ArticlePDF Available

Calculating Medication Compliance, Adherence and Persistence in Administrative Pharmacy Claims Databases

  • MedImpact Healthcare Systems, Inc.
  • MedImpact Healthcare Systems, Inc.

Abstract and Figures

Compliance, adherence, and persistence are outcomes easily measured in pharmacy claims databases. However, these measures are used with differing taxonomies and the calculations are heterogeneous. The results can then lead to spurious interpretations. Therefore, the research community would benefit from a common set of definitions and methods to calculate compliance and persistence. This paper briefly explains the definitions of compliance and persistence based on the guidance from the Medication Compliance and Persistence Special Interest Group of ISPOR, the International Society for Pharmacoeconomics and Outcomes Research, and provides analytic methods that are congruent with the preferred terminology.
Content may be subject to copyright.
Calculating Medication Compliance, Adherence, and Persistence in
Administrative Pharmacy Claims Databases
R. Scott Leslie, MedImpact Healthcare Systems, Inc., San Diego, CA
Compliance, adherence, and persistence are outcomes easily measured in pharmacy claims databases. However,
these measures are often interchanged and calculated differently. The research community should settle on a
common set of definitions and methods. This paper briefly explains compliance and persistence terminology and
methods commonly used in observational pharmacy claims research and furthermore presents a preferred manner
adopted by the Medication Compliance and Persistence Special Interest Group of ISPOR, the International Society
for Pharmacoeconomics and Outcomes Research.
Administrative pharmacy claims for the most part are cleaner and easier to work with than other administrative health
care claims databases (i.e. medical and laboratory claims). First of all, the process of adjudicating pharmacy claims
encompasses checking member eligibility and certain member and health plan restrictions. Therefore, errors are
limited because a claim won’t process unless required fields are entered correctly. Second, data fields are
straightforward and intuitive. Data includes patient and prescriber information plus drug and dosing details (days of
supply, quantity, drug name, and medication class, to name a few). An added benefit of pharmacy claims is “real
time” availability. Many pharmacy claims databases are updated daily or weekly and don’t suffer from lag time
encountered with medical records.
A downside of pharmacy databases, which is inherent for all observational databases, is actual utilization is likely to
differ from observed utilization. That is, we don’t know if patients truly consume their medications and/or adhere to
the medication dosing instructions.
Health outcomes related to pharmacy utilization include compliance, adherence, and persistence measurements. The
purpose of this paper is to describe a recommended definition of the terms and offer code that calculates these
medication utilization outcomes. This code can be a helpful start for calculating additional outcome measures.
The general term of adherence was defined by the World Health Organization in their 2001 meeting as, “the extent to
which a patient follows medical instructions”. Specific to medication adherence, a systematic literature review by
Hess et al. was conducted to gather methods used to calculate adherence in administrative pharmacy databases.
Authors used a MEDLINE search and found multiple terms used to describe the concept and identified 11 manners
used to calculate the metric.
A more comprehensive literature review on compliance, adherence and persistence was conducted by the Medication
and Compliance Special Interest Group of ISPOR, the International Society for Pharmacoeconomics and Outcomes
Research (Cramer et al.). This work group reviewed literature from 1966 to 2005 on the subject and also found
inconsistency in definitions, methodology, and computations for these concepts. Researchers found the terms were
used interchangeably and different measures were used for difference disease states. They also found arbitrary
ways for defining good and poor compliance limits, and no guidance on methodology. Hence a set of methods and
definitions would benefit the research community. At least some guidelines or a framework model for future work and
interpretation of previous work. After this review the work group spent years of discussion to develop definitions and
guidelines. They propose two distinct concepts to be used to describe patient’s medication behavior.
First, the terms compliance and adherence are to be used synonymously and to be defined as,
”the extent to which a patient acts in accordance with the prescribed interval and dose of a dosing regimen”.
Second, persistence should be defined as the,
“the duration of time from initiation to discontinuation of therapy”.
Furthermore, the Analytic Methods Working Group developed a checklist of items that should be considered or
included in a retrospective database analysis (Peterson et al). Using the ISPOR ideas and recommendations, I offer
ways to calculate these pharmacy utilization outcomes using SAS
Compliance can be calculated as both a continuous and dichotomous measure. As a continuous measure, the most
common methods, as outlined and proposed by the ISPOR working group, are by way of medication possession ratio
(MPR) and proportion of days covered (PDC). MPR is the ratio of days medication supplied to days in an time
interval (Steiner). Where PDC is the number of days covered over a time interval (Benner). This differs from MPR in
that it credits the patient with finishing the current fill of medication before starting the next refill. Some believe
compliance can be overestimated by simply summing the days supply because patients usually refill their medication
before completing the current fill.
With both methods, compliance can be made a dichotomous measure (compliant or not compliant) if a patient attains
a specified level of compliance. This level should be based on results of studies showing detrimental health
outcomes when compliance drops below a certain level. For instance, research supports a 95% compliance
threshold for antiretroviral medications (Paterson) and an 80% level for lipid lowering medications (Benner).
Because optimal compliance is associated with the pharmacokinetics and dynamics of the drug, the days supplied,
quantity supplied, and fill date fields in a claims database are the key variables required to compute MPR and PDC.
A simple calculation for MPR is to sum the days of supply for a medication over a time period using a SQL procedure.
This time period may be a fixed time interval or the length the patient used the medication, for instance date of last
claim. When using a fixed time interval one can truncate days of utilization that falls after the interval as done in the
data step below. This truncation assures that all patients are reviewed over a similar time interval (all have the same
180-day denominator).
proc sql;
create table mpr as
select distinct member_id, count(*) as num_fills, sum(days_supply) as ttldsup,
max(index_dt+days_supply,max(fill_dt+days_supply)) - index_dt as duration
from claims
group by member_id;
data mpr_adj;
set mpr;
if duration gt 180 then do;
else do;
/* if want mpr to have max of 1 and duration have max of 180 days*/
Using an example where a patient has three claims over a 180-day study period, the date of first claim (index date) is
the first day of the study period and the end of the study period is 180 days after date of first claim (Figure 1).
member_id fill_dt index_dt drug days_supply
946 603 02/17/2005 02/17/2005 a 30
947 603 06/13/2005 02/17/2005 a 30
948 603 08/11/2005 02/17/2005 a 30
Summing the days_supply field results in a total of 90 days of supply (ttldsup), but with the third claim extending
beyond the study interval, the days supplied (dsuptrnc) is 65. Therefore, MPR is 65 days over 180 days or 0.361.
