Estimating Medication Persistency Using Administrative Claims Data

July 1, 2005
Rishi Sikka, MD
Rishi Sikka, MD

,
Fang Xia, PhD
Fang Xia, PhD

,
Ronald E. Aubert, PhD, MSPH
Ronald E. Aubert, PhD, MSPH

Volume 11, Issue 7

Objectives: To review the definitions and methods for measuringmedication persistency, and to propose a uniform definition ofand calculation for persistency using pharmacy claims data.

Study Design: Literature review.

Methods: A MEDLINE search (1966 to present) was performedto identify articles detailing a definition or method of persistencymeasurement based on automated pharmacy data. Articles werescreened for relevance by title and abstract. References from identifiedarticles were used to expand the search results.

Results: The concept behind medication persistency measurementis to capture the amount of time that an individual remainson chronic drug therapy. The methods to calculate medication persistencycan be classified into 1 of 3 categories: (1) Persistency asa function of the medication possession ratio; (2) persistency as afunction of medication availability at a fixed point in time; and (3)persistency as a function of the gaps between refills.

Conclusions: The common goal of all persistency measuresshould be to reflect the continuity of medication usage and to capturethe timeliness and the frequency of refilling. The measurementof persistency as a function of the gaps between refills provides thebest assessment of refill compliance across a variety of medicationand disease states and lends itself to the well-established measurementsof survival analysis.

(Am J Manag Care. 2005;11:449-457)

Pharmaceuticals have become the primary treatmentmodality for a variety of acute and chronicmedical conditions. Unfortunately, the merereceipt of a prescription does not guarantee betterhealth. A medication's success in producing beneficialeffects depends on a patient's compliance with a therapeuticregimen. Failure to take medication appropriatelyor for the prescribed length of time could lead todetrimental health outcomes and increased healthcareexpenditures.1

Enhancing patient medication compliance hasassumed a prominent role in initiatives directed atimproving the quality of medical care. Improving compliancefirst requires the development of accurate andeasily interpretable measures of medication adherence.Traditionally, medication adherence has been based onpatient self-report, clinician perception, pill counts,pharmacologic tracers, or electronic measurementdevices. More recently, pharmacy claims data havebecome a common tool in the assessment of medicationcompliance. These large population databases affordaccess to a vast amount of information regarding medicationdosing and refilling patterns. Still, the challengeremains to convert these large quantities of claims datainto intuitive and meaningful surrogate measures ofmedication compliance.

Sclar and colleagues provided the first uniformmethodology for estimating medication compliancefrom pharmacy claims data, with the introduction of themedication possession ratio (MPR).2,3 Their work hasled to the widespread adoption of the MPR as a measurementof drug adherence.4-15 The MPR is often definedas the sum of the days' supply of medication divided bythe number of days between the first fill and the lastrefill plus the days' supply of the last refill. This calculationusually results in a ratio less than 1.0 if there arelapses in prescription refilling. Early refilling would leadto an MPR of more than 1.0; the MPR in such a case isoften truncated at the maximum value of 1.0, indicatingthe potential for perfect compliance.

Although the MPR provides insight into the availabilityof medication, it does not provide information on thetimeliness and consistency of refilling. Consider apatient who refills medication every other month duringa 1-year period. Compare this individual with someonewho refills during the first month, followed by a 6-month gap, and then continuous refilling until the endof the year. Both patients have an identical MPR. But,during the course of the entire year, the first patient ismore consistent and timelier with refilling behavior.The MPR fails to capture this particular dimension ofpatient medication adherence.

Persistency with medication refilling is calculated tofill this void in the estimation of medication compliance. Unfortunately, unlike the MPR, the literaturelacks a consistent definition of persistency and a uniformmethod of calculation. Our objective was to reviewthe various definitions and methods for measuring medication-refill persistency and to propose a uniform definitionand calculation for persistency using pharmacyclaims data.

METHODS

Literature Search

persistency

persistence

patient compliance

prescriptions,

drug

drug utilization

An Ovid MEDLINE search (1966 to present) was performedusing the key word (346 citations).A search was also conducted cross-linking the key word(27 792 citations) with the key words(27 046 citations), (12 603 citations), and (9934citations). These key words are terms with predefinedsubject headings in MEDLINE that were searchedaccording to key word. All articles were screened for relevanceby title and abstract. In addition, referencesfrom identified articles were used to expand the searchresults. Inclusion criteria were that an article had tospecifically detail a definition or method of persistencymeasurement based on automated pharmacy data.Although studies measuring medication discontinuationmay have used similar methodologies, these analyseswere not within the scope of this review.

