possession). And the second framing nearly always wins on durable behavior.

Perceived Barriers: “What is it going to cost me?”

Barriers is the believed cost of taking the action. Cost is a wide category: money, time, pain, inconvenience, side-effects, social embarrassment, the friction of scheduling, the cognitive load of figuring out where to even start. Across forty years of HBM meta-analyses — Janz and Becker’s classic 1984 review, Carpenter’s 2010 meta in Health Communication, and Champion and Skinner’s review in the 2008 Health Behavior and Health Education textbook — Perceived Barriers is the single strongest predictor of behavior. Stronger than susceptibility, severity, and benefits combined.

For designers, this finding is the most important sentence in the whole HBM literature. Most products are trying to move the wrong lever. They are pushing susceptibility (“you might be at risk”) and severity (“this could be serious”) when the variable actually blocking behavior is barriers. Five extra clicks to book an appointment. A copay the user is not sure their insurance covers. A waiting room that costs four hours. Until you have shrunk Perceived Barriers, no amount of fear messaging will tip the math.

Cues to Action (the firing pin)

Cues to Action are the triggering events that move latent intention into action. A reminder text. A symptom you can feel. A friend’s diagnosis. A news story. A doctor’s recommendation during an unrelated visit. A workplace flu-shot table. In Rosenstock’s original framing, Cues to Action sit slightly to the side of the belief variables, because they are not beliefs at all. They are stimuli that activate beliefs that were sitting dormant.

This is the lever modern behavioral design has industrialized. SMS reminders for medication adherence. Push notifications for blood-pressure logging. Automated screening-due alerts in electronic health records. The empirical evidence on cues is unusually strong: a 2009 meta-analysis by Fjeldsoe, Marshall, and Miller found mobile-phone text-message interventions produce small-to-moderate effect sizes on a range of health behaviors, with the largest effects when the cue is paired with a low-barrier action immediately available. The cue without the easy action just generates guilt.

Self-Efficacy (added 1988): “Can I actually do this?”

Self-Efficacy is the believed personal capability to perform the action. Bandura introduced the construct in 1977. Rosenstock, Strecher, and Becker grafted it onto HBM in 1988 because the original four constructs kept failing to predict sustained behavior change in chronic-disease contexts (smoking, weight loss, diabetes self-management). Smokers know smoking is bad. They know quitting helps. They are not blocked by susceptibility, severity, or benefits. They are blocked by “I have tried twice and failed twice.”

Self-Efficacy is the bridge between belief and behavior. A user who scores high on susceptibility, severity, benefits, and low on barriers will still not act if they do not believe they can pull it off. This is why self-efficacy interventions (mastery experiences, vicarious learning, verbal persuasion, somatic feedback) outperform pure information interventions for any sustained behavior.

The 1988 Self-Efficacy Addition That Saved the Model

By the mid-1980s, the Health Belief Model was in trouble. The original four constructs were predicting one-time preventive behaviors (vaccination, screening) reasonably well, but they were failing on chronic-disease self-management. Smokers, diabetics, hypertension patients, and people trying to lose weight all scored high on susceptibility, severity, and benefits, low on barriers, and still relapsed. Something was missing.

Rosenstock, working with Victor Strecher and Marshall Becker, published the 1988 amendment that added Albert Bandura’s Self-Efficacy construct as the fifth HBM variable. The paper, “Social Learning Theory and the Health Belief Model,” appeared in Health Education Quarterly. It was an unusually honest theoretical revision. Rosenstock did not claim HBM had always implicitly contained self-efficacy. He admitted the original model had been built for one-shot decisions in an era when the dominant health threats were infectious diseases requiring a single behavior change. By 1988, the dominant health threats were chronic conditions requiring sustained behavior change, and the model had to evolve.

Self-Efficacy moves the model from a static belief-evaluation snapshot to a dynamic capability assessment. The user is not just asking “is this worth doing?”. They are asking “is this worth doing AND can I sustain it?” A diabetic considering a low-carbohydrate diet is not blocked by belief in the diet’s effectiveness. They are blocked by the memory of three failed attempts to give up bread.

The addition has a second implication that has been under-explored. Self-Efficacy is the only HBM construct that the user themselves can build. Susceptibility, severity, benefits, and barriers are mostly fixed by the situation. Self-Efficacy can be deliberately constructed through Bandura’s four sources: mastery experiences (you did it once, scaled small), vicarious learning (someone like you did it), verbal persuasion (a credible person says you can), and somatic feedback (your body says you can). Every well-designed habit-formation product is, underneath, a self-efficacy machine.