Figure 1: Medication Fill Patten in 180 Day Period
2/17/05 to 8/17/05
Claim 1
6/13/2005 to 7/12/05
Claim 2
Day 117-146
8/11/05 to 9/9/05
Claim 3 XX
Day 176-180
Study Period Begins Study Period Ends
2/17/05 8/17/05
Day 151-180 Day 181-210Day 1-30 Day 31-60 Day 61-90 Day 91-120 Day 121-150
PDC looks at all days in the 180-day period to check for medication coverage. To calculate the PDC for the scenario
above, indicator variables for each day are used to flag medication coverage. The resulting compliance proportion is
similar to MPR in this scenario; however, can be different in situations with longer days of supply per claim or when
calculating compliance for more than one medication. Although PDC requires more complex code, the added benefit
is information on medication coverage specific for each day, assuming that the patient is consuming the medication
as prescribed. This is helpful when assessing compliance for multiple medications. The days of overlap may be used
to show an incremental gain by using both medications at the same time.
The first step is to transpose the data to a single observation per patient data set. This is done twice for the purposes
of detailing the fill dates and corresponding days supply for each fill. It is essential to sort the data set by patient and
fill date. Start and end dates for each subject are also calculated.
proc sort data=claims;
by member_id fill_dt;
proc transpose data = claims out=fill_dates (drop=_name_) prefix = fill_dt;
by member_id;
var fill_dt;
proc transpose data = claims out=days_supply (drop=_name_) prefix = days_supply;
by member_id;
var days_supply;
data both;
merge fill_dates days_supply;
by member_id;
format start_dt end_dt mmddyy10.;
The result of the above code creates a patient level data set, showing the medication fill pattern and days supply for
each fill. Note that missing values are given to those variables where the variable being transposed has no value in
the input data set. That is, this patient has three claims; therefore the values for fill_dt3 and fill_dt4 are missing.
Obs member_id fill_dt1 fill_dt2 fill_dt3 fill_dt4 fill_dt5
265 603 02/17/2005 06/13/2005 08/11/2005 . .
days_supply1 days_supply2 days_supply3 days_supply4 days_supply5
265 30 30 30 . .
start_dt end_dt
265 02/17/2005 08/15/2005
Next, a data step uses arrays and DO loops to find the days of medication coverage for each patient and calculates
the proportion of covered days in the review period. The first array, daydummy, creates a dummy variable for each
day in the review period. The next two arrays, groups the fill_dt and days_supply variables setting up the DO loops.
In this data set, the maximum number of fills incurred by a patient was 11 so there are 11 elements for these two
arrays. One can set the number of elements to a value beyond the reasonable amount of fills or get the maximum
number of fills in the data set from a proc contents procedure. The first do loop sets each dummy variable,
daydummy, to 0. The second do loop uses an IF statement to flag the days of the review period that the patient was
supplied the medication. Next, the variable dayscovered sums the daydummy variables. This sum is used as the
numerator in calculating p_dayscovered, the proportion of covered days in the 180 day study period.
data pdc;
set both;
array daydummy(180) day1-day180;
array filldates(*) fill_dt1 - fill_dt11;
array days_supply(*) days_supply1-days_supply11;
do ii=1 to 180; daydummy(ii)=0;end;
do ii=1 to 180;
do i = 1 to dim(filldates) while (filldates(i) ne .);
if filldates(i)<= start_dt + ii -1 <= filldates(i)+days_supply(i)-1
then daydummy(ii)=1;
drop i ii;
dayscovered=sum(of day1 - day180);label dayscovered='Total Days Covered';
p_dayscovered=dayscovered/180;label p_dayscovered='Proportion of Days Covered';
proc print data=pdc;where member_id=603;run;
The result is a data set that has 180 dummy variables, one for each day of the time period, indicating medication
coverage. Only a few of the dummy variables are displayed below. In this example the patient’s last fill_date is on
day 176 (see Figure 1 above) with a majority of the days supply for this claim extending beyond the study period. The
claim is truncated and only 5 of days of this claim are included in the days covered count.
Obs member_id day1 day2 day3 day4 day5 ***day6-day29*** day30 day31 day32 day33
265 603 1 1 1 1 1 1 1 0 0 0
***day34-day115*** day116 day117 day118 day119 ***day120-day145*** day146
265 0 0 1 1 1 1 1
***day147-day174*** day175 day176 day177 day178 day179 day180
265 0 0 1 1 1 1 1
dayscovered p_dayscovered
265 65 0.36111
In this example, a patient refills their medication before exhausting the previous fill. Figure 2 shows this scenario,
where the fourth claim, filled on 7/30/05, occurs before the end of supply of the third claim (8/5/05). Proportion of
days covered is calculated in the same manner and credits the patient for the overlapping days supply assuming the
patient is finishing the current prescription before starting the refill prescription. This code is similar to the previous
example with one extra step that identifies the overlapping days supply and shifts the fill date forward to the day after
the end of supply of the previous fill.
Using the same steps as in previous example, adding this additional DO loop adjusts fill dates by shifting them
forward. This starts with the second fill:
do u=2 to 11 while (filldates(u) ne .);
if filldates(u)<filldates(u-1)+days_supply(u-1)
then filldates(u)=filldates(u-1)+days_supply(u-1);
Shifting the fourth fill date credits the patient with 7 more days supply, increasing the days covered from 113 to 120
and increasing proportion of days covered from to 62.8% to 66.6%. Larger differences would be seen in cases where
a patient has multiple claims and multiple overlaps. These steps can be used to calculate length of therapy of
multiple medications. To assess compliance to therapy of two medications, two sets of day dummy variables would
be created and the DO loops would then flag the days that both medications were covered. Such a scenario is
illustrated below in Figure 3.
