`J Chn Epdenkl
`Cqyright
`0 1997 Elsevier Science
`Inc.
`
`1997
`
`0895-4356/97/$17.00
`PI1 SO895-4356(96)00268-S
`
`ELSEVIER
`
`REPORT
`PHARMACOEPIDEMIOLOGY
`Using Pharmacy
`The Assessment of Refill Compliance
`Records: Methods, Validity,
`and Applications
`
`John F. Steiner,‘,’ and Allan V. Prochazka2~J
`SERVICES RESEARCH,
`‘THE DEPARTMENT
`OF MEDICINE,
`UNIVERSITY
`FOR HEALTH
`‘THE CENTER
`SCIENCES CENTER,
`ANLI
`‘THE DENVER
`DEPARTMENT
`OF VETERANS
`AFFAIRS
`MEDICAL
`CENTER,
`
`OF COLORADO HEALTH
`DENVER, COLORADO 80222
`
`The
`
`ABSTRACT.
`of compli-
`as a source
`increasingly
`are used
`systems
`pharmacy
`of computerized
`records
`refill
`for assessing
`of methods
`a typology
`literature
`to develop
`the English-language
`We
`reviewed
`ance
`information.
`identify
`studies
`of compliance
`in obtaining
`medications,
`to
`the epidemiology
`(RC),
`to describe
`refill
`compliance
`to describe
`the
`clinical
`features
`that
`predicted
`RC,
`and
`to validate
`RC measures,
`to describe
`that
`attempted
`uses of RC measures
`in epidemiologic
`and health
`services
`research.
`In most
`of
`the 41 studies
`reviewed,
`patient.\
`obtained
`less medication
`than
`prescribed;
`gaps
`in
`treatment
`were
`common.
`Of
`the
`studies
`that
`assessed
`the
`validity
`of RC measures,
`most
`found
`significant
`associations
`between
`RC
`and other
`compliance
`measures,
`as
`well
`as measures
`of drug
`presence
`(e.g.,
`serum
`drug
`levels)
`or physiologic
`drug
`effects.
`Refill
`compliance
`was
`generally
`not
`correlated
`with
`demographic
`characteristics
`of study
`populations,
`was higher
`among
`drugs with
`fewer
`daily
`doses,
`and was
`inconsistently
`associated
`with
`the
`total
`number
`of drugs
`prescribed.
`We
`conclude
`that,
`though
`some methodolugic
`problems
`require
`further
`study,
`RC measures
`can he a useful
`source
`of compli-
`ance
`information
`in population-based
`studies when
`direct measurement
`of medication
`consumption
`is not
`feasi-
`hle. Colgright
`1997 Else&r
`Scirnce
`Inc.
`J CLIN EPIIIEMIOL
`50;1:105-116,
`1997.
`
`0
`
`KEY WORDS.
`
`Patient
`
`compliance,
`
`drug
`
`prescriptions,
`
`drug
`
`utilization,
`
`community
`
`pharmacy
`
`INTRODUCTION
`
`begins with appoint-
`Th e process
`compliance
`of medication
`to
`of a prescription
`by submission
`followed
`ment-keeping,
`from
`the pharmacy,
`of medications
`acquisition
`a pharmacy,
`and medication
`consumption
`[l]. Researchers
`have devel-
`oped compliance measures
`for all of the steps in this process,
`hecause
`accurate
`assessment of drug effects
`requires
`evi-
`the medication
`was obtained
`and taken
`[2]. Most
`Jence
`that
`compliance
`studies have assessed medication
`consumption,
`and some have evaluated
`appointment-keeping
`or lapses in
`obtaining
`initial
`drug prescriptions
`[3,4]. Though
`numerous
`measures of medication
`acquisition
`have been proposed,
`the
`validity and utility of these measures have not been assessed
`systematically.
`the cus-
`tracers have been
`Pill counts and pharmacologic
`clini-
`tomary measures of drug consumption
`in randomized
`record
`cal trials
`[5]. Electronic
`compliance
`monitors, which
`the actual
`time at which pill bottles are opened and medica-
`tions are presumably
`taken, have shown
`that pill counts
`typ-
`ically
`overestimate
`medication
`consumption
`and cannot
`evaluate
`the
`timing of doses, which may critically
`influence
`
`Such elec-
`the efficacy and adverse effects of treatment.
`tronic
`compliance
`monitoring
`has become
`the new
`“gold
`standard”
`for pharmacologic
`treatment
`studies
`[5--71. These
`methods
`are difficult
`to use clinically
`because of the ex-
`pense, effort, and
`time necessary
`to obtain measurements.
`Other measures,
`such as serum drug
`levels, assessment of
`physiologic
`drug effects, patient
`self-report,
`and clinician
`assessment have also been evaluated
`in clinical
`settings
`[5,8]. Even
`these measures may be of little use in studies of
`large populations,
`such as pharmacoepidemiology
`or health
`services
`research.
`In such studies, pharmacy
`refill
`records
`can provide
`otherwise
`unobtainable
`information
`about
`the
`pattern
`and
`timing
`of drug exposure,
`and
`the determinants
`and consequences
`of adherence.
`In
`this paper, we will
`(1)
`review
`the methods
`developed
`to assess medication
`refill
`compliance
`from pharmacy
`records and propose a taxonomy
`for classification
`of these measures;
`(2) describe
`the patterns
`of refill compliance
`observed
`in these studies;
`(3) summarize
`the evidence
`for the validity
`of these
`techniques;
`and
`(4)
`evaluate
`the uses and problems
`of refill compliance
`mea-
`sures.
