Population-level Directly Observed Therapy (DOT) program tracking — cohort enrollment, WHO outcome classification, and quality improvement against the 90% treatment success benchmark
National TB programs and the WHO do not evaluate treatment success one patient at a time — they track defined cohorts. A cohort is the group of all patients registered for treatment within a fixed reporting interval, usually a calendar quarter. Everyone enrolled in that window is followed forward to a standardized outcome, and the cohort's aggregate result is what appears in national and global TB reports.
A single patient's adherence journey — the day-to-day dosing, side effects, and directly observed therapy visits covered in the companion individual-patient DOT simulator — is clinically vital but tells a program almost nothing about system performance. Cohort analysis answers a different question: of everyone who started treatment in a given quarter, what fraction reached a good outcome?
This distinction matters operationally. A clinic can have excellent bedside adherence support for the patients it retains, yet still fail the population it serves if too many patients never make it back for follow-up visits at all. Cohort tracking captures both dimensions at once because it counts every enrollee, including those who disengage entirely.
WHO's cohort reporting convention groups patients by: (1) registration group — new case, relapse, treatment after failure, treatment after loss to follow-up, transferred in; and (2) drug-susceptibility category — drug-susceptible TB (6-month regimen) versus multidrug-resistant/rifampicin-resistant TB (longer, more complex regimens). Each group is evaluated against its own outcome cohort, because expected outcome distributions differ sharply between them.
At enrollment, programs capture a minimal but standardized dataset per patient: registration date, disease site (pulmonary/extrapulmonary), bacteriological confirmation status, HIV status, drug-susceptibility test result if available, and treatment regimen assigned. This record becomes the anchor for outcome assignment at the end of treatment (commonly 6 months later for drug-susceptible TB, up to 18–24 months for drug-resistant regimens).
Cohort size itself is a meaningful operational variable: a small rural clinic enrolling 20–30 patients per quarter faces very different statistical noise and program-design tradeoffs than a large urban DOT network enrolling several hundred. Small cohorts are more volatile quarter to quarter — a handful of lost-to-follow-up cases can swing the reported success rate by several percentage points — which is why WHO recommends aggregating sub-national data before drawing conclusions about program quality.
Every enrolled patient is ultimately closed out into exactly one of six internationally standardized outcome categories. This shared taxonomy — unchanged in its essentials since the 1990s DOTS strategy — is what allows a district health office, a national program, and the WHO Global TB Report to all speak the same language about what happened to a cohort.
Cured — bacteriologically confirmed TB at treatment start, smear- or culture-negative at the end of treatment with at least one previous negative result.
Treatment completed — patient completed the full treatment course, but without bacteriological confirmation of cure at the end (e.g., no end-of-treatment smear/culture was done or is unavailable). Clinically successful but not laboratory-confirmed.
Died — patient died for any reason during the course of treatment.
Treatment failed — treatment terminated or a permanent regimen change was needed because of a positive smear or culture at month 5 or later, or a bacteriologically confirmed multidrug-resistant strain detected during treatment.
Lost to follow-up — treatment was interrupted for a consecutive period defined by the program (commonly 2 months or more) without medical reason.
Not evaluated — no outcome was assigned, including patients transferred to another facility whose outcome was never reported back, and cases with unknown outcome.
"Treatment success" — the headline indicator every program is judged on — is defined as cured plus treatment completed.
Separating "cured" from "completed" preserves laboratory rigor for drug-susceptible pulmonary cases (where bacteriological confirmation is feasible) while still crediting programs for clinically successful courses in extrapulmonary or smear-negative disease, where end-of-treatment testing is often not applicable.
"Lost to follow-up" is deliberately distinguished from "not evaluated": the former means the program knows the patient disengaged, the latter means the program simply lost track of the record. Conflating them would hide data-quality problems inside what looks like a clinical outcome — a distinction that matters enormously for the quality-improvement response in Stage 5.
A cohort report that shows a high "not evaluated" fraction is usually a data-system problem, not a patient-behavior problem — transfers not reconciled, registers not closed out, or outcomes never entered. Programs distinguish this from true loss to follow-up before designing an intervention.
When a cohort's treatment success rate is low, the intuitive explanation — "patients weren't compliant" — is rarely the operative one at population scale. Decades of program evaluation data point instead to structural, program-level determinants: how treatment is delivered, what support surrounds the patient, and how strong the surrounding health system is.
Directly Observed Therapy infrastructure covers far more than the moment a dose is swallowed: it includes the density and staffing of DOT points, the reliability of drug supply chains down to the last mile, whether community health workers or digital tools extend coverage into rural or informal-settlement areas, and how quickly a missed visit triggers active tracing rather than passive record-keeping.
Cohorts served by programs with strong "case-holding" — proactive outreach the moment a patient misses a visit — consistently show lower loss-to-follow-up than programs relying on patients to self-initiate contact when problems arise.