What Rosenstock Got Right

The Health Belief Model is now 60 years old. It has weathered the rise of the Theory of Reasoned Action, the Theory of Planned Behavior, the Transtheoretical Model, Protection Motivation Theory, Social Cognitive Theory, the Fogg Behavior Model, and a generation of dual-process accounts. It is still in active use. That kind of longevity in social science is rare. Three things explain it.

It decomposed behavior into the right variables

Rosenstock’s instinct, in the 1950s, was that the choice to take a health action is not unitary. It is the product of several distinct beliefs that can vary independently and that can be intervened on separately. That decomposition felt radical at the time. The dominant lay theory was that people would act when they understood the threat. And it turned out to be correct. Modern behavior-change interventions still target HBM’s five constructs by name in trial designs, even when the trial is officially branded as Theory of Planned Behavior or COM-B (Capability, Opportunity, Motivation → Behavior).

The lasting design lesson: when behavior is not happening, the first move is to figure out which belief is failing. Not all of them. Just one or two. Then intervene on those. A one-size-fits-all message that pushes every variable is almost always weaker than a targeted message that fixes the one variable actually blocking the user.

It emphasized barriers when nobody else was

Public-health communication in the 1950s and 60s was overwhelmingly fear-based. The HBM construct of Perceived Barriers was, in effect, a quiet rebuke to the dominant approach. Rosenstock and Becker kept pointing out that even after fear messaging had worked perfectly. Even after the person believed they were susceptible to a severe condition that the recommended action would treat. They still did not act because the action was too expensive, too inconvenient, too embarrassing, or too uncertain.

This insight predates the entire behavioral-economics friction literature by half a century. Cass Sunstein and Richard Thaler’s framing of friction as a behavioral-design variable, the Behavioural Insights Team’s EAST framework’s “Easy” lever, the modern UX obsession with reducing onboarding steps. All of it is downstream of Rosenstock’s quiet 1960s observation that barriers matter more than people think.

It named cues as a separate variable

Most theories of behavior collapse the trigger into the motivation. The Health Belief Model kept them separate. Belief sets the stage. Cues fire the gun. This distinction is what allows modern behavioral designers to think clearly about when to deliver a message, not just what message to deliver. A perfect message at the wrong moment does nothing. A merely adequate message at the right moment converts.

BJ Fogg’s Behavior Model (B = MAP, where Prompt is one of the three required ingredients) is essentially a compression of this insight. So is Nir Eyal’s Hook Model (Trigger is the first step). The whole modern behavioral-design industry runs on the assumption that triggers must be designed deliberately and separately from the value proposition. That assumption is Rosenstock’s, originally.

Where the Health Belief Model Falls Apart

HBM has earned its longevity, but it has also accumulated a serious list of empirical and conceptual problems. A behavioral designer who treats the model as gospel will build the wrong things. Three failure modes matter most.

It predicts intention better than behavior

HBM is, in practice, mostly a model of intention. People who score high on its constructs tend to say they will get the screening, take the medication, change the diet. Whether they actually do is a separate question, and the model has weaker grip on the second question than on the first. Carpenter’s 2010 meta-analysis in Health Communication reviewed 18 HBM studies across 30 years and found the four core constructs explained intention reliably but explained actual behavior only modestly. The single strongest behavioral predictor was Perceived Barriers, with the other constructs contributing little independent variance once barriers were controlled.

The intention-behavior gap is not unique to HBM. Theory of Planned Behavior has the same problem. Self-Determination Theory has it. Almost every cognitive model of behavior change confronts it. But HBM has been less honest about the gap than its successors, and a designer who reads only HBM will systematically over-estimate how much of the design problem can be solved by changing what people believe.

It treats decision-making as too rational

The model assumes a quasi-rational user who consciously weighs four or five beliefs and arrives at a choice. The behavioral-economics revolution of the last 40 years has demolished that assumption. Daniel Kahneman’s dual-process account (System 1 vs System 2) implies most health decisions are made by the fast, automatic system, not the deliberative one. Implementation Intentions research from Peter Gollwitzer shows that linking a behavior to a specific context-cue (“when I leave the office, I take my pills”) routes around belief evaluation entirely.

The lesson: HBM is most useful for behaviors that are episodic and consequential enough to engage System 2. Getting a colonoscopy, starting a serious medication, choosing surgery. It is least useful for behaviors that are habitual or low-stakes-per-instance but cumulatively important. Daily flossing, walking after dinner, putting the phone down. For those behaviors, BJ Fogg’s Tiny Habits, James Clear’s habit-stacking, and Wendy Wood’s habit-formation research will outperform any belief-evaluation framework.

