Root-cause diagnosis of medication non-adherence using the WHO five-dimension model and the Necessity-Concerns Framework
Root-cause diagnosis begins not with a conversation but with data. Pharmacy claims and refill records generate a passive, continuous adherence signal that can flag a problem weeks before it manifests clinically — long before a patient volunteers "I haven't been taking it."
Adherence measurement in real-world (non-trial) settings almost always relies on pharmacy claims rather than pill counts or self-report, because claims data is continuous, objective, and already collected for billing.
Proportion of Days Covered (PDC): • PDC = (days in the measurement period "covered" by a filled supply) / (total days in period) • Preferred over Medication Possession Ratio (MPR) because PDC caps at 100% and correctly handles overlapping fills / early refills • PMA (Pharmacy Quality Alliance) and CMS Star Ratings both use PDC ≥80% as the adherence threshold for statin, RAS-antagonist, and oral diabetes medication classes • PDC <80% is associated with measurably worse outcomes: for statins, each 10-point PDC drop below 80% is associated with roughly a 5–10% relative increase in cardiovascular event risk in observational cohorts
Refill-gap detection: • A gap >7 days between the expected refill date (based on days-supply) and the actual fill date is a simple, high-sensitivity trigger for outreach • Gaps >30 days are classified as a "treatment discontinuation" event rather than intermittent non-adherence, and route to a different workflow • Claims-based triggers have high sensitivity but modest specificity — roughly 15–25% of flagged gaps reflect stockpiling, hospitalization (inpatient med not billed to the same claim), or a recent prescriber-directed dose change rather than true non-adherence, which is why the flag becomes the *entry point* to the tree, not the diagnosis itself
Why detection alone is insufficient: Knowing that adherence has dropped tells a care team nothing about mechanism. Two patients with an identical PDC of 62% can have entirely different root causes — one cannot afford the copay, another simply forgets a twice-daily dose. Applying the same intervention (e.g., an automated refill reminder) to both wastes effort on the first patient and may fully resolve the second. This is precisely the failure mode the WHO framework and its associated screening tools are designed to prevent: matching intervention to mechanism, not to the surface symptom.
Before drilling into the five WHO dimensions, every diagnostic pathway for non-adherence starts with a single, high-yield question: is this a capacity problem or a motivation problem? Horne's Necessity-Concerns Framework and the Medication Adherence Report Scale (MARS) operationalize this split.
Rob Horne's Necessity-Concerns Framework (NCF), developed from the Common-Sense Self-Regulation Model, reframes adherence as the outcome of an implicit cost-benefit calculation patients run on every medication:
• Necessity beliefs: "How much do I personally need this medicine to stay well?" — measured via the Beliefs about Medicines Questionnaire (BMQ)-Specific Necessity subscale • Concerns beliefs: "How worried am I about this medicine's long-term effects, dependence, or unpredictability?" — measured via BMQ-Specific Concerns • Necessity-Concerns Differential (NCD) = Necessity score − Concerns score; low or negative NCD predicts intentional non-adherence with meta-analytic odds ratios of roughly 2–3
The MARS-5 (Medication Adherence Report Scale, 5-item) is a short, validated self-report instrument that separates: • Unintentional items — "I forget to take it," "I miss a dose" — capacity-limited, driven by memory, routine disruption, complexity • Intentional items — "I decide to miss a dose," "I take less than instructed" — motivation-limited, driven by beliefs, side-effect experience, cost
Why the split matters clinically: An adherence aid that only addresses forgetting — pillboxes, blister packs, SMS reminders — has near-zero effect on a patient who has intentionally decided the medicine is not worth taking. Conversely, motivational interviewing aimed at "why this medicine matters" is inefficient for a patient who already believes in the treatment but has a chaotic morning routine. Because roughly a third of non-adherent patients screen positive on *both* dimensions simultaneously, the tree treats intentionality as a weighting signal that shapes which of the five WHO dimensions is explored first, rather than a strict either/or gate.
The World Health Organization's "Adherence to Long-Term Therapies: Evidence for Action" (2003) remains the reference framework for adherence root-cause work. It insists that non-adherence is never purely a patient failing — it is the product of five interacting dimensions, only one of which the patient directly controls.
1. Social and economic factors: • Cost/copay burden, insurance gaps, transportation to pharmacy, health literacy, housing instability, competing demands (caregiving, unstable work hours) • Kaiser Family Foundation and CDC survey data: roughly 1 in 12 adults report skipping or rationing doses due to cost in a given year; the figure rises sharply for the uninsured and for high-deductible plans
2. Condition-related factors: • Severity of symptoms, rate of progression, availability of effective treatments, presence/absence of comorbid depression • Asymptomatic conditions (hypertension, hyperlipidemia, early HIV) are consistently associated with the lowest adherence — there is no felt penalty for skipping a dose
3. Therapy-related factors: • Regimen complexity (pills/day, dosing frequency, dietary restrictions), duration of treatment, side-effect burden, speed of benefit onset • The Medication Regimen Complexity Index (MRCI) quantifies this; each additional daily dosing event is associated with a measurable adherence decline — once-daily regimens average ~79% adherence vs. ~50% for regimens with ≥4 daily doses (Claxton et al., 2001 meta-analysis)
4. Patient-related factors: • Beliefs about necessity and concerns (NCF/BMQ), health literacy, self-efficacy, forgetfulness, substance use, psychological distress • WHO explicitly cautions that patient-related factors are frequently over-blamed by clinicians despite explaining a minority of total adherence variance
5. Health-system and healthcare-team-related factors: • Quality of the patient-provider relationship, clarity of communication, visit length, continuity of care, pharmacy access, prior-authorization friction, care fragmentation across multiple prescribers • Poor communication about why a medicine was prescribed is one of the strongest predictors of early discontinuation in claims-based discontinuation studies
The WHO report's central and most-cited claim is that patient-related factors — the dimension clinicians instinctively blame — account for less than 10% of the variance in adherence outcomes. The other four dimensions, largely outside the patient's direct control, dominate. This is why the decision tree deliberately walks through all five dimensions rather than defaulting to "the patient is non-compliant."
