REVIEW ARTICLE

Interventions to Enhance Medication Adherencein Chronic Medical Conditions

A Systematic Review

Sunil Kripalani, MD, MSc; Xiaomei Yao, MD; R. Brian Haynes, MD, PhD

Background: Approximately 20% to 50% of patients arenot adherent to medical therapy. This review was per-formed to summarize, categorize, and estimate the effectsize (ES) of interventions to improve medication adher-ence in chronic medical conditions.

Methods: Randomized controlled trials published fromJanuary 1967 to September 2004 were eligible if they de-scribed 1 or more unconfounded interventions intendedto enhance adherence with self-administered medicationsin the treatment of chronic medical conditions. Trials thatreported at least 1 measure of medication adherence and1 clinical outcome, with at least 80% follow-up during 6months, were included. Study characteristics and resultsfor adherence and clinical outcomes were extracted. In ad-dition, ES was calculated for each outcome.

Results: Among 37 eligible trials (including 12 informa-tional, 10 behavioral, and 15 combined informational, be-

havioral, and/or social investigations), 20 studies reporteda significant improvement in at least 1 adherence measure.Adherence increasedmost consistentlywithbehavioral in-terventionsthatreduceddosingdemands(3of3studies,largeES [0.89-1.20]) and those involving monitoring and feed-back (3 of 4 studies, small to large ES [0.27-0.81]). Adher-ence also improved in 6 multisession informational trials(small to large ES [0.35-1.13]) and 8 combined interven-tions(small to largeES[absolutevalue,0.43-1.20]).Elevenstudies(4informational,3behavioral,and4combined)dem-onstrated improvement in at least 1 clinical outcome, buteffectswerevariable(verysmall to largeES[0.17-3.41])andnot consistently related to changes in adherence.

Conclusion: Several types of interventions are effectivein improving medication adherence in chronic medical con-ditions, but few significantly affected clinical outcomes.

Arch Intern Med. 2007;167:540-550

A N ESTIMATED 20% TO 50%of patients do not taketheir medications as pre-scribed and are said to benonadherent or noncom-

pliant with therapy.1-3 In the setting ofchronic medical conditions such as hy-pertension and hypercholesterolemia,medication nonadherence leads to worsemedical treatment outcomes, higher hos-pitalization rates, and increased health carecosts.4,5 Because of this, adherence has beencalled “the key mediator between medi-cal practice and patient outcomes.”6

Researchers have tested a variety of in-terventions to improve patient adher-ence, ranging from simple adjustments inthe medication regimen to complex mul-tidisciplinary interventions that addresshealth system barriers and communica-tion between patients and health care pro-fessionals.7-10 However, the overall qual-ity of the literature is poor, with widevariability in study design, including pa-tient population, outcome measure, and

duration of follow-up.11 It is desirable todraw recommendations from a smaller setof high-quality studies.

In a recent review for the Cochrane Da-tabase,12 several simple interventions ap-peared to improve adherence with short-term regimens, such as a course ofantibiotics. However, interventions to im-prove medication use for chronic condi-tions appeared less effective overall andwere often multifaceted, making it moredifficult to synthesize published evalua-tions.12 In this article, we extend the find-ings of the Cochrane review by providinga more detailed examination of interven-tions to improve adherence in chronicmedical conditions. We offer a frame-work for categorizing these interven-tions and report their relative impact usinga standardized measure of effect size (ES).The goal of this review is to help physi-cians better understand the strengthsand limitations of tested interventions forimproving long-term medication use andidentify those that are most successful.

Author Affiliations: Division ofGeneral Medicine, Departmentof Internal Medicine, EmoryUniversity School of Medicine,Atlanta, Ga (Dr Kripalani); andDepartment of ClinicalEpidemiology and Biostatistics(Drs Yao and Haynes) andDepartment of Medicine(Dr Haynes), McMasterUniversity, Hamilton, Ontario.

(REPRINTED) ARCH INTERN MED/ VOL 167, MAR 26, 2007 WWW.ARCHINTERNMED.COM540

©2007 American Medical Association. All rights reserved. on June 20, 2008 www.archinternmed.comDownloaded from

METHODS

DATA SOURCES

Electronic searches of the published lit-erature from January 1967 to Septem-ber 2004 were conducted throughMEDLINE, CINAHL, PsycINFO,SOCIOFILE, International Pharmaceu-tical Abstracts, EMBASE, and The Coch-rane Library, without language restric-tion. The exact search strategy variedacross databases and generally in-cluded terms for adherence (eg, compli-ance, noncompliance, adherence, drop-outs, treatment refusal), medication use(eg, medication, medicine, drug, treat-ment, regimen, pharmacotherapy), andclinical trial design (eg, clinical trial, in-tervention, outcome, randomized, con-trol). A full description of the searchstrategy is provided elsewhere.12

STUDY SELECTION

Reviewers screened citation titles, in-dex terms, and abstracts (if available) toidentify potentially relevant articles,which were retrieved for full-text re-view. Two reviewers independently as-sessed the articles for possible inclu-sion. Differences in assessment wereresolved by discussion or with assis-tance from a third reviewer.

Articles were selected if they re-ported a randomized controlled trial thatdescribed 1 or more unconfounded in-terventions intended to enhance adher-ence with self-administered medica-tions used in the treatment of chronicmedical conditions. Confounding wasjudged to be present if study groups weretreated unequally except for the inter-vention intended to enhance adher-ence (eg, patients in the interventiongroup received not only more encour-agement to comply but also a differentmedication or dose than those in thecontrol group). Included studies were re-quired to report at least 1 measure ofmedication adherence and 1 clinical out-come, with at least 80% follow-up of par-ticipants during 6 months if the studyresults were initially positive. Studieswith negative results and less than 6months of follow-up were included be-cause initial failure was unlikely to befollowed by success. Studies of short-term regimens were included in theCochrane review12 but excluded from thepresent analysis to permit a focus onchronic medical conditions, for whichadherence is a greater concern. Trialsthat pertained to psychiatric disorderswere excluded because adherence is gen-erally lower in patients with psychiat-ric disorders and unique challenges may

also be present, potentially limiting thegeneralizability of interventions testedin that context.13

DATA EXTRACTION

For each eligible study, 1 reviewer ex-tracted features of the patient popula-tion, study design, interventions, and con-trol, as well as results for adherence andclinical outcomes. When multiple timepoints were provided, data were ex-tracted for the longest period of follow-up, provided follow-up remained at least80%. Each extraction was confirmed byat least 1 other reviewer. Other articles onthe same trial were retrieved, and au-thors were contacted as needed for addi-tional details or verification of reportedanalyses.

DATA SYNTHESIS

Theeligible studiesdifferedsubstantiallyin patient population, intervention, ad-herence measure, and clinical outcomemeasure, making a pooling of results in-appropriate.Instead,studiesweregroupedby intervention type, using a taxonomydeveloped from other sources.8,9,14,15 In-formational interventions described cog-nitive strategies designed primarily toeducate and motivate patients by instruc-tional means, based on the concept thatpatients who understand their condi-tion and its treatment will be more in-formed, empowered, and likely to com-ply. Informational sessions conductedindividually or in a group setting, as wellas didactic and interactive approaches,were included. Examples of informa-tional interventions are face-to-face oral,telephone, written, or audiovisual edu-cation; didactic group class; and mailedinstructional material (not including re-minders or prompts to comply). Behav-ioral interventions were strategies de-signed to influence behavior throughshaping, reminding (cues), or reward-ing desired behavior (reinforcement). Ex-amples include skill building by a healthcare professional; pillboxes, calendars, achange in packaging, or other steps in-tended to remind the patient; changes indosage schedule to simplify the regimenor tailor the regimen to the patient’s dailyroutine (ie, reduce its behavioral de-mands); and rewards and reinforce-ment (eg, assessment of adherence withfeedback to the patient). Family and so-cial interventions involved social sup-port strategies, whether provided by fam-ily or another group. Examples aresupport groups and family counseling.Group sessions that were primarily di-dactic or informational, rather than sup-portive, were categorized as informa-

tional. Combined interventions includedfeatures of 2 or 3 of the preceding cat-egories. Each included study was classi-fied by 1 author using these categories,descriptors, and examples. This classifi-cation was reviewed by at least 1 otherauthor, and disagreements were re-solved by consensus.

For each outcome, an ES (Cohend) with 95% confidence interval wascalculated from information providedin the article or obtained from theauthors. The ESs compare the differ-ence in effect between study groups,divided by the standard deviation ofthis difference, resulting in standarddeviation units.16 This measure isindependent of the method of mea-surement used, thus permitting com-parison of different interventionsacross studies. In this way, ES pro-vides more information than a simpletest of significance comparing out-comes in intervention and controlgroups. The ES can be positive ornegative, depending on the effect ofthe treatment on the particular out-come measure. If the study results arestatistically significant, then the ES isgenerally significant as well.