`
`John F. Steuxr,
`for c~xrespdence:
`Piddress
`I iealth Services Research, University
`of C~INXI~
`I355 S. Cdorado
`Roulevard,
`Suite 706,
`Iknvcr,
`Accepted
`fur puhlicarion
`on 25 June, 1996.
`
`for
`Center
`ML)., M.P.H.,
`Health Sciences Center,
`CO 80222.
`
`METHODS OF LITERATURE REVIEW
`
`identify studies
`To
`ance,
`the English
`
`that employed measures of refill compli-
`language
`literature
`was
`reviewed
`from
`
`CFAD VI 1019-0001
`
`
`
`106
`
`J. F. Steiner
`
`and A. V. Prochazka
`
`searches were per-
`to June, 1994. MEDLINE
`January, 1969
`compliance,
`coopera-
`the key words: patient
`formed using
`of health
`care, patient
`patient
`acceptance
`tive behavior,
`dropouts,
`treatment
`refusal, drug prescription,
`drug utiliza-
`tion, community
`pharmacy
`services, and hospital pharmacy
`services. Additional
`papers were
`identified
`from
`the refer-
`ence
`lists of these articles and from searches of the Scientific
`Citation
`Index
`for important
`references.
`All
`studies
`that described measures of refill compliance
`were
`reviewed
`independently
`by the two study investigators
`for information
`in four categories:
`the epidemiology
`of refill
`compliance;
`comparison
`with
`other
`compliance
`measures;
`validation
`through
`association
`with measures of drug pres-
`ence
`(such as serum drug
`levels) or physiologic
`drug effects
`(such as blood pressure control
`for patients prescribed
`anti-
`hypertensive
`drugs); and
`identification
`of clinical
`features
`associated with
`variations
`in refill compliance.
`In some
`in-
`in
`stances, data presented
`the original
`papers could be re-
`analyzed
`to facilitate
`comparisons.
`To estimate
`compliance
`for an entire
`study population
`when
`refill compliance
`was
`reported
`only
`for subgroups, we calculated
`a “weighted”
`measure by multiplying
`the refill compliance
`for each sub-
`group by the number
`of patients
`in that group, and dividing
`the sum of these products
`by the
`total number
`of patients
`in the study. We also conducted
`additional
`analyses of our
`own published
`studies using
`these measures
`to allow better
`comparison
`with
`the work of other
`researchers
`[9-121. After
`the
`independent
`abstraction
`of all papers,
`the reviews were
`compared
`to attain
`a consensus. Meta-analytic
`methods
`were not used
`to aggregate
`study
`findings
`because of the
`diversity of refill compliance
`measures, clinical
`settings, and
`validation
`strategies.
`
`RESULTS
`Study Settings
`
`and Data Requirements
`
`measures
`refill compliance
`that employed
`In all, 41 studies
`States,
`the United
`King-
`were
`identified
`from
`the United
`[9-491.
`These
`studies were
`dom, Australia,
`and Finland.
`care systems that provided
`a
`generally
`conducted
`in health
`all medications
`from pharma-
`financial
`incentive
`to obtain
`cies with
`centralized
`records. Study sites
`included
`United
`States Department
`of Veterans Affairs
`(VA) Medical Cen-
`ters
`[9-12,17,21,22,26,27,29,32,49],
`health maintenance
`organizations
`(HMOs)
`[31,33,36,38,43,47],
`the administra-
`tive data bases of Medicaid
`programs
`[34,35,39,44,46,48],
`medical
`practices
`[13,16,18,28],
`single pharmacies
`[15,25,
`40-421,
`and
`insurance
`plans
`[37,45]. Three
`studies assessed
`the completeness
`of medication
`acquisition
`within
`the sys-
`tem. One study in the Group Health Cooperative
`of Puget
`Sound
`reported
`that more
`than 90% of prescriptions
`were
`filled within
`the HMO
`[31].
`In two VA
`studies, 98-100s
`of patients
`reported
`obtaining
`all medications
`at that
`facility
`PA.
`databases characteristically
`These pharmacy
`drug name, drug dosage
`(milligrams
`per pill),
`
`the
`included
`the quantity
`
`the
`fill, and
`at each pharmacy
`dispensed
`of medications
`dates of prescription
`fills. Though
`dosage
`instructions
`(pills
`per dose and
`total doses per day) were available
`in most
`pharmacy
`systems, studies
`in Medicaid
`programs
`imputed
`dosage
`instructions
`from pharmacy
`policies
`restricting
`pre-
`scription
`size to a defined
`“days’ supply”
`[48].
`
`Typology
`
`of Measures
`
`of Refill Compliance
`
`in these studies could be
`The measures of refill compliance
`(1) the distribution
`of the
`characterized
`by three attributes:
`(C) versus dichotomous
`compliance
`variable
`[continuous
`(D)];
`(2) the number of refill
`intervals evaluated
`[either sin-
`gle
`(S) or multiple
`(M)
`intervals];
`and
`(3)
`the use of the
`measure
`to assess either
`the
`time period over which medica-
`tions were available
`(A)
`to the patient
`or the
`time
`intervals
`during which
`gaps (G)
`in therapy occurred. A simple nota-
`tion
`consisting
`of a combination
`of three
`letters
`(e.g.,
`CMA)
`thus defines a typology of refill compliance measures.
