Addressing Gaps in Care: Impact of Barrier-Specific Medication Adherence Intervention
July 31, 2012, 05:06:54 PM
Richard A. Zabinski, PharmD, FAMCP; Elizabeth P. Skinner, PharmD; Erin K. Buysman, MS; and C. Ron Cantrell, PhD
Chronic diseases are the leading contributor to healthcare burden in the United States.1 They continue to increase in prevalence because of factors such as the increasing frequency of lifestyle-related risk behaviors that contribute to disease and the aging of the population.2 The excess morbidity and mortality and high healthcare costs associated with chronic diseases seem at odds with major advances during the past decade. For several chronic diseases, the benefi ts of pharmacologic and nonpharmacologic therapy are well documented; however, the potential of these therapies as demonstrated in clinical studies is seldom wholly realized in the naturalistic setting.3-5
Among the most important obstacles to improving outcomes in chronic disease is poor medication adherence, or not taking medications as prescribed (most often manifested as discontinuing therapy or taking less than the prescribed dose).6-8 Poor adherence has been linked to suboptimal outcomes in chronic conditions5,9-12 and greater use of healthcare resources.9,10,13-19 Poor adherence often exceeds 50% in chronic conditions.12,20 Citing the very low rates of adherence across chronic diseases and the potentially wide-ranging humanistic and economic benefi ts of improved adherence, healthcare providers and policy makers have called for the adoption of adherence-enhancing strategies including assessment of patients’ motivation to take medications, identifi cation of patient-specific barriers to adherence, and devotion of time in treatment to address medication adherence.21-23
Risk factors for poor adherence can be patient- and disease-specifi c and are often difficult and time consuming to identify.24-26 To aid healthcare providers in identifying patient- specific risk factors for medication nonadherence and in improving communication with patients about adherence, the ASK-20 questionnaire was developed.27,28 The ASK-20 is a brief, self-reported instrument comprising 20 clinically actionable items representing factors that affect medication adherence. The ASK-20 facilitates the provision of tailored patient education based on barriers identifi ed regardless of their medical condition. In previous research,27,28 the ASK-20 questionnaire demonstrated validity and internal consistency in identifying actionable barriers to adherence across a spectrum of chronic diseases. The study reported herein was conducted to evaluate the impact on adherence of an interactive voice response (IVR)- based telephone intervention employing items from the ASK-20 questionnaire among patients with 1 or more of 12 common chronic conditions and suboptimal medication adherence at baseline. Adherence barriers can be generalizable across conditions and the use of a simple intervention across multiple chronic conditions may be helpful in designing cost-effective tools to help identify barriers to taking medicine and develop communications to prompt behavior change.
The impact of a telephone survey intervention on medication adherence among patients with chronic conditions and suboptimal adherence was assessed in a prospective cohort study. The study comprised a 180-day baseline period during which eligible patients were identified on the basis of chronic conditions and suboptimal medication adherence, an intervention date when the telephone survey intervention was administered, and a 180-day follow-up period during which adherence outcomes were measured. The study was conducted in 3 waves (“campaigns”) of surveys beginning in October 2006. Each wave or campaign included the baseline period, the intervention date, and the follow-up period. Patients identified during the 3 campaigns were identified as nonadherent on October 13, 2006, February 2, 2007, and December 3, 2007, respectively. The 180-day period up to and including each of these dates was considered the baseline period for each respective survey campaign.
The sample was derived from a database containing eligibility, pharmacy, and medical claims from a large, geographically representative US health plan that provides coverage for physician, hospital, and prescription drug services and is affiliated with OptumHealth. Study subjects were derived from 3 large US employers (280,000 covered lives) and included those with at least 1 of 12 common chronic conditions (diabetes, hyperlipidemia, hypertension, congestive heart failure, acquired immune deficiency syndrome/human immunodeficiency virus [AIDS/HIV], hepatitis C, depression, breast cancer, myocardial infarction, coronary artery disease, epilepsy, osteoporosis) who were found to be nonadherent to 1 or more medications. All of the 12 chronic conditions of interest (those conditions studied) typically require longerterm lifestyle changes. This study used an “intent-to-treat” approach toward adherence that assumes patients should continue on the prescribed medications until the end of the study follow-up period. Nonadherence, identified using EBM Connect,29 a data analytic tool that employs evidence-based algorithms to support healthcare decision making, and administrative claims data, was defined as medication possession ratio <85% for antiretroviral and <70% for the other chronic condition therapies during the previous 6 months. A detailed description of EBM Connect is beyond the scope of this manuscript and has been previously published.29 For patients identified as nonadherent in more than 1 medication category, adherence measures were averaged over all medications to which patients were nonadherent. Patients identified as nonadherent during 1 campaign and again identified as nonadherent during a subsequent campaign were analyzed as independent observations.