Persistency Definitions

The measurement of medication persistencyattempts to capture the amount of time that an individualremains on chronic drug therapy.16-20 This dimensionof medication compliance introduces an element ofchronology that is absent from a simple MPR measure.17Under this framework, patients are classified as eitherpersistent or nonpersistent with medication therapy forsome duration of time. Individuals who are persistentwith therapy are continuous with their medication-takingbehavior during a certain period. Persistent individualsrefill their medications frequently and regularly. Incontrast, nonpersistent individuals either have sporadicrefilling practices or have discontinued refilling theirmedications completely.19,21

Although the literature often blurs the distinctionbetween nonpersistency and discontinuation, the 2concepts are not necessarily equivalent. An individualwho is nonpersistent with medication therapy may havea momentary, significant lapse in treatment, but maystill resume medication at some point in the future.22,23In contrast, medication discontinuation implies a completecessation of drug therapy with no future resumptionof treatment.22 Medication discontinuation isdifficult to determine with absolute certainty because ofthe long time horizon involved. As a result, a gap intreatment observed during a shorter, more definite periodis often used as an indicator of medication discontinuation.Although many analyses equate the presenceof these gaps to discontinuation, individuals with suchgaps may represent a degree of noncompliance and maystill resume their treatment in the near future.22,23These individuals may be appropriately classified asnonpersistent but not necessarily discontinued.

RESULTS

Persistency Measurements

In our review of medication compliance studiesmeasuring refilling persistency, we identified 3 methodsto calculate persistency with medication. Persistencymay be calculated as a function of the MPR, as a functionof medication availability at a fixed point in time,or as a function of the gaps between refills.

Persistency as a Function of the Medication

Possession Ratio.

Persistency may be defined as anMPR (or similarly calculated ratio, including proportionof days covered) greater than or equal to a predeterminedthreshold, such as 80%.18,24-29 The MPR is a continuousvariable assessing medication availability overmultiple refill intervals. The selection of an absolutecut-off for medication persistency yields a dichotomouspersistency measure for each individual: persistent ornonpersistent.

Several studies have used the concepts of MPR calculationand persistency interchangeably, without theuse of a threshold MPR value to imply persistency.30-34However, an important distinction is that the calculationof the MPR by itself does not convey informationon the timeliness of refilling or persistency. However, ifa certain minimum MPR classifies a patient as persistent,then this measure can convey information on theduration of therapy and the consistency of refilling. AnMPR of 80% is a reasonable threshold for persistencebecause it suggests very few days without drug on handand, consequently, fairly continuous medication usage.An alternate MPR threshold for persistency could beselected given an appropriate clinical or pharmacologicrationale.

The calculation of persistency as a function of theMPR is limited because of its reliance of a uniform follow-up period for all individuals. For instance, individualswith a 3-month follow-up period have fewer days ofobservation and fewer opportunities for noncompliancethan individuals with a follow-up period of 12 months.A uniform follow-up period is required to prevent individuals with shorter follow-up times from biasing theMPR upward.35 As a result, individuals with less claimsdata due to plan ineligibility or insurer switching mustbe excluded from analysis.

Persistency as a Function of Medication Possession

at a Fixed Point in Time.

This methodology measuresthe patient's possession of medication on a fixed dateafter the initial prescription. If patients possess medicationon that specific date, then they are classified as persistentfrom the initial prescription until that date, forthat entire length of time.36-45 Medication possessionmay be defined as either available medication on handor the presence of a medication refill. For example, apatient is classified as persistent at 1 year if, at 12months, he or she has a single day's supply of medicationor a refill available. Similarly, the presence of arefill within a predefined number of days of a specificdate or at any point in a given month would also classifya patient as persistent at that point in time.37,42-45

This persistency measure classifies individuals aspersistent or nonpersistent by measuring medicationpossession at a single refill interval. The timing and gapsbetween other refills is not a consideration. This failureto account for the possibility of large gaps betweenrefills yields a persistency measurement insensitive todiscrete changes in refill behavior.

For example, Figure 1 compares a patient who refillsa prescription monthly during a 12-month period withan individual who refills only 4 times during the sameperiod. Both individuals possess medication at 12months and both individuals are classified as persistentat 1 year. Intuitively, however, the first patient is moreconsistent with therapy throughout the year.

This example illustrates a limitation associated withthis method of persistency measurement. This persistencymeasure will accurately reflect the timeliness ofmedication refilling only if the gaps between refills areknown to be small and insignificant.

