Barriers to and Facilitators of Medication Adherence

Patients’ failure to adhere to recommended therapy regimens results in serious negative medical and financial sequelae.¹ Recognizing generally suboptimal adherence rates, several studies have examined barriers to medication taking. Most often these reports are condition specific. Toh and colleagues² identified complex regimens as problematic for adherence in patients with chronic heart failure. Patients ran out of medication, resulting in adverse medical consequences. Toh and colleagues proposed education, counseling, and reduced dosing frequency to improve adherence. Burley,³ working with transplant patients, offered strategies to improve adherence such as giving simple instructions to patients and family members and reiterating the importance of medications. Solomon and colleagues⁴ conducted a study to systematically assess medication use for osteoporosis, assessing barriers to treatment, specific untreated patient populations, and proven methods to increase treatment rates. They found less-than-optimal adherence, no consistent predictors of undertreatment, and limited approaches to quality improvement and interventions. Currently, no welldefined mechanism exists to identify or improve nonadherence to medication regimens within or across disease states.

Leadership within a large integrated healthcare system with a well-educated patient population was interested in identifying factors that serve as barriers to and facilitators of adherence among patients taking medications for chronic conditions. Existing rates of adherence were suboptimal, despite pharmacy being a covered benefit for most patients and the existence of readily available pharmacy services. All clinics owned by the medical group have in-clinic pharmacies, and patients are able to order medications 24 hours a day through the mail order service supported by online and call-in options. To better understand adherence, we conducted a 2-part study. In the first phase of the study, we examined adherence across 8 chronic medical conditions. In the second phase, we selected the 2 conditions with the lowest adherence rates, asthma/chronic obstructive pulmonary disease (COPD) (32% adherence) and diabetes (51%), and surveyed adherent and nonadherent patients regarding barriers to and facilitators of adherence. We report the results of the survey component of the study.

METHODSPhase 1: Study Population, Data Collection, and Identifying Adherence

Study Population. The study was conducted within a large Midwestern integrated health system serving more than 800,000 patients and included all patients 18 years or older with at least 1 of 8 medical conditions. Patients were identified using the health system’s electronic medical records and administrative databases. The selected conditions represented the most prevalent conditions treated and included those with both low- and high-cost medications. These conditions also included disease states where most care is delivered through either primary or specialty care. The 8 conditions were asthma/COPD, cancer, depression, diabetes, hypercholesterolemia, hypertension, multiple sclerosis, and osteoporosis. Patients were required to have a 12-month (365 + 15 days) record of prescription coverage and a minimum of 2 prescription fills for the medication used to treat 1 of the above-mentioned conditions.

Data Collection. Data on medication fills were obtained from January 2007 through March 2009; however, each individual’s adherence for each medication was tracked for 1 year (+ 15 days) using the most recent prescription fill information. In tracking adherence for 1 year, we allowed a grace period of 15 days, recognizing that some patients are not exact to the day in obtaining medication. We did not want to classify those with small extensions in getting refills as nonadherent. Data were linked on diagnoses for a given individual to medications associated with those diagnoses to ensure that the prescriptions corresponded to the conditions. Prescription fills were required to be at least a 28-day supply to eliminate any that might have been intended for an acute situation. We also required 2 fills and examined fills at the front and back end of the 1-year window to ensurechronicity in usage. We recognized that those who stopped medication after a short duration would be excluded but chose to focus on adherence patterns in patients attempting to take medications chronically. Primary or secondary diagnoses for any of the 8 diseases of interest were identifi ed using International Classification of Diseases, Ninth Revision codes. Prescription order data were obtained using generic product identifi er codes (Master Drug Data Base v2.0, Medi-Span, Indianapolis, Indiana) for 128 medications used to treat the conditions enumerated.

Calculation of Medication Adherence. Adherence was calculated using the medication possession ratio (MPR) and a cut-point of 80% for each medication.5-9 The number of days of study participation was determined by subtracting the first fill date from the last fill date within the 12-month (+ 15 day) study period for each included patient.