Patient demographics (age, gender, geographic region), a count of comorbid conditions, and medication use pertaining to the targeted chronic condition(s) were obtained from the database. Data on comorbid conditions were obtained from diagnoses of medical claims using the Clinical Classification Software maintained by the Agency for Healthcare Research and Quality (AHRQ).30 A count of medications related to the targeted chronic condition(s) during the baseline and follow-up periods was calculated based on prescription fills in the pharmacy claims data. A flag indicating an increase in the use of medications related to a patient’s chronic condition(s) was also created.
All eligible patients were exposed to a letter-based adherence intervention in which patients’ healthcare providers were sent mailings informing them of potential gaps in care including adherence issues; and patients were mailed letters suggesting that they discuss their care with their healthcare provider. After the letter-based adherence intervention was complete, patients from 2 of the 3 employers were invited to participate in an IVR-administered barriers survey and were categorized into 1 of 2 cohorts on the basis of whether or not they participated in the survey: the IVR cohort or the unableto- engage cohort. The IVR cohort participated in a less than 5-minute IVR intervention comprising a subset of ASK-20 questions, received “helpful hints” for overcoming the adherence barriers associated with each question, and were mailed a follow-up letter summarizing the survey and reiterating the “helpful hints” for improving adherence. Examples of such “helpful hints” are noted in Table 1.
Patients from the third employer were not given the opportunity to participate in the IVR-administered intervention and constituted the letter-only cohort (control cohort). While all 3 cohorts received the letter-based adherence intervention, neither the unable-to-engage cohort nor the letter-only cohort received the IVR-administered adherence intervention.
This study employed a subset of 6 of the ASK-20 questions27,28,31 covering adherence barriers related to the patient (lifestyle), the patient’s condition (attitudes and beliefs), social and economic factors, health system factors (talking with healthcare team), and therapy (taking medicines) (Table 1). The subset of questions were selected based on their frequency of response27,28 and importance to the employer.
The primary measure of adherence was proportion of days covered (PDC), calculated as the number of days covered by all prescription fills of interest during the 180-day follow-up period divided by the number of days between the first fill and the last day of the follow-up period. If patients were nonadherent to >1 medication, the PDC was averaged over all nonadherent medications. Other adherence measures included the frequency of adherence improvement, defined as the percentage of patients whose PDC increased during the 180-day follow-up period relative to the baseline period, and the frequency of adherence success, defined as the percentage of patients whose average PDC increased from suboptimal (<70%; <85% for antiretrovirals) to optimal (>70%; >85% for antiretrovirals) during the 180-day follow-up period.
Results were compared cohorts using analysis of variance (ANOVA) methods for continuous variables and χ2 tests for dichotomous and polychotomous variables. The analyses of the PDC and the percentage of patients with adherence improvement were based on the subset of patients who were continuously enrolled during the 180-day follow-up period. The analysis of the percentage of patients with adherence success was based on the subset of patients who were continuously enrolled during the 180-day follow-up period and had a suboptimal (<70%; <85% for antiretrovirals) baseline proportion of days covered calculated from the claims data.
Multivariate logistic regression adjusting for demographics, baseline comorbidities, disease-related medication use, and chronic conditions was carried out to determine the likelihood of adherence improvement or adherence success.
Cohort (ie, IVR cohort vs letter-only cohort) was the primary predictor of interest in these models; the unableto-engage cohort was not included in these analyses. All analyses were conducted using SAS, version 9.1 (SAS Institute, Inc, Cary, North Carolina).
All 9054 eligible patients received a letter suggesting that they discuss their care with their healthcare provider. Of 9054 eligible patients, 6834 were from the 2 employers whose patients were offered the IVR-administered intervention and 2220 were from the third employer (control) and were assigned the letter-only cohort. Of those 6834 eligible for the IVR-administered intervention, 276 chose to participate in the IVR-administered intervention (IVR cohort), and 6558 had wrong phone numbers, did not answer their phone, or chose not to participate (unable-to-engage cohort).