Figure 2: Medication Coverage with Overlapping Days Supply
4/21/05 to 5/20/05/05
Claim 1
Day 1-30
6/3/05 to 7/2/05
Claim 2
Day 44-73
7/7/05 to 8/5/05
Claim 3
Day 78-109
7/30/05 to 7/2/05
Claim 4 Overlap
Day 105-134
8/6/05 to 6/4/05
Claim 4 Shifted
Day 110-139
Study Period Begins Study Period Ends
Day 151-180Day 1-30 Day 31-60 Day 61-90 Day 91-120 Day 121-150
Figure 3. Proportion of Days Covered for Concomitant Therapy
Therapy A Claim #1 Claim #2 Claim #3 Claim #4
= 120/180 = 67%
Therapy B Claim #1 Claim #2 Claim #3 Claim #4 PDC
= 120/180 = 67%
Access to Therapy A & B
= 72/180 = 40%
Study Period Begins Study Period Ends
All Claims are of 30 days supply
Day 151-180Day 1-30 Day 31-60 Day 61-90 Day 91-120 Day 121-150
As noted, compliance is delineated from persistence, since persistence is “the duration of time from initiation to
discontinuation of therapy”. This is usually the time, measured in days, from first claim to last claim (plus the days
supply of the last claim) considering the days between refills. A limit on the days between fills, or ”gap”, should be
set based on the properties of the drug (Peterson). For medications used on seasonal basis, e.g., asthma or allergic
rhinitis medications, persistence can be measured as number for refills for a medication within a time interval to
monitor patient’s refill behavior.
Persistence can also be described as a dichotomous variable (persistent or not) if a person was continued therapy
beyond an elapsed time period, e.g. 90, 180, or 365 days.
In example 2 shown in Figure 2, persistence would be defined as days to discontinuation and measured as days from
day of first claim to day of first gap that is 30 days. Days to discontinuation would be the 130 days from 4/21/05 to
6/04/05 because the first gap of 30 days or longer occurs after the last claim. Notice the gaps between the first and
second claim and second and third claim are allowable as these gaps are less than 30 days. Therefore the patient is
not considered discontinuing therapy until after the fourth claim.
proc sort data=claims;by member_id fill_dt;run;
data pers;
set claims;
by member_id fill_dt;
if first.member_id=1 then epsd=1;
else if gap>30 then epsd+1;
format lstsply mmddyy10.;
proc sql;create table gap as
select distinct member_id, max(epsd) as epsd
from pers
group by member_id;quit;
proc freq data=gap;tables epsd;run;
proc sql;
create table persis as
select distinct member_id, max(lstsply) as lstsply format mmddyy10.,
min((max(lstsply)-index_dt), 180) as days_therapy /*days to discountinuation*/
from pers
where epsd=1
group by member_id;quit;
proc freq data=persis;tables days_therapy;run;
Administrative pharmacy claims are useful in calculating observed compliance (a.k.a. adherence) and persistence
health outcomes. Reviews of research found a variety of definitions, methods, terminology on these subjects. To
provide a model for future work and interpretation of previous work, the Medication Compliance and Persistence
Special Interest Group of ISPOR proposed methods and philosophy based on years of review by subject matter
experts and leaders in the field. This paper presents the proposed concepts and terminology for the outcomes and
offers SAS
code to compute these measures.
The appropriate level of compliance is dependent on the disease and the properties of the medication. MPR is the
most commonly used measure and provides a reasonable estimate of compliance and persistence. PDC adds more
information on medication coverage specific for each day, but assumes a patient is consuming refills as prescribed.
Every measure is fraught with an element of error and a calculated compliance based on observable data gives an
estimate of actual compliance. Reliable and valid data sources minimize the impact of error. Understanding patient’s
medication use increases the knowledge of whether therapeutic ineffectiveness or noncompliance contributes to poor
patient outcomes.
Hess, L.M., Raebel, M.A., Conner, D.A., Malone, D.C. 2006. “Measurement of Adherence in Pharmacy Administrative
Databases: A Proposal for Standard Definitions and Preferred Measures”. The Annals of Pharmacotherapy 40;1280-
International Society of Pharmacoeconomics and Outcomes Research (ISPOR) Medication Compliance and
Persistence Special Interest Group.
Cramer, J.A. et al. 2007. “Medication Compliance and Persistence: Terminology and Definitons”. Value in Health
Peterson AM et al. 2007. ”A Checklist for Medication Compliance and Persistence Studies Using Retrospective
Databases”. Value in Health 10(1):3-12.
Steiner J.F., Prochazka A.V. 1997. “The assessment of refill compliance using pharmacy records. Methods,
validation, and applications”. Journal of Clincal Epidemiology 50:105-106.
Benner JS, Pollack MF, Smith TW, et al. Long-term persistence in sue of statin therapy in elderly patients. JAMA
Paterson D.L., Swindells S., Mohr J., et al. 2000. “Adherence to protease inhibitor therapy and outcomes in patients
with HIV infection”. Ann Intern Med 133:21-30.
Leslie, R. Scott. 2007. “Using Arrays to Calculate Medication Utilization.”
Proceedings of the 2007 SAS Global Forum, Orlando, FL. Paper 043-2007.
SAS Institute Inc. 2004. “SAS Procedures: The SQL Procedure”. SAS OnlineDoc® 9.1.3. Cary, NC: SAS Institute Inc.
(April 2, 2006)
SAS Institute Inc. 2004. “Language Reference Concepts: Array Processing”. SAS OnlineDoc® 9.1.3. Cary, NC: SAS
Institute Inc.
(April 2, 2006)
SAS Institute Inc. 2004. “SAS Procedures: The TRANSPOSE Procedure”. SAS OnlineDoc® 9.1.3. Cary, NC: SAS
Institute Inc.
The author would like to thank Femida Gwadry-Sridhar for her explanation of the presented compliance and
persistence concepts and her review of this paper.