Persistency as a Function of the Gaps Between

Refills.

The most widely used method for measuringmedication persistency relies on quantifying the gapsbetween prescription refills (Figure 2).23,46-65 Each individualhas a certain grace period to obtain an additionalrefill. This grace period commences at the end of thesupply of the previous prescription and is equal to one-halfthe days' supply of 1 prescription (in this case one-halfof 30 days, or 15 days). If the patient refills theprescription by the end of the grace period, he or she isclassified as persistent. However, if a patient's refill gapexceeds the predetermined grace period, that patient isconsidered nonpersistent at that point in time.

This persistency measure is a continuous variableassessing the gaps over multiple refill intervals.Unfortunately, the literature lacks a uniform definitionfor the appropriate length of this permissible gap. Thepermissible gap can be a fixed length of time that doesnot depend on the days' supply of the previous refill.This permissible gap has a reported range between 15and 120 days after the previous refill.48,51,52,54-59,61,64Alternatively, the permissible gap may depend on thelength of the days' supply of the previous prescription.A range from one-half to 3 times the days' supply of thepreceding prescription has been used as the length ofthe permissible gap.47,49,50,53,60 In several analyses ofophthalmologic medications, the allowable grace periodfactored in the number of bottles dispensed and aknowledge of historical refilling patterns.23,62,63,65

This method of persistency measurement shares anintuitive relationship with the calculation of the MPR. Ifthe days' supply of a prescription refill is fixed, then theMPR can be made mathematically equivalent to persistencyas measured by the gaps between refills. If weconsider refills of 30-day supplies of medications and apermissible gap of 15 days, on average this measure isequivalent to an MPR of 66% (30/[30 + 15]). Similarly, ifwe assume refills of 30-day supplies of medication, thenan MPR of 80% corresponds to a grace period of about 7days between refills (30/[30 + 7]).

But compared to the MPR measure, the measurementof persistency as a function of the gaps betweenrefills inadvertently may not consider all the refillingbehavior across the observation period. Once an individualis classified as nonpersistent, refilling behavior isno longer considered in an analysis. For instance, consideran individual who misses the permissible gap byonly 1 day early in the observation period, but has nearcontinuous refilling after the classification of nonpersistence.This individual will be classified as nonpersistentearly on and the remaining refilling behavior will not beconsidered.

However, persistency measurement as a function ofthe gaps between refills possesses a unique advantageover other measures because this method readily lendsitself to survival analysis.23,46,48,49,51,52,54,57-63,65 Survivalanalysis is used to describe data defined from some originin time until the occurrence of a specific event. Withregard to medication compliance, the origin in time isthe start of an observation period and the specific eventis the first occurrence of a gap that exceeds the allowablegrace period. Individuals who are persistent untilthe end of the observation period are right-censoredbecause the event of interest, an inappropriately longgap, has not been observed.

This interpretation of refilling behavior permits theconstruction of a survival function and facilitates agraphical representation ofmedication persistency.Using Kaplan-Meier estimates,the distribution ofsurvival times can be computedto describe the persistencyof a patientpopulation. The survivalfunction can also incorporateand control for demographicand clinicalvariables that affect persistency,thus permitting ameaningful comparison ofthe correlates of complianceamong different populations.Changing theallowable grace period foran additional refill also hasa direct impact on the survivalcurves and patientpersistency. Lengtheningthe grace period providesa less stringent requirementfor refilling andtypically increases thenumber of individuals persistentwith therapy.

Comparison of Persistency Measures

Each of the discussed measures has several limitationsin common. Medication refill patterns do not necessarilyreflect the intent or directions of theprescribing physician. Gaps in therapy may appropriatelyresult as a consequence of clinical considerationssuch as abnormal laboratory results, medication titration,adverse drug interactions, or medication sideeffects. In addition, the assumption is that a new refill ofa prescription implies complete ingestion of the previousrefill. However, the possibility exists that an individualmay be providing medication to others, dumpingmedication prior to the refilling, or stockpiling medicationfor future usage. Despite all of these clinical andusage possibilities, for chronic-use medications, administrativedata are good sources for estimating medicationcompliance and persistency and identifyingpatients at risk for therapeutic failure.

Each method of persistency measurement may yielddifferent results for an individual. Figure 3 illustratesthe application of each methodology to the refilling patternof a hypothetical patient. If an MPR of 80% is usedas a criterion for persistency, then this individual isclassified as nonpersistent for the entire observationperiod. When persistency is determined by the presenceof medication possession at 12 months, then thepatient is considered persistent for 1 year. However, ifthe persistency measure is based on a refill gap greaterthan one-half of the days of the previous prescription'ssupply, then the patient is persistent up to 189 days ofthe 12-month period.