It is culturally narrow

HBM was built in a 1950s American public-health context that framed health decisions as individual choices made by individual actors. The model’s susceptibility, severity, benefits, and barriers are all scored at the individual level. That framing breaks in collectivist cultures where the health decision is family-level (an elderly parent in a multigenerational household), community-level (a village health-worker program), or kin-network-level (where the cost of an action includes the social cost to relatives).

Studies of HBM in Chinese, South Asian, Sub-Saharan African, and Latin American contexts consistently find that adding family-level or community-level variables substantially improves predictive power. The model’s individualist scaffolding is a feature where it fits the culture and a bug where it does not. Designers building for global audiences need to keep that in mind, especially for conditions where stigma, marriage prospects, or family economics are part of the decision.

What is Really Happening Inside the Brain

Rosenstock did not have access to neuroscience. He was inferring belief structures from survey data and behavioral observations. Modern affective neuroscience has filled in much of what the model was reaching for. The picture is messier and more interesting than the original boxes-and-arrows diagrams suggest.

Perceived Susceptibility and Perceived Severity together form what neuroscientists call threat appraisal. This process is anchored in the amygdala, which evaluates incoming stimuli for personal salience and danger within roughly 100 milliseconds of perception. The amygdala does not do statistics. It does pattern-matching against stored experience. This is why personalized cues (a friend’s diagnosis, a felt symptom) move susceptibility while population-level statistics largely do not. The amygdala has nothing to match the statistic against.

Perceived Benefits and Perceived Barriers are evaluated downstream, in the ventromedial prefrontal cortex (vmPFC). The vmPFC is where the brain computes subjective value across competing options. Brian Knutson’s neuroimaging work on choice prediction shows that vmPFC activation tracks the integrated value signal that predicts action better than self-reported preferences do. When the vmPFC computes that the expected benefit exceeds the expected cost, action is more likely. The Perceived-Benefits-minus-Perceived-Barriers calculation Rosenstock proposed is, neurobiologically, a real thing. It is just done unconsciously and very quickly.

Self-Efficacy maps onto activation in the anterior cingulate cortex (ACC) and dorsolateral prefrontal cortex (dlPFC) regions associated with effort assessment and motor-plan generation. When self-efficacy is low, the dlPFC computes the action as effortful and the ACC flags conflict between intention and capacity. The result is what Bandura predicted: the action gets initiated less, abandoned faster, and re-attempted less after failure.

Cues to Action work because they hijack attentional networks. A push notification arriving at 8:00 AM does not change your underlying beliefs about cholesterol. It captures bottom-up attention, briefly elevating the relevant memory traces, and creates a momentary window in which the integrated value signal in the vmPFC actually gets evaluated. Without that window, the calculation sits unscored in working memory.

The synthesis worth carrying out of the neuroscience: HBM is not wrong about the belief variables. It is wrong about when they get computed. The brain does not constantly score susceptibility, severity, benefits, barriers, and self-efficacy. It scores them in brief windows triggered by cues. Designers who treat cues as the dependent variable rather than the independent variable are inverting the actual causal flow.

HBM vs Other Behavior-Change Theories

The Health Belief Model is one of perhaps a dozen mainstream behavior-change theories. A designer choosing which to use should know what each one captures that HBM does not.

HBM vs Theory of Planned Behavior (Ajzen, 1991)

The Theory of Planned Behavior (TPB) predicts behavioral intention from three variables: attitudes toward the behavior, subjective norms (what people who matter to you think you should do), and perceived behavioral control (similar to self-efficacy). TPB is more general than HBM. It applies to any behavior, not just health behaviors. And it explicitly includes social norms, which HBM under-weights. The trade-off is that TPB has no explicit threat appraisal. For health behaviors where the threat is doing the motivational work, HBM is the cleaner tool. For social behaviors where norms are the motivational engine (voting, recycling, organ donation), TPB is the better fit.

HBM vs Transtheoretical Model (Prochaska & DiClemente, 1983)

The Transtheoretical Model (TTM, the “Stages of Change” model) frames behavior change as a sequence: precontemplation, contemplation, preparation, action, maintenance, and termination. Where HBM is a cross-sectional snapshot of beliefs at one moment, TTM is a longitudinal map of how a person moves through different psychological states over time. The two are complementary. HBM tells you which beliefs are blocking action at any moment. TTM tells you which intervention type (consciousness-raising vs commitment-building vs counter-conditioning) is appropriate at the user’s current stage.

