Skip to main content

Measuring medication adherence in patients with incident hypertension: a retrospective cohort study

Abstract

Background

Though pharmacy claims data are commonly used to study medication adherence, there remains no standard operational definition for adherence especially for patients on multiple medications. Even when studies use the same terminology, the actual methods of calculating adherence can differ drastically. It is unclear whether the use of different definitions results in different conclusions regarding adherence and associated outcomes. The objective of our study was to compare adherence rates and associations with mortality using different operational definitions of adherence, and using various methods of handling concurrent medication use.

Methods

We conducted a cohort study of patients aged ≥65 years from Manitoba, Canada, with incident hypertension diagnosed in 2004 and followed to 2009. We calculated adherence rates to anti-hypertensive medications using different operational definitions of medication adherence (including interval and prescription based medication possession ratios [MPR] and proportion of days covered [PDC]). For those on concurrent medications, we calculated adherence rates using the different methods of handling concurrent medication use, for each definition. We used logistic regression to determine the association between adherence and mortality for each operational definition.

Results

Among 2199 patients, 24.1% to 90.5% and 71.2% to 92.7% were considered adherent when using fixed interval and prescription-based interval medication possession ratios [MPRi and MPRp] respectively, depending on how concurrent medications were handled. Adherence was inversely associated with death, with the strongest association for MPRp measures. This association was significant only when considering adherence to any anti-hypertensive [aOR 0.70, 95% CI 0.51, 0.97], or when the mean of the class-specific MPRp’s [adjusted OR 0.71, 95% CI 0.53, 0.95] was used. No significant association existed when the highest or lowest class-specific MPRp was used as the adherence estimate.

Conclusion

The range of adherence estimates varies widely depending on the operational definition used. Given less variation in adherence rates and their stronger association against mortality, we recommend using prescription-based MPR’s to define medication adherence.

Peer Review reports

Background

Medication adherence is defined as the extent to which a patient takes prescribed medications according to the dosage and frequency recommended by the provider [1, 2]. Non-adherence to prescribed medications is associated with poor treatment outcomes, increased hospitalizations, and increased cost to the health care system for chronic diseases such as hypertension, diabetes, and heart failure [35]. Despite the presence of a clinical definition of medication adherence, there remains no standard operational definition for medication adherence in health research, especially when using pharmacy claims data [6].

In the literature, there are a wide number of terms and operational definitions used to assess medication adherence (see Table 1 for a list of definitions used in the hypertension literature). Even when studies use the same terms, the operationalized or practical definition presented, the time frame considered, and the method of calculating adherence often differ. Most studies use the “medication possession ratio” (MPR) [1, 7] or similar related adherence measures, generally defined as the proportion of a time period where a medication supply is available [1]. Variations in MPR calculations stem from different denominators used, which can either be a fixed time interval or a variable period between prescriptions. The former is termed “interval based MPR” [MPRi] and the latter, “prescription based MPR” [MPRp] [8]. MPR can be reported as a continuous or as a dichotomized measure, where patients with an MPR above a certain threshold are considered “adherent”. While the level of optimal adherence may differ for different clinical conditions, a threshold of 0.80 [1] is conventionally used.

Table 1 Terminology and operational definitions of medication adherence used in hypertension literature

Despite wide acceptance and usage of MPR, there remains significant variation not only in its specific calculation as mentioned above, but also in methods of managing multiple concurrent medications (“polytherapy”) to obtain a single measure of adherence for each patient (see Table 2). Methods of calculating an MPR estimate for patients on polytherapy can consider either adherence to individual drug classes or adherence overall to any medication. In addition, a related measure to MPR termed Proportion of Days Covered [PDC] is often preferred in polytherapy due to its lower risk of overestimation [9]; PDC is defined as the proportion of days in a fixed observation period where at least one of multiple medications is available.

Table 2 Methods of handling use of multiple concurrent anti-hypertensive medications when measuring adherence in hypertension literature

The many different operational definitions and methods used to handle polytherapy may result in very different adherence estimates and highly variable conclusions [8], highlighting the need to harmonize operational definitions of adherence used in health research. Our study objective was to compare adherence rates using the different operational definitions (MPRi, MPRp, and PDC) and methods of handling concurrent medications, and to determine the strengths of association between medication adherence using these different operational definitions with all-cause mortality in a population of Canadians with newly diagnosed hypertension.