Although analyses use these measures to report persistency,each method of persistency measurement maybe actually measuring different phenomena. A usefulmethod to elicit these distinctions involves the use ofthe Steiner typology for refill compliance measurement.9 According to Steiner, refill compliance measurementscan be assessed according to 3 characteristics:(1) the distribution of the compliance variable as eithercontinuous or dichotomous; (2) the evaluation of eithersingle or multiple refill intervals; and (3) the measurementof either medication availability or medicationgaps (Table 1).9

The measurement of persistency as a function of theMPR classifies individuals as persistent or nonpersistentby evaluating multiple refill intervals. The selection ofan appropriate MPR cut-off directly affects this measure'saccuracy of providing information on the continuityof medication usage. In addition, all individualsmust have a uniform follow-up period to prevent biasesin calculating the MPR.35

The measurement of persistency as a function ofmedication possession at a fixed point in time is also adichotomous variable. Individuals are classified as persistentor nonpersistent based on medication on handor the presence of a refill at a fixed time interval. Thismethod accurately measures persistency if the refillgaps are small and infrequent.

The measurement of persistency as a function ofrefill gaps distinguishes itself by assessing refill gaps overmultiple intervals. The length of the follow-up periodacross a cohort of individuals may vary. This measurealso can be applied to a population without knowing thesize of their refill gaps in advance. However, a disadvantageof the measurement of persistency as a function ofrefill gaps is that consistent refilling behavior that occursafter the designation of nonpersistency is not consideredin the analysis.

Further analyses are needed to determine which persistencymeasurement possesses superior external validationto costs or clinical outcomes. However, themeasurement of persistency as a function of the gapsbetween refills has been utilized most frequently in theliterature. The determination of survival times providesadditional information for individuals with an intermediateMPR. The underlying medication compliance of thepopulation also does not restrict the implementation orthe interpretation of this measure. Measuring the gapsbetween refills captures the duration of therapy and theconsistency attributes of persistency and also informs usof the moment when medication refilling becomes irregular.The determination of the specific time when apatient becomes nonpersistent also allows the applicationof the well-developed techniques of survival analysis.

Implementation of a Persistency Measurement

Which Measure to Use to Assess Compliance?

Persistency measurement is only 1 component of theprocess of describing an individual's medication-takingbehavior. The use of this measure does not preclude theuse of other measures (including the MPR) that describedifferent aspects of medication compliance. Painting themost accurate picture of an individual'smedication compliance mayrequire multiple different measuresand techniques.

As noted, the measurement of persistencyas a function of the gapsbetween refills is used in most persistencystudies. However, thisapproach does not negate the valueof other methods of persistencymeasurement. Competing methodologiesmay require less sophisticatedstatistical training and may beeasier to implement. Other methodologiesmay be used as long as theresults generated are interpreted inlight of their discussed limitations.

All 3 techniques of persistencymeasurement have similar datarequirements for implementation. But the use of survivalanalysis in the measurement of persistency as afunction of the gaps between refills may pose some additionalhurdles. Organizations without the technical orpersonnel resources can still perform these analyseswith add-on packages for common spreadsheet softwareand software downloaded from the Internet. Theremainder of this section discusses specific aspectsassociated with the implementation of a persistencymeasure as a function of the gaps between refills.

Medication Compliance Analysis Using Persistency

Measure as a Function of the Gaps Between

Refills.

The first steps in a medication complianceanalysis are identification of the population and medicationof interest and the selection of an analytic startdate. Persistency measurement can be applied to apopulation of new medication users, current users, ora combination of both. A persistency analysis also typicallyconsiders the compliance of 1 medication ormedication class at a time. If new medication users arebeing studied, then the analytic start date should bethe first identified prescription. If the persistencyanalysis involves current users, then the start time ofthe analytic period can be either a fixed date or thedate of the first refill identified after a fixed date.

Once the population of interest and the analytic startdate have been chosen, the next most important issueto be determined is the selection of the length of thepermissible gap between refills. The length of the graceperiod between refills can reflect issues related tomedication half-life, clinical efficacy, dosage titration,or source of refilling. Often, a grace period may be selectedthat is a function of the number of days in the previousprescription's supply. For instance, a retail claimwith a supply of 30 days or less will be assigned a graceperiod of one-half the number of days in the previousprescription's supply. This strategy may be problematicwith mail-order prescriptions, which typicallyinvolve a 90-day drug supply. An allowable gap of onehalfthe previous days' supply for mail-order claimsmay be too generous, often resulting in a 45-day periodfor an additional refill. A stricter but somewhat arbitrarygrace period for refilling of mail-order claims maybe 30 days.