HBM vs Protection Motivation Theory (Rogers, 1975)

Protection Motivation Theory (PMT) was explicitly developed as a refinement of HBM. Rogers split threat appraisal (susceptibility plus severity) and coping appraisal (response efficacy plus self-efficacy minus response costs) into separate streams that interact. PMT also gave fear a formal role that HBM never quite resolved. The trade-off is that PMT is more focused on fear-appeal communication design and less general than HBM. If you are designing public-health messaging specifically around threats, PMT often outperforms HBM. For broader behavior-change product design, HBM’s wider construct set is more useful.

HBM vs Fogg Behavior Model (Fogg, 2009)

BJ Fogg’s Behavior Model collapses behavior into three variables: Motivation, Ability, and Prompt (B = MAP). Fogg’s framework is faster to apply and more useful at the moment of behavior. HBM is more useful in the strategic phase, when you are deciding what to communicate, what to remove, and what to build. The two models are not in competition. Fogg tells you what conditions must be present in the user’s mind at the moment of behavior. HBM tells you which of those conditions are likely to be missing, and why, for a given health context.

HBM vs COM-B (Michie, van Stralen & West, 2011)

COM-B (Capability, Opportunity, Motivation → Behavior) is the UK Behavioural Insights industry standard. It is broader than HBM. Capability and Opportunity both include physical and environmental constraints HBM does not address. And it is more directly tied to a 19-intervention-type taxonomy (the Behaviour Change Wheel). For policy and large-program design, COM-B is often the right framework. For product-level health interventions where the question is “what does this user believe?”, HBM is sharper. The two play well together: use COM-B to identify which broad domain is failing, then use HBM to diagnose the specific beliefs inside the Motivation domain.

HBM in the Real World

Vaccination campaigns: COVID-19 and beyond

The COVID-19 vaccination rollout was the largest natural experiment in HBM application in living memory. The CDC, WHO, and most national public-health agencies built communication strategies that mapped almost directly onto the model. Susceptibility messaging emphasized variants and breakthrough infections. Severity messaging emphasized hospitalization and long-COVID. Benefits messaging emphasized return-to-normal. Barriers were attacked through free shots, walk-in clinics, mobile vaccination units, and employer paid time off. Cues to Action included reminder texts, employer mandates, and pharmacy notifications.

The interesting finding from the post-2022 academic literature: across multiple studies, the variable that best predicted uptake was not susceptibility or severity but a barrier-reduction variable — physical access plus trust in the local provider. Communities where the vaccine was given at a familiar pharmacy by a trusted pharmacist outperformed communities where the vaccine was equally accessible but distributed through unfamiliar mass-vaccination sites. Once again, barriers (in the broad sense including social-trust friction) beat threat appraisal.

Cancer screening

Mammography, cervical screening, colorectal screening, and prostate screening have all been studied through the HBM lens for decades. The pattern is consistent: information campaigns that raise susceptibility and severity have small effects on screening rates. Interventions that reduce barriers (free or subsidized screening, weekend hours, mobile units, automated scheduling, transportation reimbursement) have substantially larger effects. The most effective single intervention in the literature is the “patient navigator” model: a person who walks the patient through scheduling, transport, financial questions, and follow-up. The patient navigator is, in HBM terms, a walking barrier-reduction machine.

Medication adherence

Chronic medication adherence is one of the most expensive failure modes in healthcare. Global estimates put non-adherence at hundreds of billions of dollars annually in avoidable downstream costs. HBM analyses of adherence consistently find that susceptibility and severity are not the limiting variables (patients know what they have), but barriers and self-efficacy are. Side-effect concerns, cost, complexity of the regimen, and prior failed attempts all dominate. Modern adherence apps (Medisafe, MyTherapy, Mango Health) are HBM machines with self-efficacy as the explicit design target: streaks, mastery experiences, vicarious learning through peer communities.

Mental-health help-seeking

One of the more interesting modern applications of HBM is mental-health stigma reduction. The classic failure mode in mental-health help-seeking is that susceptibility and severity are often over-perceived (the user knows something is wrong) but barriers dominate. Social stigma, employer concerns, fear of being told they are “weak,” and the friction of finding a therapist who takes their insurance and is accepting new patients. Modern teletherapy products (BetterHelp, Talkspace, Cerebral) explicitly attack the barrier side: privacy, asynchronous communication, fast onboarding, predictable monthly cost. The threat appraisal is already done. The product wins by being easier than the existing alternatives.

Wearables and personalized risk

The Apple Watch, Oura Ring, Whoop, and Fitbit ecosystems have inadvertently become some of the most successful HBM-aligned products ever built. They do not lecture about susceptibility. They show your resting heart rate, your heart rate variability, your sleep, your activity trend. Personalized data does what population statistics could not: it moves the susceptibility belief by giving the user something the amygdala can pattern-match against. The Apple Heart Study alone identified atrial fibrillation in users who would not have sought screening on their own: a textbook case of a wearable acting as both a susceptibility-raising agent and a Cue to Action.

