Patient Compliance

Abstract

Patient compliance is the extent to which a person's behavior (taking medication, attending appointments, following a diet) coincides with the recommendations of a health-care provider. Quantitative reviews place average non-adherence near a quarter of patients, and higher still for the long-term regimens of chronic disease, making the gap between what is prescribed and what is done one of medicine's largest and most consistent problems. Because that gap is weakly a matter of knowledge and strongly one of belief, memory, and competing demands, compliance belongs to the cognitive psychology of goal pursuit as much as to clinical practice. This article covers how compliance is defined and measured, why the field moved from compliance toward adherence and concordance, its determinants, consequences, and interventions. Three demonstrations model where adherence breaks down, the belief balance that predicts it, and the modest effect interventions achieve.

Keywords: medication adherence, concordance, necessity-concerns framework, adherence taxonomy, intention-behavior gap

Patient compliance is where the clinic's recommendation meets the patient's life. A prescription is an instruction issued at one moment for a behavior to be carried out repeatedly over days, months, or years, and the distance between the writing of that instruction and its faithful execution is filled with forgetting, doubt about whether the drug is needed, worry about its harms, cost, side effects, and the ordinary friction of a complicated life. The question that organizes the field is why so many patients who sought care, accepted a diagnosis, and left with a plan they endorsed nonetheless depart from it, and the answer has moved steadily away from blaming the patient's will and toward understanding non-compliance as a predictable product of beliefs, cognition, and circumstances that can be measured and, within limits, reshaped.

What Patient Compliance Is

Patient compliance, in the usage adopted by the United States National Library of Medicine, is the extent to which a patient follows the advice or prescription of a health-care provider. The oldest and most literal term in the field, compliance frames the clinician as the source of a correct instruction and the patient as the party who does or does not carry it out, and for that reason it has drawn sustained criticism as implying a one-way, obedience-based relationship. A comprehensive review of three decades of research traced how the field came to see that framing as both inaccurate and counterproductive, because it mislocates the cause of non-adherence in patient fault rather than in the interaction between the patient's beliefs and circumstances and the regimen they were asked to follow (Vermeire et al., 2001). The preferred modern term, adherence, keeps the same measurable behavior — the match between recommendation and action — while dropping the connotation of obedience; concordance, a still later term, refers specifically to the negotiation and agreement between prescriber and patient that precedes the behavior rather than to the behavior itself (Chakrabarti, 2014).

The size of the problem is what gives the terminology its weight. A quantitative review of fifty years of research found that, averaged across conditions and regimens, about a quarter of patients fail to adhere to medical recommendations, with wide variation by disease, by the complexity of the regimen, and by how adherence is measured (DiMatteo, 2004). An influential clinical review put the same point in operational terms for the physician: adherence to long-term therapy for chronic illness in developed countries averages only about half, so the problem is not a minority of difficult patients but a routine feature of care that any treatment plan must anticipate (Osterberg & Blaschke, 2005). Figure 1 shows the behavior as a sequence of stages, each of which a patient must complete for a prescription to translate into a correctly taken medicine.

Types of Patient Compliance

In the MeSH classification, Patient Compliance sits beneath the broader heading of Patient Acceptance of Health Care and is divided into two narrower descriptors, each naming a distinct compliance behavior rather than a different kind of patient. The two are orthogonal: they pick out different acts a patient is asked to perform — taking a medicine and keeping an appointment — so a single person can adhere to one and not the other, and a clinic can have a problem with one while the other is well managed. The division is an indexing classification, a way of filing the literature, not a theory of the patient; the same cognitive machinery of belief, memory, and competing demand runs beneath both. Both narrower descriptors below are live articles on this site.