Typically, once an individual exceeds the permissiblegap, his or her refilling behavior after the classificationof nonpersistency is not considered in an analysis.This limitation may be partially corrected by either asensitivity analysis or a secondary analysis of the refillingbehavior of nonpersistent individuals. In a sensitivityanalysis, the length of the permissible gap can bevaried in small increments. If a small change in the permissiblegap results in a large change in the length oftime an individual is persistent, this result would indicatean individual who has fairly continuous refillingafter the initial nonpersistent event. In addition, a secondaryanalysis could be performed of individuals whoare initially classified as nonpersistent. The refillingbehavior of this cohort could be studied separately todetermine the true extent of poor or intermittent complianceafter the initial nonpersistent event.

Table 2 illustrates the calculation of persistencyfrom the claims of an individual new to medical therapy.The administrative data for each claim consist of astart date and the number of days' supply of medication.This information is used to determine the startand the end dates of each medication claim and to evaluatethe gap between the end of the previous supplyand the beginning of the subsequent claim. If a patientrefills early, prior to the end date of the previous claim(as in claim number 3), the end date for the currentrefill is extended to account for the overlapping days ofmedication. The evaluation continues through subsequentclaims within the observation period until theoccurrence of the first prolonged gap exceeding a predeterminedthreshold value. In this particular analysis,the grace period was 15 days after a refill with 30 days'supply and 30 days after a refill with 90 days' supply. Thesurvival time for an individual is then calculated by takingthe difference in days from the start date of the firstclaim to the end date of the last claim preceding the prolongedgap.

This process is repeated on all patients. Patientswho do not have a prolonged gap during the observationperiod and who remain on treatment beyond theobservation period contribute information only untilthe last day of the observation period. These individualsare right-censored because their exact survivaltime becomes incomplete during the follow-up, orright side, of the observation period. Using Kaplan-Meier estimates, the distribution of survival timescan be computed to describe the persistency of apatient population. Plotting the survival functionalong a time scale creates a survival curve depictingthe proportion of patients persistent at any givenpoint in time.

The information from a survival analysis can beused to assess the effect of an intervention aimed atimproving medication persistency. Clinical and demographicfactors between the control and interventiongroups can be controlled in the construction of thesurvival function. An intervention to improve medicationpersistency would be expected to result in a largerproportion of individuals categorized as persistentat the end of the observation period, compared withthe control group. Also, the percentage of people classified as persistent would be greater in the interventiongroup at any given point in time.

CONCLUSIONS

Compared with other techniques for measuringpatient medication compliance, the use of administrativeclaims databases is relatively new. Within thisdomain, the MPR has become an easily applied and validatedtool for measuring refill compliance. However,the MPR does not provide information on the continuityof medication usage. The measurement of medicationpersistency is an attempt to remedy this limitation.A persistency measurement should capture informationregarding the duration of consistent and timely medication-taking behavior. An intuitive starting point is tomeasure the gaps between medication refills.Establishing a threshold for failure to refill in a timelyfashion allows us to draw upon the techniques of survivalanalysis to measure persistency.

The combination of an MPR and a persistency metriccould provide timely information on the dynamics ofpatients' medication compliance. This strategy couldpermit targeted interventions for individuals and populationsat risk for medication noncompliance. The creationof survival curves could also be used to assess theeffect of an intervention on medication compliance. Aspharmacy claims data assume a more prominent role inassessing the quality of care, the techniques of persistencymeasurement will require additional validation asa means of measuring the continuity of medicationrefilling behavior.

From the Department of Medical Affairs, Medco Health Solutions, Franklin Lakes, NJ(RS, FX, REA) and the Department of Emergency Medicine, Boston Medical Center,Boston, Mass (RS).

A portion of this manuscript was presented at the International Society ofPharmacoeconomic and Outcomes Research (ISPOR) 7th Annual European Congress,Hamburg, Germany, October 24-26, 2004.

Address correspondence to: Ronald E. Aubert, MSPH, PhD, Medco Health Solutions,Inc, 100 Parsons Pond Drive, MS F2-4, Franklin Lakes, NJ 07417. E-mail:ronald_aubert@medco.com.

Clin Ther.