Your comments and questions are valued and encouraged. Contact the author at:
R. Scott Leslie
MedImpact Healthcare Systems, Inc.
10680 Treena Street
San Diego, CA 92131
Work Phone: 858-790-6685
Fax: 858-689-1799
SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS
Institute Inc. in the USA and other countries. ® indicates USA registration.
Other brand and product names are trademarks of their respective companies.
... Therefore, we truncated PDCs greater than 100% at 100%. We adopted the computational calculation from Leslie (2007) and Leslie et al. (2008) [25,26]. ...
... Therefore, we truncated PDCs greater than 100% at 100%. We adopted the computational calculation from Leslie (2007) and Leslie et al. (2008) [25,26]. ...
Full-text available
Background Suboptimal patient adherence to pharmacological therapy of type 2 diabetes may be due in part to pill burden. One way to reduce pill burden in patients who need multiple medications is to use fixed-dose combinations. Our study aimed to compare the effects of fixed-dose combination versus loose-dose combination therapy on medication adherence and persistence, health care utilization, therapeutic safety, morbidities, and treatment modification in patients with type 2 diabetes over three years.Methods Using administrative data, we conducted a retrospective controlled cohort study comparing type 2 diabetes patients who switched from monotherapy to either a fixed-dose combination or a loose-dose combination. Adherence was assessed as the primary endpoint and calculated as the proportion of days covered with medication. After using entropy balancing to eliminate differences in observable baseline characteristics between the two groups, we applied difference-in-difference estimators for each outcome to account for time-invariant unobservable heterogeneity.ResultsOf the 990 type 2 diabetes patients included in our analysis, 756 were taking a fixed-dose combination and 234 were taking a loose-dose combination. We observed a statistically significantly higher change in adherence (year one: 0.22, p
... To adjust for potential confounders, all of the models included prespecified covariates of age, sex, comorbidities, pre-index exacerbations, GINA step, continuous measure of SABA fills, interaction between GINA and SABA, payer type, baseline health care costs, and adherence to maintenance treatment (proportion of days covered). [27][28][29] ...
BACKGROUND: Despite the availability of effective treatments, patients with asthma, regardless of severity, remain at risk of severe exacerbations resulting in significant burden to patients, the health care system, and insurance providers. OBJECTIVE: To examine severe exacerbations, treatment patterns, health care resource utilization (HCRU), and costs across all asthma severities. METHODS: In this retrospective study, patients aged 4 years and older filling 1 short-acting (β2-agonist (SABA) and at least 1 maintenance fill or at least 2 SABAs with or without maintenance fills were identified from administrative claims data from the IBM MarketScan Commercial and IBM MarketScan Multistate Medicaid Research databases (January 2010 to December 2017). Patients were indexed on a random SABA fill (2011-2016) and had 12 months of continuous eligibility pre-index and post-index. Patients were classified into Global Initiative for Asthma (GINA) 2018 severity steps and by asthma control, as measured by SABA fill use in the 12 months pre-index: low (1 SABA fill per year), medium (2-3 SABA fills per year), and high (≥ 4 SABA fills per year); well controlled, not well controlled, and very poorly controlled, respectively. Severe asthma exacerbation events, health care costs, and asthma-related HCRU and costs were assessed relative to asthma severity and asthma control post-index. RESULTS: Of 1,005,522 patients, 50.3% filled GINA Step 1; 19.7% GINA Step 2; 10.9% GINA Step 3; and 19.1% GINA Steps 4-5 treatments. Overall, 953,337 severe exacerbation events occurred (approximately 0.95 events per patient), equating to 0.96, 0.67, 0.83, and 1.28 events per patient for patients filling GINA Step 1 through Steps 4-5, respectively. GINA Step 1 had the highest proportion of patients experiencing at least 1 event (57.0%), followed by GINA Steps 4-5 (55.2%), GINA Step 3 (45.0%), and GINA Step 2 (41.9%) treatments (P < 0.05). For GINA Step 1, 64.4% of well-controlled patients experienced at least 1 exacerbation event vs 50.4% of not well-controlled and 53.0% of very poorly controlled patients (P < 0.05). For patients filling GINA Step 2-5 treatments, a greater proportion of very poorly controlled patients experienced at least 1 exacerbation event vs well-controlled patients (P < 0.05). The average total annual health care cost per patient was $7,148 and total annual asthma-related costs were $1,741. Each additional SABA fill was associated with a 26.0%, 10.8%, and 34.6% increase in incidence of total exacerbations, all-cause costs, and asthma-related costs, respectively (P < 0.05). CONCLUSIONS: In this real-world database study, increased SABA fills and occurrence of exacerbations were correlated and associated with higher all-cause and asthma-related costs across all severities. New treatment paradigms, particularly for rescue therapies, are warranted to improve clinical and cost outcomes in these patients. DISCLOSURES: This analysis was funded by AstraZeneca. Michael Pollack, Hitesh Gandhi, and Ileen Gilbert are employees and stockholders of AstraZeneca and contributed to the design and conduct of the study. AstraZeneca was given an opportunity to review the final version of the manuscript. At the time of the study, Joseph Tkacz was an employee of IBM Watson Health, which received funding from AstraZeneca to conduct this study. Miguel Lanz has received research funding from AstraZeneca, Optinose, and Regeneron and consulting fees and honoraria from ALK, Amgen, AstraZeneca, Novartis, Sanofi, and Regeneron. Njira Lugogo received consulting fees for advisory board participation from Amgen, AstraZeneca, Genentech, GlaxoSmith-Kline, Novartis, Regeneron, Sanofi, and Teva; honoraria for nonspeaker's bureau presentations from GlaxoSmithKline and AstraZeneca; and travel support from AstraZeneca. Her institution received research support from Amgen, AstraZeneca, Avillion, Gossamer Bio, Genentech, GlaxoSmithKline, Regeneron, Sanofi, and Teva.
... We consider the cut-off point of 66% for the rate of adherence or therapeutic compliance according to the literature [5,40,41]. Like compliance, persistence or not was considered as a dichotomous variable [42] and was measured taking into account whether or not the person completed the 4 months of intervention, considering the grace period. ...