We calculated ESs according to es-tablished methods.16,17 For studies thatprovided range instead of a standard de-viation, we converted range to stan-dard deviation.18 The absolute value ofeach statistically significant ES was cat-egorized as very small (�0.20), small(0.20 to �0.50), medium (0.50 to�0.80), or large (�0.80).16 We calcu-lated ESs between each intervention andcontrol group. Some studies had morethan 2 arms and provided only an over-all test of significance comparing re-sults across experimental groups. Inthese cases, the significance of the ES forintervention-control pairs may have dif-fered from the overall significance of re-sults. Also, if study authors did notspecify the number of patients in-volved in the final analysis, we used thebaseline sample size of each group to cal-culate the ES. Calculating the ES in thisway, analogous to an intention-to-treatanalysis, may underestimate the trueeffect of the intervention.

RESULTS

The electronic searches returned13 061 citations (including 458 re-view articles), 955 of which werejudged to merit full-text review. Atotal of 38 articles describing 37 ran-domized controlled trials met all in-clusion criteria, including 2 articlesfrom the same trial.19,20 Among the in-

(REPRINTED) ARCH INTERN MED/ VOL 167, MAR 26, 2007 WWW.ARCHINTERNMED.COM541

©2007 American Medical Association. All rights reserved. on June 20, 2008 www.archinternmed.comDownloaded from

cluded studies, 13 reported on 12 in-formational interventions,19-31 10 onbehavioral interventions,32-41 and 15on combined interventions.42-56 Nonedescribed purely family or social in-terventions (Figure).

Most studies reported a singlemeasure of adherence. However,the choice of measure variedwidely, from self-report scales tomore objective measures obtainedby auditing refill rates, performingpill counts, and assessing the tim-ing of prescription bottle openingby Medication Event MonitoringSystem caps. Approximately half ofthe studies reported a single clini-cal outcome, which also variedgreatly, depending on the patientpopulation.

The ES could be calculated in 30studies for adherence outcomes andin 30 studies for clinical outcomes.Most studies (71% for adherence and100% for clinical outcomes) that didnot provide enough information tocalculate the ES were not statisti-cally significant, which illustrated aform of reporting bias.

INFORMATIONALINTERVENTIONS

The studies of informational inter-ventions each compared intensiveeducation to limited education or

usual care (Table 1). Sample sizevaried from 46 to 350. The interven-tions ranged in duration from a singlesession of less than 1 hour to severalhours over many sessions. Counsel-ing was provided by a physician,nurse, health educator, or pharma-cist. Both individual and group ses-sions were described, sometimes incombination, and they were often ac-companied by written materials.

Half (6 of 12) of these studies re-ported a significant increase in at least1 measure of adherence.19-21,23,25,28,30

However, 3 of these 6 trials had mixedresults. Gallefoss et al19,20 reported animprovement among only patientswith asthma not those with chronicobstructive pulmonary disease. Levyet al25 demonstrated improvement inadherence only for severe, not mild,asthma attacks. In the study bySchaffer and Tian30 of 3 educationalstrategies, 2 were effective.

Among the 10 studies for whichan ES for adherence outcomescould be calculated,19-25,28-31 a largeES was observed only in a trialconducted by Pradier et al28 ofintensive, multisession counselingfor human immunodeficiency virus(HIV) treatment adherence (ES,1.13 for self-reported adherence).Four other studies19-21,23,25 that pro-vided counseling over multiple ses-sions had a small to medium ES

(range, 0.35-0.68), and the remain-der were not significant.

Most informational interventions(8of12)didnot improveclinicalout-comes. Of the 4 studies with at least1positiveclinical result,2hadasmallES (Pradier et al28: ES,−0.39 for HIVRNA level; Levy et al25: ES, 0.34 forasthma symptom scores). The other2 trials had medium to large ESs(Gallefossetal19,20:ES,0.52 for forcedexpiratory volume in 1 second in pa-tients with asthma; Cote et al22: ES,−0.64 for urgent asthma-related vis-itsand3.41forpeakexpiratory flow).Eachof these4studiesprovidedfairlyintensive counseling with reinforce-ment over time. Only the interven-tions used by Cote et al22 and Pradieret al28 consistently led to better clini-cal outcomes, despite the study byCote et al having no effect on adher-ence owing to a high rate of adher-ence in the control group (97%).Gallefoss et al19,20 again sawimprove-ment only among the asthma sub-group, and Levy et al25 had mixed re-sults based on the clinical outcomemeasure.

BEHAVIORALINTERVENTIONS

The sample size of the behavioralstudies ranged from 27 to 497(Table 2). The most common andeffective forms of behavioral inter-vention were dosage simplifica-tion32,34,36 and repeated assessmentof medication use with feed-back.35,37-39 Other trials tested spe-cialized packaging,33 directly ob-served therapy,40 and cognitivebehavior therapy,41 but none of thesetechniques significantly affected ad-herence or clinical outcomes.

All3 trialsofdosagesimplificationdemonstratedanimprovement inad-herence, whether switching from 2doses to 1 dose per day or from 4 to2.32,34,36 The ESs that could be calcu-latedwerelarge(range,0.89-1.20).34,36

The impact on clinical outcomes wasmixed. Brown et al34 found a signifi-cantly higher percentage of pa-tients reached their low-density li-poprotein cholesterol goal whenprescribed extended-release niacincompared with the short-acting form(medium ES of 0.66). However, thestudy by Baird et al32 of once-dailymetoprolol demonstrated no

13 061 Citations Identified by Electronic Searches

12 106 Citations Excluded on Review of Title, Medical Subject Headings, and/or Abstract

955 Full-Text Articles Reviewed

917 Excluded (Not Randomized, Patients Not Prescribed Medication for Chronic Medical Disorder, Intervention Not Intended to Affect Adherence With Prescribed Self-administered Medications, Confounding, No Measurement of Medication Adherence, No Measurement of Treatment Outcomes, <80% Follow-up, <6 Months of Follow-up)

38 Articles Included in Review

10 Articles Describing Behavioral Interventions

15 Articles Describing Combined Interventions

13 Articles Describing 12 Informational Interventions

Figure. Study flow.

(REPRINTED) ARCH INTERN MED/ VOL 167, MAR 26, 2007 WWW.ARCHINTERNMED.COM542

©2007 American Medical Association. All rights reserved. on June 20, 2008 www.archinternmed.comDownloaded from

Table 1. Informational Interventions

SourcePopulation

and Sample Sizes Intervention/ControlAdherence Measures

and Results, %ES (95% CI)

for AdherenceClinical Outcome Measures

and Results, %ES (95% CI) for

Clinical Outcome

Cantode Cetinaet al,21 2001

Women seekingcontraception,I = 175, C = 175

I: Structured pretreatmentcounseling on DMPA withaudiovisual materials,including mode of action,expected benefits, and sideeffects (emphasized noharm to health), reinforcedat each follow-up visit.

C: Routine information onexpected side effects.

Continued DMPAinjections (clinicattendance, 12 mo):I = 82.9, C = 56.6,� = 26.3, P�.05

Medium, 0.59(0.38 to 0.80)

Pregnancy rate: I = 0, C = 0,� = 0, P = NS

None, 0(−0.21 to 0.21)

Cote et al,22

2001Asthma, I = 33, C = 30 I: Structured education: limited

education plus structurededucational program(PRECEDE modeladdressed beliefs, attitudes,knowledge, and socialsupport), reinforcement at6-mo visit.

C: Limited education oninhaler technique andself-management planbased on PEF.

Had prescription forsteroid inhaler(self-report, 12 mo):I = 100, C = 97, � = 3,P = .70

None, 0.35(−0.15 to 0.84)

(1) Urgent visits for asthmaexacerbation (6-12 mo)*:I = 9, C = 34, � = −25,P = .03;

(2) PEF (mean ± SEM,12 mo): I = 424 ± 25L/min, C = 349 ± 19 L/min,� = 75 L/min, P = .03

(1) Medium, −0.64(−1.13 to −0.14);

(2) large, 3.41(2.64 to 4.18)

Gallefosset al,19,20

1999

Asthma (I = 39, C = 39)and COPD (I = 31,C = 31)

I: Educational booklet, groupsessions (2 2-h sessions)by physician andpharmacist, 1 or 2individual sessions (40 min)from both a nurse and aphysiotherapist,self-management plan.

C: Usual care.

Received �75% ofprescribed doses ofinhaled steroid (refillaudit, 12 mo):

(1) asthma: I = 57, C = 32,� = 25, P = .04;

(2) COPD: I = 50, C = 58,� = −8, P = NS

(1) Medium, 0.51(0.04 to 0.98);

(2) none, −0.16(−0.70 to 0.38)

Change in FEV1 (mean ± SD,12 mo):

(1) asthma: I = 112 ± 386 mL,C = −83 ± 383 mL, � = 195mL, P�.05;

(2) COPD: I = 97 mL, C = 50mL, � = 47 mL, P = NS†

(1) Medium, 0.52(0.04 to 0.99);

(2) NA‡

Hill et al,23

2001Rheumatoid arthritis,

I = 51, C = 49I: Individual education

(7 30-min sessions).C: Usual care (drug

information leaflet).