Methods

Sources of data

Administrative data extracted from hospital discharge abstracts and physician claims were collected in Manitoba, Canada from April 1, 1997 to March 31, 2009. We linked this database with vital statistics for mortality data, the health insurance registry for demographic data, and the Manitoba Pharmacare prescription database for drug data. Drug data from April 1, 2002 to March 31, 2009 were used for this study. Given the universality of Manitoba Pharmacare, this database contains dispensing records for all outpatient prescription drug claims for all Manitoban residents, with the exception of First Nations and Inuit persons, inmates incarcerated in federal penitentiaries, military personnel, RCMP, and veterans; these subgroups are covered by the federal drug benefit plan [10]. Institutional ethics board approval was obtained from the University of Calgary Conjoint Health Research Ethics Board. Manitoba Health granted access to the administrative databases used in this study. Direct patient consent was not obtained, as the data being analyzed were routinely collected administrative health data that had been de-identified prior to receipt of data.

Study population

The 25 diagnosis fields in hospital discharge abstracts and the 1 field in physician claims were searched for the relevant International Classification of Diseases, Ninth and Tenth Revisions (ICD-9 and ICD-10) codes for hypertension (401.x-405.x, 110.x-115.x). The case definition for incident hypertension in the databases was one hospitalization or two physician claims within 2 years with a hypertension ICD code. We determined incidence with a three-year washout period [11]. Index date of diagnosis was the first date at which the case definition was fulfilled. Patients without at least a single prescription refill within one year after the first prescription fill in any of the five antihypertensive medication classes of interest (thiazide-type diuretics, beta blockers [BB], calcium channel blockers [CCB], angiotensin converting enzyme inhibitors or angiotensin receptor blockers [ACEI/ARB], or a combination containing at least one of the above classes; see Appendix 1), and patients who died within one year of the first prescription fill were excluded to ensure that each operational definition could be calculated for each patient. The study population included a random sample of Manitoba residents aged ≥65 years with incident hypertension, with an index date of diagnosis between April 1, 2004 and March 31, 2005.

Sample size calculation

Using conservative assumptions, the target sample size, assuming that the mortality rate of patients over the age of 65 years old with hypertension who are non-adherent to medications is 17% [12], to achieve a power of 80%, a level of significance of 5%, and to detect a relative risk ratio of mortality of 0.72 [13] for patients who are adherent to medications compared with patients who are non-adherent to medications, is 2124. To account for the exclusion of approximately 30% of the entire sample either because they have not had had a single fill or refill of any anti-hypertensive medications in the entire follow-up period, or die within the first year of diagnosis, 3000 patients with incident hypertension were randomly selected for analysis. Random sampling was performed using the probability proportional to size method based on age, sex, and comorbidities.

Measures of medication adherence

We calculated MPR using the following 2 formulae (interval-based and prescription-based MPR respectively) for an observation period of 1 year:

$$ \mathrm{MPRi}=\frac{\mathrm{Days}\ \mathrm{supply}\ \mathrm{of}\ \mathrm{medication}}{365\ \mathrm{days}},\mathrm{capped}\ \mathrm{at}\ 1 $$
$$ \mathrm{MPRp}=\frac{\mathrm{Days}\ \mathrm{supply}\ \mathrm{of}\ \mathrm{medication}\ \mathrm{excluding}\ \mathrm{supply}\ \mathrm{from}\ \mathrm{last}\ \mathrm{refill}}{\mathrm{Last}\ \mathrm{refill}\ \mathrm{date}\hbox{-} \mathrm{First}\ \mathrm{refill}\ \mathrm{date}},\ \mathrm{capped}\ \mathrm{at}\ 1 $$

Medications within the same medication class (Appendix 1) were considered interchangeable.

For patients on polytherapy in our study, we calculated four different MPRi’s and MPRp’s: a) MPR considering adherence to any antihypertensive, obtained by first summing the days supply of each anti-hypertensive prescription in the numerator [9], then dividing by the denominator as stated above; b) average of the MPR’s specific to each anti-hypertensive medication class [8, 14]; c) calculating the MPR’s specific to each medication class, then taking the highest of the class-specific MPR’s [15]; and d) calculating the MPR’s specific to each medication class, then taking the lowest of the class-specific MPR’s [16]. We termed these methods “sum MPR”, “mean MPR”, “high MPR” and “low MPR” respectively. Of note, for sum MPR, no medication class-specific MPRs are calculated. The denominator used for sum MPRp is the last refill date for any of the medication classes minus the first refill date for any of the medication classes (even if this medication class is not the same one used for the last refill date). We also calculated PDC for each patient by dividing the number of days where at least one medication was available (each day is considered individually and is a binary measure) [9] by 365 days.