The Elephant in the Room

The honest critique of the Health Belief Model is not that it is wrong. It is that the entire field of public-health communication has been using it as cover for a strategy that does not work, while the actual variable that moves behavior keeps getting ignored.

For 60 years, public-health agencies have produced posters, ads, brochures, websites, and TV spots aimed primarily at susceptibility and severity. Don’t smoke or you’ll get cancer. Get your kids vaccinated or they could die. Lose weight or you’ll have a heart attack. The strategy assumes that if people just understood the threat, they would act. The Health Belief Model is then cited as the theoretical backing: “we are addressing perceived susceptibility and perceived severity,” the campaign report says. As though naming the construct counts as targeting it correctly.

The problem is that the same Health Belief Model’s own evidence base, going back to Janz and Becker’s 1984 meta-analysis and continuing through Carpenter’s 2010 work, says clearly that Perceived Barriers is the strongest predictor and susceptibility and severity are among the weakest. The literature has known this for forty years. The campaigns keep targeting the weak variables.

Why? Because moving susceptibility and severity is cheap. A poster costs nothing. A 30-second ad has a low marginal cost. Telling people that diabetes is serious can be done from headquarters. Reducing barriers, by contrast, is expensive. It requires changing how the clinic is staffed, how the appointment is scheduled, how the insurance is billed, how the medication is delivered, how the patient is followed up. Barrier reduction is operations work. Threat communication is marketing work. Most public-health agencies are better at the second.

The same pattern repeats inside consumer health products. The wellness app that pushes more daily reminders, the fitness brand that runs another inspirational ad, the supplement company that adds another fear-driven landing page. They are all targeting susceptibility and severity because that is what the marketing team can ship. The barrier-reduction work would require redesigning onboarding, simplifying the regimen, integrating with insurance, building a real coaching layer, and answering the user’s specific questions at the moment the question comes up. Most companies are not staffed for that.

Yu-kai’s design rule on this is direct: if your behavior-change product is mostly producing more communication, it is targeting the wrong HBM variable. The user’s susceptibility and severity beliefs are not the problem. The barriers between intention and action are. Spend the design budget on shrinking the barrier set, not on producing more content that explains the threat the user already knows about.

How to Apply HBM with the Octalysis Framework

The Health Belief Model maps onto the Octalysis Framework with unusual cleanness. Each HBM construct is dominated by a specific Core Drive, which means the design move for each construct is not abstract. It has a named technique, a known reward pattern, and a White-Hat or Black-Hat orientation. Below is the per-Core-Drive mapping I use when applying HBM through Octalysis on real product work.

Core Drive 1 (CD1): Epic Meaning & Calling (the long-arc anchor)

HBM does not have a CD1 construct, and that is one of its real gaps. People sustain hard health behaviors much longer when the behavior is connected to a meaning larger than personal benefit. The grandparent who walks an hour a day not to lower cholesterol but to be alive for the grandchildren. The cancer survivor who follows the screening regimen not for themselves but to model the behavior for their daughter. CD1 is the variable HBM under-represents and the variable that nearly always shows up in qualitative interviews with long-term adherent patients.

Design move: where the action will require sustained behavior over months or years, build a CD1 layer the user can actually see. Not “improve your health”. That is too abstract. Something like a private dashboard the user shows their family, a milestone celebration the family is invited to, or an explicit framing of the behavior as “for them.” This adds a motivational reservoir HBM does not capture.

Core Drive 2 (CD2): Development & Accomplishment (where Benefits live)

Perceived Benefits is dominated by CD2. The user is implicitly asking “what will I have achieved by taking this action?” The answer must be visible, specific, and ladder-able. Visible: progress is shown. Specific: the benefit is a concrete future state, not an abstract risk reduction. Ladder-able: there are intermediate wins, not just a single distant outcome.

Design move: every benefit should have a progress indicator the user can see today. A blood-pressure tracker that shows the trend line. A medication-adherence streak. A weight-loss graph. The graph itself is doing CD2 work. It converts the abstract benefit into a felt accomplishment.

Core Drive 3 (CD3): Empowerment of Creativity & Feedback (where Self-Efficacy is built)

Self-Efficacy is built primarily through CD3. The user gains capability belief through doing the action, getting feedback, adjusting, and trying again. This is exactly Bandura’s “mastery experience” mechanism. The design implication is that any health product that wants to build self-efficacy needs to give the user something they can actually do and get real-time feedback on. Reading a brochure about diabetes management does not build self-efficacy. Logging a meal and seeing the predicted glucose response does.