Adherence �85% (plasmalevels ofpharmacologicalmarker, 24 wk): I = 85,C = 55, � = 30, P = .01

Medium, 0.68(0.22 to 1.13)

(1) Articular index (mean,range): I = 15.9 (0-41),C = 13.7 (0-58), � = 2.2,P = .33;

(2) morning stiffness (inminutes): I = 49 (0-308),C = 50 (0-190), � = −1,P = .41;

(3) pain score:I = 2.38 (1-3.28),C = 2.55 (1.14-4.50),� = −0.17, P = .44†

(1) None, −0.19(−0.64 to 0.27)§;

(2) none, 0.02(−0.44 to 0.47)§;

(3) none, 0.25(−0.20 to 0.71)§

Laporte et al,24

2003Thromboembolic

disease, I = 43,C = 43

I: Intensive education (writtenmaterial, daily nurse andphysician visits while inhospital, daily tests abouteducation, emphasis onadherence).

C: Standard education (basicinformation, no emphasison adherence).

(1) % Correct bottleopenings (MEMS caps,[mean ± SD, 3 mo]):I = 84.5 ± 17.7,C = 80.7 ± 19.4,� = 3.8, P = .36;

(2) % Of prescribed pillstaken (pill count, 3 mo):I = 99.7 ± 15.1,C = 100.6 ± 18.2,� = −0.9, P = .82

(1) None, 0.23(−0.34 to 0.54);

(2) NA‡

(1) % of INRs in target range(mean ± SD, 3 mo):I = 51.3 ± 17.5,C = 55.9 ± 16.1, � = −4.6,P = .23;

(2) % of time in target INRrange (mean ± SD, 3 mo):I = 61.6 ± 17.6,C = 65.8 ± 17.6, � = −4.2,P = .28

(1) None, −0.09(−0.53 to 0.34);

(2) none, −0.09(−0.52 to 0.35)

Levy et al,25

2000Asthma, I = 103,

C = 108I: Individual education by

nurse (1-h session, thentwo 30-min sessions at6-wk intervals),self-management planbased on PEF or symptommonitoring.

C: Usual care.

Increased use of inhalersfor asthma attacks(self-report, 6 mo):

(1) inhaled steroids formild attacks: I = 47,C = 23, � = 24, P = .11;

(2) rescue medications formild attacks: I = 89,C = 82, � = 7, P = .67;

(3) inhaled steroids forsevere attacks: I = 51,C = 21, � = 24,P�.001;

(4) rescue medications forsevere attacks: I = 89,C = 76, � = 13, P�.05

(1) None, 0.51(−0.10 to 1.12);

(2) none, 0.20(−0.41 to 0.81);

(3) medium, 0.64(0.30 to 0.98);

(4) small, 0.35(0.01 to 0.69)

(1) PEF (6 mo): � = 20.1L/min, P�.05;

(2) severe attacks (6-wkperiod): I = 34, C = 42,� = −8, P = NS;

(3) symptom scores (mean ±SD, 6 mo): I = 45.7 ± 22.9,C = 38.1 ± 22.0, � = 7.6,P�.001;

(4) routine physician visits(median, range):I = 1 (1-6), C = 1 (1-23),� = 0, P�.05;

(5) emergency departmentvisits (median, range):I = 0 (1-7), C = 0 (1-7),� = 0, P = NS

(1) None, 0.15(−0.14 to 0.44);

(2) none, −0.17(−0.46 to 0.13);

(3) small, 0.34(0.06 to 0.62);

(4) none, 0(−0.34 to 0.34)§||;

(5) none, 0(−0.34 to 0.34)§||

Morice andWrench,26

2001

Asthma, I = 40, C = 40 I: Individual inpatienteducation by nurse(two 30-min sessions),booklet, self-managementplan based on PEF.

C: Usual care.

Self-reported adherence(6 mo): similar inI and C

NA‡ Urgent visits, call-outs, orreadmission (N): I = 25,C = 25, � = 0, P = NS

None, −0.29(−0.77 to 0.20)

(continued)

(REPRINTED) ARCH INTERN MED/ VOL 167, MAR 26, 2007 WWW.ARCHINTERNMED.COM543

©2007 American Medical Association. All rights reserved. on June 20, 2008 www.archinternmed.comDownloaded from

significant improvement in bloodpressure, whereas Girvin et al36 ac-tually found a trend toward in-creased blood pressure in the once-daily therapy group.

The 4 trials that involved assess-ment and feedback asked patients tomonitor and report their medica-tion use and/or blood pressure.35,37-39

The 3 studies that then provided pa-tients with tailored feedback, rein-forcement, or rewards demon-strated a significant improvement inadherence (ESs small to large; range,

0.27-0.81).35,37,39 The fourth study,which simply made home bloodpressure values available to the pa-tient and physician, did not im-prove adherence.38 Only 1 study inthis group produced a clear improve-ment in clinical outcomes (MarquezContreras et al39: ES,0.89 for totalcholesterol and 0.78 for low-density lipoprotein cholesterol).Friedman and colleagues35 re-ported a decrease in the diastolicblood pressure only after control-ling for baseline blood pressure in

adjusted analyses (small ES of 0.29);they found no significant change forsystolic blood pressure.

COMBINED INTERVENTIONS

Most (13 of 15) combined interven-tions included informational andbehavioral components,42-45,47-55 and2 joinedsocial support strategieswitheither informationalorbehavioralele-ments (Table 3).46,56 Sample sizeagain varied widely, from 37 to 1113.Of the 13 studies with informational

Table 1. Informational Interventions (cont)

SourcePopulation

and Sample Sizes Intervention/ControlAdherence Measures

and Results, %ES (95% CI)

for AdherenceClinical Outcome Measures

and Results, %ES (95% CI) for

Clinical Outcome

Petersonet al,27 2004

Dyslipidemia, I = 45,C = 49

I: Home visits by pharmacist(monthly for 6 mo);included counseling onmedication and lifestylechanges, assessment ofside effects, point-of-carecholesterol testing.

C: Usual care.

Self-reported adherence(6 mo): no change

NA‡ Cholesterol levels (mean ±SD): I = 4.4 ± 0.6 mmol/L,C = 4.6 ± 0.8 mmol/L,� = 0.2 mmol/L, P = .24

None, −0.29(−0.72 to 0.15)

Pradier et al,28

2003HIV, I = 124, C = 122 I: Individually delivered

session focused oncognitive, emotional, socialand behavioral determinantsaffecting adherence(3 45- to 60-min sessions).

C: Usual care.

100% Adherence(self-report, 6 mo):I = 75, C = 61, � = 14,P = .04

Large, 1.13(0.86 to 1.41)

Change in HIV RNA frombaseline (mean ± SD,6 mo): I = −0.22 ± 0.86 logcopies/mL, C = 0.12 ± 0.90log copies/mL, � = −0.34log copies/mL, P = .002

Small, −0.39(−0.64 to −0.14)

Rawlingset al,29 2003

HIV, I = 96, C = 99 I: Small group sessions led byhealth care professional,focusing on patientempowerment, HIVtreatment, and medicationmanagement (weekly for 4wk); used flip charts,videotapes, patientlogbooks, and patientworkbooks, plus routineadherence counseling.

C: Routine adherencecounseling.

% of Prescribed dosestaken (MEMS caps,6 mo): I = 70, C = 74� = −4, P = NS

None, −0.09(−0.37 to 0.19)

% With HIV RNA �40copies/mL (24 wk): I = 60,C = 55, � = 5, P = .53

None, 0.10(−0.25 to 0.46)

Schaffer andTian,30 2004

Asthma, I1 = 10, I2 = 12,I3 = 11, C = 13

I: Audiotape alone (I1),booklet alone (I2), oraudiotape plus booklet (I3).

C: Usual care (standardprovider education).

% of Doses filled (refillaudit, 6 mo): I1 = 48,I2 = 77, I3 = 77, C = 40;

(1) I1-C: � = 8, P = .17;(2) I2-C: � = 37, P = .02;(3) I3-C: � = 34, P = .04

(1) None, 0.16(−0.66 to 0.99);

(2) none, 0.77(−0.01 to 1.56)¶

(3) none, 0.77(−0.03 to 1.57)¶

Asthma controlquestionnaire# (6 mo):I1 = 1.47, I2 = 1.30,I3 = 1.30, C = 1.25;

(1) I1-C: � = 0.22, P = .50;(2) I2-C: � = 0.05, P = .40;(3) I3-C: � = 0.05, P = .80

(1) None, 0.30(−0.53 to 1.13);

(2) none, 0.36(−0.43 to 1.15);

(3) none, 0.11(−0.69 to 0.91)

van Es et al,31

2001Asthma, N = 112 I: Discussion of

self-management plan withpediatrician, individualeducation by nurse(4 30-min sessions), groupeducation (3 90-minsessions).

C: Usual care (pediatricianevery 4 mo).

Self-reported adherence(10-point scale, mean ±SD, 12 mo):I = 7.8 ± 1.6,C = 7.3 ± 1.8, � = 0.5,P = .14

None, 0.30(−0.13 to 0.73)

(1) FEV1 (% predicted, mean ±SD, 12 mo)†:I = 96.2 ± 17.0,C = 96.9 ± 13.4, � = −0.7,P = .83;

(2) FEV1/FVC (mean ± SD,12 mo)†: I = 80.9 ± 10.8,C = 84.5 ± 8.1, � = −3.6,P = .10

(1) None, −0.05(−0.47 to 0.37);

(2) none, −0.38(−0.82 to 0.06)

Abbreviations: C, control; CI, confidence interval; COPD, chronic obstructive pulmonary disease; DMPA, depot-medroxyprogesterone acetate; ES, effect size;FEV1, forced expiratory volume in 1 second; FVC, forced vital capacity; HIV, human immunodeficiency virus; I, intervention; INR, international normalized ratio;MEMS, medication event monitoring system; NA, not available; NS, not significant; PEF, peak expiratory flow; PRECEDE, Predisposing, Reinforcing, Enabling, Causes in,Educational Diagnosis and Evaluation; �, difference between intervention and control groups.