Measuring and Defining Adherence

The central difficulty in measuring adherence is that the word once covered several different behaviors at once, blurring where a regimen actually broke down. A landmark consensus project resolved this by splitting adherence into three distinct, separately measurable phases: initiation, whether the patient takes the first dose at all; implementation, how closely the actual dosing over time matches what was prescribed; and persistence, how long the patient continues before stopping. This ABC taxonomy replaced a single blurred notion of compliance with a vocabulary precise enough to say which part of adherence a given patient or intervention affects (Vrijens et al., 2012). The distinction matters clinically because the remedies differ: a patient who never initiates needs a different response from one who persists for a year but misses a third of doses.

No single measurement method captures all three phases well, and each has characteristic errors. Pill counts and pharmacy refill records are cheap but can overstate adherence, because a collected or absent pill is not a swallowed one; electronic monitors that timestamp each opening of a container capture implementation in fine detail but are costly; self-report is easy but biased toward overstatement; and drug-level assays are objective but only sample a single moment (Osterberg & Blaschke, 2005). Because every method is imperfect in a different direction, the measured adherence rate depends heavily on how it was measured, which is one reason reported rates vary so widely. The first demonstration lets the reader set the three phases independently and see how they multiply into an overall adherence that no single phase reveals on its own.

Set each phase for a cohort of 1000 patients started on a chronic medicine. Because a patient must clear every phase, overall adherence is the product of the three, not their average, so losses compound down the chain.

90% × 80% × 60% = effective adherence 43.2% — 432 of 1000 fully adhering

At the default 90% / 80% / 60% the cohort reproduces the Table 2 baseline: 43.2%, or 432 fully adhering. Notice that the phase with the most room to improve returns the most, which is why locating where a regimen fails matters more than a general push to comply. Illustrative values within the ranges reported for long-term therapy.

Why Patients Depart From a Plan

The determinants of adherence have been catalogued repeatedly, and a review of systematic reviews sorted the hundreds of reported factors into a handful of families: characteristics of the patient, of the condition, of the therapy, of the patient's social and economic situation, and of the health system and its relationship with the patient. The same review found that no single factor dominates and that the factors interact, which is why interventions aimed at one cause in isolation tend to disappoint (Kardas et al., 2013). A scoping review of patients' own accounts added the texture that the quantitative reviews flatten: patients describe forgetting, fear of side effects, doubt about whether a drug is working, the burden of complex regimens, cost, and a sense of not being heard by their prescriber, reasons that are a mix of the practical and the cognitive rather than a simple failure of will (Kvarnström et al., 2021).

The most explanatory cognitive account is the necessity-concerns framework, which holds that whether a patient takes a long-term medicine is governed by the balance between two beliefs: how far they judge the medicine necessary for their health, and how much they are concerned about its potential harms. A meta-analytic review across many conditions found that stronger necessity beliefs and weaker concerns each independently predicted better adherence, and that the two combined — the necessity-minus-concerns differential — predicted adherence better than either alone (Horne et al., 2013). A second meta-analysis confirmed the pattern across diverse illnesses, establishing the belief balance as one of the most robust psychological predictors of medication-taking (Foot et al., 2016). This belief-based account has deeper roots in two classic models of health behavior: the health belief model, which frames action as a weighing of perceived threat against the perceived benefits and barriers of acting (Rosenstock, 1974), and Leventhal's common-sense model, in which patients act on their own lay representation of an illness — its identity, cause, timeline, consequences, and controllability — rather than on the clinician's model (Leventhal et al., 1992). The second demonstration lets the reader adjust a patient's necessity and concern beliefs and see where the balance places them and how it bears on predicted adherence.

Set how far a patient judges a long-term medicine necessary and how much they are concerned about its harms. Their position in the belief space predicts adherence: the difference between the two, not knowledge, is the proximal driver.

differential (necessity − concerns) +4 — Accepting, predicted adherence 80%

A patient who judges the medicine necessary and holds few concerns sits in the accepting quadrant and is most likely to adhere; one whose concerns outweigh a weak sense of necessity sits in the skeptical quadrant and least likely (Horne et al., 2013). The curve mapping the differential to adherence is illustrative, not a fitted model; the point is that the balance of beliefs, not information, moves adherence.

