A Tool to Compare Different Types of Adherence Measures

 

Background

There are several types of measures to calculate medication adherence. Andrade et al. [1] in their systematic literature review that investigated different adherence measures concluded that identification, aptness and selection of measures for adherence should be determined by the objectives of the study, and limitations and benefits of the measures should be considered. Adherence refers to the extent to which a patient’s behaviour to take the prescribed medications aligns with the instructions and recommendations from the prescriber [2]. A widely-used measure of adherence is defined in terms of a proportion-of-days-covered model, which calculates the proportion of days within a fixed interval that the patient has an available supply of medication [3], and reported using the medication possession ratio (MPR) [1].

 

After reviewing 77 studies, Andrade et al [1] reported that the two most frequently used definitions for MPR are:

 

                                 (1)

and

                                          (2)

 

Based on my experience in analysing prescribing data, to determine the MPR for patients after a required classification (such as hypertension), there are several temporal considerations (Figure 1) that need to be taken into account for an accurate measure of MPR.

 

Figure 1: Temporal relations to consider when calculating MPR. Prx denotes a prescription where x is a sequence number. ● represents a patient classification and TotGap indicates the total lapse in medication that needs to be considered when determining medication possession during the period of interest. Scenario (i) is a typical case where the medication lapse occurs during the evaluation period (EP); (ii) is a case with a lapse running into the EP; (iii) is an on-going lapse at the end of the EP; (iv) is a case where only part of the total lapse is needed; and (v) and (vi) represent cases where there is no medication at all during the EP, but in (vi), only part of the EP needs to be considered.

 

It needs to be acknowledged that the definitions in equations (1) and (2) are commonly used for MPR and may be adequate for the majority of cases, however, in this research I attempt to offer a more refined definition that considers a patient’s medication availability at the edges of the EP, as well as the time of classification. For example, consider scenario (i) in Figure 1. If definition (2) is used to calculate MPR for this prescribing scenario, only the second prescription (denoted by Pr2) will be included, which is a rather incomplete picture of medication possession for this patient over the EP. In terms of the definition in (1) for scenario (i), failure to account for Pr1 (because it is not ‘obtained during the observation period’) tends to underestimate medication possession; conversely, the medication possession from Pr3 will be overestimated. Moreover, if a denominator of the entire duration of the EP is used for a scenario such as (iv) where Pr2 is the patient’s very first prescription of that drug after being classified, the resulting MPR would be misleadingly low. If a period of considerable duration (such as a year) is considered, there are likely to be many patients who are newly diagnosed during the EP, and, as such, if low MPRs are used as the basis to identify non-adherent patients, cases with a prescribing pattern similar to scenario (iv) will contribute towards false-positives resulting in low specificity. Similar issues arise when definition (1) and/or (2) is applied to other scenarios.

 

Therefore, for the work presented herein, MPR is calculated by including the boundary prescriptions such that if scenario (i) in Figure 1 is considered, only those parts of Pr1 and Pr3 coverages that fall within the EP are included in the numerator of the MPR calculation. Furthermore, a run-in period prior to the EP is used (this has been discussed previously) so that prescriptions such as Pr1 (prescribed prior to beginning of EP that run-into EP) can be correctly accounted for.

 

The following definitions have been used for MPR calculations:

 

If patient was classified before beginning of the EP:

                                                               (3)

Else:

             (4)

 

where Total gap duration refers to the sum of all medication lapses as determined after various temporal considerations shown in Figure 5.1.

 

It should be noted that several possession based adherence measures have been identified in the literature. A recent study by Karve et al [4] compared 11 different adherence measures based on a systematic literature review on long-term medication adherence measures by Hess et al [5]. The authors indicate that two measures –proportion of days covered (PDC) uncapped and PDC capped (at a maximum of 100%) should be considered first when selecting among different adherence measures. The definitions for these two measures (adapted from Karve et al [4] and Hess et al [5]) are shown in Table 1.

 

Table 1: Definitions of PDC uncapped and PDC capped

Adherence Measure

Formula

PDC uncapped

PDC capped

 

The definition of MPR is somewhat similar to that of PDC capped since the maximum value of MPR is 100% which is the case when the patient is fully adherent. MPR does not account for medication oversupply explicitly so there is no room for gradual accumulation of medication resulting in a reasonable period of oversupply which can shadow non-adherence. Further, I have included the special case (v) in Figure 5.1 so that patients who have not been on any medication at all during the EP can also be identified.