*Exact values not provided in article. Results are estimated from a figure.†Results obtained or confirmed by personal communication with authors.‡Could not be calculated with available data.§The ES was calculated by converting range into standard deviation.||The ES was calculated by using median instead of mean.¶The 95% CI may be inaccurate owing to small sample size.#Lower score indicates better asthma control.

(REPRINTED) ARCH INTERN MED/ VOL 167, MAR 26, 2007 WWW.ARCHINTERNMED.COM544

©2007 American Medical Association. All rights reserved. on June 20, 2008 www.archinternmed.comDownloaded from

Table 2. Behavioral Interventions

SourcePopulation andSample Sizes Intervention/Control

Adherence Measuresand Results, %

ES (95% CI)for Adherence

Clinical OutcomeMeasures and Results, %

ES (95% CI) forClinical Outcome

Baird et al,32

1984Hypertension,

I = 196,C = 193

I: Metoprolol once a day.C: Metoprolol twice a day.

% of Prescribed pillstaken (pill count, 10wk): P = .009

NA* BP: P = NS NA*

Becker et al,33

1986Hypertension,

I = 86,C = 85

I: Special reminder (blister)packaging ofanti-hypertensives.

C: Traditional pill bottles.

(1) 100% Adherence(self-report): I = 56.0,C = 54.1, � = 1.9,P = NS;

(2) �80% of prescribedpills taken (pill count):I = 84, C = 75.3,� = 8.7, P = NS

(1) None, 0.04(−0.26 to 0.34);

(2) none, 0.22(−0.08 to 0.52)

DBP (3 mo): I = 85.3 mm Hg,C = 88.8 mm Hg, � = 3.5 mm Hg,P = NS

NA*

Brown et al,34

1997Dyslipidemia,

N = 31(crossoverstudy)

I: Controlled release niacintwice a day.

C: Regular niacin 4 times a day.

% of Prescribed pillstaken (pill count, 8 mo):I = 96, C = 85, � = 11,P = .01

Large, 0.89(0.38 to 1.41)

Achieved LDL-C �100 (8 mo):I = 83, C = 52, � = 31, P�.01

Medium, 0.66(0.15 to 1.18)

Friedmanet al,35

1996

Hypertension,N = 299(resultsavailablefor I = 133,C = 134)

I: Telephone-linked computersystem to assess adherence,side effects, knowledge, andBP (weekly for 6 mo); feedbackgiven to patient and physician.

C: Usual care.

Change in adherence (pillcount, 6 mo): I = 17.7,C = 11.7, � = 6, P = .03

Small, 0.27(0.03 to 0.51)

(1) Decrease in SBP (6 mo): I = 11.5mm Hg, C = 6.8 mm Hg, � = 4.7mm Hg, P = .20†;

(2) decrease in DBP (6 mo): I = 5.2mm Hg, C = 0.8 mm Hg, � = 4.4mm Hg, P = .02†

(1) None, 0.16(−0.08 to 0.40);

(2) small, 0.29(0.05 to 0.53)

Girvin et al,36

1999Hypertension,

N = 27(crossoverstudy)

I: Enalapril, 20 mg once daily.C: Enalapril, 10 mg twice daily.

% of Prescribed dosestaken (mean with 95%CI, 16 wk):

(1) pill count:I = 99.0 (97.6 to 100.4), C = 94.9 (92.8-97.0),� = 4.1, P�.01;

(2) MEMS:I = 101.2(98.9 to 103.6),C = 90.1 (85.4 to 94.8),� = 11.1, P�.001

(1) Large, 0.92(0.34 to 1.50);

(2) large, 1.20(0.60 to 1.80)

(1) SBP (mean with 95% CI, 16wk)‡: I = 139.1 mm Hg(135.0-143.2 mm Hg), C = 133.8mm Hg (129.4-138.3 mm Hg),� = 5.3 mm Hg, P = .07;

(2) DBP (mean with 95% CI,mm Hg, 16 wk)‡:I = 86.4 (83.7-89.2),C = 85.4 (82.6-88.1), � = 1.0,P = .09

(1) None, 0.54(−0.03 to 1.10);

(2) none, 0.51(−0.06 to 1.07)

Hayneset al,37

1976

Hypertension,I = 20,C = 19

I: Self-monitoring of pills andblood pressure, tailoredmedication schedule, homevisits by research assistant,rewards.

C: Usual care.

% of Prescribed dosestaken (pill count, 12thmo): I = 65.8, C = 43.2,� = 22.6, P = .02

Medium, 0.78(0.12 to 1.44)

DBP (mean ± SD, 12 mo):I = 93.1 ± 1.3 mm Hg,C = 96.4 ± 1.3 mm Hg, � = −3.3mm Hg, P = .12

None, −0.60(−1.25 to 0.05)

Johnsonet al,38

1978

Hypertension,I1 = 35,I2 = 35,I3 = 35,C = 35

I: Self-monitoring of bloodpressure at home (I2), monthlyhome visits by researchassistant to monitor bloodpressure (I3), or both (I1);physician and patient givenresults.

C: Usual care.

% of Prescribed dosestaken (pill count andinterview, 6 mo):I1 = 76.3, I2 = 78,I3 = 69.3, C = 68.5;

(1) I1-C: � = 7.8, P = NS;(2) I2-C: � = 9.5, P = NS;(3) I3-C: � = 0.8, P = NS

(1) None, 0.17(−0.30 to 0.65);

(2) none, 0.22(−0.26 to 0.69);

(3) none, 0.02(−0.46 to 0.50)

DBP (mean, 6 mo): I1 = 95.9mm Hg, I2 = 94.1 mm Hg,I3 = 95.2 mm Hg, C = 95.7mm Hg;

(1) I1-C: � = 0.2 mm Hg, P = NS;(2) I2-C: � = −1.6 mm Hg, P = NS;(3) I3-C: � = −0.5 mm Hg, P = NS

(1) None, 0.02(−0.45 to 0.49);

(2) none, −0.15(−0.62 to 0.33);

(3) none, −0.14(−0.62 to 0.34)

MarquezContreraset al,39

2004

Dyslipidemia,I = 63,C = 63

I: 3 Telephone calls (at 7-10 d,2 mo, and 4 mo) to assessadherence, provide feedback,and give a tailoredrecommendation.

C: Usual care.

% of Prescribed dosestaken (pill count, mean± SD, 6 mo):I = 93.0 ± 8.2,C = 84.4 ± 12.8,� = 8.6, P�.001

Large, 0.81(0.43 to 1.19)

(1) Decrease in total cholesterol(mean ± SD, 6 mo): I = 64.7 ± 35mg/dL, C = 33.1 ± 37 mg/dL,� = 31.6 mg/dL, P�.005;

(2) decrease in LDL-C (mean ± SD,6 mo): I = 63.2 ± 27 mg/dL,C = 35.6 ± 42 mg/dL, � = 27.6mg/dL, P = .001

(1) Large, 0.89(0.50 to 1.27);

(2) medium, 0.78(0.41 to 1.16)

Walley et al,40

2001Tuberculosis,

I1 = 170,I2 = 165,C = 162

I: Direct observation of therapyby health workers (I1), ordirect observation of treatmentby family members (I2).

C: Usual care (self-administeredtreatment).

Compliant with collectingmedications fromhealth center (8 mo)§:I1 = 73, I2 = 68, C = 67;

(1) I1-C: � = 6, P = NS;(2) I2-C: � = 2, P = NS

(1) None, 0.13(−0.09 to 0.36);

(2) none, 0.02(−0.20 to 0.24)

(1) Cure rate (8 months): I1 = 64,I2 = 55, C = 62;

(1A) I1-C: � = 2, P = .73;(1B) I2-C: � = −7, P = .23(2) Cure or completed treatment (8

mo): I1 = 67, I2 = 62, C = 65;(2A) I1-C: � = 2, P = .75;(2B) I2-C: � = −3, P = .65

(1A) None, 0.04(−0.18 to 0.27);

(1B) none, −0.14(−0.36 to 0.08);

(2A) none, 0.04(−0.18 to 0.27);

(2B) None, −0.06(−0.28 to 0.16)

Weber et al,41

2004HIV, I = 32,

C = 28I: Cognitive behavior therapy

(3-25 sessions during 1 y).C: Usual care (monthly clinic

visits).

% of Prescribed dosestaken (MEMS, 10-12mo): I = 92.8, C = 88.9,� = 3.9, P = .15||

None, 0.14(−0.41 to 0.68)

(1) Undetectable viral load: I = 72.4,C = 79.2, � = −6.8, P = NS;

(2) CD4 count (median with range):I = 407.5 (203-955),C = 490 (95-1796), � = −82.5,P = NS

(1) None, −0.16(−0.70 to 0.38);

(2) none, −0.26(−0.80 to 0.28)¶

Abbreviations: BP, blood pressure; C, control; CI, confidence interval; DBP, diastolic blood pressure; ES, effect size; HIV, human immunodeficiency virus;I, intervention; LDL-C, low-density lipoprotein cholesterol; MEMS, medication event monitoring system; NA, not available; NS, not significant; SBP, systolic bloodpressure; �, difference between intervention and control groups.