MPR’s and PDC’s were additionally dichotomized using the standard threshold of 0.80. For patients on anti-hypertensive therapy prior to the index date, adherence measures were calculated from the first prescription fill starting from 2002. If prescription refills were obtained prior to exhaustion of the supply of the previous fill for the same medication class, the date of the refill was prorated to start on the day after exhaustion of the previous supply.

Mortality

The primary outcome measure was mortality from the period 2005 to 2009 (a maximum of 5 years of follow-up). Any patients moving out-of-province or reaching the end of the observation period were censored.

Statistical analysis

Means and medians for continuous medication adherence variables were reported. These variables were also dichotomized, and Fisher’s exact test was used to compare the proportion of adherent and non-adherent patients who had died. We also stratified this analysis by age, sex, income, and number of comorbidities. Multiple logistic regression models were employed to model the odds of death for those who are adherent to medications, compared to those who were non-adherent to medications, using the different operational definitions for adherence. The threshold used to define adherence was an MPR or PDC ≥ 0.80. Unadjusted and adjusted Cox proportional hazards regressions were performed for each operational definition of adherence, to assess time to death. Given their potential to confound the association between medication adherence and mortality, we controlled for demographic and comorbidity variables in our adjusted logistic regression and Cox proportional hazards regression models; these variables included age, sex, Charlson comorbidity index, income quintile, and previous hospitalization for cardiovascular disease within three years prior to the index date of hypertension diagnosis. Data on health-related behaviours, such as diet and activity levels, were not available given the nature of data used in this study and thus no adjustment for these variables were undertaken. However, previous studies have shown weak correlations and non-statistically significant associations between health-related behaviours and medication adherence [1720]. Health-related behaviours are therefore unlikely to be a significant confounder in the associations between medication adherence and mortality. All analyses were conducted using SAS Version 9.4 (SAS Institute, Cary, NC).

Sensitivity analysis

Because there is no robust evidence base to suggest that an MPR or PDC threshold of 0.80 used to classify those who are adherent versus non-adherent to medications is optimal or superior to other thresholds, we conducted sensitivity analysis, using various other adherence thresholds. For this sensitivity analysis, adjusted logistic regression as described above was also performed for thresholds of 0.70, 0.75, 0.85, and 0.90, for each of the operational definitions studied.

Results

Baseline characteristics

From a total of 5189 eligible patients aged 65 years or older with incident hypertension diagnosed in the 2004 fiscal year, a random sample of 3000 patients were selected for analysis. Of this sample of 3000 patients, we excluded those with no prescription filled for any anti-hypertensive in the entire follow-up period (n = 652), those who died within one year of their first anti-hypertensive prescription fill (n = 51), and those without at least one medication refill within one year of their first prescription fill (n = 98). Our final cohort comprised of 2199 patients. The median and mean follow-up durations were 4.41 years (interquartile range 4.11, 4.72) and 4.13 years (standard deviation 0.99) respectively.

Baseline characteristics can be found in Table 3. Overall, mean (SD) age was 75.2 (7.0) years, 45.4% were male, 12.2% had a previous hospitalization for cardiovascular disease, 33.7% were new users of anti-hypertensives, and 64.7% were on monotherapy. Comparison of baseline characteristics between non-adherent and adherent groups depended on the adherence measure being used (Table 3).

Table 3 Demographics and descriptive characteristics of medication use, for adherent versus non-adherent users as defined by three different operational definitions

Adherence rates by operational definition

For patients on monotherapy (n = 1422), MPRi adherence estimates (mean 0.83, SD 0.23) were similar though consistently lower than the equivalent MPRp estimates (mean 0.87, SD 0.19, see Table 4). The PDC and MPRi methods gave identical adherence estimates for those on monotherapy only. These findings were consistent even with stratification by age, sex, income, and comorbidities.

Table 4 Mean and median adherence rates, and proportion with adherence ≥0.80 for patients on monotherapy (n = 1422)

In polytherapy (n = 777), the range of adherence rates varied widely, based on whether MPRi or MPRp was used, and also depending on the method used to manage polytherapy (Table 5). The lowest overall estimate of adherence was with using the “low MPRi” method, or taking the lowest class-specific MPRi for each patient (mean 0.47, SD 0.32). The highest overall estimate of adherence was with using the “sum MPRi” method, or by taking the sum of the days supply from each anti-hypertensive medication class as the numerator in the calculation of MPRi (mean 0.95, SD 0.13). “Mean MPRi” (0.66, SD 0.23) and “high MPRi” (0.84, SD 0.21) resulted in adherence estimates between these two extremes. Similarly, when adherence measures were dichotomized, the proportion of adherent users ranged from 24.1% with “low MPRi” to 90.5% with “sum MPRi”. In contrast to patients on monotherapy, PDC and MPRi (regardless of the method used to handle polytherapy) do not give the same adherence estimates for patients on polytherapy.