Design move: give the user a small, low-stakes version of the target behavior they can practice with rapid feedback. The self-efficacy gain transfers to the harder real behavior. This is the entire mechanism behind every successful behavior-change product from Duolingo to Couch-to-5K to MyFitnessPal.

Core Drive 4 (CD4): Ownership & Possession (the personal-data anchor)

Perceived Benefits also draws on CD4 when the benefit is framed as something the user already owns and could lose, or something they would gain ownership of. “Your” data. “Your” health. “Your” recovery. Personalized dashboards, exportable health reports, and user-named tracked metrics all activate CD4. The Apple Health app is a CD4 machine. The user perceives the data as theirs, which makes the underlying health metrics feel like a possession worth protecting.

Design move: anywhere the user’s own data is involved, the framing must lean into ownership. “Your sleep this week.” “Your blood-pressure trend.” Never the impersonal third-person. CD4 is also the drive that makes a saved partial form feel like an asset worth coming back to finish, which directly attacks Perceived Barriers in onboarding.

Core Drive 5 (CD5): Social Influence & Relatedness (where Cues to Action live)

Cues to Action are dominated by CD5. The most reliable cue across the HBM literature is a recommendation from someone the user trusts. A doctor, a family member, a friend who recently had the same condition. Social Cues are not just notifications. They are signals from a trusted in-group that the behavior is normative and expected.

Design move: where the budget allows, embed a real human cue (a coach, a navigator, a peer mentor) rather than relying on automated reminders. Where the budget does not allow a human, make the automated reminder feel as much like a personal message as possible. The “your doctor recommends” framing outperforms the “you should” framing by a wide margin in adherence studies.

Core Drive 6 (CD6): Scarcity & Impatience (use sparingly)

CD6 in health design is dangerous. The “limited time” framing works for promotional behaviors (book your screening this month) but corrodes trust if used for chronic conditions. A diabetes-management app that uses countdown timers and limited-time offers feels predatory, because the user knows their condition is not going away. CD6 should be used surgically. At moments of natural scarcity (vaccine availability window, an enrollment period, a true seasonal opportunity). And never atmospherically across the product.

Core Drive 7 (CD7): Unpredictability & Curiosity (the discovery cue)

CD7 powers a subset of Cues to Action: the unexpected finding. A wearable that surprises the user with an unusual heart-rate pattern. A blood-test result that uncovers something the user did not know to look for. A new symptom that does not fit the user’s existing mental model. CD7 cues are powerful because they generate a curiosity loop the user wants to close: “what is going on with me?” That curiosity converts to action more reliably than a planned reminder, because the user is now self-motivated rather than externally prompted.

Design move: where the product has access to user data, surface anomalies rather than just trends. An app that says “your sleep dropped 22 minutes this week, unusual for you” is doing CD7 work. The novelty captures attention; the personal-data framing converts attention to action.

Core Drive 8 (CD8): Loss & Avoidance (where Susceptibility and Severity live)

Perceived Susceptibility and Perceived Severity are pure CD8 levers. They both ask the user to evaluate “what bad thing might happen to me?” CD8 is the most powerful Core Drive in the short term and the most exhausting in the long term. A health product that runs on CD8 atmospherically. Constantly reminding the user of what they might lose. Burns out the user within weeks. A health product that uses CD8 surgically. At the moment of a real risk event, calibrated to the actual stakes. Converts.

Design move: deploy CD8 at the cue, not as the ambient mood of the product. The doctor visit, the test result, the family-history moment, the symptom flare-up. Then immediately follow CD8 with CD3 (Empowerment of Creativity & Feedback) and CD4 (Ownership & Possession) so the user does not stay in fear. They move from fear to capability to ownership.

The synthesis: a per-construct design move

The reason HBM and Octalysis work so well together is that HBM names the variable that is failing and Octalysis names the design move that fixes it. The mapping I use:

• Susceptibility low? Personalize the data (CD4-flavored CD8). Population statistics will not move it.
• Severity low? Use a vivid concrete consequence (CD8), not a generic warning.
• Benefits perceived as small? Show progress (CD2) and ownership of the future state (CD4).
• Barriers too high? Cut friction in the workflow, not in the messaging. Operations work, not marketing.
• Cues missing? Trusted-source social cues (CD5) outperform generic reminders. Anomaly-based cues (CD7) outperform scheduled ones.
• Self-Efficacy low? Build it through real practice with rapid feedback (CD3). Brochures will not build it.