*Could not be calculated with available data.†Adjusted for age, sex, baseline BP, and baseline adherence; unadjusted analyses showed no significant difference.‡Night-time SBP and DBP also showed a trend toward higher values in the intervention group.§Nonadherence was defined as failing to collect medications for 2 consecutive months.||Adherence also compared by a 10-point visual analog scale, I = 9.93, C = 9.80, � = 0.13, P = .012.¶The ES was calculated by using median instead of mean and by converting range to standard deviation.

(REPRINTED) ARCH INTERN MED/ VOL 167, MAR 26, 2007 WWW.ARCHINTERNMED.COM545

©2007 American Medical Association. All rights reserved. on June 20, 2008 www.archinternmed.comDownloaded from

and behavioral elements, 5 demon-strated improvement in all adher-ence measures,42,47,48,50,51 and 3 othersreported mixed or nearly significantresults.43,45,53 The significant effectsranged in size from small to large (ab-solute ES values of 0.43-1.20). Nostudies in this group were able todemonstrate a significant change inall measured clinical parameters, al-though several trials noted a signifi-cant improvement in at least 1 clini-caloutcome42,48,51,55 orastrongtrend.53

The ESs ranged widely across stud-ies (very small to large ES [0.17-1.58]). No consistent relationshipswere observed between interven-tion characteristics or disease stateand the success of the interventions.For example, 3 trials implementedrelatively similar interventionsamong patients with asthma, buttheir reported effects varied sub-stantially.42,45,47 Regarding the 2 stud-ies with strong social support ele-ments, both successfully improvedadherence (unable to calculate ES),but neither led to better clinical out-comes.46,56

COMMENT

In this systematic review of random-ized controlled trials designed to en-hance medication adherence inchronic medical conditions, 16 of 37trials reported a consistent improve-ment in adherence.* Four otherstudies noted a significant change inadherence among a particular pa-tient subgroup, with some but notall of the adherence measures used,or in a certain arm of a study test-ing multiple interventions.19,20,25,30,43

Nine of the 20 trials that had someimpact on adherence were able todemonstrate an improvement insome19,20,25,35,42,48,51 or all28,34,39 of themeasured clinical outcomes. Of the17 studies that did not show an im-provement in adherence, 2 im-proved clinical outcomes; the ap-parent lack of effect on adherence inthese studies was probably due to in-sensitive adherence measures or aceiling effect.22,55

Previous reviews have foundfew if any consistent relationships

between the characteristics andeffectiveness of interventions toimprove adherence.7 In the pres-ent study, categorization of thetrials by intervention type (infor-mational, behavioral, or combinedinformational, behavioral, and/orsocial support) and calculation ofESs permit a few general observa-tions. Behavioral interventionsthat reduced the dosing demandsof individual therapies consis-tently improved adherence with alarge ES (0.89-1.20). Apart fromthis, other successful interven-tions usually contained multipleelements delivered over time. Forexample, 6 informational inter-ventions that provided educa-tional counseling over a few ses-sions and addressed self-careissues improved adherence withsmall to large ESs (0.35-1.13).Successful behavioral interven-tions based in monitoring andfeedback and the most successfulcombined interventions alsoincluded various elements, suchas self-management plans, rein-forcement, and occasional lyrewards.

Despite the ability of these inter-ventions to improve adherence,however, a positive effect on clini-cal outcomes was demonstrated in-frequently, and ESs pertaining toclinical outcomes were highly vari-able (very small to large ES [0.17-3.41]). Moreover, results for clini-cal outcomes were not consistentlyrelated to the magnitude of adher-ence results or the characteristics ofthe intervention. Unfortunately,many of the studies were relativelysmall and not sufficiently poweredto detect differences in clinical out-comes. Thus, the results of many ofthe studies are indeterminate forclinical outcome, rather than clearlynegative.

The adherence literature and pre-sent study had several additionallimitations. First, because we in-cluded only published trials, the re-ported findings may overestimate thetrue effect of such interventions dueto publication bias. Second, our in-clusion criteria were somewhat strin-gent. Many studies with shorter fol-low-up or no measurement ofclinical outcomes were excluded,and they are listed elsewhere.12

Third, our ability to categorize thetrials into distinct modes of inter-vention was limited by the fre-quently complex nature of the in-terventions, requiring us to makejudgments about the predominantcomponents. Fourth, although wecalculated ES and can make gen-eral comments about the magni-tude of adherence and clinical out-come differences, the includedstudies were too heterogeneous towarrant pooling in a formal meta-analysis. Fifth, the literature re-view was completed in September2004 and trials published since thenmay differ in their results, al-though they would be unlikely tosubstantially change the conclu-sions drawn from the 37 trials in-cluded in this review.

Because most of the availableliterature does not separate outthe effects of the individual com-ponents of multifaceted interven-tions, it is not possible to drawdefinit ive conclusions aboutwhich features of combined inter-ventions are most beneficial.Additional research is needed toclarify which features are mostresponsible for changes in adher-ence and clinical outcomes, withthe caveat that individual compo-nents may not prove powerfulenough to show important effects.Future studies should also examinethe effect of varying the intensity ofinterventions to determine dose-response relationships. Such find-ings would have important im-plications for health systemsconsidering the implementation ofpatient adherence programs on alarge scale. Investigations shouldbe conducted with clinically mean-ingful outcomes as the primary endpoints and be sufficiently poweredto detect a difference in these mea-sures. Most important, futureresearch should seek to understandthe determinants of adherencebehavior and to develop and testinnovative ways to help peopleadhere to prescribed medicationregimens, rather than persistingwith existing approaches.

In summary, we found that sev-eral types of interventions are ef-fective in improving medicationadherence, but few were able todemonstrate an impact on clinical

*References 21, 23, 28, 32, 34-37, 39, 42,46-48, 50, 51, 56.

(REPRINTED) ARCH INTERN MED/ VOL 167, MAR 26, 2007 WWW.ARCHINTERNMED.COM546

©2007 American Medical Association. All rights reserved. on June 20, 2008 www.archinternmed.comDownloaded from

Table 3. Combined Interventions

SourcePopulation andSample Sizes Intervention/Control

Adherence Measuresand Results, %

ES (95% CI)for Adherence

Clinical OutcomeMeasures and Results, %

ES (95% CI) forClinical Outcome

Bailey et al,42

1990Asthma, I = 132,

C = 135I: Educational pamphlets,

workbook, tailoredself-management plan,individual counseling session,support group, telephonefollow-up, reinforcement.

C: Usual care (educationalpamphlets, routine physicianencouragement).

(1) Self-reported adherence toinhaler (�5 of 6 items, 12mo): I = 86.1, C = 55.9,� = 30.2, P = .001;

(2) self-reported adherence tomedications (�5 of 6 items,12 mo): I = 96.0, C = 82.7,� = 13.3, P = .005;

(3) adherence rated excellentby staff (12 mo): I = 78.3,C = 51.0, � = 27.3, P = .001

(1) Medium, 0.69(0.43 to 0.95);

(2) small, 0.46(0.19 to 0.72);

(3) medium, 0.58(0.32 to 0.84)

Self-report (12 mo):(1) severe symptom in past 7 d:

I = 11.4, C = 27.0, � = −15.6,P = .002;

(2) bothered by asthma in past 7 d:I = 64.6, C = 75.0, � = −10.4,P = .11;

(3) �5 episodes in past 3 mo:I = 17.5, C = 47.4, � = −29.9,P�.001;

(4) asthma interfered with life inpast 3 mo: I = 52.9, C = 59.1,� = −6.2, P = .13

(1) Small, −0.40(−0.67 to −0.14);

(2) none, −0.23(−0.49 to 0.04);

(3) medium, −0.66(−0.92 to −0.39);

(4) none, −0.12(−0.39 to 0.14)

Berrienet al,43

2004

HIV, I = 20,C = 17

I: 8 Structured nurse home visitsduring 3 mo to educate,identify and resolve barriers,and provide incentives;included notebooks andpill-swallowing training.

C: Usual care (medicationcounseling at clinic every3 mo, visual aids, pillboxes,reminders, emotional support,nurse telephone call asneeded, 1 home visit ifadherence poor).

(1) Refill score (4 points):I = 2.7, C = 1.7, � = 1.0,P = .002;

(2) self-reported adherence(change from baseline,mean ± SEM, 3 mo):I = 2.7 ± 0.88,C = 0.2 ± 0.96, � = 2.5,P = .07

(1) Large, 1.20(0.46 to 1.93);

(2) none, 0.67(−0.03 to 1.36)

Undetectable viral load (6-11 mo):I = 45.0, C = 23.5, � = 21.5,P = NS

None, 0.46(−0.20, to 1.11)

Brus et al,44

1998Rheumatoid

arthritis,I = 29, C = 31

I: 6 Group patient educationalmeetings (4 2-h meetingsduring the first months,reinforcement meetings at 4and 8 mo); partners invited;included counseling onadherence, energyconservation, joint protection,physical activity education andtraining, behavioral contract,and feedback.