Consequences

The consequences of non-adherence are measured first in health. A meta-analysis pooling decades of studies found that good adherence was associated with a substantial improvement in treatment outcomes — the odds of a good outcome were roughly three times higher in adherent than in non-adherent patients — a relationship that held across conditions and that establishes adherence as a determinant of outcome in its own right, not merely a marker of healthier patients (DiMatteo et al., 2002). The clinical cost is concrete: an unadhered regimen is an uncontrolled blood pressure, an unsuppressed infection, a transplant rejected, a relapse that a taken medicine would have prevented.

The economic cost compounds the clinical one. A broad clinical review characterized non-adherence as a problem that drives avoidable hospitalizations, emergency visits, and premature death, and estimated the associated waste in the hundreds of billions of dollars annually in the United States alone, costs that fall on payers and the whole system rather than only the individual patient (Kleinsinger, 2018). An earlier overview had already framed non-adherence as a problem whose scale the health system systematically underestimated, precisely because its costs are diffuse and appear downstream as the failure of treatments that were, on paper, prescribed correctly (Brown & Bussell, 2011). The combination of worse outcomes for the patient and large avoidable expenditure for the system is what justifies the sustained effort to raise adherence rather than treating it as an unalterable fact of care.

Interventions

The sobering finding of the intervention literature is how modest and effortful improvement is. The authoritative Cochrane systematic review of interventions to enhance medication adherence concluded that even the most effective interventions were complex and labour-intensive, that their effects on both adherence and clinical outcomes were inconsistent and generally small, and that no single approach could be recommended across the board — a conclusion notable for its restraint given the size of the problem (Nieuwlaat et al., 2014). A later narrative review for clinicians reached a compatible verdict: interventions that work tend to be multi-component, tailored to the specific barriers a patient faces, and sustained rather than one-off, so improving adherence resembles ongoing management more than a single fix (Kini & Ho, 2018).

Where the effects can be pooled, they are real but small. A meta-analysis of 771 intervention trials found a statistically reliable but modest average improvement in adherence, confirming both that interventions do something and that the something is limited (Conn & Ruppar, 2017). One lever with better-than-average support is the clinician relationship itself: a meta-analysis found that patients of physicians who communicate well are significantly more adherent, and estimated that training physicians in communication measurably raises adherence — evidence that concordance, the agreement reached in the consultation, is not merely a softer word for compliance but a modifiable cause of it (Haskard Zolnierek & DiMatteo, 2009). The third demonstration models a cohort of patients against a clinical adherence threshold and lets the reader apply an intervention of a chosen effect size, making concrete why a modest average effect moves only a modest number of patients across the line.

Twenty patients carry baseline adherence values spread from low to high; the clinical target is 80% of doses covered. Apply an intervention effect and watch how many patients cross the threshold. A square turns green when that patient reaches 80%.

6 of 20 at or above the 80% threshold

With no intervention, 6 of 20 patients clear the threshold. A modest effect lifts only a few more across it, because most of the gain accrues to patients already near the line while those far below remain short — the concrete reason the intervention literature reports real but small average effects (Nieuwlaat et al., 2014; Conn & Ruppar, 2017). The cohort is a fixed illustrative spread, not trial data.

Worked Example

The three-phase taxonomy can be made quantitative with a simple illustrative model. Because a patient must clear every phase for a regimen to succeed, overall adherence is the product of the three phase probabilities: effective adherence = initiation × implementation × persistence. Model initiation as the fraction of patients who take the first dose, implementation as the fraction of prescribed doses actually taken among those who start, and persistence as the fraction still on therapy at the end of the period. The values below are illustrative, chosen to sit within the ranges the literature reports for long-term therapy, not figures drawn from any one study; their purpose is to show how the arithmetic behaves and why a regimen that looks adequate at each separate stage can still fail overall.