Binary indicators of adherence utilized an MPR of 0.80 or higher. If the MPR was lower than 0.80, the patient was considered nonadherent. Medication adherence was calculated individually for each patient for each medication and for each disease of interest. Patients on more than 1 medication for a single disease were evaluated for adherence to each individual medication and deemed nonadherent to their regimen if they did not achieve the 80% MPR for any of the prescribed medications for that condition. From the 8 conditions of interest, we identified the 2 conditions with the lowest adherence rates: diabetes (adherence 51%) and asthma/COPD (32%). These 2 conditions became the focus for the survey of adherence barriers and facilitators.

For both conditions, we calculated adherence using pharmacy claims data that we considered robust. For asthma patients we included medications used on a scheduled basis because we could not track medications used on an as-needed basis. We examined data on patients with diabetes both including and excluding those who take insulin solely for treatment of their diabetes. We also ran analyses on these patients using an adjusted MPR, as has been done by others.¹⁰ The percentage of adherent patients did not vary a great deal regardless of approach. In the end, we included all diabetes patients in our study.

Phase 2: The Survey

Sampling for the Survey. Among patients with a single condition (asthma/COPD or diabetes) we created 4 groups for the survey component: (1) asthma/COPD adherent for each single condition (n = 413); (2) diabetes adherent for each single condition (n = 587); (3) asthma/COPD nonadherent for each single condition (n = 896); and (4) diabetes nonadherent for each single condition (n = 585). Additional groups of patients having 2 or 3 conditions were also created: (1) adherent multiple-condition patients (n = 1972) and (2) nonadherent multiple condition patients (n = 4242). The multiple-condition patients had 2 or 3 of the following conditions: diabetes, hypertension, and hypercholesterolemia. From the 4 single-condition groups, we randomly selected 250 per group, and from the multiple-condition groups we randomly selected 100 per group for a total of 1200 targeted survey participants.

The Survey. The ASK-20 questionnaire served as the basis of our survey. The questions focused on potential barriers to adherence based on lifestyle, attitudes, support, and general medication taking.¹¹ The survey also asked about patterns of medication taking. We added 2 questions about barriers to and facilitators of taking medications. For each, we included several closed-ended options based on items found in the literature^2,5-7,11,12 and our pharmacy leadership expertise. We also provided space for open-ended comments. Finally, we inquired about where patients obtain their medications (ie, in-clinic pharmacy, plan-provided mail order service, outside the system), interactions with pharmacy personnel, and demographics.

The survey was sent out in 2 waves. The first 1000 surveys were mailed to the patients with 1 condition. Subsequently, 200 surveys were sent to the patients with multiple conditions. Surveys sent to patients with 1 condition allowed for participant replacement if either the address or telephone number was problematic. Due to limited resources, there was no replacement for surveys sent to those with multiple conditions. All surveys in both waves were followed with a second survey 3 weeks after the initial mailing and then with telephone follow-up approximately 3 weeks later. There were 10 attempts to contact each subject before classifying that individual as a nonrespondent. No incentives were offered for survey completion. The study was approved by the health system’s internal review board.

Analysis. Survey items were summarized to show the proportions with each response option separately by condition and by adherence status. The primary analysis compared adherent and nonadherent patients within a condition on survey items using either independent samples with t tests or contingency tables with the Pearson χ² or Fisher exact test. There was no adjustment for multiple statistical tests (type I error). Our secondary analysis included comparisons of responses from patients with asthma/COPD with those from patients with diabetes. Statistically signifi cant differences for these comparisons are reported here. To limit table size, results of the surveys of patients with multiple conditions are summarized in the text when results differ substantially from what was found with single-condition patients.

RESULTS

Survey response rates were 72.8% in the first wave of 1000 and 58% in the second wave of 200. Where participant replacement was allowed, 11% of the initial patients targeted had to be replaced due to inadequate addresses or phone numbers. Refusal rates were 9.4% and 13.5%, respectively, and the percentages never reached after 10 attempts were 17.7% and 28.5%.

Table 1

presents demographic information for single-condition participants. Ages ranged from 18 to more than 70 years with balanced representation across age groups. Nonadherent asthma/COPD patients were younger than those deemed adherent. Overall, males represented 54% of participants, although more males provided information on asthma/COPD and more females provided information on diabetes. The majority of the population was white (88.5%) and employed full or part time; more than one-third were retired. The majority had at least completed high school and approximately 65% were married. The 116 individuals with multiple conditions were 56% female and older (51% were aged >70 years); thus, more were retired (61%) and 18% were widowed.