Table 2 shows demographics and baseline comorbidities of the 3 cohorts. The IVR cohort compared with the other cohorts tended to be older and had a higher proportion of females. The mean number of AHRQ comorbidities (any comorbidity, not limited to prospectively defined targeted chronic conditions of interest) was 6.4 across the sample and was highest in the IVR cohort (7.0) and lower in the other 2 cohorts (6.4 unable-to-engage cohort, 6.2 letter-only cohort) (P = .0058). Among the 12 chronic conditions prospectively identified as being of interest in this study, diabetes, hypertension, AIDS/HIV, depression, and coronary artery disease significantly differed in frequency across cohorts; the frequency of the remaining chronic conditions of interest (hyperlipidemia, congestive heart failure, hepatitis C, breast cancer, myocardial infarction, epilepsy, and osteoporosis) did not significantly differ across cohorts.
Survey Results: Adherence Barriers
Among the 276 patients in the IVR cohort, the mean number of adherence barriers before intervention was 1.30 ± 1.23. The most common barrier to adherence was the inconvenience of taking medicine more than once daily (44.0% agree or strongly agree) followed by forgetting to take medicine (25.7%), cost of medicine (25.4%), and lack of confidence that medications will help them (18.5%) (Table 3). The least common barriers to adherence were not thinking that the medicine was working and lack of understanding of healthcare provider instructions about medication. The majority of respondents indicated that they had not skipped or stopped taking a medicine because they did not think it was working (89% disagree) and that they understood their healthcare provider’s instructions about their medicines (93.0% strongly agree or agree) (Table 3).
Impact of Intervention on Adherence
Adherence outcomes were assessed for the subset of the sample that was continuously enrolled in the 180- day follow-up period (n = 256 in IVR cohort, n = 5953 in unable-to-engage cohort, n = 1726 in letter-only cohort). The mean ± SD number of unique medications (regardless of whether they were being taken for a chronic condition of interest) was highest in the IVR cohort and the unable-to-engage cohort and lower in the letter-only cohort during both the baseline period (6.9 ± 4.8 IVR letter-only cohort, 5.9 ± 3.9 unableto- engage cohort; P <.001) and the 180-day follow-up period (6.4 ± 4.2 IVR cohort, 6.1 ± 4.4 unable-to-engage cohort, 5.9 ± 4.1 letter-only cohort; P = .0386). The mean number of unique medications being taken for a chronic condition of interest differed among cohorts at baseline (2.3 ± 1.8 IVR cohort, 2.4 ± 2.0 unable-toengage cohort, 2.3 ± 1.9 letter-only cohort; P = .0043) but not during the 180-day follow-up period (2.0 ± 1.9 IVR cohort, 2.1 + 1.9 unable-to-engage cohort, 2.0 ± 1.8 letter-only cohort; P = .2253).
Average Proportion of Days Covered
During the baseline period, the average PDC was slightly higher in the IVR cohort (0.62) than in the unableto- engage cohort (0.59) or the letter-only cohort (0.58) (P = .0007) (Table 4). During the 180-day follow-up period, the PDC significantly differed among the IVR (0.60), unableto- engage (0.53), and letter-only cohorts (0.52) (P = .0023) (Table 4).
Adherence Improvement and Adherence Success
The frequency of adherence improvement was highest in the IVR cohort (55.5%) and lower in the unableto-engage- (45.4%) and letter-only cohorts (45.9%) (P =.0067) (Figure 1). In the multivariate analysis, patients in the IVR cohort were 42% more likely than patients in the letter-only cohort to demonstrate adherence improvement from baseline to follow-up (adjusted odds ratio 1.416 [95% confidence interval [CI] 1.068-1.877]).
The frequency of adherence success was highest in the IVR cohort (37.6%) and lower in the unable-to-engage- (30.2%) and letter-only cohorts (28.7%) (P = .0462) (Figure 2). In the multivariate analysis, patients in the IVR cohort were 45% more likely than patients in the letter-only cohort to demonstrate adherence success from baseline to follow-up (adjusted odds ratio 1.448 [95% CI 1.025-2.046]).