`Calculation
`of each
`type of measure
`is demonstrated
`using
`hypothetical
`refill data
`(Table
`l), beginning
`with an initial
`fill on day 0 and ending on day 450 after a total of six medi-
`cation
`fills. Figure
`1 demonstrates
`the
`time span of each
`prescription
`for these hypothetical
`data, periods of overlap,
`treatment
`gaps, and
`fluctuations
`in two
`typical compliance
`In
`indices.
`the sample data,
`the
`initial
`three prescriptions
`are for a 30-day supply, while
`the subsequent
`three prescrip-
`tions are for 90-day supplies.
`If the final day of the measure-
`ment period
`is defined by the date of the
`last medication
`fill in the series (Day 250 in Table
`I), the calculated
`compli-
`ance
`indices
`can be defined as “embedded”
`in
`the series of
`refills.
`If the end of the period
`of observation
`is identified
`by an arbitrary
`date
`(Day 450
`in Table
`l),
`the compliance
`measures
`include
`a “terminal
`gap” after
`the
`final prescrip-
`tion
`fill.
`
`(1) CONTINUOUS,
`
`OF MEDI-
`MEASURES
`SINGLE-INTERVAL
`the hypothetical
`data
`Using
`(CSA).
`AVAILABILITY
`CATION
`for each of the six refill
`intervals
`in Table 1, a CSA measure
`the days’ supply obtained
`during
`is calculated
`by dividing
`each
`interval by the total days in the interval.
`For example,
`in
`interval
`3 of the hypothetical
`data, 30 days of medica-
`tions are obtained
`over a period
`of 90 days, for a CSA of
`0.33.
`
`(2) CONTINUOUS,
`
`OF MEDI-
`MEASURES
`SINGLE-INTERVAL
`time periods
`identify
`Such measures
`(CSG).
`GAPS
`CATION
`by assuming
`is unlikely
`during which medication
`exposure
`that
`the medication
`is taken exactly as prescribed
`until
`the
`supply
`is exhausted,
`though patients may in fact be partially
`compliant
`with
`their drugs
`throughout
`such an interval.
`In
`interval
`3 of Table
`1, the 30-day supply
`is presumed
`to have
`been depleted
`by day 90,
`leaving a 60-day medication
`gap
`until
`the next
`fill on day 150, so that CSG = 60/90 = 0.67.
`When
`no gap occurs, as in intervals
`1, 2, 4, and 5 of Table
`1, CSG = 0. During
`intervals
`in which
`gaps occur, CSA
`=
`(1 - CSG).
`
`CFAD VI 1019-0002
`
`
`
`Refill Compliance Using Pharmacy Records
`
`TABLE
`
`1. Hypothetical
`
`refill
`
`compliance
`
`data
`
`and
`
`calculation
`
`of continuous
`
`compliance
`
`indices
`
`Prescription
`interval
`
`Days’
`supply
`obtained
`
`Day
`of
`fill
`
`in
`
`Days
`interval”
`
`Single
`interval
`compliance
`(CW
`
`Cumulative
`days’
`SUPPlY
`obtained
`
`1
`2
`3
`4
`5
`6
`
`0
`30
`60
`150
`200
`250
`450'
`
`30
`30
`30
`90
`90
`90
`-
`
`30
`30
`90
`50
`50
`200
`
`1.00
`1 .oo
`0.33
`1.80
`1.80
`0.45
`
`30
`60
`90
`180
`270
`360
`
`Continuous
`measure
`medication
`acquisition
`(CM&
`
`1 .oo
`1.00
`0.60
`0.90
`1.08
`0.80
`
`of
`
`with
`Days
`treatment
`gap
`h
`interval
`
`Single
`interval
`gap
`(CSG)
`
`of
`
`Total
`days
`treatment
`gap’
`
`0
`0
`60
`0
`0
`110
`
`0.00
`0.00
`0.67
`0.00
`0.00
`0.55
`
`0
`0
`60
`60
`60
`90
`
`107
`
`of
`
`Continuous
`measure
`medication
`gaps
`(CMG)
`
`0.00
`0.00
`0.40
`0.30
`0.24
`0.20
`
`C&&tion
`Cumulative
`interval)/Days
`of ohservatwn
`S’Defined
`hAfter
`‘Arbitrary
`
`cwnpl~ancc
`refill
`of cuntinwus
`days’ supply &tained/T<)tal
`In
`interval,
`CSG
`= 0.00
`period.
`as days
`currectlon
`day
`
`CSA
`(ahhreviations):
`indices
`of observation
`fill
`or end
`to next
`days
`if days
`in
`interval
`5 days’
`supply
`
`= L?ays’
`period;
`obtained;
`
`supply
`CSG
`CMG
`
`obtained
`=
`(Days
`= Total
`
`interval/Days
`of
`at heginning
`~ days’
`supply
`in mterval
`days
`of
`treatment
`gaps/Total
`
`m
`ohrained
`days
`
`=
`CMA
`interval;
`of
`at beginning
`to next
`fill
`or end
`
`period.
`in previous
`
`umxvals.
`
`until
`br
`ending
`
`fill or end
`next
`oversupplies
`any
`observation
`
`of observation
`obtained
`period.
`
`OF
`MEASURES
`MULTIPLE-INTERVAL
`CONTINUOUS,
`(3)
`This measure
`is gener-
`MEDICATION
`AVAILABILITY
`(CMA).
