Medication Adherence

Medication Adherence Phases, Dispensing Measures and Drug Costs in Economic Models

Medication adherence describes whether a patient who has been prescribed a medicine collects it, takes the doses at the intended intervals and keeps taking it for as long as treatment should last. In health economics it links the effect a drug shows in trials to the health gain and drug spending it produces in routine care. It is the medicine-specific case of adherence, which also covers diet and lifestyle advice. Analysts mostly meet it through dispensing records and model assumptions, and those choices decide how much of a drug's value a model credits and how much of its cost it charges.

Two definitions in current use

Two published definitions shape how studies measure the concept. Cramer and colleagues (2008) treated medication compliance, with adherence as a synonym, and medication persistence as two different constructs. Compliance in that sense is conformity with the timing, dosage and frequency of the regimen, persistence is the time between initiation and discontinuation of therapy, and no overarching term combines the two.

The ABC taxonomy of Vrijens and colleagues (2012) takes the wider view now common in adherence research. It defines adherence to medications as the process by which patients take their medications as prescribed, in three phases. Initiation is the first dose, implementation is the extent to which actual dosing corresponds to the prescribed regimen, and discontinuation ends treatment, with persistence the time between initiation and the last dose. The authors regard patient compliance as synonymous with medication adherence. A study reporting "adherence" may therefore describe dose taking alone or the whole course of treatment.

NICE guidance on medicines adherence (clinical guideline CG76, 2009) adds the clinical frame. It asks professionals to accept that a patient with capacity has the right to decide not to take a medicine, and to discuss whether non-adherence comes from beliefs and concerns (intentional) or practical problems (unintentional). A dispensing gap can therefore reflect an informed decision as well as a barrier.

What each phase looks like in real data

The phases fail at different rates and in different ways. Fischer and colleagues linked 195,930 electronic prescriptions to pharmacy claims and found that 78% were filled, falling to 72% for new medicines. Primary non-adherence for newly prescribed medicines was 28.4% for hypertension, 28.2% for hyperlipidaemia and 31.4% for diabetes, and the authors note that earlier studies of non-adherence failed to capture these prescriptions.

Electronic monitoring shows what happens after the first dose. In the dosing histories of 4,783 patients with hypertension taking a once-daily drug, Vrijens and colleagues (2008) found that about half had stopped within one year. On any day, patients still engaged with the regimen omitted about 10% of scheduled doses, and 48% took a drug holiday of more than 78 hours at least once a year.

The ABC taxonomy therefore treats initiation and persistence as time-to-event data for standard survival analysis, such as Kaplan-Meier curves or the proportion persistent at a stated time. Implementation is quantified by comparing prescribed and actual dosing histories, for example as the proportion of prescribed drug taken. The authors conclude that these differences preclude a single, quantitatively useful parameter covering all three phases.

Measuring adherence from dispensing records

For studies that use claims data or pharmacy records, Raebel and colleagues (2013) proposed standard terms. Primary adherence asks whether the patient received the first prescription within a defined number of days after it was ordered, commonly 30 or 60. Secondary adherence asks whether refills followed as prescribed among patients who collected the first one.

Secondary adherence is usually summarised by the proportion of days covered or the medication possession ratio, each described on its own page. Raebel and colleagues list a weakness shared by these measures, namely that they cannot be calculated for patients with primary non-adherence and so systematically exclude the most non-adherent patients. Discontinuation in a database can only be assessed against a pre-specified number of days without medication, and they report 180 days as often used. The usual 80% cut-point for adequate adherence has been applied by convention and only rarely with supporting evidence.

Among 264,789 statin initiators, Franklin and colleagues found that a six-group group-based trajectory model summarised observed monthly adherence better than the proportion of days covered (C statistic 0.938 against 0.881). In clinical practice NICE CG76 suggests asking about missed doses over a specific period such as the past week, and considering records of prescription re-ordering, pharmacy medication records and return of unused medicines to identify patients needing support.