Table 5 Mean and median adherence rates and proportion with adherence ≥0.80 for patients on polytherapy (n = 777)

The range of adherence was much narrower for MPRp adherence estimates, with “low MPRp” giving again the lowest adherence estimate (mean 0.84, SD 0.21) and “sum MPRp” giving the highest estimate (mean 0.96, SD 0.12). The variation in the proportion of adherent users was also much smaller using MPRp, ranging from 71.2% with “low MPRp” to 92.7% with “sum MPRp”. PDC gave a more conservative measure compared to both sum MPRi and sum MPRp; 83.0% of the sample was classified as adherent using PDC. Even with stratification across age, sex, income, and comorbidities, the same patterns could be seen where MPRi’s gave much wider ranges of adherence estimates compared with MPRp’s, with “low MPR” giving the lowest estimates and “sum MPR” giving the highest estimates regardless of whether the prescription or interval based MPR method was used.

Association between adherence and mortality

A total of 387 (17.6%) deaths occurred in the follow-up period. Greater medication adherence was consistently associated with lower odds of death, after risk adjustment (see Table 6). Overall, the strength of association was stronger for MPRp than for MPRi or PDC (adjusted odds ratio [aOR] for “sum MPRp” was 0.70 [95% CI 0.51, 0.97] for “sum MPRi” aOR 0.78 [95% CI 0.58, 1.05], for PDC aOR 0.80 [95% CI 0.60, 1.07]). Of the numerous methods of handling concurrent medication use, “sum MPRp” and “mean MPRp” had the strongest associations with mortality with these being the only measures to reach statistical significance (aOR 0.70 [95% CI 0.51, 0.97] and 0.71 [95% CI 0.53, 0.95] respectively).

Table 6 Adjusted logistic regression modeling the odds ratio of death, for patients who are adherent versus non-adherent to anti-hypertensive medication

Similarly, in monotherapy, the association between medication adherence and mortality was stronger for MPRp measures than for MPRi and PDC measures based on point estimates, despite none reaching statistical significance. In polytherapy patients, the same inverse association could be seen between medication adherence as measured by MPRp and death, with the strongest associations seen again with “sum” and “mean” MPRp, though again none reached statistical significance.

Unadjusted and adjusted Cox proportional hazards regression provided similar results to logistic regression. However, testing of the proportional hazards assumption indicated its violation. Kaplan-Meier survival curves revealed a divergence of survival in adherent versus non-adherent groups at approximately 3.5 to 4 years post index date of hypertension diagnosis for MPRp definitions (with a less prominent trend using MPRi and PDC definitions).

Sensitivity analysis

Similar results were seen when adjusted logistic regression was performed using different adherence thresholds for each of the operational definitions (Appendix 2). Across all adherence thresholds, adherence estimates as measured by MPRp were more strongly associated with mortality than as measured by MPRi. The MPRp definitions that showed a statistically significant association between adherence and mortality varied depending on the adherence threshold used. Statistical significance for this association was reached for mean MPRp when using adherence thresholds of 0.70, 0.75, and 0.80; for sum MPRp, the adherence thresholds were 0.80 and 0.85. For low MPRp, statistical significance was reached when the adherence threshold used was 0.70.

Discussion

In a cohort of 2199 patients with incident hypertension, we found that different definitions of medication adherence resulted not only in different baseline characteristics of “adherent” and “non-adherent” groups but also vast differences in estimated adherence rates. Similar to previous studies, adherence measures based on a fixed observation period (MPRi and PDC) resulted in lower adherence estimates compared with variable observation periods based on prescription refill dates (MPRp) [8, 21]. In addition, different methods of calculating adherence for patients on polytherapy provided very different estimates of adherence [8]. For example, if only those with an MPRi of ≥0.80 for each and every medication class were considered adherent, only 24.1% would be classified as being adherent. Conversely, if we considered those with an overall MPRi of ≥0.80, when all medication classes were grouped together, as adherent, over 90% of the same sample population would be classified as adherent. Therefore, given this wide variation in adherence estimates and their implications on study conclusions, it is imperative that future adherence studies are transparent in providing information regarding: 1) specific operational definitions of adherence used and the numerator and denominator used in these calculations; and 2) the method used to manage polytherapy. Because adherence measure names and corresponding calculations are inconsistent across studies [7], we note that adherence measure terms alone are inadequate in describing the differences in calculations.