Suppose a clinic starts 1,000 patients on a chronic medication. Table 2 works four scenarios through the model, holding two phases and improving one at a time.

The arithmetic makes two things concrete. First, because the phases multiply, the overall figure is far lower than any single phase suggests: a regimen that looks tolerable at 90%, 80%, and 60% leaves fewer than half of patients fully adhering, because the losses compound rather than average. Second, the phase with the most room to improve yields the most: lifting persistence from 0.60 to 0.85 recovers 180 patients, far more than perfecting initiation, which is why identifying where in the sequence a regimen fails matters more than exhorting patients to comply in general. The product rule is also why a single intervention rarely transforms the headline rate — it usually touches only one factor in a chain of three.

Discussion

The study of compliance has followed a clear arc: away from a model that located the fault in the patient's character and toward one that locates the behavior in a web of beliefs, cognitive failures, and practical barriers that systems can measure and partly correct. The early framing treated non-compliance as the patient's failure to obey a correct instruction, a framing that both misdescribed the behavior and predicted it poorly because it looked for the cause in the patient's will rather than in the interaction between the patient's beliefs and the regimen (Vermeire et al., 2001). The vocabulary itself carried the shift: from compliance, with its implication of obedience, to adherence, which keeps the behavior but drops the blame, to concordance, which names the prior agreement that makes adherence likely (Chakrabarti, 2014).

What follows from this relocation is a strategy of understanding and targeting rather than exhortation. Because the determinants are multiple and interacting, and because patients act on their own beliefs about necessity and risk, the productive response is to identify the specific barrier a given patient faces — a belief, a phase of the regimen, a practical obstacle — and address that, rather than to apply a generic push toward compliance (Kardas et al., 2013) (Horne et al., 2013). The gap between the clarity of this diagnosis and the modest size of the available remedies is the field's defining tension (Nieuwlaat et al., 2014): the causes of non-adherence are now well understood, yet improving it remains slow, complex work, because there is no single lever when there is no single cause.

Cognitive Implications

Non-adherence is, at its psychological core, an instance of the intention-behavior gap operating repeatedly over time. A patient who accepts a prescription has formed an intention to take it, yet a formed intention is a weak predictor of the eventual act, and the weakness is compounded for a medication regimen because the intention must be re-enacted every day for months or years, surviving forgetting, doubt, and competing demands at each repetition. This is why non-adherence is so weakly a knowledge problem: the patient rarely lacks the information or the initial intention but fails to convert it into the recurring act. The necessity-concerns framework describes the belief side of this gap — a patient whose concerns come to outweigh perceived necessity revises the intention itself — while the common-sense model describes how patients continually reinterpret their illness and adjust their behavior to a lay representation that may diverge from the clinical one (Leventhal et al., 1992) (Horne et al., 2013).

The most directly applicable cognitive remedy is the implementation intention, a concrete if-then plan that specifies when, where, and how an intended action will be performed and so delegates its initiation to an anticipated situational cue. Forming such a plan has been shown to close a substantial part of the intention-behavior gap across many behaviors, because the pre-decided cue triggers the action automatically rather than requiring the person to remember and decide anew each time (Gollwitzer, 1999). Linking a daily dose to an existing routine — a specific meal, the act of brushing teeth — is an implementation intention in clinical form, and it addresses precisely the implementation and persistence phases where everyday adherence erodes. Patient compliance thus sits squarely within the cognitive psychology of goal pursuit, and the interventions that work are those that help a patient bridge the recurring gap between a goal they genuinely hold and the specific future acts that would fulfil it.