That mapping turns HBM from a survey instrument into a design surface. The diagnosis stays HBM. The intervention becomes Octalysis-specific. Together they produce a product that does not just communicate threat but actually changes behavior.

Practical Steps: Running an HBM Audit on Your Product

A behavioral designer should be able to run an HBM audit on any health-adjacent product in a half day. Six steps.

1. Name the target behavior precisely. Not “improve health”: “log one blood-pressure reading per day before breakfast.” HBM scoring only works against a specific behavior. Vague behaviors produce vague diagnoses.
2. Score the five constructs from the user’s perspective. For each construct, write one sentence in the user’s voice. Susceptibility: “Yes / no, I believe this could happen to me because ___.” Severity: “If it happened, it would be ___ bad because ___.” Repeat for benefits, barriers, self-efficacy.
3. Identify the lowest-scoring construct. That is the variable blocking the behavior. Most products are blocked by either Barriers (operational) or Self-Efficacy (capability). Susceptibility and Severity are usually adequate.
4. Match the failing construct to its Core Drive intervention. Use the HBM-to-Octalysis mapping above. Barriers fail? Cut friction in the workflow. Self-Efficacy fails? Add a low-stakes practice loop with feedback.
5. Design ONE intervention against the failing construct. Not five. The HBM literature is full of multi-construct interventions that move nothing because they spread the design budget too thin. Pick the lowest construct and concentrate.
6. Add a Cue to Action calibrated to the moment of natural arousal. Without a cue, the redesigned belief structure sits dormant. With the wrong cue (mid-meeting push notification) it generates negative affect. The cue should land when the user has both attention and capacity to act.

I have run this audit on dozens of products. The single most common finding is that the design team has been working on the wrong construct for months. The screening reminder app that needs better booking flow. The diabetes coach that needs simpler logging. The mental-health app that needs cheaper sessions. The HBM audit makes the misallocation visible in an afternoon.

HBM Was the Beginning, Not the End

The Health Belief Model is older than the laptops on which most behavioral designers will read this. It was built for a public-health world that no longer exists, in service of behavior-change problems that have largely been replaced by harder ones. It has been correctly criticized for being too rational, too individualistic, too cross-sectional, and too eager to predict intention rather than behavior.

And it is still the cleanest belief-decomposition framework in the literature. Sixty years on, when a behavioral designer wants to answer the question “why isn’t the user doing the thing they say they want to do?”, HBM remains one of the first tools to reach for. Not because the model is perfect, but because the questions it forces you to ask — what does the user believe about susceptibility, severity, benefits, barriers, and their own capability? — are the right questions.

What the modern behavioral designer should do is treat HBM as the diagnostic, not the intervention. Use it to find the failing belief. Then reach for Octalysis, Self-Determination Theory, the Fogg Behavior Model, COM-B, the Behaviour Change Wheel. Whatever framework gives you the right intervention design surface for the specific belief that is failing. The combined toolkit produces work HBM alone cannot.

The most lasting thing Rosenstock built was not the model itself. It was the discipline of decomposing behavior into separable beliefs you can intervene on one at a time. That discipline survives every refinement of every successor theory. It is what makes the Health Belief Model worth reading in 2026, sixty years after its first formal statement.

Frequently Asked Questions

Who created the Health Belief Model?

The Health Belief Model was developed in the 1950s by social psychologists at the U.S. Public Health Service. Irwin Rosenstock, Godfrey Hochbaum, Stephen Kegels, and Howard Leventhal. To explain why people were not participating in free tuberculosis screening programs. Rosenstock’s 1974 paper in Health Education Monographs is the canonical formal statement.

What are the constructs of the Health Belief Model?

The original four constructs are Perceived Susceptibility, Perceived Severity, Perceived Benefits, and Perceived Barriers, with Cues to Action acting as the trigger and Modifying Variables (age, sex, socioeconomic status, knowledge) acting as moderators. Self-Efficacy was added as the fifth construct in 1988 by Rosenstock, Strecher, and Becker.

Which Health Belief Model construct predicts behavior most strongly?

Across multiple meta-analyses including Janz and Becker (1984) and Carpenter (2010), Perceived Barriers is the strongest predictor of actual behavior. Susceptibility and Severity, which are the constructs most public-health campaigns target, are among the weakest. The practical lesson: shrink the cost of action before amplifying the threat.

When was Self-Efficacy added to the Health Belief Model?

Self-Efficacy was added in 1988 by Irwin Rosenstock, Victor Strecher, and Marshall Becker in their paper “Social Learning Theory and the Health Belief Model” in Health Education Quarterly. The addition borrowed Albert Bandura’s construct because the original four variables under-predicted sustained behavior change in chronic-disease contexts.