Full-text available
Background: Computer-based programs have been implemented from a psychosocial approach for the care of people with dementia (PwD). However, several factors may determine adherence of older PwD to this type of treatment. The aim of this paper was to identify the sociodemographic, cognitive, psychological, and physical-health determinants that helped predict adherence or not to a "GRADIOR" computerized cognitive training (CCT) program in people with mild cognitive impairment (MCI) and mild dementia. Method: This study was part of a randomized clinical trial (RCT) (ISRCTN: 15742788). However, this study will only focus on the experimental group (n = 43) included in the RCT. This group was divided into adherent people (compliance: ≥60% of the sessions and persistence in treatment up to 4 months) and non-adherent. The participants were 60-90 age and diagnosed with MCI and mild dementia. We selected from the evaluation protocol for the RCT, tests that evaluated cognitive aspects (memory and executive functioning), psychological and physical health. The CCT with GRADIOR consisted of attending 2-3 weekly sessions for 4 months with a duration of 30 min Data analysis: Phi and Biserial-point correlations, a multiple logical regression analysis was obtained to find the adherence model and U Mann-Whitney was used. Results: The adherence model was made up of the Digit Symbol and Arithmetic of Wechsler Adult Intelligence Scale (WAIS-III) and Lexical Verbal Fluency (LVF) -R tests. This model had 90% sensitivity, 50% specificity and 75% precision. The goodness-of-fit p-value of the model was 0.02. Conclusions: good executive functioning in attention, working memory (WM), phonological verbal fluency and cognitive flexibility predicted a greater probability that a person would be adherent.
... We calculated the adherence to ART using the Proportion of Days Covered (PDC), assessed by the total number of days a patient was exposed to antiretrovirals (days covered), divided by the total time period (Leslie et al., 2008). For ART in single-tablet regimens, we assumed individuals were taken pills once daily. ...
Full-text available
Pharmacy dispensing data are useful for estimating adherence to therapy. Here, we implement multiple adherence measures to antiretroviral therapy (ART) and provide an online tool for visualising results. We conducted a cohort study for 2,042 people dispensed ART in Australia. We assessed adherence using the Proportion of Days Covered (PDC) within 360 days of follow-up as a continuous measure and dichotomised (PDC ≥80%). We defined a covered day as the 1) exposure to ≥3 antiretrovirals at the same time 2) exposure to any antiretroviral 3) lowest number of days covered per antiretroviral 4) average of days covered over all antiretrovirals 5) highest number of days covered per antiretroviral. For each method, we conducted sensitivity analyses. The median PDC ranged between 93.3%−98.3%. Between 67.0%−87.7% of individuals were classified as adherent, with higher values for measure 2 (85.5%−89.7%) and lower values for measure 3 (67.0%−70.9%). Censoring loss to follow-up had a higher impact on adherence estimates than considering a grace period. The variation in adherence estimates can be substantial, especially when dichotomising adherence. Researchers should consider operationalising multiple measures to estimate adherence bounds and identify a range of people at risk of non-adherence for targeted interventions. ARTICLE HISTORY
... Erenumab dispensing patterns included initial dose prescribed (70 mg vs 140 mg), switches in dose and time to switch (in days), and persistence. Persistence was evaluated using 30-day and 45-day allowable gaps in therapy as well as the proportion of days covered (PDC) method using stockpiling algorithm [12]. PDC was calculated as the ratio of number of days the patient was covered by erenumab in a period to the total number of days in the period. ...
Full-text available
IntroductionErenumab is indicated for migraine preventive treatment in adults. The objective of this study was to provide descriptive information on real-world use of erenumab including patient profile and treatment patterns.Methods We completed a retrospective review of US data (through May 2019) from the IBM MarketScan® Early View Databases, identifying adult patients newly treated with erenumab with a migraine claim in the year prior to first erenumab claim (index) and at least 1 year of continuous pre-index medical and pharmacy insurance coverage, to assess pre- and post-erenumab migraine characteristics, comorbidities, healthcare resource utilization, and associated costs. All data were summarized using descriptive statistics.ResultsA total of 9753 patients met inclusion criteria. The average (SD) age was 46 (12) years, 85% of patients were female, and 64% had at least one claim for chronic migraine; 70% of erenumab users had an initial dose of 70 mg; 77% of patients in the 6-month follow-up sample (n = 4437) remained on their initial erenumab dose. Persistence at 6-month follow-up was 47.3% with a mean (95% CI) proportion of days covered of 0.68 (0.67, 0.68). In the post-erenumab period, claims for comorbidities of non-migraine headaches and anxiety were reduced and there was a shift to decreased use of acute and preventive medications. Reductions in overall use and associated cost of healthcare resources such as inpatient hospitalization and outpatient office visits were minimal, with slightly more pronounced reductions in the subgroup of patients that were persistent to erenumab.Conclusions We observed reductions in claims for important migraine characteristics, comorbidities, and a shift to decreased use of acute and preventive migraine medications—observations indicative of the real-world effectiveness of erenumab. Further examination is required as persistence to erenumab, which may be influenced by dose titration, appears to be an important factor in changes to healthcare resource utilization and costs.
... Medication compliance was quantitatively measured according to the recommendation of the Medication Compliance and Persistence Special Interest Group from the International Society for Pharmacoeconomics and Outcomes Research (ISPOR). 11 Medication persistence and adherence was defined by continuing medication for the prescribed ...