How medication adherence enters a decision model

Each phase acts on a different model quantity. Initiation sets the share of the prescribed cohort that receives any treatment, persistence sets how long treatment costs and benefits accrue, and implementation sets dose exposure while on treatment. One adherence percentage applied to everything treats these as interchangeable.

The ISPOR Economics of Medication Compliance Working Group (Hughes and colleagues, 2007) found that compliance and persistence were not included routinely in pharmacoeconomic analyses. It noted that Markov models allow persistence to be included directly as patients move between states each cycle, and that discrete event simulation may offer more flexibility.

Two further choices move results. The effect estimate from an intention-to-treat analysis already reflects the adherence seen in that trial, so reducing it again for non-adherence risks counting part of the shortfall twice. The link between dose exposure and benefit is a dose-response question for each drug, and scaling the effect in proportion to doses taken is an assumption, not a default. Drug costs follow what is dispensed, so medicine supplied but not taken is still a cost to the payer.

Worked example with an illustrative cohort of 1,000 patients

All numbers in this example are illustrative. A once-daily medicine costing GBP 1.00 per day is prescribed to 1,000 patients for one year (365 doses each). Of these, 80% collect the first prescription. Among initiators, the share still on treatment at the start of each quarter is 100%, 85%, 72% and 60%, anyone who stops does so at a quarter end, and each quarter lasts 91.25 days. Patients on treatment take 90% of their doses, and every day on treatment is dispensed.

$$D = I \times \bar{T} \times m$$

where $D$ is the expected number of doses taken per prescribed patient over the year, $I$ is the proportion who initiate, $\bar{T}$ is the mean number of days on treatment per initiator within the year, and $m$ is the implementation proportion, the share of doses taken while on treatment.

Step 1. Mean days on treatment. $\bar{T} = 91.25 \times (1.00 + 0.85 + 0.72 + 0.60) = 91.25 \times 3.17 = 289.26$ days.

Step 2. Doses taken. $D = 0.80 \times 289.26 \times 0.90 = 208.27$ doses per prescribed patient, or 57.1% of the 365 prescribed.

Step 3. Where the 156.73 missing doses are lost. Non-initiation loses 0.20 × 365 = 73.00 doses. Early discontinuation loses 0.80 × 75.74 = 60.59, where 75.74 is 365 minus 289.26. Missed doses while on treatment lose 0.80 × 289.26 × 0.10 = 23.14. The three sum to 156.73, which equals 365 minus 208.27.

Step 4. What a dispensing measure would report. Among initiators, supply covers 289.26 of 365 days on average, a mean proportion of days covered of 79.25%. The 80% cut-point applies patient by patient, so the 60% who stay on treatment all year count as adherent and the 40% who stop do not. The dose-taking share among initiators is lower, at 289.26 × 0.90 / 365 = 0.713, or 71.3%, because dispensing cannot see missed doses, and the 200 patients who never collected the medicine are absent from the measure altogether.

Step 5. Drug cost in the model. Dispensed cost is 0.80 × 289.26 days × GBP 1.00 = GBP 231.41 per prescribed patient. A model that charged only for doses taken (208.27 of 365) would charge GBP 208.27 per patient, understating spending by GBP 23.14 per patient, the value of medicine dispensed but not taken.

Interpretation. Here most lost doses arise before treatment starts or after it stops, so an intervention aimed only at missed doses while on treatment would address under 15% of the shortfall. A single figure of 57%, or a mean proportion of days covered of 79%, would conceal that split.

Cost sharing and prescription charges

Out-of-pocket cost is one route by which health financing reaches medicine taking. Eaddy and colleagues reviewed 160 articles and found that 85% of those examining changes in cost sharing showed a higher patient share of medication costs significantly associated with lower adherence, and most studies of outcomes linked higher adherence with better outcomes. NICE CG76 asks professionals to check whether prescription charges are a problem for the patient and, if so, to consider options to reduce costs. Design features that set the patient's share are covered under copayment and cost sharing.