Classifying the dominant WHO dimension is necessary but not sufficient — "therapy-related factors" is still too broad to act on. The final branch of diagnosis narrows to a single, specific, addressable cause using structured interview technique, illustrated here through five representative patient cases.
Once a dimension is identified as dominant (via screener responses, claims patterns, and brief structured interview), a "5 Whys"-style drilldown converts a broad category into an actionable leaf node. Five illustrative branches, one per dimension:
• Social/Economic → Copay Burden: patient's out-of-pocket cost rose after a formulary tier change; refill gaps cluster precisely around the monthly billing cycle • Condition-Related → Asymptomatic Hypertension: patient reports "I feel fine, so I stop when I run low and don't rush the refill" — a classic asymptomatic-condition pattern • Therapy-Related → Regimen Complexity: patient is on 14 pills/day across 6 medications with three different dosing windows — MRCI score in the highest decile • Patient-Related → Low Necessity Belief: BMQ-Specific shows Necessity score below the Concerns score (negative NCD) — patient doubts the medicine is doing anything • Health-System-Related → Poor Provider Communication: patient never received a clear explanation of what the medication treats or what happens if stopped; discontinued after a change in prescriber
Each leaf is deliberately concrete enough to map onto exactly one evidence-based intervention family — which is the entire point of running the tree instead of applying a generic "improve adherence" instruction.
The 2014 Cochrane systematic review by Nieuwlaat and colleagues remains the largest synthesis of adherence-intervention trials — 182 RCTs — and its central finding is sobering: most interventions have small, inconsistent effects, and effect size depends heavily on whether the intervention actually targets the patient's specific barrier.
The Cochrane review's most actionable conclusion is not "adherence interventions work" (effects are modest and heterogeneous overall) but that matching intervention type to barrier type consistently outperforms one-size-fits-all approaches:
• Copay burden → Patient Assistance Programs (PAPs), $4 generic substitution, formulary tier appeal, 90-day mail-order fills to reduce per-fill friction — trials of PAP enrollment report adherence (PDC) increases in the range of 15–25 percentage points among enrolled patients • Asymptomatic condition → Motivational interviewing anchored in perceived risk communication, home BP self-monitoring feedback loops, teach-back on silent-disease pathophysiology • Regimen complexity → Simplification: switching to once-daily dosing or a fixed-dose combination pill; the single most reliably effective structural intervention across trials, because it removes the barrier rather than compensating for it • Low necessity belief / high concerns → Necessity-Concerns counseling using BMQ feedback, addressing specific misconceptions and side-effect fears directly (not generic education) • Health-system communication gaps → Structured care-coordination follow-up calls within 7–14 days of a new prescription or prescriber change, teach-back confirmation of indication and stop-risk
Critically, interventions that ignore the diagnosed dimension underperform: reminder packaging given to an intentionally non-adherent, cost-burdened patient shows near-zero incremental effect in subgroup analyses, because the barrier was never memory in the first place.
A root-cause diagnostic tree is only useful if it is closed by measurement. Ninety days after the matched intervention, PDC is recalculated from the same claims pipeline that generated the original flag, and the observed lift is benchmarked against pooled trial effect sizes.
The final stage closes the diagnostic loop rather than assuming success. Three outcomes are possible on re-measurement:
1. PDC rises to ≥80% and holds — the intervention resolved the diagnosed barrier; the patient exits the active-monitoring pathway but remains subject to routine claims surveillance 2. PDC rises partially but stays <80% — common when a second, undiagnosed barrier co-exists (e.g., cost was resolved but an unaddressed side-effect concern remains); this routes the case back into the tree for a second branch evaluation rather than repeating the same intervention 3. PDC does not improve — signals either a misclassification (wrong dimension/root cause identified) or an intervention implementation failure (e.g., patient never actually enrolled in the assistance program); triggers a fidelity check before re-diagnosis
Across the five illustrative case types modeled here, matched interventions produce adherence lifts in the 9–19 percentage-point range, consistent with the upper end of Cochrane-reviewed effect sizes for well-targeted interventions — substantially larger than the roughly 2-point average lift seen when interventions are applied without a preceding root-cause diagnosis. Durability is the remaining challenge: only about 60% of interventions that succeed at 90 days retain a meaningful effect at 12 months, which is why most adherence programs now treat the decision tree as a recurring process rather than a one-time diagnostic event.
The single largest determinant of whether an adherence intervention works is not its intrinsic strength but whether it targets the patient's actual, diagnosed barrier. A weak intervention correctly matched routinely outperforms a strong intervention misapplied — the entire rationale for running the WHO five-dimension decision tree before acting.