How does the Health Belief Model differ from the Theory of Planned Behavior?

HBM is health-specific and built around threat appraisal. The Theory of Planned Behavior (Ajzen, 1991) is general and built around attitudes, subjective norms, and perceived behavioral control. HBM is sharper for health threats. TPB is sharper for social behaviors where group norms drive action. The two are complementary, not competing.

What are the criticisms of the Health Belief Model?

The main criticisms: it predicts intention better than actual behavior; it treats decision-making as too rational; it under-weights emotion, habit, and social influence; it is culturally narrow (individualist framing); and it is largely cross-sectional, providing a static snapshot rather than a longitudinal account of how beliefs change over time.

Can the Health Belief Model be applied outside of health?

Yes, with care. The model has been adapted to financial behavior (retirement saving), safety behavior (seat belts), cybersecurity, environmental behavior, and educational engagement. The translation works when the target behavior involves a perceived threat, a perceived response, and a barrier-versus-benefit calculation. It works less well for purely social or hedonic behaviors where threat appraisal is not the engine.

How does the Health Belief Model fit with the Octalysis Framework?

HBM constructs map cleanly onto Octalysis Core Drives: Susceptibility and Severity are Core Drive 8 (Loss & Avoidance) levers; Perceived Benefits live in Core Drive 2 (Development & Accomplishment) and Core Drive 4 (Ownership & Possession); Self-Efficacy is built through Core Drive 3 (Empowerment of Creativity & Feedback); Cues to Action sit primarily in Core Drive 5 (Social Influence & Relatedness) and Core Drive 7 (Unpredictability & Curiosity). Together, HBM diagnoses the failing belief and Octalysis prescribes the design intervention.

References

1. Rosenstock, I. M. (1966). Why people use health services. Milbank Memorial Fund Quarterly, 44(3), 94–127.
2. Rosenstock, I. M. (1974). Historical origins of the Health Belief Model. Health Education Monographs, 2(4), 328–335.
3. Becker, M. H. (Ed.). (1974). The Health Belief Model and personal health behavior. Health Education Monographs, 2(4) (special issue).
4. Janz, N. K., & Becker, M. H. (1984). The Health Belief Model: A decade later. Health Education Quarterly, 11(1), 1–47.
5. Rosenstock, I. M., Strecher, V. J., & Becker, M. H. (1988). Social learning theory and the Health Belief Model. Health Education Quarterly, 15(2), 175–183.
6. Carpenter, C. J. (2010). A meta-analysis of the effectiveness of Health Belief Model variables in predicting behavior. Health Communication, 25(8), 661–669.
7. Champion, V. L., & Skinner, C. S. (2008). The Health Belief Model. In K. Glanz, B. K. Rimer, & K. Viswanath (Eds.), Health Behavior and Health Education: Theory, Research, and Practice (4th ed., pp. 45–65). Jossey-Bass.
8. Hochbaum, G. M. (1958). Public participation in medical screening programs: A sociopsychological study. PHS Publication No. 572. U.S. Government Printing Office.
9. Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84(2), 191–215.
10. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211.
11. Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude change. Journal of Psychology, 91(1), 93–114.
12. Prochaska, J. O., & DiClemente, C. C. (1983). Stages and processes of self-change of smoking: Toward an integrative model of change. Journal of Consulting and Clinical Psychology, 51(3), 390–395.
13. Michie, S., van Stralen, M. M., & West, R. (2011). The behaviour change wheel: A new method for characterising and designing behaviour change interventions. Implementation Science, 6, 42.
14. Fogg, B. J. (2009). A behavior model for persuasive design. In Proceedings of the 4th International Conference on Persuasive Technology, Article 40.
15. Fjeldsoe, B. S., Marshall, A. L., & Miller, Y. D. (2009). Behavior change interventions delivered by mobile telephone short-message service. American Journal of Preventive Medicine, 36(2), 165–173.
16. Gollwitzer, P. M. (1999). Implementation intentions: Strong effects of simple plans. American Psychologist, 54(7), 493–503.
17. Chou, Y.-K. (2015). Actionable Gamification: Beyond Points, Badges, and Leaderboards. Octalysis Media.

Related Reading

• The Octalysis Framework. The 8-Core-Drive design surface for translating HBM diagnoses into Octalysis interventions.
• Sludge (Sunstein). The inverse of HBM’s Perceived Barriers: the friction that blocks beneficial action.
• The Behavioral Framework Library. The full library of behavioral and psychological frameworks every S-Tier designer should know.

WOULD YOU LIKE YU-KAI CHOU TO WORK WITH YOUR ORGANIZATION?