Adjusted logistic regression revealed that regardless of the operational definition used for medication adherence, there was a non-statistically significant trend between adherence to anti-hypertensive medications and lower risk of death. The strength of association was greater in prescription-based measures compared with interval-based measures. However, both methods are not without their faults. While prescription-based measures likely overestimate adherence by not accounting for patients who inappropriately discontinue their medications [22], interval-based measures likely underestimate adherence, by interpreting medication switches as non-adherence.

Our sensitivity analysis revealed that the operational definitions that demonstrate the strongest and most significant associations between adherence and mortality vary depending upon the adherence threshold used. When using low MPRp, adherence above the threshold of 0.70 is significantly associated with lower risk of mortality; this significance is lost when higher thresholds are used. For mean MPRp measures, adherence using the thresholds of 0.70 to 0.80 is associated with reduced risk of death, but not when using higher thresholds of 0.85 or 0.90. For sum MPRp, adherence above the threshold of 0.80 and 0.85 is associated with reduced risk of death. Therefore, the optimal adherence threshold may differ based on the method used to handle polytherapy. When using stringent measures of adherence (such as when adherence is measured based on the medication with lowest adherence in the case of low MPRp), a lower adherence threshold such as 0.70 may be preferred. In contrast, when using a more liberal measure of adherence (such as considering a patient adherent even if he or she adheres to only one of numerous medications, such as in the case of sum MPRp), a higher adherence threshold such as 0.80 or 0.85 may be preferred.

For the standard adherence threshold of 0.80, the adherence-mortality association reached statistical significance only when using mean and sum prescription-based MPR measures. These measures likely reflect “overall” adherence to a patient’s full regimen of anti-hypertensive medications. Being adherent to only one medication and non-adherent to others (that is, taking into account only highest MPR) is not significantly associated with reduced risk of death, as it may not reflect global adherence to all medications. In contrast, adherence based on “low MPR” (that is, all medication classes must have adherence ≥0.80) is likely too stringent a criterion, and its high specificity would result in many patients being considered non-adherent. The mean and sum MPRp are likely a balance between these two extremes, where non-adherence to at least one medication class in a regimen can profoundly affect these adherence estimates. Our study suggests that mean and sum MPRp are the preferred operational measures of adherence, when using the adherence threshold of 0.80, given their significant association with mortality.

Certain limitations in our study deserve consideration. First, to allow calculation of medication adherence over a one-year observation period, we excluded patients who died or did not have at least one prescription refill within one year of their first prescription fill. We have therefore likely excluded those who are sickest and most dependent on chronic medication therapy, and those who are most non-adherent. As a result, adherence is likely overestimated and the association between adherence and mortality weakened. Second, a maximum of five years of follow-up may be insufficiently long to assess association between adherence and mortality. This concern is supported by Kaplan Meier survival curves showing divergence for the non-adherent and adherent groups beginning only at 3.5 to 4 years from index date of diagnosis. Third, it is possible that the older age of the sample population could have confounded the association between medication adherence and mortality. A Cochrane systematic review showed that in the elderly aged 80 years or older, anti-hypertensive therapy did not reduce total mortality, though it did reduce cardiovascular mortality and morbidity [23]. Because 25% of our sample population is aged 80 years or older, in whom there may not be mortality benefit of anti-hypertensive therapy, the overall associations between anti-hypertensive adherence and mortality may have been weakened by this subgroup. Fourth, because we did not directly compare the different operational definitions of adherence, we cannot state the superiority of one definition over another and therefore cannot recommend a single standardized operational definition for adherence. Rather, our findings suggest that different operational definitions of adherence result in very different estimates of adherence, and certain definitions are associated with long-term outcomes in patients with hypertension, while others are not. Lastly, we used adherence thresholds of 0.80 and a one-year observation period due to its wide usage throughout the literature, despite limited evidence to support these parameters. However, our sensitivity analysis shows similar results and conclusions when using varying adherence thresholds. We recommend confirmation of our findings in a separate large cohort of patients with hypertension, to ensure the validity and generalizability of our conclusions.