1. Kennedy JJ, Coyne JS, Sclar DA. Drug affordability and prescription noncompliancein the United States: 1997-2002. 2004;26:607-614.

Clin Ther.

2. Sclar DA, Chin A, Skaer TL, Okamoto MP, Nakahiro RK, Gill MA. Effect ofhealth education in promoting prescription refill compliance among patients withhypertension. 1991;13:489-495.

Am J Med.

3. Sclar DA, Skaer TL, Chin A, Okamoto MP, Gill MA. Utility of a transdermaldelivery system for antihypertensive therapy. Part 2. 1991;91:57S-60S.

J Clin Pharm Ther.

4. Skaer TL, Sclar DA, Markowski DJ, Won JKH. Effect of value-added utilities onprescription refill compliance and Medicaid health care expenditures—a study ofpatients with non-insulin-dependent diabetes mellitus. 1993;18:295-299.

J Hum

Hypertens.

5. Skaer TL, Sclar DA, Markowski DJ, Won JKH. Effect of value-added utilities onprescription refill compliance and health care expenditures for hypertension. 1993;7:515-518.

J Hum Hypertens.

6. Skaer TL, Sclar DA, Markowski DJ, Won JKH. Utility of a sustained-release formulationfor antihypertensive therapy. 1993;7:519-522.

J Int Med Res.

7. Sclar DA, Robison LM, Skaer TL, et al. Antidepressant pharmacotherapy: economicevaluation of fluoxetine, paroxetine, and sertraline in a health maintenanceorganization. 1995;23:395-412.

Clin Ther.

8. Okano GJ, Rascati KL, Wilson JP, Remund DD, Grabenstein JD, Brixner DI.Patterns of antihypertensive use among patients in the US Department of Defensedatabase initially prescribed an angiotensin-converting enzyme inhibitor or calciumchannel blocker. 1997;19:1433-1445.

J Clin Epidemiol.

9. Steiner JF, Prochazka AV. The assessment of refill compliance using pharmacyrecords: methods, validity, and applications. 1997;50:105-116.

Am J Manag Care.

10. Hurley JS, Frost EJ, Trinkaus KM, Buatti MC, Emmett KE. Relationship of compliancewith hormone replacement therapy to short-term healthcare utilization in amanaged care population. 1998;4:1691-1698.

J

Manag Care Pharm.

11. Blanford L, Dans PE, Ober JD, Wheelock C. Analyzing variations in medicationcompliance related to individual drug, drug class, and prescribing physician. 1999;5:47-51.

Am

Heart J.

12. Roe CM, Motheral BR, Teitelbaum F, Rich MW. Angiotensin-convertingenzyme inhibitor compliance and dosing among patients with heart failure. 1999;138:818-825.

Diabetes Educ.

13. Sclar DA, Robison LM, Skaer TL, Dickson WM, Kozma CM, Reeder CE.Sulfonylurea pharmacotherapy regimen adherence in a Medicaid population: theinfluence of age, gender, and race. 1999;25:531-538.

Pharmacotherapy.

14. Lawrence M, Guay DRP, Benson SR, Anderson MJ. Immediate-release oxybutyninversus tolterodine in detrusor overactivity: a population analysis.2000;20:470-475.

Am J Manag Care.

15. Xuan J, Duong PT, Russo PA, Lacey MJ, Wong B. The economic burden ofcongestive heart failure in a managed care population. 2000;6:693-700.

Curr Hypertens

Rep.

16. Urquhart J. Some economic consequences of noncompliance. 2001;3:473-480.

Manag Care.

17. Dezii CM. Persistency with drug therapy: a practical approach using administrativeclaims data. 2001;10(2):42-45.

JAMA.

18. Benner JS, Glynn RJ, Mogun H, Neumann PJ, Weinstein MC, Avorn J. Long-termpersistence in use of statin therapy in elderly patients. 2002;288:455-461.

Curr Hypertens Rep.

19. Gerth WC. Compliance and persistence with newer antihypertensive agents.2002;4:424-433.

Clin Ther.

20. Richter A, Anton SE, Koch P, Dennett SL. The impact of reducing dose frequencyon health outcomes. 2003;25:2307-2335.

Curr Hypertens

Rep.

21. Dusing R. Adverse events, compliance, and changes in therapy. 2001;3:488-492.

Drug Regimen Compliance. Issues in Clinical Trials

and Patient Management.

22. Metry J-M. Measuring compliance in clinical trials and ambulatory care. In:Metry J-M, Meyer UA, eds. New York: John Wiley & Sons; 1999:1-22.