Response to Survey Questions

Tables 2A

and

2B

present the survey responses for those with a single condition broken out by condition. More than 90% of patients reported understanding provider instructions about medications and medication labels. Patients could call their provider with questions and were confi dent their medications helped them. About 22% reported taking medications more or less often than prescribed and 21% didn’t always have medication available when needed. Patients were unlikely to skip or stop using a medication because they didn’t think it was working, it made them feel bad, or it cost too much. Nonadherent asthma/COPD patients were more likely than adherent asthma/COPD patients to report worrying about effects on sexual health (14% vs 7%, P <.05) and not having medication with them when needed (23% vs 12%, P <.01). Among diabetic patients, nonadherent patients were more likely to report forgetting to take their medicine (42% vs 28%, P <.05), taking a medicine more or less often than prescribed (28% vs 19%, P <.05), skipping or stopping medication when they didn’t think it was working (6.4% vs 2.2%, P <.05), or skipping or stopping medication because of cost (8% vs 3%, P <.05).

Diabetes patients were less likely than asthma/COPD patients to feel confi dent that their medication would help them (adherent 76% vs 86%, P <.05; nonadherent 71% vs 82%, P <.05). Diabetes patients were more likely to report taking too many medications daily (adherent 31% vs 19%, P <.01; nonadherent 38% vs 12%, P <.05). Among those with multiple conditions, levels of agreement with survey items and patterns of differences between those adherent and nonadherent (data not shown) were similar.

Barriers to and Facilitators of Medication Adherence

Reported barriers to taking medications included frequency, forgetting, multiple medications, not obtaining refills in time, worrying about side effects, and diffi culty swallowing. From the listed barriers, those most commonly cited (

Table 3

) were having an irregular schedule (22%) and having to take pills with food (13%). Cost (18%) was cited by nonadherent patients having multiple conditions. Both single-condition, nonadherent asthma/ COPD patients (23% vs 12%, P <.01) and nonadherent patients with multiple conditions (28% vs 11%, P <.05) were more likely to cite irregular schedules as a barrier than those who were adherent. Single-condition, nonadherent asthma/COPD patients were also more likely than those who were adherent to report stopping medications when feeling better (20% vs 4%, P <.01), as were nonadherent patients with multiple conditions (18% vs 0%, P <.01). Among adherent patients, those with diabetes were more likely than asthma/COPD patients to cite irregular schedules (24% vs 12%, P <.01). Among nonadherent patients, those with diabetes were more likely to cite having to take too many pills (12% vs 4%, P <.05).

Taking medications at the same time every day was the most frequently cited facilitator (93% single condition and 91% multiple conditions). Using a weekly pill container was also reported (47% single condition and 67% multiple conditions). Among those with multiple conditions, taking medications at the same time (97% vs 84%, P <.05) and having reminder systems were more commonly cited by adherent versus nonadherent patients (12% vs 0%).

Pharmacy-Related Information

Although the pharmacy-related questions were of greatest interest to the health plan’s Pharmacy Administration Division, it was interesting to learn that incentivizing patients to use the mail order option (a 2-month cost for a 3-month supply) had both advantages and disadvantages. Participants made several comments on how they appreciated the cost savings. However, many commented that this option eliminated the face-to-face interaction with pharmacy personnel, thus decreasing the opportunity to ask questions as readily as they might in an in-person setting.

DISCUSSION

Patients who do not follow prescribed medication regimens face serious negative consequences for themselves, their providers, and their health plans. One might also consider the burden medication nonadherence places on employers and society. The aging population and increasing numbers of patients with chronic conditions are expanding reliance on medications to treat these conditions to improve quality of life and prevent morbidity and mortality. Nonetheless, even with awareness ofthe benefits medication can provide, many patients are nonadherent.^2,13-16 Although comprehensive data systems make it possible to track prescription patterns, the extraction of such information does little to explain issues surrounding adherence.¹⁷ Therefore, to better understand patient behaviors where covered pharmacy services are readily available, we conducted the current survey.