Current Directions

The most active front is the measurement and targeting of adherence at the level of its three phases. The ABC taxonomy made it possible to ask which phase an intervention affects, and newer work uses electronic monitoring, pharmacy data, and digital tools to detect in near-real time whether a patient has failed to initiate, is implementing poorly, or has stopped, so that support can be directed to the specific failure rather than applied generically (Vrijens et al., 2012) (Kvarnström et al., 2021). The promise is precision; the recurring hazard, well recognized in the adherence literature, is that richer monitoring can slide into surveillance and erode the concordance it was meant to support if it is used to police patients rather than to understand and help them.

A second direction refines interventions toward the multi-component, tailored, and sustained forms the evidence favours, and away from the one-off reminders that the Cochrane evidence found wanting (Nieuwlaat et al., 2014) (Kini & Ho, 2018). The convergence of the belief-based and the phase-based strands points toward a model of adherence support that identifies a patient's necessity-concerns balance and the phase at which their regimen is failing, then matches the response to both — strengthening a necessity belief here, addressing a concern there, embedding an implementation intention to protect daily dosing — and reserves intensive, sustained contact for the patients whose non-adherence signals a genuine clinical or social need rather than a lapse of memory (Conn & Ruppar, 2017).

Common Misconceptions

Non-adherent patients do not care about their health.

Much non-adherence is a failure to convert a genuine intention into a recurring action, or the product of a reasoned worry about the medicine, rather than indifference: adherence is governed by the balance a patient strikes between a drug's perceived necessity and their concerns about it, so a patient can care deeply and still not take it (Horne et al., 2013).

Better-informed patients automatically adhere.

Adherence is only weakly a knowledge problem; patients act on their own lay representation of the illness and on their beliefs about the medicine, so simply supplying more information rarely changes behavior unless it shifts those beliefs (Leventhal et al., 1992) (Kvarnström et al., 2021).

Compliance and adherence and concordance mean the same thing.

They mark a deliberate progression: compliance implies the patient obeying a clinician's instruction, adherence keeps the same measurable behavior but drops the obedience framing, and concordance refers to the prior agreement negotiated between prescriber and patient rather than to the behavior at all (Chakrabarti, 2014).

There is an intervention that reliably fixes non-adherence.

The authoritative systematic review found that even the most effective interventions are complex and labour-intensive and that their effects are inconsistent and generally small; improving adherence is sustained, tailored work, not a single reliable fix (Nieuwlaat et al., 2014) (Kini & Ho, 2018).

Glossary

Adherence.

The extent to which a patient's behavior matches agreed recommendations; the preferred modern term for compliance, keeping the behavior while dropping the implication of obedience.

Beliefs about Medicines Questionnaire (BMQ).

A validated instrument that measures a patient's perceived necessity of a medicine and their concerns about it, operationalizing the necessity-concerns framework.

Common-sense model.

Leventhal's account in which patients act on a lay representation of their illness — its identity, cause, timeline, consequences, and controllability — rather than on the clinician's model.

Compliance.

The degree to which a patient follows a provider's advice or prescription; the oldest term in the field, criticized for implying a one-way, obedience-based relationship.

Concordance.

The agreement reached between prescriber and patient about whether and how to treat; a prior negotiation rather than the medication-taking behavior itself.

Health belief model.

A framework in which health action follows from a weighing of perceived threat against the perceived benefits of and barriers to acting.

Implementation intention.

A concrete if-then plan specifying when, where, and how an action will be performed, used to protect daily dosing by linking it to a situational cue.

Implementation.

The phase of adherence describing how closely a patient's actual dosing, once started, matches the prescribed regimen.

Initiation.

The phase of adherence describing whether a patient takes the first dose of a newly prescribed medicine at all.

Intention-behavior gap.

The discrepancy between what a person intends and what they do; non-adherence is a clinical instance of it, re-enacted at every dose.

Necessity-concerns framework.

The account holding that adherence is governed by the balance between how necessary a patient judges a medicine and how concerned they are about its harms.

Non-adherence.

Departure from an agreed regimen, whether by never starting, by dosing incorrectly, or by stopping early; averaged across conditions it affects roughly a quarter of patients.