C: Brochure on rheumatoidarthritis.

% of Prescribed doses taken(pill count, mean ± SD,6-12 mo): I = 89 ± 16,C = 84 ± 21, � = 5, P = NS

None, 0.15(−0.38 to 0.68)

(1) Disease activity score (12 mo):P = NS;

(2) C-reactive protein (12 mo):P = NS;

(3) functional score (12 mo):P = NS;

(4) range of motion (12 mo):P = NS

NA*

Cote et al,45

1997Asthma, N = 188

(I1 = 50,I2 = 45, C = 54completed thestudy)

I: Asthma education program(1 h) with book and writtenaction plan based on peak flowmonitoring with remindingand reinforcing at each visit(I1) or symptom monitoring(I2).

C: Usual care (physicianinstruction on medication use,symptom triggers, inhalertechnique, and possibly oralaction plan).

�60% of Prescribed dosestaken (weight of usedcanisters, 12 mo): P = .06

NA* All are decrease from baseline,mean ± SEM, 12 mo:

(1) hospitalizations:I1 = 0.20 ± 0.07, I2 = 0.31 ± 0.19,C = 0.17 ± 0.07, P = .60;

(2) emergency department visits:I1 = 1.6 ± 0.4, I2 = 1.2 ± 0.4,C = 1.5 ± 0.4, P = .50;

(3) oral corticosteroid course:I1 = 0.5 ± 0.3, I2 = 0.4 ± 0.2,C = 0.8 ± 0.3, P = .20;

(4) days lost from work or school:I1 = 6.6 ± 3.1, I2 = 3.5 ± 1.9,C = 4.5 ± 3.2, P = .60

(1A) I1-C: Small, 0.43(0.04 to 0.82);

(1B) I2-C: large, 1.03(0.60 to 1.45);

(2A) I1-C: none, 0.25(−0.13 to 0.64);

(2B) I2-C: medium, −0.76(−1.17 to −0.35);

(3A) I1-C: large, −1.01(−1.42 to −0.60);

(3B) I2-C: large, −1.56(−2.01 to −1.11);

(4A) I1-C: medium, 0.67(0.28 to 1.07);

(4B) I2-C: none, −0.38(−0.78 to 0.02)

Coull et al,46

2004Ischemic heart

disease,I = 165,C = 154

I: Monthly informational orsupportive group meetings(2 h) led by lay healthmentors.

C: Usual care.

Self-reported adherence(12 mo): P�.01

NA* Total events (mean, 12 mo):I = 0.38, C = 0.39, � = −0.01,P = NS

NA*

Farber andOliveria,47

2004

Asthma, I = 28,C = 28

I: Basic asthma education,written self-management plan,prescriptions, holdingchamber, and 3 follow-up callsto reinforce plan (at 1-2 wk,3-4 wk, and 3 mo).

C: Usual care.

Dispensing events ofcontrolling medications(eg, corticosteroids) (mean,6 mo): I = 2.1, C = 0.63,� = 1.47, P = .004

Large, 0.87(0.29 to 1.44)

Urgent visits for asthmaexacerbation: I = 43, C = 36,� = 7, P = .56

None, 0.14(−0.41 to 0.70)

Knobelet al,48

1999

HIV, I = 65,C = 121

I: Individual counseling, detailedinformation aboutmedications, tailoredmedication schedule,telephone support, monthlyclinic visits.

C: Usual care.

�90% of Prescribed dosestaken on schedule (pillcount, 24 wk): I = 76.7,C = 52.7, � = 24, P = .002

Medium, 0.51(0.19 to 0.82)

(1) Undetectable viral load (24 wk):I = 65.0, C = 54.4, � = 10.6,P = .18;

(2) reduction in viral load (mean ±SD, 24 wk): I = 1.98 ± 0.78log10/mL, C = 1.02 ± 0.5log10/mL, � = 0.96 log10/mL,P = .04;

(3) increase in CD4 cell count(mean ± SD, 24 wk):I = 96.3 ± 56, C = 55.1 ± 33,� = 41.2, P�.001†

(1) None, 0.22(−0.10 to 0.53);

(2) large, 1.58(1.22 to 1.93);

(3) large, 0.98(0.64 to 1.31)

(continued)

(REPRINTED) ARCH INTERN MED/ VOL 167, MAR 26, 2007 WWW.ARCHINTERNMED.COM547

©2007 American Medical Association. All rights reserved. on June 20, 2008 www.archinternmed.comDownloaded from

Table 3. Combined Interventions (cont)

SourcePopulation andSample Sizes Intervention/Control

Adherence Measuresand Results, %

ES (95% CI)for Adherence

Clinical OutcomeMeasures and Results, %

ES (95% CI) forClinical Outcome

Nazarethet al,49

2001

Elderlywith �4medications,I = 165,C = 151

I: Home visit by pharmacist 7-14 d afterhospital discharge (includedadherence assessment, counseling ofpatient and caregivers, disposing ofexcess medications, contactingphysician as needed).

C: Usual care.

Self-reported adherence(mean ± SD, 6 mo):I = 0.78 ± 0.3,C = 0.78 ± 0.3, � = 0,P = NS

None, 0(−0.36 to 0.36)

(1) Subsequent hospitalization (6 mo):I = 27.9, C = 28.4, � = −0.5, P = NS;

(2) death (6 mo): I = 16.1, C = 12.6,� = 3.5, P = NS;

(3) outpatient department attendance:I = 28.5, C = 26.5, � = 2, P = NS;

(4) general practitioner attendance:I = 71, C = 70.7, � = 0.3, P = NS

(1) None, −0.01(−0.25 to 0.23);

(2) none, 0.10(−0.14 to 0.34);

(3) none, 0.05(−0.20 to 0.29);

(4) none, 0.01(−0.23 to 0.25)

Petersonet al,50

1984

Epilepsy, I = 27,C = 26

I: Counseling, tailoring of medicationschedule to lifestyle, leaflet, specialmedication container, self-monitoringof medication use and seizures,mailed reminders for appointmentsand missed refills.

C: Usual care.

(1) Plasma phenytoin level(6 mo): I = 9.9µmol/L/mg/kg, C = 7.1µmol/L/mg/kg, � = 2.8µmol/L/mg/kg, P�.05;

(2) refilled medication within1 wk of due date (refillaudit, 6 mo): I = 88,C = 50, � = 38, P�.01

(1) Medium, 0.72(0.08 to 1.36);

(2) large, 0.86(0.31 to 1.42)

Self-reported seizures (median, 6 mo):I = 2.5, C = 3.5, � = 1.0, P = NS

NA*

Pietteet al,51

2000

Diabetes,I = 137,C = 143

I: Automated telephone assessment andtailored education with reminders,reinforcement, and nurse follow-up.

C: Usual care.

Self-reported medicationnonadherence (12 mo):I = 48, C = 69, � = −21,P = .003

Small, −0.43(−0.68 to −0.18)

(1) HbA1c: I = 8.1, C = 8.4, � = −0.3,P = .10;

(2) serum glucose: I = 180 mg/dL,C = 221 mg/dL, � = −41 mg/dL,P = .002;

(3) diabetes-related symptoms (N):I = 4.0, C = 5.4, � = −1.4, P�.001

(1) None, 0.21(−0.04 to 0.46);

(2) small, 0.40(0.15 to 0.65);

(3) small, 0.43(0.17 to 0.68)

Sackettet al,52

1975

Hypertension,I1 = 37,I2 = 28,I3 = 44,C = 25

I: 2 � 2 factorial design: care atworksite by occupational healthphysicians (I1), programmededucation and adherence reminders(I2), or both interventions (I3).

C: Usual care.

(1) �80% of Prescribeddoses taken (pill count,6 mo): P = NS;

(2) urine metabolites:P = NS;

(3) characteristic change inserum electrolytes:P = NS

NA* DBP: P = NS NA*

Tuldraet al,53

2000

HIV, I = 55,C = 61

I: Psychoeducative intervention toimprove knowledge and self-efficacy,tailored medication schedule,telephone number provided, oralreinforcement at clinic visits.

C: Usual care.

�95% Self-reportedadherence (48 wk): I = 58,C = 41, � = 17, P = .06

None, 0.34(−0.05 to 0.73)

Undetectable viral load (48 wk): I = 58,C = 45, � = 13, P = .06

None, 0.26(−0.13 to 0.65)

Volumeet al,54

2001

Elderlywith �3medications,I = 159,C = 204

I: Comprehensive pharmaceutical careservices including 30- to 45-minindividual assessment by pharmacistand frequent follow-upcommunication, with standarddocumentation of contacts.

C: Usual care (pharmacist-patientcontact triggered by prescriptions).