`ally calculated
`as the sum of the days’ supply obtained
`over
`a series of intervals, divided by the total days from the begin-
`ning
`to end of the
`time period.
`The hypothetical
`compli-
`ante data
`in Table
`1 demonstrate
`a common
`discrepancy
`between
`the
`“embedded”
`CMA measure
`(270 days worth
`
`= 1.08)
`over 250 days, CMA
`obtained
`of medications
`1 and pre-
`between
`prescription
`which
`can be calculated
`(360 days
`lower estimate of CMA
`scription
`6, and the much
`worth
`of medications
`obtained
`by the end of observation
`on day 450, CMA
`= 0.8)
`if the
`terminal
`gap after interval
`6 is included
`in the calculation.
`In some studies,
`the number
`of refills obtained
`(rather
`than
`the days’ supply obtained)
`
`1. The
`the
`time
`fills
`drug
`
`FIGURE
`depicts
`prescription
`of
`periods
`treatment
`and
`portrays
`panel
`lower
`in
`compliance
`tions
`(a continuous,
`CMA
`measure
`interval
`tion
`availability)
`continuous
`measure
`over
`the
`
`of
`same
`
`panel
`upper
`of
`six
`span
`Table
`1,
`from
`availability
`The
`gaps.
`fluctua.
`indices
`multiple-
`of medica-
`and CMG
`multiple-interval
`treatment
`time
`
`90
`
`SO
`
`SO
`
`Fill 6
`
`Fill 6
`
`Fill 4
`
`Fill 3
`
`Fill 2
`
`Fill 1
`
`Days Supply
`
`Available
`
`30
`
`10
`
`30
`
`SO
`
`Treatment
`
`Gaps
`
`60
`
`270
`
`30
`
`0
`
`too
`
`zoo
`
`300
`
`400
`
`Time
`
`(Days)
`
`(a
`
`gaps)
`period.
`
`CFAD VI 1019-0003
`
`
`
`108
`
`J. F. Steiner and A. V. Prochazka
`
`of refills
`ordinal
`
`time
`in a given
`CMA
`measure
`
`number
`as an
`
`by the expected
`divided
`span was
`calculated
`[17,22,27,321.
`in dif-
`oversupplies
`CMA measures described medication
`accumu-
`ferent ways. For example,
`in Table
`1, the patient
`prescrip-
`lates an 80-day oversupply
`of medications
`during
`tion
`intervals
`4 and 5. Such an oversupply
`can be subtracted
`from 1.0 to reflect variance
`from perfect compliance
`[23],
`or the proportion
`of medications
`obtained
`can be greater
`than 1 .O as exemplified
`in Table
`1, where CMA
`= 1.08 at
`the end of interval
`5 [9,16].
`
`(4)
`
`CONTINUOUS,
`
`OF
`MEASURES
`MULTIPLE-INTERVAL
`total
`divides
`the
`This measure
`(CMG).
`GAPS
`MEDICATION
`treatment
`gaps by the duration
`of the
`of days in
`number
`time period of interest.
`In Table
`1,90 days without medica-
`tion occurred
`over 450 days of observation,
`so that CMG
`=
`0.20. Some CMG measures adjust
`for oversupplies
`obtained
`during previous prescription
`intervals which
`reduces
`the du-
`ration
`of treatment
`gaps [9]. In Table
`1, the 9O-day supply
`obtained
`on Day 250 and
`the 80-day surplus already avail-
`able are presumed
`to be depleted
`by Day 450,
`the end of
`the observation
`period,
`resulting
`in 200
`-
`(90 + 80) = 30
`days of treatment
`gap during
`interval 6, rather
`than
`the 1 IO-
`day gap if only data
`from
`interval
`6 were used.
`the last
`The end date of the analysis period may be either
`refill date
`[9] or an arbitrary
`date, such as the end of a calen-
`dar year [39,46,48].
`In the first case, all gaps are “embedded”
`within
`a series of fills. In the second case, a “terminal
`gap”
`may be present after
`the
`last refill. Both measures assume
`that gaps are due
`to reduced compliance
`rather
`than
`to cli-
`nician
`instructions
`for temporary
`(in
`the case of “embed-
`ded” gaps) or permanent
`(in
`the case of “terminal”
`gaps)
`drug cessation,
`or to acquisition
`of drugs outside
`the phar-
`macy system. When
`a measure of embedded
`gaps
`is used,
`the CMG
`index can no longer be directly
`derived
`from
`the
`value of the CMA
`index
`(unlike
`the
`relationship
`between
`CSA and CSG), because periods of oversupply
`can be inter-
`spersed with periods
`of medication
`depletion,
`as occurs at
`the end of interval
`5 in Table
`1. When
`a terminal
`gap
`is
`=
`sufficiently
`long
`to deplete
`all oversupplies,
`CMA
`(1 -
`CMG),
`as illustrated
`at the end of interval
`6 in Table
`1.
`
`(5-H)
`
`DICHOTOMOUS,
`
`SINGLE-,
`
`OR MULTIPLE-INTERVAL
`
`MEASURES
`
`OF MEDICATION
`
`FOR MEDICATION
`AVAILABILITY
`These measures used dichoto-
`DMG).
`DMA,
`DSG,
`(DSA,
`GAPS
`“compliant”
`from
`“partially
`to distinguish
`cutoffs
`mous
`compliant”
`individuals.