Persistence.

The phase of adherence describing how long a patient continues a medicine before discontinuing it.

Proportion of days covered (PDC).

A pharmacy-based adherence measure giving the fraction of days in a period for which a patient had the medication available, commonly judged against an 80% threshold.

Three-phase (ABC) taxonomy.

The consensus division of adherence into initiation, implementation, and persistence, which allows a regimen's failure to be located in a specific phase.

Key Researchers

Peter M. Gollwitzer

(New York University and University of Konstanz). Developed the theory of implementation intentions, which explains how a concrete if-then plan closes the recurring gap between intending to take a medicine and taking it. ORCID · Wikipedia · Scholar

R. Brian Haynes

(McMaster University). A founder of evidence-based medicine and senior author of the Cochrane review on interventions to enhance medication adherence, whose central finding is how modest and labour-intensive even the effective interventions are. ORCID · Faculty · Wikipedia

Rob Horne

(University College London). Formalized the necessity-concerns framework and the Beliefs about Medicines Questionnaire, establishing the balance between a patient's perceived necessity and concerns as a proximal driver of adherence. Faculty · Wikipedia

Howard Leventhal

(Rutgers University). Developed the common-sense model of self-regulation, which explains why patients act on their own representation of an illness rather than the clinician's, a central reason sensible-seeming regimens are abandoned. Faculty · Institute

Bernard Vrijens

(AARDEX Group and University of Liège). A biostatistician who led the consensus that split adherence into initiation, implementation, and persistence — the ABC taxonomy now standard in the field. Repository

Frequently Asked Questions

What is patient compliance?

Patient compliance is the extent to which a person's behavior (taking medication, attending appointments, following a diet or lifestyle change) matches the recommendations of a health-care provider. The term is increasingly replaced by adherence to avoid the implication that the patient is simply obeying an instruction (Vermeire et al., 2001).

What is the difference between compliance, adherence, and concordance?

Compliance implies the patient following a clinician's instruction; adherence keeps the same measurable behavior but drops the obedience framing; and concordance refers to the agreement negotiated between prescriber and patient before the behavior, not to the behavior itself (Chakrabarti, 2014).

How common is non-adherence?

Averaged across conditions, about a quarter of patients fail to adhere to recommendations, and adherence to long-term therapy for chronic disease averages only around half, so non-adherence is a routine feature of care rather than a minority problem (DiMatteo, 2004) (Osterberg & Blaschke, 2005).

Why do patients not take medicines they agreed to?

Whether a patient takes a long-term medicine is governed largely by the balance between how necessary they judge it and how concerned they are about its harms, together with forgetting, cost, side effects, and regimen complexity, a mix of belief and circumstance rather than a lack of will (Horne et al., 2013) (Kvarnström et al., 2021).

What are the three phases of adherence?

A consensus taxonomy divides adherence into initiation (taking the first dose), implementation (dosing correctly over time), and persistence (continuing rather than stopping early); a regimen can fail at any one, and the remedy differs for each (Vrijens et al., 2012).

Does non-adherence actually affect outcomes?

Yes. A meta-analysis found the odds of a good treatment outcome roughly three times higher in adherent than in non-adherent patients, establishing adherence as a determinant of outcome and not merely a marker of healthier patients (DiMatteo et al., 2002).

Can anything reliably improve adherence?

Interventions help, but the authoritative evidence is sobering: the most effective are complex and labour-intensive, their average effect is modest, and no single approach works across the board, so improving adherence is sustained, tailored work (Nieuwlaat et al., 2014) (Conn & Ruppar, 2017).

Does the clinician relationship matter?

Considerably. A meta-analysis found that patients of physicians who communicate well are significantly more adherent, and that training physicians in communication measurably raises adherence, evidence that the agreement reached in the consultation is a modifiable cause of adherence (Haskard Zolnierek & DiMatteo, 2009).

References