Am J Ophthalmol.

23. Reardon G, Schwarz GF, Mozaffari E. Patient persistency with topical ocularhypotensive therapy in a managed care population. 2004;137(suppl):S3-S12.

Clin Ther.

24. Rizzo JA, Simons WR. Variations in compliance among hypertensive patientsby drug class: implications for health care costs. 1997;19:1446-1457.

Clin Ther.

25. Degli Esposti L, Degli Esposti E, Valpiani G, et al. A retrospective, populationbasedanalysis of persistence with antihypertensive drug therapy in primary carepractice in Italy. 2002;24:1347-1357.

Am J Cardiol.

26. Kopjar B, Sales AEB, Pineros SL, Sun H, Li YF, Hedeen AN. Adherence withstatin therapy in secondary prevention of coronary heart disease in veterans administrationmale population. 2003;92:1106-1108.

J Clin Hypertens (Greenwich).

27. Esposti LD, Di Martino M, Saragoni S, et al. Pharmacoeconomics of antihypertensivedrug treatment: an analysis of how long patients remain on various antihypertensivetherapies. 2004;6(2):76-82.

Am Heart J.

28. Simpson E, Beck C, Richard H, Eisenberg MJ, Pilote L. Drug prescriptions afteracute myocardial infarction: dosage, compliance, and persistence. 2003;145:438-444.

J Manag Care Pharm.

29. Perwien A, Hall J, Swensen A, Swindle R. Stimulant treatment patterns andcompliance in children and adults with newly treated attention-deficit/hyperactivitydisorder. 2004;10:122-129.

Health Serv Res.

30. Crystal S, Sambamoorthi U, Merzel C. The diffusion of innovation in AIDStreatment: zidovudine use in two New Jersey cohorts. 1995;30:593-614.

JAMA.

31. Avorn J, Monette J, Lacour A, et al. Persistence of use of lipid-lowering medications:a cross-national study. 1998;279:1458-1462.

J

Gen Intern Med.

32. Crystal S, Sambamoorthi U, Moynihan PJ, McSpiritt E. Initiation and continuationof newer antiretroviral treatments among Medicaid recipients with AIDS. 2001;16:850-859.

CMAJ.

33. Pilote L, Beck C, Richard H, Eisenberg MJ. The effects of cost-sharing onessential drug prescriptions, utilization of medical care and outcomes after acutemyocardial infarction in elderly patients. 2002;167:246-252.

Eur Urol.

34. Verhamme KM, Dieleman JP, Bleumink GS, Bosch JL, Stricker BH,Sturkenboom MC. Treatment strategies, persistence of drug use and treatment discontinuationin men with LUTS suggestive of benign prostatic hyperplasia: theTriumph project. 2003;44:539-545.

Am J Manag Care.

35. Johnson ES, Mozaffari E. Measuring patient persistency with drug therapy usingmethods for the design and analysis of natural history studies. 2002;8(suppl):S249-S254.

J Hypertens

Suppl.

36. Caro JJ. Stepped care for hypertension: are the assumptions valid? 1997;15(7):S35-S39.

Clin Ther.

37. Bloom BS. Continuation of initial antihypertensive medication after 1 year oftherapy. 1998;20:671-681.

J Hypertens Suppl.

38. Caro JJ, Speckman JL. Existing treatment strategies: does noncompliance makea difference? 1998;16(7):S31-S34.

Psychiatr Serv.

39. Fairman KA, Drevets WC, Kresiman JJ, Teitelbaum F. Course of antidepressanttreatment, drug type, and prescriber's specialty. 1998;49:1180-1186.

CMAJ.

40. Caro JJ, Salas MM, Speckman JL, Raggio G, Jackson JD. Persistence with treatmentfor hypertension in actual practice. 1999;160:31-37.

CMAJ.

41. Caro JJ, Speckman JL, Salas M, Raggio G, Jackson JD. Effect of initial drugchoice on persistence with antihypertensive therapy: the importance of actualpractice data. 1999;160:41-46.

AIDS Patient Care STDS.

42. Ostrop NJ, Gill MJ. Antiretroviral medication adherence and persistence withrespect to adherence tool usage. 2000;14:351-358.

Diabetes Care.

43. Boccuzzi SJ, Wogen J, Fox J, Sung JCY, Shah AB, Kim J. Utilization of oralhypoglycemic agents in a drug-insured US population. 2001;24:1411-1415.

Clin Ther.