Background:Non-compliance with angiotensin receptor blockers (ARB) or statin is one of the major hurdles to optimal medical treatment. This study investigated whether fixed-dose combination (FDC) improved compliance to medication compared with traditional free combination (FC). Methods and Results:In this retrospective nationwide cohort study, medication persistency, medication adherence measured by proportion of days covered (PDC), and all-cause death of 123,992 patients who started ARB and stain were investigated for 540 days. Patients had a mean age of 63 years and 48% were male. Persistency, PDC, and proportion of PDC ≥80% of FDC (N=34,776) were higher than those for FC (N=89,216) in both unadjusted analysis (54.5% vs. 27.8%; 84.1% vs. 63.1%; 75.5% vs. 48.1%) and propensity-score matched analysis (P<0.001, all). Death risk for the investigation period (0–540 days) was lower in FDC in unadjusted (1.8% vs. 2.6%, P<0.001) and adjusted cohort (P<0.05). In landmark analyses at days 180 and 360, there was no significant difference of death risk between FDC and FC (P>0.05). Conclusions:In this real-world data analysis, patients taking FDC of ARB and statin showed higher medication persistence and adherence compared to patients taking FC of ARB and statin up to 540 days. The risk of all-cause death was not different between FDC and FC despite better medication compliance in the FDC patients.
... Research. This review defined adherence as "the extent to which a patient takes treatment in accordance with the prescribed interval and dose of a dosing regimen" [12]. Similarly, the World Health Organisation (WHO) defines adherence as "the extent to which the patient follows medical instructions" [13]. ...
Full-text available
Background: Nicotine replacement therapy (NRT) has proven effect in assisting smoking cessation. However, its effectiveness varies across studies and population groups. This may be due to differences in the rate of adherence. Hence, this review aims to examine the level of adherence to NRT and to assess if the level of adherence to NRT affects success of smoking cessation. Methods: A systematic review and meta-analysis was conducted using studies retrieved from five electronic databases (MEDLINE, Scopus, EMBASE, Web of science, and PsycINFO) and grey literature. Pooled analysis was conducted using Stata version 16 software. Methodological quality and risk of bias were assessed using the NIH Quality Assessment Tool. Analyses were done among those studies that used similar measurements to assess level of adherence and successful smoking cessation. Heterogeneity of studies was assessed using the Higgins' I2 statistical test. Funnel plots and Egger's regression asymmetry test were used to affirm presence of significant publication bias. Results: A total of 7521 adult participants of 18 years old and above from 16 studies were included in the analysis. Level of adherence to NRT among participants of randomised controlled trials were found to be 61% (95% CI, 54-68%), p-value of < 0.001 and I2 = 85.5%. Whereas 26% of participants were adherent among participants of population-based studies with 95% CI, 20-32%, p-value of < 0.001 and I2 = 94.5%. Level of adherence was the lowest among pregnant women (22%) with 95% CI, 18-25%, p-value of 0.31 and I2 = 15.8%. Being adherent to NRT doubles the rate of successful quitting (OR = 2.17, 95% CI, 1.34-3.51), p-value of < 0.001 and I2 = 77.6%. Conclusions: This review highlights a low level of adherence to NRT among participants of population-based studies and pregnant women as compared to clinical trials. Moreover, the review illustrated a strong association between adherence and successful smoking cessation. Hence, it is recommended to implement and assess large scale interventions to improve adherence. Health programs and policies are recommended to integrate the issue of adherence to NRT as a core component of smoking cessation interventions. Trial registration: PROSPERO registration number: CRD42020176749 . Registered on 28 April 2020.
... The medication adherence rate was determined cumulatively by assessing the proportion of the time that medications were taken, based on prescription refill dates. In this period, we calculated the proportion of days covered (PDC) [17], taking into account possible overlapping prescriptions. The first step of the procedure was to shift the refill date forward to the day after the end of the supply of the previous refill, and then medication possession was computed as [18] total days' supply over the number of days in the observation period. ...
Background This study evaluates, in a real-world setting, to what extent the recommended therapies by international guidelines, are prescribed after a first hospitalization for heart failure (HF), and to analyse adherence and persistence, and the effect of treatment adherence on mortality and re-hospitalization. Methods From the Lombardy healthcare administrative database, we analysed patients discharged after their incident HF, from 2000 to 2012. Adherence was defined as the proportion of days covered (PDC) ≥80% adjusted for hospitalizations and persistence as the absence of discontinuation of therapy for >30 days. A logit model was used to determine the effect of patients' adherence on mortality and readmissions. Results Of 100,422 HF patients (52% males, age 75 ± 12 years), 86,846 (87%) had a prescription for angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers (ACE/ARBs), 64,135 (64%) for beta-blockers (BB), and 36,893 (37%) for mineralocorticoid receptor antagonists (MRAs), as mono-, bi- or tri-therapy. In patients on monotherapy, PDC was 78 ± 22% for ACE/ARBs, 69 ± 29% for BB and 54 ± 29% for MRAs; in those on bi-therapy, PDC was 63 ± 31% for ACEI/ARBs+BB, 41 ± 29% for ACEI/ARBs+MRAs, and 40 ± 26% for MRAs+BB; for patients on tri-therapy, PDC was 42 ± 28%. Medication persistence was present in 47% of patients treated with ACEI/ARBs, in 35% of patients treated with BB and in 14% of patients treated with MRAs. Re-hospitalizations and in mortality were significantly reduced in adherent patients (p < 0.000). Conclusions Polypharmacy is associated with an increased rate of non-adherence and non-persistence in incident HF. Non-adherence is associated with an increased risk of mortality and re-hospitalizations.
Objectives The Sub-Saharan African (SSA) region now has the highest estimated effect size of hypertension for stroke causation worldwide. An urgent priority for countries in SSA is to develop and test self-management interventions to control hypertension among those at highest risk of adverse outcomes. Thus the overall objective of the Phone-based Intervention under Nurse Guidance after Stroke II study (PINGS-2) is to deploy a hybrid study design to assess the efficacy of a theoretical-model-based, mHealth technology-centered, nurse-led, multi-level integrated approach to improve longer term blood pressure (BP) control among stroke survivors. Materials and methods A phase III randomized controlled trial involving 500 recent stroke survivors to be enrolled across 10 Ghanaian hospitals. Using a computer-generated sequence, patients will be randomly assigned 1:1 into the intervention or usual care arms. The intervention comprises of (i) home BP monitoring at least once weekly with nurse navigation for high domiciliary BP readings; (2) medication reminders using mobile phone alerts and (3) education on hypertension and stroke delivered once weekly via audio messages in preferred local dialects. The intervention will last for 12 months. The control group will receive usual care as determined by local guidelines. The primary outcome is the proportion of patients with systolic BP <140 mm Hg at 12 months. Secondary outcomes will include medication adherence, self-management of hypertension, major adverse cardiovascular events, health related quality of life and implementation outcomes. Conclusion An effective PINGS intervention can potentially be scaled up and disseminated across healthcare systems in low-and-middle income countries challenged with resource constraints to reduce poor outcomes among stroke survivors.
