Tuberculosis Adherence Research Is Measuring Different Things Under the Same Name

Tuberculosis remains one of the deadliest infectious diseases on the planet, and curing it depends on something deceptively simple: patients taking their medicines for months on end. Yet a sweeping new analysis suggests that the research community charged with understanding why people do or do not complete treatment has been talking past itself for years. A scoping review published in PLOS Global Public Health by João Pedro Ramos and colleagues at the Institute of Public Health of the University of Porto examined how observational studies define and measure tuberculosis treatment adherence, and the verdict is stark: the field is riddled with construct conflation, meaning the same word describes fundamentally different behaviours from one study to the next.

The stakes of this ambiguity are far from academic. When a patient interrupts tuberculosis treatment, the consequences cascade outward: treatment failure, relapse, continued transmission of the bacterium, and, most dangerously, the emergence of drug-resistant strains that are far costlier and harder to cure. Poor adherence worsens individual prognosis and undermines population-level control of the disease, imposing heavy burdens on health systems, particularly in low- and middle-income countries where tuberculosis remains the infectious disease at greatest risk of escalation. If researchers cannot agree on what adherence actually is, the evidence base for interventions designed to support it rests on shifting sand.

The terminology itself has a complicated history. Early research spoke of compliance, a word that carried a paternalistic undertone, framing patients as subjects who ought to obey medical orders and implicitly assigning blame when behaviour diverged from clinical instruction. The World Health Organization later championed adherence, a term intended to soften that blame and imply agreement between patient and provider. But, as the Porto team notes, this linguistic shift was never accompanied by a clearly articulated theoretical framework. The result is that compliance and adherence are now used interchangeably across much of the literature, without any genuine conceptual distinction between them.

Contemporary theory has moved well beyond this vocabulary problem. Adherence is now understood as a complex, multidimensional and dynamic phenomenon, shaped not only by individual behaviour but by socioeconomic conditions, health system factors, disease characteristics and the nature of the treatment itself. Modern taxonomies distinguish between initiation, the point at which a prescribed regimen is first taken; implementation, the extent to which actual dosing matches the prescription over time; and persistence, the length of time before premature discontinuation. In the tuberculosis field specifically, concept analyses describe adherence as a fluctuating spectrum, with patients moving between more and less adherent states as resources, relationships and circumstances change. The question the review posed was whether observational research actually measures that construct.

To find out, the researchers searched eight bibliographic databases from inception to February 2026, covering Web of Science, SCOPUS, PsycInfo, Sociology Source Ultimate, Medline, EMBASE, LILACS and CINAHL. From more than 60,000 records identified, deduplication and multi-stage screening, with inter-reviewer agreement formally assessed using Cohen’s Kappa, yielded 103 observational studies of drug-susceptible tuberculosis in adults, drawn from 32 countries. The African Region contributed the largest share of studies, followed by the Americas and the Western Pacific. The team then applied an inductive content analysis, mapping the labels authors used onto the constructs their measurement strategies implicitly presupposed, and generating alluvial plots to visualise the pathways from label to instrument to operational definition.

The findings reveal a field where definition is the exception rather than the rule. Sixty-eight percent of the included studies provided no explicit definition of adherence at all, allowing operational criteria to serve as de facto definitions. Another 15.5 percent cited external definitions, most often from the WHO or national tuberculosis programmes, but then failed to distinguish between the conceptual level and the operational one, describing adherence as concordance with treatment while analysing programmatic indicators. Only a small subset, 8.7 percent, acknowledged adherence as a multifactorial and complex phenomenon involving behavioural, cognitive or psychosocial processes such as effort, intention and commitment.

The construct mapping exposed just how far labels drift from what is actually measured. Among studies using the word adherence, the most frequent implied construct was behavioural execution, at 39.8 percent, but a substantial share instead operationalised it as dose ingestion, at 22.3 percent, treatment persistence, at 23.3 percent, or programmatic compliance. The label compliance mapped to a similarly scattered range of constructs. By contrast, terms such as default, abandonment and loss to follow-up clustered more tightly around treatment persistence, typically defined as treatment interruption exceeding one or two months. In other words, studies that appear statistically comparable may be measuring entirely different phenomena, a situation the authors describe as conceptually incommensurable even when the numbers look commensurable.

Measurement practices were equally heterogeneous. Nearly two-thirds of studies relied on clinician assessment or self-report, most commonly through validated questionnaires such as the Morisky Medication Adherence Scale, the Medication Adherence Report Scale, or study-specific instruments. Yet even multidimensional scales were routinely collapsed: 19.4 percent of studies dichotomised their outcomes using thresholds, 17.5 percent retained ordinal categories, and only two analysed adherence as a continuous variable. Pill counts and attendance at directly observed therapy visits featured in a further group of studies, with binary definitions classifying patients as non-adherent if a single dose was missed or if intake fell below cut-offs ranging from 75 to 95 percent. A substantial minority mined secondary data from tuberculosis registers and surveillance systems, where treatment outcome categories stood in for behaviour, and pharmacy dispensing records yielded proportion-of-days-covered indicators. Strikingly, only five studies employed direct measures, all using urine drug metabolite testing to detect recent ingestion, and even these provided snapshot assessments rather than longitudinal pictures.

The authors argue that these patterns reflect an unresolved ontological ambiguity rather than mere measurement variation. When WHO definitions were cited, their application frequently diverged from the conceptual breadth implied, and the conceptual ambitions articulated in study introductions were often diluted at the point of analysis. The shift from compliance to adherence, they suggest, has been largely rhetorical and linguistic rather than methodological. The problem is compounded by study design: more than 70 percent of the included studies were cross-sectional, reinforcing a static representation of what is in reality a dynamic, months-long process. Dichotomisation strips away the behavioural, relational and structural dimensions captured by the instruments, reconstituting adherence as a stable individual trait rather than a trajectory, and blurring the distinction between behaviour and programmatic outcome.

The review does carry limitations worth noting. As a scoping review, it characterised conceptual and measurement practices rather than the strength of empirical associations, and assigning implied constructs inevitably involved analytic judgement, though ambiguous cases were conservatively retained as unclear. The scope was restricted to observational studies of drug-susceptible tuberculosis in adults, so practices in drug-resistant or paediatric contexts, and in intervention trials or digital adherence technology research, may differ. Grey literature was excluded, and database indexing may have under-captured research from some regions. Still, the authors’ prescription is clear: studies should explicitly define the dimension of treatment-taking behaviour under investigation; analytic approaches should preserve the temporal and multidimensional nature of adherence rather than reducing it to binary scales; and the WHO and national programmes should standardise reporting by requiring explicit distinction between conceptual definitions, measurement strategies and analytic classifications. Without such changes, they warn, people may be labelled non-adherent on the basis of wildly heterogeneous processes, from an occasional missed dose to structural disruption of their lives, and interventions may be misdirected away from the social and structural mechanisms that truly shape whether tuberculosis treatment succeeds.

Subject of Research: Measurement and definition of tuberculosis treatment adherence in observational research

Article Title: Measurement of Tuberculosis treatment adherence: A methodological scoping review of observational studies

Article References: Measurement of Tuberculosis treatment adherence: A methodological scoping review of observational studies. (n.d.). https://doi.org/10.1371/journal.pgph.0007422

Image Credits: AI Generated

DOI: 10.1371/journal.pgph.0007422

Keywords: tuberculosis, treatment adherence, scoping review, measurement methods, construct validity, public health, drug-susceptible TB, self-report, treatment persistence, WHO, observational studies, PLOS Global Public Health