Reporting standards for adherence studies

Hughes and colleagues (2007) reported that a lack of rigour and consistency in definitions often limited the usefulness of the analyses they reviewed. The ESPACOMP Medication Adherence Reporting Guideline (EMERGE), published by De Geest and colleagues in 2018, recommends standard reporting based on an accepted taxonomy. It was derived from a literature review and a Delphi study with 26 adherence experts, and has four minimum reporting criteria and 17 best-practice items. A study that states which phase it measured, and how, is far easier to place in a model.

Boundaries with nearby concepts

Adherence and treatment adherence are the general forms for any agreed or recommended treatment, and medication adherence is the case for prescribed medicines. Compliance is the older term for the same behaviour, often avoided for its connotation of obedience. Medication persistence is the duration component, which the ABC taxonomy places inside adherence and the Cramer definitions place beside it. The proportion of days covered and the medication possession ratio measure supply, not the behaviour itself.

Common misreadings in economic work

Most errors come from collapsing the phases into one number or from treating supply as use. Two such errors are listed below.

1. Dropping primary non-adherers. Refill measures begin at the first fill, so they omit the patients with the lowest adherence.
2. Using one percentage for every phase. Initiation, implementation and persistence act on different model quantities.

Sources

• Cramer JA, Roy A, Burrell A, Fairchild CJ, Fuldeore MJ, Ollendorf DA, Wong PK. Medication compliance and persistence: terminology and definitions. Value in Health. 2008;11(1):44-47. https://pubmed.ncbi.nlm.nih.gov/18237359/
• De Geest S, Zullig LL, Dunbar-Jacob J, Helmy R, Hughes DA, Wilson IB, Vrijens B. ESPACOMP Medication Adherence Reporting Guideline (EMERGE). Annals of Internal Medicine. 2018;169(1):30-35. https://doi.org/10.7326/M18-0543
• Eaddy MT, Cook CL, O'Day K, Burch SP, Cantrell CR. How patient cost-sharing trends affect adherence and outcomes: a literature review. P&T. 2012;37(1):45-55. https://pmc.ncbi.nlm.nih.gov/articles/PMC3278192/
• Fischer MA, Stedman MR, Lii J, Vogeli C, Shrank WH, Brookhart MA, Weissman JS. Primary medication non-adherence: analysis of 195,930 electronic prescriptions. Journal of General Internal Medicine. 2010;25(4):284-290. https://pubmed.ncbi.nlm.nih.gov/20131023/
• Franklin JM, Shrank WH, Pakes J, Sanfélix-Gimeno G, Matlin OS, Brennan TA, Choudhry NK. Group-based trajectory models: a new approach to classifying and predicting long-term medication adherence. Medical Care. 2013;51(9):789-796. https://pubmed.ncbi.nlm.nih.gov/23685406/
• Hughes D, Cowell W, Koncz T, Cramer J; ISPOR Economics of Medication Compliance Working Group. Methods for integrating medication compliance and persistence in pharmacoeconomic evaluations. Value in Health. 2007;10(6):498-509. https://pubmed.ncbi.nlm.nih.gov/17970932/
• National Institute for Health and Care Excellence. Medicines adherence: involving patients in decisions about prescribed medicines and supporting adherence. Clinical guideline CG76. Published 28 January 2009, last updated September 2019. https://www.nice.org.uk/guidance/cg76
• Raebel MA, Schmittdiel J, Karter AJ, Konieczny JL, Steiner JF. Standardizing terminology and definitions of medication adherence and persistence in research employing electronic databases. Medical Care. 2013;51(8 Suppl 3):S11-S21. https://pmc.ncbi.nlm.nih.gov/articles/PMC3727405/
• Vrijens B, De Geest S, Hughes DA, et al. A new taxonomy for describing and defining adherence to medications. British Journal of Clinical Pharmacology. 2012;73(5):691-705. https://pmc.ncbi.nlm.nih.gov/articles/PMC3403197/
• Vrijens B, Vincze G, Kristanto P, Urquhart J, Burnier M. Adherence to prescribed antihypertensive drug treatments: longitudinal study of electronically compiled dosing histories. BMJ. 2008;336(7653):1114-1117. https://pmc.ncbi.nlm.nih.gov/articles/PMC2386633/