The results of this study show that an IVR-administered intervention was associated with significant improvements in both continuous and categorical measures of PDC compared with controls. Because the distribution of PDC has a tendency to be skewed, it is important to examine both the mean and median values in analysis.
Although mean PDC declined in each cohort, the IVR group had the smallest decline, and the median PDC of the IVR group increased by 4 points (6.7%) whereas the median PDC of the other 2 cohorts declined by 1 point (–1.6%), revealing a net effect in favor of the IVR group of 5 points, or 8.3%. This finding reflects the extreme challenge of correcting adherence after it has become apparent in claims data. Although our intervention was not designed to employ a prevention strategy, a preferred approach may be to address maladherence before it becomes a gap in care. Indeed, research indicates that it is much easier to prevent than correct adherence gaps.32 However, prevention entails broader scale outreach to more individuals and thereby requires use of scalable, cost-efficient technology, such as IVR.
In multivariate analyses, patients in the IVR cohort compared with the letter-only cohort were 42% more likely to demonstrate adherence improvement (defined as an increase in the proportion of days covered) and 45% more likely to demonstrate adherence success (defined as an increase in the average proportion of days covered from <70% to >70%; 85% for antiretrovirals) from baseline to follow-up.
These results are encouraging in that adherence improvement was achieved through the use of a scalable technology that delivers customized advice based on items derived from the ASK- 20 instrument. Previous research suggests that adherenceimproving interventions are most successful when they are tailored to the patient’s individual adherence barriers; however, the resource demands of individualized adherence interventions can render such tailoring unfeasible in the clinical setting.33 The ASK-20-derived telephone intervention employed in the current study addresses both the challenge of resourcing and that of providing individualized adherence intervention. First, IVR-administered delivery of adherence interventions, unlike many interventions designed to improve adherence,33 does not require intensive healthcare provider or pharmacist resources. Moreover, the use of ASK-20 items facilitated provision of individualized interventions by focusing the “helpful hints” on patient specific adherence barriers as revealed by their responses. The results of the current study are consistent with other research suggesting the promise of IVR-administered systems for improving various aspects of healthcare.34
Enrollment in the IVR administered portion was lower than expected. This may be due to the inability to reach individuals by phone. We adjusted the call times and the number of outbound calls to try to increase the enrollment rates of the second and third cohorts. Future studies should incorporate additional research to analyze ways of improving study enrollment. Among patients who participated in the telephone survey in the current study, the most common barrier to adherence was that “taking medicines more than once a day is inconvenient,” a statement with which 44% of respondents agreed or strongly agreed. The importance of dosing frequency as a determinant of adherence in chronic diseases was borne out in a systematic review of 20 studies that determined, in patients with chronic diseases, the impact of dosing frequency on adherence assessed by medication event monitoring systems.35 Adherence rates were higher in all studies among patients using less frequently dosed medications compared with those using more frequently dosed medications, and differences were statistically significant in 15 out of 20 (75%) of the studies. With oncedaily dosing compared with 3-times-daily dosing and twice-daily dosing, the frequency of adherent days was higher by 22% to 41% and 2% to 44%, respectively. Other common barriers to adherence in the current study included “I just forget to take my medicines some of the time,” and “I skipped, stopped, did not refill, or took less medicine because of the cost”—each of which had “agree” or “strongly agree” ratings among approximately 25% of respondents. The adherence survey results corroborate previous research suggesting that the barriers to medication adherence in patients with chronic diseases are numerous and diverse.36,37 The primary barriers to adherence observed in the current study have also been cited in other studies in which other prominent barriers to adherence with medication additionally included low perceived disease severity, young age, lack of confidence in the healthcare provider or the medication, and lack of motivation to change behavior.10,11,22,24-26,36 The heterogeneity of barriers illustrates their patient-specific nature and underlines the importance of barrier-tailored adherence education.