`In many cases, dichotomous
`mea-
`sures were created
`from continuous
`indices, using various
`cutoffs with no clinical
`or pharmacological
`rationale
`offered
`for the choice of a particular
`threshold
`value. Alternatively,
`patients were defined as noncompliant
`if a gap of a specified
`length
`was
`identified
`(over a period
`of multiple
`refills
`[20,26,29,37,41].
`For example,
`the patient
`obtaining
`the hy-
`pothetical
`refill data
`in Table
`1 would be classified as non-
`compliant
`by some DMG
`definitions
`because of the exis-
`
`tence of two prolonged
`CMA
`or CMG.
`
`treatment
`
`gaps,
`
`regardless
`
`of his
`
`Epidemiology
`
`of Refill Compliance
`
`the
`that described
`of studies
`findings
`the
`2 presents
`Table
`using CMA
`or DMA mea-
`epidemiology
`of refill compliance
`sures, while Table 3 describes
`the epidemiology
`of multiple-
`interval medication
`gaps (CMG
`or DMG). Mean CMA was
`less than
`I .O in 17 of the 20 studies
`in which
`it was assessed.
`Thus, most patients
`obtained
`less of their medication
`than
`was prescribed
`over
`time periods
`ranging
`from 2 to 24
`months. The wide
`range
`in refill compliance
`among
`individ-
`uals
`is indicated
`by
`the
`large standard
`deviation
`around
`these mean values. The compliance
`distributions
`for CMA
`measures generally were unimodal,
`bell-shaped,
`and skewed
`toward
`reduced
`refill compliance
`[l I]. The
`three studies
`in
`which CMA was greater
`than 1.0, indicating
`acquisition
`of
`more medication
`than prescribed,
`were conducted
`in VA
`Medical
`Centers which
`routinely
`dispensed
`90-
`to loo-day
`medication
`supplies
`for inexpensive,
`long-term
`drugs. One
`study demonstrated
`that distribution
`of such
`large medica-
`tion supplies
`increased
`overall medication
`acquisition
`and
`reduced
`gaps in treatment
`[12].
`that al-
`Six studies
`in Table
`2 reported CMA measures
`lowed assessment of drug stockpiling,
`defined as acquisition
`in
`of a 10% surplus or more. The prevalence
`of stockpiling
`The
`these studies was between
`4.8% and 35.1%
`[9-13,161.
`were
`characteristics
`of patients who stockpiled medications
`not described,
`and no attempt
`was made
`to determine
`whether
`stockpiling
`led
`to overconsumption
`or simply
`hoarding
`of drugs.
`that
`four studies using CMG measures
`In Table
`3, the
`gaps [9- 121 reported gaps in treatment
`assessed “embedded”
`only about half as large as the studies
`that
`included
`“termi-
`nal gaps”
`[39,46].
`
`to
`
`Refill Compliance
`Between
`Associations
`Other Adherence
`Measures
`
`Measures
`
`and
`
`comparisons
`statistical
`reported
`review
`in our
`Five studies
`and other compliance
`between measures of refill compliance
`between
`refill compli-
`behaviors
`(Table
`4). The association
`ance and appointment-keeping
`was statistically
`significant,
`but weak
`(r = 0.20),
`in one study [22]. Of the
`four studies
`that correlated
`refill compliance
`with measures of self-re-
`ported medication
`consumption,
`two
`[16,25]
`reported
`sig-
`nificant
`associations, while
`one
`[29] did not.
`In the
`fourth
`study [I I], the overall correlation
`between CMA
`and CMG
`refill compliance
`measures and self-reported
`compliance,
`as
`measured
`by a validated
`four-item
`self-reported
`scale
`[50],
`was not statistically
`significant.
`However,
`patients
`provid-
`ing “noncompliant”
`responses
`to all four questions
`had sig-
`nificantly
`lower CMA
`(0.89 5 0.14)
`than
`those who gave
`“compliant”
`responses
`to one or more questions
`(1.06 +
`
`CFAD VI 1019-0004
`
`
`
`109
`
`of
`
`Prevalence
`drug
`stockpiling
`(CMA
`2 110%)
`
`Refill
`
`Compliance
`
`Using
`
`Pharmacy
`
`Records
`
`TABLE
`
`2. The
`
`epidemiology
`
`of
`
`refill
`
`compliance:
`
`Multi-interval
`
`measures
`
`of medication
`
`availability
`
`Study
`
`[I31
`
`[I41
`
`:t;;
`
`119,231
`
`WI
`
`I241
`[421
`I271
`
`PI
`
`[321
`[lOI
`IllI
`1341
`[351
`
`t:';;
`[381
`[I21
`
`n
`
`58
`62
`59
`
`324
`
`171
`
`Setting
`
`British
`
`general
`
`practice
`
`hospital
`general
`practice
`
`University
`British
`VAMC
`
`Finnish
`
`population
`
`survey
`
`VAMC
`
`claims
`
`care
`
`plans
`
`Duration
`(months)
`
`11
`4
`12
`24
`6
`6
`2
`
`6
`
`14
`
`4
`
`7
`
`Medication(s)
`
`CMA
`
`(*SD)
`
`supplements
`
`prescribed
`All
`Prenatal
`iron
`Atropine
`All
`prescribed
`Non-PRN
`drugs
`PRN
`drugs
`Antihypertensives
`
`0.24
`i
`2 0.24
`
`0.84
`0.67
`0.56
`ND
`0.64
`0.40
`ND
`
`Arthritis
`
`drugs
`
`0.64
`
`? 0.32
`
`drugs
`
`drugs
`
`supplements
`
`+ 0.36
`
`5 0.21
`2 0.23
`
`!I 0.47
`k 0.25
`
`? 0.25
`? 0.28
`
`1058
`276
`30
`30
`52
`73
`93
`85
`118
`1135
`8894
`453
`19029
`170
`176
`114
`2289
`78
`119
`
`pharmacies
`100 private
`hospital
`University
`consultation
`VAMC-pharmacist
`VAMC-no
`consultation
`VAMC
`VAMC
`VAMC
`VAMC
`VAMC
`Medicaid
`Medicaid
`3 HMOs
`Insurance
`HMO
`10 VAMCs
`VAMC
`Managed
`Medicaid
`HMO
`
`tz
`1471
`
`12
`3
`6
`6
`-+ 4
`15
`14 ?