Conclusion

This study is an important contribution to the literature, by providing a better understanding of commonly used medication adherence definitions and their association with outcomes. We recommend the use of prescription-based denominators when calculating medication possession ratios and related measures given their narrower range of estimates and stronger associations with long-term outcomes. In patients using concurrent medications, we recommend using the methods of handling polytherapy that are most strongly associated with mortality. For a standard adherence threshold of 0.80, these include the mean medication possession ratio (measured as an average of individual class-specific MPRs) and “sum” medication possession ratio (measured as adherence to any medications in the relevant medication classes). We have shown clearly that different operational definitions of medication adherence and different methods of handling polytherapy can result in a wide range of adherence estimates and therefore conclusions reached about adherence, highlighting the need to harmonize these definitions. Given these implications, it is of utmost importance that future adherence studies using pharmacy claims data carefully select and describe the medication adherence definitions used, especially for patients using multiple concurrent medications.

Abbreviations

ACEI:

Angiotensin converting enzyme inhibitor

aOR:

Adjusted odds ratio

ARB:

Angiotensin receptor blocker

BB:

Beta blocker

CCB:

Calcium channel blocker

CI:

Confidence interval

ICD-10:

International Classification of Diseases, Tenth Revision

ICD-9:

International Classification of Diseases, Ninth Revision

MPR:

Medication possession ratio

MPRi:

Interval based medication possession ratio

MPRp:

Prescription based medication possession ratio

OR:

Odds ratio

PDC:

Proportion of days covered

References

  1. Andrade SE, Kahler KH, Frech F, Chan KA. Methods for evaluation of medication adherence and persistence using automated databases. Pharmacoepidemiol Drug Saf. 2006;15(8):565–74. discussion 575–567.

    Article  PubMed  Google Scholar 

  2. Cramer JA, Roy A, Burrell A, et al. Medication compliance and persistence: terminology and definitions. Value Health. 2008;11(1):44–7.

    Article  PubMed  Google Scholar 

  3. Dunbar-Jacob J, Mortimer-Stephens MK. Treatment adherence in chronic disease. J Clin Epidemiol. 2001;54 Suppl 1:S57–60.

    Article  PubMed  Google Scholar 

  4. Sokol MC, McGuigan KA, Verbrugge RR, Epstein RS. Impact of medication adherence on hospitalization risk and healthcare cost. Med Care. 2005;43(6):521–30.

    Article  PubMed  Google Scholar 

  5. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med. 2005;353(5):487–97.

    Article  CAS  PubMed  Google Scholar 

  6. Cramer JA, Benedict A, Muszbek N, Keskinaslan A, Khan ZM. The significance of compliance and persistence in the treatment of diabetes, hypertension and dyslipidaemia: a review. Int J Clin Pract. 2008;62(1):76–87.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Sattler EL, Lee JS. Perri M, 3rd. Medication (re)fill adherence measures derived from pharmacy claims data in older Americans: a review of the literature. Drugs Aging. 2013;30(6):383–99.

    Article  PubMed  Google Scholar 

  8. Choudhry NK, Shrank WH, Levin RL, et al. Measuring concurrent adherence to multiple related medications. Am J Manag Care. 2009;15(7):457–64.

    PubMed  PubMed Central  Google Scholar 

  9. Martin BC, Wiley-Exley EK, Richards S, Domino ME, Carey TS, Sleath BL. Contrasting measures of adherence with simple drug use, medication switching, and therapeutic duplication. Ann Pharmacother. 2009;43(1):36–44.

    Article  PubMed  Google Scholar 

  10. Health Canada. Federal Public Drug Benefit Programs. October 1, 2004; http://www.hc-sc.gc.ca/hcs-sss/pharma/acces/fedprog-eng.php. Accessed 8 2014.

  11. Quan H, Khan N, Hemmelgarn BR, et al. Validation of a case definition to define hypertension using administrative data. Hypertension. 2009;54(6):1423–8.

    Article  CAS  PubMed  Google Scholar 

  12. Robitaille C, Dai S, Waters C, et al. Diagnosed hypertension in Canada: incidence, prevalence and associated mortality. Can Med Assoc J. 2012;184(1):E49–56.

    Article  Google Scholar 

  13. Shin S, Song H, Oh SK, Choi KE, Kim H, Jang S. Effect of antihypertensive medication adherence on hospitalization for cardiovascular disease and mortality in hypertensive patients. Hypertens Res. 2013;36(11):1000–5.

    Article  PubMed  Google Scholar 

  14. Lau DT, Nau DP. Oral antihyperglycemic medication nonadherence and subsequent hospitalization among individuals with type 2 diabetes. Diabetes Care. 2004;27(9):2149–53.