The results of this study should be interpreted in the context of its limitations. First, claims data are collected for payment purposes rather than research purposes and are subject to coding errors, factors that could impinge on the accuracy with which the data reflected patients’ medical and pharmacy utilization. Second, the use of prescription claims data to reflect medication use could introduce inaccuracy in cases in which prescribed drugs were not taken as prescribed. In addition, all of the chronic conditions studied typically require longer-term lifestyle changes, which are not observable in claims, in combination with pharmacologic therapy. This study was further limited in our ability to follow patients’ longer term, due to sample size limitations and loss to follow-up. Third, the study examined use of the specific medications to which a patient was nonadherent in the baseline period; it did not account for switching to a comparable medication in a different medication class. This study used an “intent-to-treat” approach toward adherence that assumes patients should continue on the prescribed medications
until the end of the study follow-up period. Thus, patients who switched to a different, but comparable, medication between the baseline period and the follow-up period were counted as being nonadherent regardless of whether or not they were adherent with the new medication regimen during the follow-up period. This phenomenon could also further explain the overall drop in PDC rates across all cohorts. However, the primary measures (percent of patients improving adherence and the percent reaching adherence success) were compared across cohorts, a practice that should help minimize the overall impact. Fourth, selection bias could have been operating, as patients who agreed to participate in the survey were self-selected rather than a random sample of those eligible for the survey. Patients who agreed to participate in the survey and responded to the IVR were indeed different from those who did not participate in systematic ways. This group was slightly older, with more females, more comorbidities, and better adherence at baseline. Importantly, improvement in adherence among telephone survey participants over those who did not participate in the survey was found in multivariate analyses that controlled for observable differences in baseline characteristics; however, these analyses could not control for unobservable or unknown baseline differences between cohorts.
These limitations notwithstanding, this study demonstrates the promise of an IVR-administered intervention using items from the ASK-20 questionnaire. Through IVR-administered interventions, patient-specific barriers to medication adherence were identified, and an adherence-improving helpful hints program was delivered. The IVR-administered intervention compared with a letterbased intervention was associated with the highest adherence across several measures during a 180-day follow-up period in patients with chronic diseases and suboptimal adherence at baseline.
1. Nugent R. Chronic diseases in developing countries: health and economic burdens. Ann N Y Acad Sci. 2008;1136:70-79.
2. Centers for Disease Control and Prevention. Trends in aging—United States and worldwide. MMWR Morb Mortal Wkly Rep. 2003;52(6):101-104.
3. McIntyre RS, O’Donovan C. The human cost of not achieving full remission in depression. Can J Psychiatry. 2004;49(3)(suppl 1):10S-16S.
4. Piecoro LT, Potoski M, Talbert JC, Doherty DE. Asthma prevalence, cost, and adherence with expert guidelines on the utilization of health care services and costs in a state Medicaid population. Health Serv Res. 2001;36(2):357-371.
5. Maddox TM, Ho PM. Medication adherence for the patient with coronary artery disease: challenges for the practitioner. Curr Opin Cardiol. 2009;24(5):468-472.
6. World Health Organization. Adherence to long-term therapies: evidence for action.http://www.who.int/chp/knowledge/publications/adherence_introduction.pdf. Published 2003. Accessed June 25, 2007.
7. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med. 2005;353(5): 487-497.
8. Lehane E, McCarthy G. Medication non-adherence—exploring the conceptual mire. Int J Nurs Pract. 2009;15(1):25-31.
9. Donnelly LA, Morris AD, Evans JM; DARTS/MEMO collaboration. Adherence to insulin and its association with glycaemic control in patients with type 2 diabetes. QJM. 2007;100(6):345-350.
10. DiMatteo MR. Variations in patients’ adherence to medical recommendations: a quantitave review of 50 years of research. Med Care. 2004;42(3):200-209.
11. Rabenda V, Hiligsmann M, Reginster JY. Poor adherence to oral bisphosphonate treatment and its consequences: a review of the evidence. Expert Opin Pharmacother. 2009;10(14):2303-2315.
12. Pompili M, Serafini G, Del Casale A, et al. Improving adherence in mood disorders: the struggle against relapse, recurrence and suicide risk. Expert Rev Neurother. 2009;9(7):985-1004.
13. Cantrell CR, Eaddy MT, Shah MB, Regan TS, Sokol MC. Methods for evaluating patient adherence to antidepressant therapy: a real world comparison of
adherence and economic outcomes. Med Care. 2006;44(4):300-303.
14. Kennedy S, McIntyre R, Fallu A, Lam R. Pharmacotherapy to sustain the fully remitted state. J Psychiatry Neurosci. 2002;27(4):269-280.
15. White TJ, Vanderplas A, Ory C, Dezii CM, Chang E. Economic impact of patient adherence with antidepressant therapy within a managed care organization. Dis Manag Health Outcomes. 2003;11(12):817-822.