`5
`ND
`12
`t
`12
`12
`6
`20
`4
`925
`14
`t
`12
`12
`12
`
`prescribed
`All
`Cardiovascular
`Theophylline
`Theophylline
`Phenytoin
`Antihypertensives
`Lithium
`carbonate
`All
`prescribed
`Antihypertensives
`Clonidine
`Antihypertensives
`Atenolol
`12 selected
`Pentoxifylline
`All prescribed
`Digoxin
`Potassium
`Glyburide
`Theophylline
`Inhaled
`steroids
`Inhaled
`cromolyn
`
`0.54
`0.62
`0.96
`0.76
`0.91
`1.02
`0.73
`1.09
`0.96
`0.67
`0.62
`0.49
`0.72
`0.58
`0.92
`1.04
`0.74
`0.58
`0.79
`0.54
`0.44
`
`k 0.07
`-t 0.34
`2 0.43
`? 0.34
`
`Ahhreviations:
`
`ND
`
`= no data,
`
`VAMC
`
`= Veterans
`
`Affairs
`
`Medical
`
`Center,
`
`HMO
`
`= health
`
`maintenance
`
`organization
`
`12.1%
`4.8%
`ND
`11.9%
`ND
`ND
`(>lOO%
`11.5’S
`compliance)
`4.1%
`(>lOO%
`compliance)
`ND
`ND
`ND
`ND
`15.4%
`33.0%
`ND
`27.1%
`23.9%
`ND
`ND
`ND
`ND
`ND
`13.0%
`35.1%
`ND
`ND
`ND
`ND
`ND
`
`0.26,
`reported
`a DMG
`but
`not
`weakly
`study
`
`p = 0.02).
`compliance
`measure
`CMA
`with
`[ 111.
`In
`
`small
`In one
`than
`were
`refill
`compliance
`of
`calculated
`with
`provider
`assessments
`the
`one
`study
`that
`
`the
`
`individuals
`more
`23%
`study,
`as compliant
`identified
`A CMG
`measure,
`[29].
`data,
`correlated
`same
`of
`compliance
`in
`correlated
`refill
`compliance
`
`by
`
`one
`
`pill
`
`with
`strongly
`< 0.001).
`of partial
`tecting
`could
`
`CMA
`[16],
`counts
`pill
`with
`associated
`in
`a graph
`From
`(~80%)
`compliance
`partial
`compliance
`be estimated
`as 53%,
`
`was
`period
`two-year
`a
`over
`(r = 0.68,
`p
`compliance
`count
`the
`sensitivity
`that
`publication,
`de-
`refills
`for
`in
`obtaining
`consuming
`them
`(~80%)
`in
`with
`a specificity
`of 93%.
`No
`
`TABLE 3. The
`
`epidemiology
`
`of
`
`refill
`
`compliance:
`
`Multi-interval
`
`measures
`
`of medication
`
`gaps
`
`Study
`
`[91
`
`1101
`[Ill
`[I21
`
`n
`
`52
`73
`85
`118
`176
`114
`2440
`7247
`
`Setting
`
`VAMC
`VAMC
`VAMC
`VAMC
`10 VAMCs
`VAMC
`Medicaid
`Medicaid
`
`Duration
`(months)
`
`15 z 4
`14 2 5
`12
`14 2 4
`9-+5
`14 s 7
`12
`12
`
`Medication(s)
`
`Phenytoin
`Antihypertensives
`All
`prescribed
`Antihypertensives
`All prescribed
`Digoxin
`drugs
`Glaucoma
`Heart
`failure
`drugs
`
`CMG
`of
`
`Proportion
`medication
`
`without
`time
`( f SD)
`
`k 0.18
`0.16
`-+ 0.12
`0.10
`-t 0.09
`0.14
`z 0.14
`0.13
`0.15
`i
`0.18
`5 0.14
`0.10
`2 0.31
`0.31
`0.30 2 0.30
`
`reviations:
`Ahh
`“Studies
`
`using
`
`= na data,
`ND
`a specific date,
`
`VAMC
`rather
`
`= Veterans
`than
`the
`
`Affairs
`last medicanon
`
`Medical
`
`fill,
`
`Center.