Consistent with prior reports, we found patients value having systems or daily routines making them less likely to forget to take their medications.^18,19 The single mostreported barrier (22%) was having an irregular schedule. Taking medications at the same time every day was the most frequently mentioned facilitator (93%), followed by using pill containers (47%) and having family members remind them (16%). Patients also reported several systems they use to facilitate adherence: diaries or checklists recording when medications are taken, consistent places where medications are kept (kitchen, bathroom, car), and using a spice rack to monitor supply. Smart phone alarms were mentioned as particularly helpful when one is out of town or away from the normal routine.

Interaction with a physician has been reported as an important factor in adherence.^20,21 Overall, patients in this study seemed to have a positive, interactive relationship with their providers. More than 87% claimed “My doctor and I work together to make decisions,” regardless of condition or adherence status. Similarly, social support has been suggested as an important variable affecting adherence.^19,22,23 In our population, 16% claimed family members remind them to take their medications. Receiving reminders from family members did not affect adherence in patients with asthma/ COPD. In patients with diabetes, however, adherent patients were more likely to report family members reminding them (24%) than the nonadherent counterparts (15%) (P <.05).

Although many patients have reported cost as a major barrier,^16,21,24,25 our patients had mixed comments. Differences were found between adherent and nonadherent patients with multiple conditions, with the nonadherent patients more likely to report cost as a barrier. Overall, however, more mentioned how grateful they were to have pharmacy coverage and an incentive to use the mail order pharmacy service.

As previously reported, it is evident that there is no 1 clear solution to nonadherence.³ The World Health Organization considers 5 dimensions of adherence. These include factors related to social and economic issues, the health system, the condition, the therapy, and the patient. All aspects must be taken into account and approaches individualized to target patient needs.²⁶

Limitations and Strengths

This study was conducted in a single healthcare system that includes a pharmacy benefi t, so fi ndings might not be representative of all patients. Moreover, the population was predominantly white (nearly 90%) and well educated, with fewer than 25% with a high school education or less, further limiting the generalizability. However, the system is fairly large, with more than 800,000 covered lives, and the comprehensive databases allow the acquisition of information linking diagnoses and pharmacy data. Not all systems have the capacity to obtain and link such data. Further, the issues faced by patients appear quite similar to the issues others have reported. In addition, in determining adherence for phase 1 of the overall study, we assumed that obtaining a prescription was equivalent to actually taking the medication. Thus, we may be presenting a more optimistic assessment of adherence than what is actually true. We omitted fills for less than a 28- day supply (7%) and those patients on a given medication for fewer than 12 months. We did these things to ensure that we did not categorize anyone whose initial therapy did not work for them as nonadherent. Our goal was to focus on adherence patterns in patients attempting to take medications chronically. For this initial survey and assessment, we chose to focus on the least complex patients. Although we did include a small proportion of patients with multiple conditions, their numbers were fewer and studying these patients was not the primary objective of our analysis. It is well known that adherence is poorer as the number of conditions increases.Yet even with minimal complexity, in a system with easy access, readily available pharmacy service, and patients with covered benefi ts for care and for medication, there were conditions for which intervention is sorely needed.

Within the survey component we did not ask about perceived severity of condition. Prior research has indicated that patients are more adherent if they perceive their condition to be serious.²⁷ Having this information would have been beneficial as we begin to consider various interventions to improve adherence. The Ask-20 questionnaire includes 2 questions that specifically mention pill taking. One asks if the patient is able to read and understand pill bottle labels, and the other asks if it is hard to swallow pills. For some participants, these questions would not be relevant. It is also possible that patients answered survey questions about their nonprescribed or nonscheduled medications even though our adherence assessments were calculated based on prescribed, scheduled medications. Nonetheless, we found patient responses quite informative.

Despite the limitations of the study, the comprehensive information we obtained allowed us to identify those conditions most in need of attention and has provided direction for future interventions.

CONCLUSION AND IMPLICATIONS

It appears that patients most adherent with their medication taking have created sustainable routine or systems that serve as ongoing reminders for them. Identifying approaches to address the issue of medication nonadherence remains an important priority for the health plan. The findings of this study have helped guide discussions with leadership to develop mechanisms to assist patients in efforts to adhere to prescribed medication regimens and thus maximize treatment outcomes.