16. Denhaerynck K, Manhaeve D, Dobbels F, et al. Prevalence and consequences of nonadherence to hemodialysis regimens. Am J Crit Care. 2007;16(3):222-225; quiz 236.
17. Leucht S, Heres S. Epidemiology, clinical consequences, and psychosocialtreatment of nonadherence in schizophrenia. J Clin Psychiatry. 2006;67(suppl 5): 3-8.
18. Sokol MC, McGuigan KA, Verbrugge RR, Epstein RS. Impact of medication adherence on hospitalization risk and healthcare cost. Med Care. 2005;43(6): 521-530.
19. Sorensen SV, Baker T, Fleurence R, et al. Cost and clinical consequence of antibiotic non-adherence in acute exacerbations of chronic bronchitis. Int J Tuberc Lung Dis. 2009;13(8):945-954.
20. Vinker S, Shani M, Baevsky T, Elhayany A. Adherence with statins over 8 years in a usual care setting. Am J Manag Care. 2008;14(6):388-392.
21. Elliott R. Non-adherence to medicines: not solved but solvable. J Health Serv
Res Policy. 2009;14(1):58-61.
22. Julius RJ, Novitsky MA Jr, Dubin WR. Medication adherence: a review of the literature and implications for clinical practice. J Psychiatr Pract. 2009;15(1): 34-44.
23. Warriner AH, Curtis JR. Adherence to osteoporosis treatments: room for improvement. Curr Opin Rheumatol. 2009;21(4):356-362.
24. Masand PS. Tolerability and adherences issues in antidepressant therapy. Clin Ther. 2003;25(8):2289-2304.
25. Bruce ML, Wells KB, Miranda J, Lewis L, Gonzalez JL; NIMH Affective Disorders Workgroup. Barriers to reducing burden of affective disorders. Ment Health Serv Res. 2002;4(4):187-197.
26. Chao J, Nau DP, Aikens JE. Patient-reported perceptions of side effects of antihyperglycemic medication and adherence to medication regimens in persons with diabetes mellitus. Clin Ther. 2007;29(1):177-180.
27. Hahn SR, Park J, Skinner EP, et al. Development of the ASK-20 adherence barrier survey. Curr Med Res Opin. 2008;24(7):2127-2138.
28. Matza L, Yu-Isenberg KS, Coyne K, et al. Further testing of the reliability and validity of the ASK-20 adherence barrier questionnaire in a medical center outpatient population. Curr Med Res Opin. 2008;24(11):3197-3206.
29. Ingenix. Symmetry EBM Connect. http://www.optuminsight.com/content/ attachments/SymmetryEBMConnectproductsheet.pdf. Published 2009. Accessed February 24, 2010.
30. Clinical Classification Software for ICD-9-CM. Comorbidity Version 3.2. Rockville, MD: Agency for Healthcare Research and Quality. http://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp. Revised March 27, 2012. Accessed June 12, 2007.
31. Matza LS, Park J, Coyne KS, Skinner EP, Malley KG, Wolever RQ. Derivation and validation of the ASK-12 adherence barrier survey. Ann Pharmacother. 2009; 43(10):1621-1630.
32. Chang A, Liberman JN, Moyna A, et al. Improving persistency for maintenance medication therapy through an interactive voice response program. Poster presented at: 22nd Academy of Managed Care Pharmacy Annual Meeting; April 7-10, 2010; San Diego, CA.
33. Banning M. A review of interventions used to improve adherence to medication in older people. Int J Nurs Stud. 2009;46(11):1505-1515.
34. Oake N, Jennings A, van Walraven C, Forster AJ. Interactive voice response systems for improving delivery of ambulatory care. Am J Manag Care. 2009;15(6): 383-391.
35. Saini SD, Schoenfeld P, Kaulback K, Dubinsky MC. Effect of medication dosing frequency on adherence in chronic diseases. Am J Manag Care. 2009;15(6): e22-e33.
36. Odegard PS, Capoccia K. Medication taking and diabetes: a systematic review of the literature. Diabetes Educ. 2007;33(6):1014-1029; discussion 1030-1031.
37. Odegard PS, Gray SL. Barriers to medication adherence in poorly controlled diabetes mellitus. Diabetes Educ. 2008;34(4):692-697.