Self-reported adherence(4-item scale, mean ± SD,12-13 mo)‡:I = 0.56 ± 0.75,C = 0.47 ± 0.69, � = 0.09,P = NS

None, 0.13(−0.08 to 0.33)

HRQOL (SF-36, mean ± SD, 12-13mo):

(1) mental component score:I = 56.14 ± 8.30, C = 54.55 ± 8.65,� = 1.59, P = NS;

(2) physical component score:I = 36.87 ± 11.62,C = 38.39 ± 11.44, � = −1.52,P = NS

(1) None, 0.19(−0.02 to 0.40);

(2) none, −0.13(−0.34 to 0.08)

Weinbergeret al,55

2002

Asthmaand COPD,I = 447,C1 = 363,C2 = 303

I: Pharmaceutical care programincluding review of monthly PEF data,adherence, and health care utilization;patient education materials; and otherresources to facilitate pharmacistintervention.

C: C1: monthly PEF self-monitoring (datanot provided to pharmacist);C2: usual care (no PEF metersprovided).

(1) Self-reportednonadherence (12 mo):I = 22.5, C1 = 22.7,C2 = 23.3, P = .22;

(2) self-reported adherence(4-item scale, mean ±SD, 12 mo)‡:I = 0.8 ± 1.1,C1 = 0.8 ± 1.1,C2 = 0.8 ± 1.0, P = .57

(1A) I-C1: None,−0.01(−0.16 to 0.15);

(1B) I-C2: None,−0.02(−0.18 to 0.14):

(2A) I-C1: none, 0(−0.15 to 0.15);

(2B) I-C2: None, 0(−0.16 to 0.16)

(1) Urgent visits for breathing-relatedproblem (12 mo):

(1A) COPD: P = .34;(1B) asthma: P�.001§;(2) HRQOL (mean ± SEM, 12 mo):(2A) COPD: I = 4.42 ± 0.06,

C1 = 4.30 ± 0.08, C2 = 4.31 ± 0.08,P = NS;

(2B) asthma�: I = 4.97 ± 0.06,C1 = 4.93 ± 0.06, C2 = 4.83 ± 0.07,P = NS;

(3) PEF rate (% predicted, mean ±SEM, 12 mo): I = 63.72 ± 0.58,C1 = 64.56 ± 0.65,C2 = 61.82 ± 0.71, P = .006

(1) NA*;(2A) I-C1: none, 0.14

(−0.11 to 0.39);I-C2: none, 0.16(−0.09 to 0.41);

(2B) I-C1: none, 0.05(−0.15 to 0.24);I-C2: none, 0.17(−0.05 to 0.38);

(3A) I-C1: none, −0.08(−0.23 to 0.08);

(3B) I-C2: very small,0.17 (0.01 to 0.34)

Wysockiet al,56

2001

Diabetes,I1 = 38,I2 = 40,C = 41

I: I1: Behavior-family systems therapy(10 sessions), monetary incentive;I2: education and support (10 groupeducational and social supportmeetings), monetary incentive.

C: Usual care (pediatric endocrinologyfollow-up, self-managementtraining).

Self-care inventory (changefrom baseline, mean ±SD, 12 mo)¶: I1 = 3.3,I2 = −1.2, C = −5.4;

(1) I1-C: � = 8.7;(2) I2-C: � = 4.2; overall

P�.05

NA* HbA1c (change from baseline, 12 mo):I1 = 0.9, I2 = 0.3, C = 1.1;

(1) I1-C: � = −0.2;(2) I2-C: � = −0.8; overall P = NS

NA*

Abbreviations: C, control; CI, confidence interval; COPD, chronic obstructive pulmonary disease; DBP, diastolic blood pressure; ES, effect size; HbA1c, hemoglobin A1c;HIV, human immunodeficiency virus; HRQOL, health-related quality of life; I, intervention; NA, not available; NS, not significant; PEF, peak expiratory flow;SF-36, 36-Item Short-Form Health Survey; �, difference between intervention and control groups.

*Could not be calculated with available data.†P value recalculated from results in article.‡Lower score indicates better adherence.§Intervention group had more urgent visits than control groups.||Results obtained or confirmed by personal communication with authors.¶Higher score indicates better adherence.

(REPRINTED) ARCH INTERN MED/ VOL 167, MAR 26, 2007 WWW.ARCHINTERNMED.COM548

©2007 American Medical Association. All rights reserved. on June 20, 2008 www.archinternmed.comDownloaded from

outcomes. Based on this review ofthe literature, the most effective in-terventions appear to be those thatsimplify dosing demands. Inter-ventions that involve monitoringand feedback, as well as informa-tional interventions delivered overmultiple sessions, are probablyalso effective.

Accepted for Publication: August23, 2006.Correspondence: Sunil Kripalani,MD, MSc, Division of General Medi-cine, Department of Internal Medi-cine, Emory University School ofMedicine, 49 Jesse Hill Jr Dr SE,Atlanta, GA 30303 (skripal@emory.edu).Author Contributions: Dr Kripalanihad full access to all of the data inthe study and takes responsibility forthe integrity of the data and the ac-curacy of the data analysis. Studyconcept and design: Kripalani andHaynes . Acquis i t ion o f data :Kripalani, Yao, and Haynes. Analy-sis and interpretation of data:Kripalani, Yao, and Haynes. Draft-ing of the manuscript: Kripalani, Yao,and Haynes. Critical revision of themanuscript for important intellec-tual content: Kripalani and Yao. Sta-tistical analysis: Kripalani and Yao.Obtained funding: Haynes. Adminis-trative, technical, and material sup-port: Yao and Haynes. Study super-vision: Kripalani and Haynes.Financial Disclosure: None re-ported.Funding/Support: Dr Kripalani wassupported by a Mentored Patient-Oriented Research Career Develop-ment Award (grant K23 HL077597)and formerly by the Emory Men-tored Clinical Research Scholars Pro-gram (National Institutes of Health/National Center for ResearchResources grant K12 RR017643). DrHaynes is supported by the MichaelGent Professorship.Role of the Sponsors: The fundershad no role in the design and con-duct of the study; collection, man-agement, analysis, and interpreta-tion of the data; or the preparation,review, or approval of the manu-script.Acknowledgment: We thankStephen Walter, PhD, McMasterUniversity, for his assistance with EScalculations and comments on an

early draft of the manuscript. Wealso thank Heather McDonald, MSc,Aqeel Degani, MSc, and Amit Garg,MD, PhD, for their contributions toprevious Cochrane reviews.

REFERENCES

1. DiMatteo MR. Variations in patients’ adherence tomedical recommendations: a quantitative reviewof 50 years of research. Med Care. 2004;42:200-209.

2. Osterberg L, Blaschke T. Adherence to medication.N Engl J Med. 2005;353:487-497.

3. Haynes RB, Taylor DW, Sackett DL, eds. Compli-ance in Health Care. Baltimore, Md: Johns Hop-kins University Press; 1979.

4. DiMatteo MR, Giordani PJ, Lepper HS, CroghanTW. Patient adherence and medical treatment out-comes: a meta-analysis. Med Care. 2002;40:794-811.

5. Sokol MC, McGuigan KA, Verbrugge RR, EpsteinRS. Impact of medication adherence on hospital-ization risk and healthcare cost. Med Care. 2005;43:521-530.

6. Kravitz RL, Melnikow J. Medical adherence re-search: time for a change in direction? Med Care.2004;42:197-199.

7. McDonald HP, Garg AX, Haynes RB. Interven-tions to enhance patient adherence to medica-tion prescriptions: scientific review. JAMA. 2002;288:2868-2879.

8. Peterson AM, Takiya L, Finley R. Meta-analysis oftrials of interventions to improve medicationadherence. Am J Health Syst Pharm. 2003;60:657-665.

9. Roter DL, Hall JA, Merisca R, Nordstrom B, Cre-tin D, Svarstad B. Effectiveness of interventionsto improve patient compliance: a meta-analysis.Med Care. 1998;36:1138-1161.

10. Krueger KP, Felkey BG, Berger BA. Improving ad-herence and persistence: a review and assess-ment of interventions and description of steps to-ward a national adherence initiative. J Am PharmAssoc. 2003;43:668-678.

11. Nichol MB, Venturini F, Sung JC. A critical evalu-ation of the methodology of the literature on medi-cation compliance. Ann Pharmacother. 1999;33:531-540.

12. Haynes RB, Yao X, Degani A, Kripalani S, Garg A,McDonald HP. Interventions to enhance medica-tion adherence. Cochrane Database Syst Rev.2005;(4):CD00011.

13. Haynes RB. Determinants of compliance: the dis-ease and the mechanisms of treatment. In: HaynesRB, Taylor DW, Sackett DL, eds. Compliance inHealth Care. Baltimore, Md: Johns Hopkins Uni-versity Press; 1979:49-62.

14. Krousel-Wood M, Hyre A, Muntner P, Morisky D.Methods to improve medication adherence in pa-tients with hypertension: current status and fu-ture directions. Curr Opin Cardiol. 2005;20:296-300.

15. Walter I, Nutley SM, Davies HTO. Developing a tax-onomy of interventions used to increase the im-pact of research. http://www.st-andrews.ac.uk/~ruru/Taxonomy%20development%20paper%20070103.pdf. Accessed November 29, 2005.

16. Cohen J. Statistical Power Analysis for the Be-havioral Sciences. 2nd ed. Hillsdale, NJ: LawrenceErlbaum Associates; 1988.