    Article  CAS  PubMed  Google Scholar 

  15. Rolnick SJ, Pawloski PA, Hedblom BD, Asche SE, Bruzek RJ. Patient characteristics associated with medication adherence. Clin Med Res. 2013;11(2):54–65.

    Article  PubMed  PubMed Central  Google Scholar 

  16. Yang W, Chang J, Kahler KH, et al. Evaluation of compliance and health care utilization in patients treated with single pill vs. free combination antihypertensives. Curr Med Res Opin. 2010;26(9):2065–76.

    Article  CAS  PubMed  Google Scholar 

  17. Levesque A, Li HZ, Pahal JS. Factors related to patients’ adherence to medication and lifestyle change recommendations: Data from Canada. International Journal of Psychological Studies. 2012;4(2):42.

    Article  Google Scholar 

  18. Broadbent E, Donkin L, Stroh JC. Illness and treatment perceptions are associated with adherence to medications, diet, and exercise in diabetic patients. Diabetes Care. 2011;34(2):338–40.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Aggarwal B, Mosca L. Lifestyle and psychosocial risk factors predict non-adherence to medication. Ann Behav Med. 2010;40(2):228–33.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Halava H, Korhonen MJ, Huupponen R, et al. Lifestyle factors as predictors of nonadherence to statin therapy among patients with and without cardiovascular comorbidities. CMAJ. 2014;186(12):E449–456.

    Article  PubMed  PubMed Central  Google Scholar 

  21. Kozma CM, Dickson M, Phillips AL, Meletiche DM. Medication possession ratio: implications of using fixed and variable observation periods in assessing adherence with disease-modifying drugs in patients with multiple sclerosis. Patient Prefer Adherence. 2013;7:509–16.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Karve S, Cleves MA, Helm M, Hudson TJ, West DS, Martin BC. An empirical basis for standardizing adherence measures derived from administrative claims data among diabetic patients. Med Care. 2008;46(11):1125–33.

    Article  PubMed  Google Scholar 

  23. Musini VM, Tejani AM, Bassett K, Wright JM. Pharmacotherapy for hypertension in the elderly. Cochrane Database Syst Rev. 2009;4, CD000028.

    Google Scholar 

  24. Hansen RA, Voils CI, Farley JF, et al. Prescriber Continuity and Medication Adherence for Complex Patients. Ann Pharmacother. 2015;49(3):293–302.

    Article  PubMed  Google Scholar 

  25. Krousel-Wood M, Islam T, Webber LS, Re RN, Morisky DE, Muntner P. New medication adherence scale versus pharmacy fill rates in seniors with hypertension. Am J Manag Care. 2009;15(1):59–66.

    PubMed  PubMed Central  Google Scholar 

  26. Heisler M, Hogan MM, Hofer TP, Schmittdiel JA, Pladevall M, Kerr EA. When more is not better: treatment intensification among hypertensive patients with poor medication adherence. Circulation. 2008;117(22):2884–92.

    Article  PubMed  Google Scholar 

  27. Baggarly SA, Kemp RJ, Wang X, Magoun AD. Factors associated with medication adherence and persistence of treatment for hypertension in a Medicaid population. Res Social Adm Pharm. 2014;10(6):e99–e112.

    Article  PubMed  Google Scholar 

  28. Bramley TJ, Gerbino PP, Nightengale BS, Frech-Tamas F. Relationship of blood pressure control to adherence with antihypertensive monotherapy in 13 managed care organizations. J Manag Care Pharm. 2006;12(3):239–45.

    PubMed  Google Scholar 

  29. Holmes HM, Luo R, Hanlon JT, Elting LS, Suarez-Almazor M, Goodwin JS. Ethnic disparities in adherence to antihypertensive medications of medicare part D beneficiaries. J Am Geriatr Soc. 2012;60(7):1298–303.

    Article  PubMed  PubMed Central  Google Scholar 

  30. Poon I, Lal LS, Ford ME, Braun UK. Racial/ethnic disparities in medication use among veterans with hypertension and dementia: a national cohort study. Ann Pharmacother. 2009;43(2):185–93.

    Article  PubMed  Google Scholar 

  31. Dickson M, Plauschinat CA. Compliance with antihypertensive therapy in the elderly: a comparison of fixed-dose combination amlodipine/benazepril versus component-based free-combination therapy. Am J Cardiovasc Drugs. 2008;8(1):45–50.