TB disproportionately affects lower-income households, and the WHO estimates roughly 40–50% of TB-affected households face catastrophic costs — expenses exceeding 20% of annual household income — driven substantially by transport to clinics and lost wages from repeated visits, not drug costs (which are usually free under national programs).
Nutritional support matters clinically as well as operationally: undernutrition is both a TB risk factor and a driver of poor treatment tolerance, and food insecurity is a frequently cited reason patients disengage mid-course. Cash transfer, transport voucher, and nutritional supplementation programs have been associated with measurably higher cohort completion rates in multiple national program evaluations.
Stigma, housing instability, incarceration history, substance use, and migration all independently predict loss to follow-up at the cohort level, and their prevalence varies enormously by setting — meaning two programs with identical clinical protocols can post very different completion rates purely because of the populations and social conditions they serve.
Underlying all of this is general health-system strength: workforce density, electronic health record continuity across facilities (critical for the "transferred in / transferred out" reconciliation that prevents patients from becoming "not evaluated"), and the fiscal stability of the national TB program itself. Weak health systems amplify every other risk factor; strong ones buffer against them.
The WHO End TB Strategy, adopted by the World Health Assembly in 2014, sets a treatment success rate target of at least 90% for every reporting cohort — cured plus treatment completed, divided by all patients registered in the cohort. Programs are judged not against an individual patient's progress but against how their entire quarterly intake performed against this fixed line.
A benchmark comparison is only meaningful when the denominator is right: the target applies to cured + completed as a share of everyone registered, including patients who died, failed, defaulted, or were never evaluated — not just a share of patients who "finished" some version of care. This is why a cohort can look reasonably well-run on the ground (attentive staff, good adherence support for the patients still coming in) and still fall short of 90% if enough patients drop out of the denominator entirely.
Global averages sit close to but typically just under the 90% line for drug-susceptible TB cohorts, while multidrug-resistant and rifampicin-resistant TB cohorts — treated with longer, harder-to-tolerate regimens — commonly report success rates in the 55–65% range, a gap the End TB Strategy explicitly tracks separately.
Using a single fixed threshold, rather than ranking programs against each other, is a deliberate design choice: it gives every program — regardless of size, setting, or starting point — an absolute, unambiguous target to plan against, and it makes trend tracking over years directly comparable across the entire global TB reporting system.
Programs that clear 90% still monitor closely for early warning signs (a rising lost-to-follow-up trend, for instance) rather than treating benchmark achievement as a stopping point, since a single quarter's good result can mask a slower structural decline.
The End TB Strategy's 90% success target sits alongside two other 2035 milestones — a 90% reduction in TB deaths and an 80% reduction in incidence relative to 2015 — reflecting that treatment completion is one lever among several, but it is the lever most directly under a program's day-to-day operational control.
When a cohort's treatment success rate falls short of the 90% benchmark, the response is not to admonish individual patients but to run a structured, program-level quality-improvement cycle: identify which outcome category is driving the shortfall, investigate why, and deploy interventions matched to the actual root cause.
The first quality-improvement step is disaggregating the shortfall: is the gap to 90% mostly loss to follow-up, mostly "not evaluated" records (a data problem, not a patient problem), or a genuine rise in deaths or treatment failure that may signal drug-resistance or HIV-coinfection issues needing clinical rather than programmatic response? Each driver calls for a different fix, and misdiagnosing the cause wastes the QI cycle.
In most below-target cohorts globally, loss to follow-up is the single largest driver of the gap between measured performance and the 90% target — which is why programmatic attention concentrates heavily on retention rather than on the clinical regimen itself.
Enhanced patient support — transport reimbursement, nutritional supplementation, and social protection cash transfers — directly targets the catastrophic-cost pathway to disengagement identified in Stage 3.
Digital adherence technologies — including 99DOTS (medication sleeves with hidden codes revealed on dose-taking, reported by SMS) and video-observed therapy (VOT) — reduce the need for daily clinic attendance while preserving observation, particularly valuable for patients facing transport or workplace-schedule barriers to in-person DOT.
Community health worker (CHW) programs extend case-holding into the home and community, catching missed doses within days rather than waiting for a patient to reappear (or not) at the next scheduled visit — directly addressing the case-holding gap described in Stage 3.
Programs typically structure the response as a Plan–Do–Study–Act (PDSA) cycle run each reporting quarter: Plan the intervention against the identified root cause; Do — implement it in the next intake cohort; Study — compare the new cohort's outcome distribution against the benchmark and the prior cohort; Act — scale what worked, adjust what didn't, and re-run the cycle.
Because cohort outcomes are only knowable months after enrollment (up to 24 months for drug-resistant regimens), the feedback loop is inherently slow — which is exactly why getting the root-cause diagnosis right in step one matters: a misdirected intervention costs an entire reporting cycle before the program can tell whether it worked.
Programs that track lost-to-follow-up as a distinct, monitored indicator — rather than folding it into an undifferentiated "treatment not successful" bucket — consistently identify and correct completion problems faster, because the QI team knows exactly which lever to pull.