`to define
`
`a “termmal
`
`gap”
`
`in
`
`compliance
`
`CFAD VI 1019-0005
`
`
`
`110
`
`J. F. Steiner and A. V. Prochazka
`
`TABLE
`
`4. Associations
`
`between
`
`refill
`
`compliance
`
`and
`
`other
`
`compliance
`
`measures
`
`Study
`
`[I61
`
`ts:;
`WI
`1111
`
`R
`
`59
`
`171
`49
`62
`59
`
`Refill
`
`compliance
`method
`
`Duration
`(months)
`
`Medication(s)
`
`CMA
`
`CMA
`DMG
`DMA
`CMA
`
`CMG
`
`24
`
`All prescribed
`
`6
`12
`12
`14 k 4
`
`drugs
`Arthritis
`Anti-epileptics
`Psychiatric drugs
`Antihypertensives
`
`14 +- 4
`
`Antihypertensives
`
`Other
`compliance
`
`measure
`
`Pill count compliance
`Self-report
`Appointment-keeping
`Self-report
`Self-report
`Self-report
`Provider assessment
`Self-report
`Provider assessment
`
`Association
`
`(1, < 0.001)
`r = 0.68
`(p < 0.001)
`I = 0.47
`(p = 0.005)
`r = 0.20
`(p < 0.001)
`x’ = 25.42
`“No significant association”
`r = 0.11 (I, = 0.44)
`(p = 0.11)
`r = 0.22
`r =
`-0.05
`(p = 0.69)
`r =
`-0.30
`(p = 0.03)
`
`.
`Ahhreviations: CMA =
`t’
`availability
`multqde-interval
`LOII mucus
`multiple-interval availability measure, DMG
`= dichotomous
`
`= continuous
`CMG
`measure,
`multiple-interval gap measure, ND
`
`multiple-interval
`= no data.
`
`gap measure,
`
`DMA
`
`= dichotomous
`
`con-
`with medication
`refill compliance
`studies correlated
`sumption
`as measured
`by electronic medication
`monitors.
`
`the
`spite
`[38,481.
`
`reduction
`
`in drug costs due
`
`to poor compliance
`
`Association
`Presence
`
`Between
`or Effect
`
`Refill Compliance
`
`and Drug
`
`Characteristics
`Clinical
`Compliance
`
`Associated
`
`with Refill
`
`or CMG
`of CMA
`studies assessed the correlation
`Three
`5). All
`levels
`(Table
`measures with
`serum or urine
`drug
`found statistically
`significant
`associations,
`with
`correlation
`coefficients
`ranging
`from 0.2 1 to 0.47
`[9,11,14]. Of the five
`studies
`that correlated
`refill
`compliance
`with measures of
`drug effect such as blood pressure
`(in patients on antihyper-
`tensive drugs) or pulse
`rate (in patients
`on beta-adrenergic
`blockers),
`statistically
`significant
`associations were observed
`in four
`(Table
`5) [9,11-14,17,26].
`Only
`two studies evalu-
`ated
`the persistence
`of associations
`between
`refill compli-
`ance and drug presence
`or effect after multivariate
`adjust-
`ment
`for other predictors
`of drug action,
`such as prescribed
`dose, body weight,
`and renal
`function
`[12,17].
`In both cases,
`refill compliance
`remained
`a significant
`predictor
`of drug
`effect.
`
`Associations
`Outcomes
`
`Between
`
`Refill Compliance
`
`and Health
`
`was
`of measures of refill compliance
`validity
`The predictive
`assessed in a few studies by the association
`of these measures
`with clinical
`outcomes, health
`services utilization,
`or health
`care costs. Maronde
`et al.
`[30]
`found a substantial
`rise
`in
`hospitalizations
`for uncontrolled
`hypertension
`among
`indi-
`viduals with prolonged
`gaps in acquisition
`of antihyperten-
`sive drugs. Steiner et al. [ 1 l] d emonstrated
`that previous par-
`tial
`compliance
`with
`antihypertensive
`drugs
`predicted
`success in subsequent
`clinical
`efforts
`to reduce unnecessary
`medications.
`Psaty et al. [31] showed
`that gaps
`in
`therapy
`with beta-adrenergic
`blockers were associated with a time-
`dependent
`increase
`in
`the
`rate of acute myocardial
`in-
`farction,
`suggesting
`a previously
`unrecognized
`drug with-
`drawal
`effect. Two
`studies
`found an increase
`in the overall
`costs of medical
`care
`for partially
`compliant
`patients,
`de-
`
`in Tables 6 and 7 assessed associations
`The studies described
`of the patient
`or treatment
`regimen
`between
`characteristics
`In many of these studies, a large num-
`and refill compliance.
`ber of potential
`predictors
`were
`tested
`for association with
`refill compliance,
`increasing
`the
`likelihood
`of “false-posi,
`tive”
`associations.
`Conversely,
`studies may not have
`re-
`ported
`clinical
`features with
`statistically
`insignificant
`as-
`sociations
`with
`refill
`compliance.
`Finally, multivariate
`statistical
`analyses
`to identify
`independent
`predictors
`of re-
`fill compliance
`were often not performed.
`behaviors
`Similar
`to studies evaluating
`other compliance
`[51], no consistent
`relationship
`was observed between
`socio-
`demographic
`variables
`or disease characteristics
`and
`refill
`compliance
`(Table
`6). A number
`of studies observed
`sub-
`stantial
`differences
`in refill compliance
`among drugs (Table
`7)
`[17,22,24,35,36,37,43,45,47,48].
`In some cases, drugs
`with
`lower
`refill compliance
`appeared
`to be intended
`for as-
`needed
`(PRN) use [ 17,241. Substantial
`differences were also
`observed
`in refill compliance
`among drugs
`in the same ther-
`apeutic
`class which
`were
`intended
`for daily use.
`[22,35,
`36,43,45,47,481.