44. Conlin PR, Gerth WC, Fox J, Roehm JB, Boccuzzi SJ. Four-year persistencepatterns among patients initiating therapy with the angiotensin II receptor antagonistlosartan versus other antihypertensive drug classes. 2001;23:1999-2010.

Can J Cardiol.

45. Marentette MA, Gerth WC, Billings DK, Zarnke KB. Antihypertensive persistenceand drug class. 2002;18:649-656.

Pharmacotherapy.

46. Kotzan JA, Martin BC, Wade WE. Persistence with estrogen therapy in a postmenopausalMedicaid population. 1999;19:363-369.

Manag Care.

47. Dezii CM. A retrospective study of persistence with single-pill combinationtherapy vs concurrent two-pill therapy in patients with hypertension. 2000;9(9 suppl):2-6.

Can J Cardiol.

48. Sheehy O, LeLorier J. Patterns of amlodipine and felodipine use in an elderlyQuebec population. 2000;16:1109-1117.

Arch Intern Med.

49. Catalan VS, Couture JA, LeLorier J. Predictor of persistence of use of thenovel antidiabetic agent acarbose. 2001;161:1106-1112.

Clin Ther.

50. Dailey G, Kim MS, Lian JF. Patient compliance and persistence with antihyperglycemicdrug regimens: evaluation of a Medicaid patient population with type2 diabetes mellitus. 2001;23:1311-1320.

Obstet Gynecol.

51. Pilon D, Castilloux AM, LeLorier J. Estrogen replacement therapy: determinantsof persistence with treatment. 2001;97:97-100.

MedGenMed.

52. Becker SL, Dezii CM, Burtcel B, Kawabata H, Hodder S. Young HIV-infectedadults are at greater risk for medication nonadherence. 2002;4(3):21.

J Int

Med Res.

53. Dailey G, Kim MS, Lian JF. Patient compliance and persistence with antihyperglycemiatherapy: evaluation of a population of type 2 diabetic patients. 2002;30:71-79.

Am J Manag Care.

54. Dasgupta S, Oates V, Bookhart BK, Vaziri B, Schwartz GF, Mozaffari E.Population-based persistence rates for topical glaucoma medications measuredwith pharmacy claims data. 2002;8(suppl):S255-S261.

South

Med J.

55. Dezii CM, Kawabata H, Tran M. Effects of once-daily and twice-daily dosingon adherence with prescribed glipizide oral therapy for type 2 diabetes. 2002;95:68-71.

J Hum Hypertens.

56. Hasford J, Mimran A, Simons WR. A population-based European cohort studyof persistence in newly diagnosed hypertensive patients. 2002;16:569-575.

Br J Clin Pharmacol.

57. Larsen J, Andersen M, Kragstrup J, Gram LF. High persistence of statin use ina Danish population: compliance study 1993-1998. 2002;53:375-378.

Am J Manag Care.

58. Shaya FT, Mullins CD, Wong W, Cho J. Discontinuation rates of topical glaucomamedications in a managed care population. 2002;8(suppl):S271-S277.

Eur J Clin Pharmacol.

59. Abraha I, Montedori A, Stracci F, Rossi M, Romagnoli C. Statin compliancein the Umbrian population. 2003;59:659-661.

Am J Manag Care.

60. Aubert RE, Fulop G, Xia F, Thiel M, Maldonato D, Woo C. Evaluation of adepression health management program to improve outcomes in a first or recurrentepisode depression. 2003;9:374-380.

Rheumatology (Oxford).

61. Moride Y, Ducruet T, Rochon S, Lavoie F. Persistency of use of COX-2-specificinhibitors and non-specific non-steroidal anti-inflammatory drugs (NSAIDs) inQuebec. 2003;42(suppl 3):iii17-iii22.

Clin Ther.

62. Reardon G, Schwartz GF, Mozaffari E. Patient persistency with ocularprostaglandin therapy: a population-based, retrospective study. 2003;25:1172-1185.

Eur J Ophthalmol.

63. Reardon G, Schwartz GF, Mozaffari E. Patient persistency with pharmacotherapyin the management of glaucoma. 2003;13(suppl 4):S44-S52.

J

Manag Care Pharm.

64. Wogen J, Kreilick CA, Livornese RC, Yokoyama K, Frech F. Patient adherencewith amlodipine, lisinopril, or valsartan therapy in a usual-care setting. 2003;9:424-429.

Am J Ophthalmol.

65. Schwartz GF, Reardon G, Mozaffari E. Persistency with latanoprost or timololin primary open-angle glaucoma suspects. 2004;137(suppl):S13-S16.