 

The two adherence measures in Table 1 use the entire duration of the EP as the denominator and studies that use these definitions for adherence calculations usually include only patients who were classified prior to EP. “Most patients with hypertension will require two or more antihypertensive medications to achieve goal blood pressure” [6], however most studies only consider patients on monotherapy in an attempt to simply adherence calculations [4, 7-9] and this is likely to exclude patients with greater disease severity [4].

 

Figure 2 shows an example scenario where values produced by each of the measures can be directly compared. For comparison purposes, assume a hypothetical patient who was classified with hypertension prior to the beginning of the EP. Also assume that the patient was initially on ACE-Inhibitor monotherapy (denoted by Pr1, Pr2, Pr3 and Pr5 prescriptions) and then a diuretic was added during the EP (denoted by Pr4 and Pr6) to intensify therapy.

 

Figure 2: Prescribing patterns for a hypothetical patient. The gray rectangular boxes denote prescriptions (Pr) and the numerical values within a box indicate the prescription duration. Relevant temporal durations required for adherence calculations are also shown. All durations are shown in days.

 

Table 2 shows the overall antihypertensive medication adherence rates based on the different measures.

 

Table.2: Different adherence calculations

Measure

Adherence Calculation

PDC uncapped

90 x 5 / 365 = 123%

PDC capped

90 x 5 / 365 [capped at 100% if PDC>100%] = 100%

MPR

[365 - (30+35)] / 365 = 82%

 

It should be clear that the patient in Figure 2 is not fully adherent to medication as suggested by PDC capped and PDC uncapped. The scenario in Figure 2 is a fairly realistic ‘real-world’ example and the purpose of this was to make the reader aware of some of the important temporal relationships that need to be considered. If the end date of Pr1 in Figure 2 was before the beginning of the EP for example, there would be a treatment gap at the beginning of the EP as well, but still PDC capped and PDC uncapped would result in the same values where as the MPR calculation used in this thesis would result in a more realistic value of [365 – (90 + 35)] / 365 = 66%.

 

The ‘Tool’

The tool I have developed as part of the ChronoMedIt framework can be used to compare different types of adherence measures. At some stage we’ll be looking at carrying out a detailed comparison based on data from several practices. Some screenshots of the tool are shown below.

 

The data can also be exported into Excel for further comparison if required:

 

References

1.            Andrade, S.E., K.H. Kahler, F. Frech, et al., Methods for evaluation of medication adherence and persistence using automated databases. Pharmacoepidemiol Drug Saf, 2006. 15(8): p. 565-74; discussion 575-7.

2.            Adherence to Long-term Therapies: Evidence for action. 2003 [cited 30 Novemver 2009]; Available from: http://www.emro.who.int/ncd/Publications/adherence_report.pdf.

3.            Burnier, M., Medication adherence and persistence as the cornerstone of effective antihypertensive therapy. Am J Hypertens, 2006. 19(11): p. 1190-6.

4.            Karve, S., M.A. Cleves, M. Helm, et al., An empirical basis for standardizing adherence measures derived from administrative claims data among diabetic patients. Med Care, 2008. 46(11): p. 1125-33.

5.            Hess, L.M., M.A. Raebel, D.A. Conner, et al., Measurement of adherence in pharmacy administrative databases: a proposal for standard definitions and preferred measures. Ann Pharmacother, 2006. 40(7-8): p. 1280-88.

6.            Chobanian, A.V., G.L. Bakris, H.R. Black, et al., Seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. Hypertension, 2003. 42(6): p. 1206-52.

7.            Bloom, B.S., Continuation of initial antihypertensive medication after 1 year of therapy. Clin Ther, 1998. 20(4): p. 671-81.

8.            Bramley, T.J., P.P. Gerbino, B.S. Nightengale, et al., Relationship of blood pressure control to adherence with antihypertensive monotherapy in 13 managed care organizations. J Manag Care Pharm, 2006. 12(3): p. 239-45.

9.            Elliott, W.J., C.A. Plauschinat, G.H. Skrepnek, et al., Persistence, adherence, and risk of discontinuation associated with commonly prescribed antihypertensive drug monotherapies. J Am Board Fam Med, 2007. 20(1): p. 72-80.