Full-text available
Background: Chronic opioid use is associated with poorer clinical outcomes in inflammatory bowel disease. Aims: To investigate an association between chronic opioid use and persistence with biologic agents in management of inflammatory bowel disease. Methods: A total of 16 624 patients diagnosed with inflammatory bowel disease and receiving a first-time biologic prescription from 2011 to 2016 were identified retrospectively from the Truven MarketScan Database. A cohort of 1768 patients were identified as chronic opioid users utilising outpatient prescription claims. Utilisation patterns of biologic therapies were assessed from inpatient administration and outpatient claims data, including persistence calculations. Information on healthcare utilisation and common comorbidities was also collected. A Cox regression model was constructed to assess the hazard of chronic opioid use on early discontinuation of biologic therapy controlling for disease severity. Results: A mean 1.5 different biologic agents were utilised by inflammatory bowel disease patients with chronic opioid use (vs 1.37 in the comparator group; P < 0.0001). A lower proportion of the chronic opioid use cohort persisted on biologic therapies to the end of the study period (16.2% vs 33.5% P < 0.0001). Inflammatory bowel disease patients with chronic opioid use utilised more healthcare resources and had a higher rate of comorbidities than the reference cohort. Patients with chronic opioid use were 23% more likely (hazard ratio 1.23; 95% CI [1.16-1.31]) to be non-persistent with biologic therapy while accounting for relevant markers of disease acuity. Conclusions: Chronic opioid use is associated with increased hazard of biologic discontinuation in inflammatory bowel disease. Symptoms of opioid withdrawal may mimic IBD flares thereby leading providers to inappropriately switch biologic therapies and compromise disease control.
Full-text available
While clinical trials demonstrate the benefits of blood pressure and cholesterol reduction, medication adherence in clinical practice is problematic. We hypothesized that a single-pill would be superior to a 2-pill regimen for achieving adherence. In this retrospective, cohort study based on pharmacy claims data, patients newly initiated on a calcium channel blocker (CCB) or statin simultaneously or within 30 days, regardless of sequence, were followed (N=4703). Adherence was measured over 6 months as proportion of days covered (PDC). At baseline, mean age was 63.0 years, 51.6% were female, and mean number of other medications was 7.8. Overall, 16.9% of patients were on single-pill amlodipine/atorvastatin, 15.6% amlodipine + atorvastatin, 24.7% amlodipine + other statin, 13.9% other CCB + atorvastatin, 28.9% other CCB + other statin. Percentages of patients achieving adherence (PDC >or= 80%) were: 67.7% amlodipine/atorvastatin; 49.9% amlodipine + atorvastatin; 40.4% amlodipine + other statin; 46.9% other CCB + atorvastatin; 37.4% other CCB +other statin. After adjusting for treatment selection and cohort differences, odds ratios for adherence with amlodipine/atorvastatin were 1.95 (95% confidence interval [CI], 1.80-2.13) vs amlodipine + atorvastatin, 3.10 (95% CI, 2.85-3.38) vs amlodipine + other statin, 2.06 (95% CI, 1.89-2.24) vs other CCB + atorvastatin, 2.85 (95% CI, 2.61-3.10) vs other CCB + other statin (all p<0.0001). Single-pill amlodipine/atorvastatin may provide clinical benefits through improving adherence, offering clinicians a practical solution for cardiovascular risk management.
Full-text available
Combination antiretroviral therapy with protease inhibitors has transformed HIV infection from a terminal condition into one that is manageable. However, the complexity of regimens makes adherence to therapy difficult. To assess the effects of different levels of adherence to therapy on virologic, immunologic, and clinical outcome; to determine modifiable conditions associated with suboptimal adherence; and to determine how well clinicians predict patient adherence. Prospective, observational study. HIV clinics in a Veterans Affairs medical center and a university medical center. 99 HIV-infected patients who were prescribed a protease inhibitor and who neither used a medication organizer nor received their medications in an observed setting (such as a jail or nursing home). Adherence was measured by using a microelectronic monitoring system. The adherence rate was calculated as the number of doses taken divided by the number prescribed. Patients were followed for a median of 6 months (range, 3 to 15 months). During the study period, 45,397 doses of protease inhibitor were monitored in 81 evaluable patients. Adherence was significantly associated with successful virologic outcome (P < 0.001) and increase in CD4 lymphocyte count (P = 0.006). Virologic failure was documented in 22% of patients with adherence of 95% or greater, 61% of those with 80% to 94.9% adherence, and 80% of those with less than 80% adherence. Patients with adherence of 95% or greater had fewer days in the hospital (2.6 days per 1000 days of follow-up) than those with less than 95% adherence (12.9 days per 1000 days of follow-up; P = 0.001). No opportunistic infections or deaths occurred in patients with 95% or greater adherence. Active psychiatric illness was an independent risk factor for adherence less than 95% (P = 0.04). Physicians predicted adherence incorrectly for 41% of patients, and clinic nurses predicted it incorrectly for 30% of patients. Adherence to protease inhibitor therapy of 95% or greater optimized virologic outcome for patients with HIV infection. Diagnosis and treatment of psychiatric illness should be further investigated as a means to improve adherence to therapy.