    Article  CAS  PubMed  Google Scholar 

  32. Krousel-Wood M, Holt E, Joyce C, et al. Differences in cardiovascular disease risk when antihypertensive medication adherence is assessed by pharmacy fill versus self-report: the Cohort Study of Medication Adherence among Older Adults (CoSMO). J Hypertens Feb. 2015;33(2):412–20.

    Article  CAS  Google Scholar 

  33. Stewart K, Mc Namara KP, George J. Challenges in measuring medication adherence: experiences from a controlled trial. Int J Clin Pharm. 2014;36(1):15–9.

    Article  PubMed  Google Scholar 

  34. Schectman JM, Bovbjerg VE, Voss JD. Predictors of medication-refill adherence in an indigent rural population. Med Care. 2002;40(12):1294–300.

    Article  PubMed  Google Scholar 

  35. Han E, Suh DC, Lee SM, Jang S. The impact of medication adherence on health outcomes for chronic metabolic diseases: A retrospective cohort study. Res Social Adm Pharm. 2014;10(6):e87–98.

    Article  PubMed  Google Scholar 

  36. Li P, McElligott S, Bergquist H, Schwartz JS, Doshi JA. Effect of the Medicare Part D coverage gap on medication use among patients with hypertension and hyperlipidemia. Ann Intern Med. 2012;156(11):776–84. W-263, W-264, W-265, W-266, W-267, W-268, W-269.

    Article  PubMed  Google Scholar 

  37. Chapman RH, Petrilla AA, Benner JS, Schwartz JS, Tang SS. Predictors of adherence to concomitant antihypertensive and lipid-lowering medications in older adults: a retrospective, cohort study. Drugs Aging. 2008;25(10):885–92.

    Article  PubMed  Google Scholar 

  38. GuildCare. MedScreen Programs. 2013; https://www.rxscreen.com.au/ProtocolDocs/MedScreen/GuildCare%20MedScreen%20Programs%20Protocol.pdf. Accessed 9 Feb 2017.

  39. Jackson 2nd KC, Sheng X, Nelson RE, Keskinaslan A, Brixner DI. Adherence with multiple-combination antihypertensive pharmacotherapies in a US managed care database. Clin Ther. 2008;30(8):1558–63.

    Article  PubMed  Google Scholar 

  40. Elliott WJ, Plauschinat CA, Skrepnek GH, Gause D. Persistence, adherence, and risk of discontinuation associated with commonly prescribed antihypertensive drug monotherapies. J Am Board Fam Med. 2007;20(1):72–80.

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

None.

Funding

This project was funded by Canadian Institutes of Health Research (grant number: MOP-97823). The researchers had independence from the funding agency.

Availability of data and materials

The datasets generated and analyzed during the current study are not publicly available for data privacy and ethical considerations, but are available from the corresponding author upon reasonable request.

Authors’ contributions

Each of the three authors meets the authorship requirements as established by the International Committee of Medical Journal Editors in the Uniform Requirements for Manuscripts Submitted to Biomedical Journals. KT, HQ, and DR conceived of the study. KT performed the statistical analyses, and all authors interpreted this analysis. KT drafted the manuscript, and all performed critical revision of the manuscript. All authors have read and approved the manuscript.

Competing interests

K.T. is supported by fellowship awards from the Canadian Institutes of Health Research and Alberta Innovates- Health Solutions, H.Q. receives salary support from Alberta Innovates-Health Solutions, and D.R. is supported by the Population Health Investigator award from Alberta Innovates- Health Solutions. This study is based in part on de-identified data provided by Canadian provincial health ministries and do not represent the views of the provincial governments. The results and conclusions are those of the authors and no official endorsement by Manitoba Health is intended nor should be inferred.

Consent for publication

Not applicable.

Ethics approval and consent to participate

Institutional ethics board approval was obtained from the University of Calgary Conjoint Health Research Ethics Board (E18889). Manitoba Health granted access to the administrative databases used in this study. Direct patient consent was not obtained, as the data being analyzed were routinely collected administrative health data that had been de-identified prior to receipt of data.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Karen L. Tang.

Appendices

Appendix 1

Table 7 Antihypertensive medication classes

Appendix 2

Table 8 Sensitivity analysis for odds ratios modeling death, for patients who are adherent versus non-adherent to anti-hypertensive medication, using different thresholds to define adherence

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Tang, K.L., Quan, H. & Rabi, D.M. Measuring medication adherence in patients with incident hypertension: a retrospective cohort study. BMC Health Serv Res 17, 135 (2017). https://doi.org/10.1186/s12913-017-2073-y

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12913-017-2073-y

Keywords