`among different
`refill compliance
`Three studies compared
`but reached dif-
`drugs [35,37,48],
`classes of antihypertensive
`the magnitude
`of compliance
`and
`ferent conclusions
`about
`of drugs
`in different
`pharmaco-
`about
`the relative
`ranking
`logic classes. Two
`studies of refill compliance
`with antide-
`pressant drugs
`in separate samples of patients
`from
`the same
`HMO
`[36,43] showed
`the highest
`compliance
`with
`fluoxe-
`tine, but an inconsistent
`rank order of compliance
`for other
`antidepressant
`drugs. Assignment
`to treatment was not ran-
`dom
`in any of these studies, and only one demonstrated
`persistent
`differences among drugs after adjustment
`for char-
`acteristics
`of the patient
`or the
`treatment
`regimen
`[22].
`In recent years, studies
`tising electronic medication
`moni-
`
`CFAD VI 1019-0006
`
`
`
`&
`‘j
`E
`P
`4
`6
`3
`r
`‘U
`
`gap measure, ND = no data.
`
`multiple-interval
`
`> 1.10
`
`compliance
`
`with mean
`
`patients
`
`excluded
`
`analysis
`
`Subgroup
`
`11.10
`
`compliance
`
`with mean
`
`-
`
`patients
`
`excluded
`
`analysis
`
`Subgroup
`
`(p = 0.04)
`
`propranolol
`
`for prescribed
`
`dose
`Adjustment
`
`-
`
`(p < 0.05)
`
`reported
`
`test not
`
`Statistical
`
`and
`
`weight,
`
`dose
`
`digoxin
`for creatinine,
`
`prescribed
`
`Adjustment
`
`>l.lO
`
`compliance
`
`with mean
`
`patients
`
`excluded
`
`analysis
`
`Subgroup
`
`(p = 0.005)
`
`>l.lO
`
`compliance
`
`with mean
`
`patients
`
`excluded
`
`analysis
`
`Subgroup
`
`-
`
`Comments
`
`analyses
`
`or
`
`in
`
`= dichotomous
`
`DMG
`
`gqz measurr,
`
`nudtiplc-interval
`
`ND
`
`(p = 0.02)
`
`r = -0.49
`
`rate
`
`r = 0.33 (p = 0.02)
`
`(p = 0.15)
`
`r = 0.17
`
`r = -0.30
`
`ND
`
`(p = 0.23)
`
`r = -0.14
`
`0.89 (p < 0.001)
`
`to
`
`4 = 0.65
`
`Cramer’s
`
`r = -0.41
`
`(p > 0.05)
`
`r = -0.27
`
`ND
`
`r = 0.63 (p < 0.05)
`
`ND
`
`significant
`
`Not
`
`effects
`
`drue
`
`r = 0.25 (p = 0.03)
`
`(0 = 0.06)
`
`r = 0.21
`
`-0.42
`
`I =
`
`(p = 0.004)
`
`-0.40
`
`r =
`
`r = 0.37 (p = 0.01)
`
`(p = 0.03)
`
`r = 0.31
`
`ND
`
`r = 0.47 (p < 0.05)
`
`subgroup
`
`adjusted
`Association
`
`association
`
`Crude
`
`levels
`
`drug
`
`or urine
`
`effects
`
`drug
`
`or physiologic
`
`= contmuws
`
`measure, CMG
`
`availahiltty
`
`pulse
`
`Resting
`
`pressure
`
`blood
`
`Diastolic
`
`pressure
`
`blood
`pressure
`
`Diastolic
`blood
`
`Systolic/diastolic
`
`rate
`pressure
`
`Pulse
`
`blood
`
`concen-
`in hemo-
`
`Diastolic
`tration
`globin
`
`Change
`
`B. Phvsiologic
`
`levels
`
`digoxin
`
`Serum
`
`levels
`
`phenytoin
`
`Serum
`
`levels
`
`phenytoin
`
`Serum
`
`metabolites
`
`atropine
`
`Urine
`
`Beta-blockers
`
`Antihypertensives
`
`Antihypertensives
`
`Antihypertensives
`
`Propranolol
`
`Hydrochlorothiazide
`
`supplements
`
`Iron
`
`Dtgoxm
`
`Phenytoin
`
`Phenytoin
`
`Atropine
`
`or effect
`of drug
`
`presence
`Measure
`
`Medication(s)
`
`A. Serum
`
`levels
`
`drug
`
`serum
`
`and
`
`compliance
`
`multiple-interval
`
`= continuous
`
`CMA
`
`Ahhteviations:
`
`-C 4
`
`12
`
`14 ? 5
`
`14 ? 5
`
`12
`
`6
`
`6
`
`4
`
`925
`
`15 + 4
`
`15 ?4
`
`12
`
`(months)
`Duration
`
`CMA
`
`CMG
`
`CMA
`
`DMG
`
`CMA
`
`CMA
`
`25
`
`[II]
`
`49
`
`55
`
`27
`
`25
`
`[91
`
`[26]
`
`[17]
`
`CMA
`
`62
`
`[13]
`
`CMA
`
`CMG
`
`CMA
`
`CMA
`
`method
`compliance
`Refill
`
`86
`
`[12]
`
`44
`
`59
`
`n
`
`[91
`
`]141
`
`Study
`
`refill
`
`between
`
`5. Association
`
`TABLE
`
`CFAD VI 1019-0007
`
`
`
`112
`
`J. F. Steiner
`
`and A. V. Prochazka
`
`TABLE 6.