Full-text available
A variety of measures have been developed to calculate refill adherence from administrative data such as pharmacy claims databases. These measures have focused on improving the accuracy of adherence measures or clarifying the evaluation time frame. As a result, there are many measures used to assess adherence that may or may not be comparable or accurate. To compare available refill adherence measures. A systematic literature review was conducted to identify current or recently used measures of calculating adherence from administrative data. A MEDLINE search (January 1990-March 2006) was undertaken using the search terms adherence or compliance in the title combined with administrative, pharmacy, or records in any field, including subheadings medical, nursing, and hospital records. Non-English articles were excluded. Seven hundred fifteen articles were available for review. Review articles and letters were excluded from measure selection, but were included in the search terms and used to identify additional research articles. Adherence measures were excluded if they were incompletely described, produced non-numeric values, or were duplicates. Eleven refill adherence measures were identified and compared using data from the LOSE Weight (Long-term Outcomes of Sibutramine Effectiveness on Weight) study. Measures compared include Continuous Measure of Medication Acquisition (CMA); Continuous Multiple Interval Measure of Oversupply (CMOS); Medication Possession Ratio (MPR); Medication Refill Adherence (MRA); Continuous Measure of Medication Gaps (CMG); Continuous, Single Interval Measure of Medication Aquisition (CSA); Proportion of Days Covered (PDC); Refill Compliance Rate (RCR); Medication Possession Ratio, modified (MPRm); Dates Between Fills Adherence Rate (DBR); and Compliance Rate (CR). The results suggest that the CMA, CMOS, MPR, and MRA are identical in terms of measuring adherence to prescription refills throughout the study period, each with a value of 63.5%; CMG and PDC are slightly lower (63.0%) and are equivalent to MRA when oversupply is truncated. CR, MPRm, RCR, and CSA result in higher adherence values of 84.4%, 86.6%, 104.8%, and 109.7%, respectively. Five measures produce equivalent results for measuring prescription refill adherence over the evaluation period. Of these, MRA has the fewest calculations, is easily truncated if one desires to exclude surplus medication issues, and requires the least amount of data. MRA is therefore recommended as the preferred measure of adherence using administrative data.
Context Knowledge of long-term persistence with 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitor (statin) therapy is limited because previous studies have observed patients for short periods of time, in closely monitored clinical trials, or in other unrepresentative settings.Objective To describe the patterns and predictors of long-term persistence with statin therapy in an elderly US population.Design, Setting, and Patients Retrospective cohort study including 34 501 enrollees in the New Jersey Medicaid and Pharmaceutical Assistance to the Aged and Disabled programs who were 65 years of age and older, initiated statin treatment between 1990 and 1998, and who were followed up until death, disenrollment, or December 31, 1999.Main Outcome Measures Proportion of days covered (PDC) by a statin in each quarter during the first year of therapy and every 6 months thereafter; predictors of suboptimal persistence during each interval (PDC <80%) were identified using generalized linear models for repeated measures.Results The mean PDC was 79% in the first 3 months of treatment, 56% in the second quarter, and 42% after 120 months. Only 1 patient in 4 maintained a PDC of at least 80% after 5 years. The proportion of patients with a PDC less than 80% increased in a log-linear manner, comprising 40%, 61%, and 68% of the cohort after 3, 12, and 120 months, respectively. Independent predictors of poor long-term persistence included nonwhite race, lower income, older age, less cardiovascular morbidity at initiation of therapy, depression, dementia, and occurrence of coronary heart disease events after starting treatment. Patients who initiated therapy between 1996-1998 were 21% to 25% more likely to have a PDC of at least 80% than those who started in 1990.Conclusions Persistence with statin therapy in older patients declines substantially over time, with the greatest drop occurring in the first 6 months of treatment. Despite slightly better persistence among patients who began treatment in recent years, long-term use remains low. Interventions are needed early in treatment and among high-risk groups, including those who experience coronary heart disease events after initiating treatment.
Objective: The aim of the study is to provide guidance regarding the meaning and use of the terms "compliance" and "persistence" as they relate to the study of medication use. Methods: A literature review and debate on appropriate terminology and definitions were carried out. Results: Medication compliance and medication persistence are two different constructs. Medication compliance (synonym: adherence) refers to the degree or extent of conformity to the recommendations about day-to-day treatment by the provider with respect to the timing, dosage, and frequency. It may be defined as "the extent to which a patient acts in accordance with the prescribed interval, and dose of a dosing regimen." Medication persistence refers to the act of continuing the treatment for the prescribed duration. It may be defined as "the duration of time from initiation to discontinuation of therapy." No overarching term combines these two distinct constructs. Conclusions: Providing specific definitions for compliance and persistence is important for sound quantitative expressions of patients' drug dosing histories and their explanatory power for clinical and economic events. Adoption of these definitions by health outcomes researchers will provide a consistent framework and lexicon for research.
The refill records of computerized pharmacy systems are used increasingly as a source of compliance information. We reviewed the English-language literature to develop a typology of methods for assessing refill compliance (RC), to describe the epidemiology of compliance in obtaining medications, to identify studies that attempted to validate RC measures, to describe clinical features that predicted RC, and to describe the uses of RC measures in epidemiologic and health services research. In most of the 41 studies reviewed, patients 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 methodologic problems require further study, RC measures can be a useful source of compliance information in population-based studies when direct measurement of medication consumption is not feasible.
The increasing number of retrospective database studies related to medication compliance and persistence (C&P), and the inherent variability within each, has created a need for improvement in the quality and consistency of medication C&P research. This article stems from the International Society of Pharmacoeconomics and Outcomes Research (ISPOR) efforts to develop a checklist of items that should be either included, or at least considered, when a retrospective database analysis of medication compliance or persistence is undertaken. This consensus document outlines a systematic approach to designing or reviewing retrospective database studies of medication C&P. Included in this article are discussions on data sources, measures of C&P, results reporting, and even conflict of interests. If followed, this checklist should improve the consistency and quality of C&P analyses, which in turn will help providers and payers understand the impact of C&P on health outcomes.
Language Reference Concepts: Array Processing
  • Sas Institute Inc
SAS Institute Inc. 2004. "Language Reference Concepts: Array Processing". SAS OnlineDoc® 9.1.3. Cary, NC: SAS Institute Inc.
Using arrays to calculate medication utilization
  • R S Leslie
Leslie, R. Scott. 2007. "Using Arrays to Calculate Medication Utilization." Proceedings of the 2007 SAS Global Forum, Orlando, FL. Paper 043-2007.