17. Hedges LV, Olkin I. Statistical Methods for

Meta-analysis. Orlando, Fla: Academic Press; 1985.18. Walter S, Yao X. Effect sizes can be calculated for

studies reporting ranges for outcome variables.J Clin Epidemiol. In press.

19. Gallefoss F, Bakke PS. How does patient educa-tion and self-management among asthmatics andpatients with chronic obstructive pulmonary dis-ease affect medication? Am J Respir Crit Care Med.1999;160:2000-2005.

20. Gallefoss F, Bakke PS, Rsgaard PK. Quality of lifeassessment after patient education in a random-ized controlled study on asthma and chronic ob-structive pulmonary disease. Am J Respir Crit CareMed. 1999;159:812-817.

21. Canto de Cetina TE, Canto P, Ordonez Luna M.Effect of counseling to improve compliance inMexican women receiving depot-medroxypro-gesterone acetate. Contraception. 2001;63:143-146.

22. Cote J, Bowie DM, Robichaud P, Parent JG, Bat-tisti L, Boulet LP. Evaluation of two different edu-cational interventions for adult patients consult-ing with an acute asthma exacerbation. Am JRespir Crit Care Med. 2001;163:1415-1419.

23. Hill J, Bird H, Johnson S. Effect of patient educa-tion on adherence to drug treatment for rheuma-toid arthritis: a randomised controlled trial. AnnRheum Dis. 2001;60:869-875.

24. Laporte S, Quenet S, Buchmuller-Cordier A, et al.Compliance and stability of INR of two oral anti-coagulants with different half-lives: a ran-domised trial. Thromb Haemost. 2003;89:458-467.

25. Levy ML, Robb M, Allen J, Doherty C, Bland JM,Winter RJ. A randomized controlled evaluation ofspecialist nurse education following accident andemergency department attendance for acuteasthma. Respir Med. 2000;94:900-908.

26. Morice AH, Wrench C. The role of the asthma nursein treatment compliance and self-management fol-lowing hospital admission. Respir Med. 2001;95:851-856.

27. Peterson GM, Fitzmaurice KD, Naunton M, VialJH, Stewart K, Krum H. Impact of pharmacist-conducted home visits on the outcomes of lipid-lowering drug therapy. J Clin Pharm Ther. 2004;29:23-30.

28. Pradier C, Bentz L, Spire B, et al. Efficacy of aneducational and counseling intervention on ad-herence to highly active antiretroviral therapy:French prospective controlled study. HIV ClinTrials. 2003;4:121-131.

29. Rawlings MK, Thompson MA, Farthing CF, et al;NZTA 4006 Helping to Enhance Adherence to An-tiretroviral Therapy (HEART) Study Team.Impact of an educational program on efficacy andadherence with a twice-daily lamivudine/zidovudine/abacavir regimen in underrepre-sented HIV-infected patients. J Acquir Immune De-fic Syndr. 2003;34:174-183.

30. Schaffer SD, Tian L. Promoting adherence: ef-fects of theory-based asthma education. Clin NursRes. 2004;13:69-89.

31. van Es SM, Nagelkerke AF, Colland VT, ScholtenRJ, Bouter LM. An intervention programme usingthe ASE-model aimed at enhancing adherence inadolescents with asthma. Patient Educ Couns.2001;44:193-203.

32. Baird MG, Bentley-Taylor MM, Carruthers SG, et al.A study of efficacy, tolerance and compliance ofonce-daily versus twice-daily metoprolol (Beta-loc) in hypertension. Clin Invest Med. 1984;7:95-102.

33. Becker LA, Glanz K, Sobel E, Mossey J, Zinn SL,

(REPRINTED) ARCH INTERN MED/ VOL 167, MAR 26, 2007 WWW.ARCHINTERNMED.COM549

©2007 American Medical Association. All rights reserved. on June 20, 2008 www.archinternmed.comDownloaded from

Knott KA. A randomized trial of special packag-ing of antihypertensive medications. J Fam Pract.1986;22:357-361.

34. Brown BG, Bardsley J, Poulin D, et al. Moderatedose, three-drug therapy with niacin, lovastatin,and colestipol to reduce low-density lipoproteincholesterol �100 mg/dl in patients with hyper-lipidemia and coronary artery disease. Am JCardiol. 1997;80:111-115.

35. Friedman RH, Kazis LE, Jette A, et al. A telecom-munications system for monitoring and counsel-ing patients with hypertension. Impact on medi-cation adherence and blood pressure control. AmJ Hypertens. 1996;9:285-292.

36. Girvin B, McDermott BJ, Johnston GD. A com-parison of enalapril 20 mg once daily versus 10mg twice daily in terms of blood pressure lower-ing and patient compliance. J Hypertens. 1999;17:1627-1631.

37. Haynes RB, Sackett DL, Gibson ES, et al. Improve-ment of medication compliance in uncontrolledhypertension. Lancet. 1976;1:1265-1268.

38. Johnson AL, Taylor DW, Sackett DL, Dunnett CW,Shimizu AG. Self-recording of blood pressure inthe management of hypertension. CMAJ. 1978;119:1034-1039.

39. Marquez Contreras E, Casado Martinez JJ,Corchado Albalat Y, et al. Efficacy of an interven-tion to improve treatment compliance inhyperlipidemias. Aten Primaria. 2004;33:443-450.

40. Walley JD, Khan MA, Newell JN, Khan MH.Effectiveness of the direct observation compo-nent of DOTS for tuberculosis: a randomisedcontrolled trial in Pakistan. Lancet. 2001;357:664-669.

41. Weber R, Christen L, Christen S, et al. Effect of

individual cognitive behaviour intervention on ad-herence to antiretroviral therapy: prospective ran-domized trial. Antivir Ther. 2004;9:85-95.

42. Bailey WC, Richards JM Jr, Brooks CM, SoongSJ, Windsor RA, Manzella BA. A randomized trialto improve self-management practices of adultswith asthma. Arch Intern Med. 1990;150:1664-1668.

43. Berrien VM, Salazar JC, Reynolds E, McKay K;H. I. V. Medication Adherence Intervention Group.Adherence to antiretroviral therapy in HIV-infected pediatric patients improves with home-based intensive nursing intervention. AIDS Pa-tient Care STDS. 2004;18:355-363.

44. Brus HL, van de Laar MA, Taal E, Rasker JJ, Wieg-man O. Effects of patient education on compli-ance with basic treatment regimens and health inrecent onset active rheumatoid arthritis. AnnRheum Dis. 1998;57:146-151.

45. Cote J, Cartier A, Robichaud P, et al. Influence onasthma morbidity of asthma education pro-grams based on self-management plans follow-ing treatment optimization. Am J Respir Crit CareMed. 1997;155:1509-1514.

46. Coull AJ, Taylor VH, Elton R, Murdoch PS, Har-greaves AD. A randomised controlled trial of se-nior Lay Health Mentoring in older people withischaemic heart disease: The Braveheart Project.Age Ageing. 2004;33:348-354.

47. Farber HJ, Oliveria L. Trial of an asthma educa-tion program in an inner-city pediatric emer-gency department. Pediatr Asthma AllergyImmunol. 2004;17:107-115.

48. Knobel H, Carmona A, Lopez JL, et al. Adherenceto very active antiretroviral treatment: impact ofindividualized assessment. Enferm Infecc Micro-biol Clin. 1999;17:78-81.

49. Nazareth I, Burton A, Shulman S, Smith P, HainesA, Timberal H. A pharmacy discharge plan for hos-pitalized elderly patients—a randomized con-trolled trial. Age Ageing. 2001;30:33-40.

50. Peterson GM, McLean S, Millingen KS. A ran-domised trial of strategies to improve patient com-pliance with anticonvulsant therapy. Epilepsia.1984;25:412-417.

51. Piette JD, Weinberger M, McPhee SJ, Mah CA,Kraemer FB, Crapo LM. Do automated calls withnurse follow-up improve self-care and glycemiccontrol among vulnerable patients with diabetes?Am J Med. 2000;108:20-27.

52. Sackett DL, Haynes RB, Gibson ES, et al. Ran-domised clinical trial of strategies for improvingmedication compliance in primary hypertension.Lancet. 1975;1:1205-1207.

53. Tuldra A, Fumaz CR, Ferrer MJ, et al. Prospec-tive randomized two-arm controlled study to de-termine the efficacy of a specific intervention toimprove long-term adherence to highly active an-tiretroviral therapy. J Acquir Immune Defic Syndr.2000;25:221-228.

54. Volume CI, Farris KB, Kassam R, Cox CE, Cave A.Pharmaceutical care research and educationproject: patient outcomes. J Am Pharm Assoc(Wash). 2001;41:411-420.

55. Weinberger M, Murray MD, Marrero DG, et al.Effectiveness of pharmacist care for patients withreactive airways disease: a randomized con-trolled trial. JAMA. 2002;288:1594-1602.

56. Wysocki T, Greco P, Harris MA, Bubb J, WhiteNH. Behavior therapy for families of adolescentswith diabetes: maintenance of treatment effects.Diabetes Care. 2001;24:441-446.

(REPRINTED) ARCH INTERN MED/ VOL 167, MAR 26, 2007 WWW.ARCHINTERNMED.COM550

©2007 American Medical Association. All rights reserved. on June 20, 2008 www.archinternmed.comDownloaded from