NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Viswanathan M, Golin CE, Jones CD, et al. Closing the Quality Gap: Revisiting the State of the Science (Vol. 4: Medication Adherence Interventions: Comparative Effectiveness). Rockville (MD): Agency for Healthcare Research and Quality (US); 2012 Sep. (Evidence Reports/Technology Assessments, No. 208.4.)

Cover of Closing the Quality Gap: Revisiting the State of the Science (Vol. 4: Medication Adherence Interventions: Comparative Effectiveness)

Closing the Quality Gap: Revisiting the State of the Science (Vol. 4: Medication Adherence Interventions: Comparative Effectiveness).

Show details

Discussion

This chapter summarizes key findings and strength of evidence for each Key Question (KQ), followed by a summary of the limitations of the review, limitations of the evidence base, gaps in the evidence that may benefit from future research, and overall conclusions.

Key Findings and Strength of Evidence

Key Question 1. Effect of Patient, Provider, or Systems Interventions on Medication Adherence and Other Outcomes

Overview

Overall, the evidence from 57 trials in 63 articles included in this comparative effectiveness review suggests that numerous pathways provide opportunities to improve medication adherence across clinical conditions. These approaches include relatively low-cost, low-intensity telephone and mail interventions. They also include some relatively intense interventions, such as care coordination and case management (requiring close and ongoing monitoring of patients) and collaborative care; such interventions often require some, or even a good deal of restructuring of typical approaches to health care delivery in the United States.

Despite such evidence about promising approaches to improving medication adherence, only a subset of these effective interventions relate better adherence with better health outcomes or other important end results. We found relatively little evidence linking improved adherence to improvements in other outcomes, such as biomarkers, morbidity, mortality, quality of life, quality of care, patient satisfaction, health care utilization, and costs.

Findings Specific to Clinical Conditions

The volume of evidence regarding improving medication adherence differs sharply by clinical condition. We found the greatest amount of evidence, in terms of numbers of trials or studies or numbers of subjects (or both), for hypertension and depression, followed by hyperlipidemia, asthma, and diabetes (Table 75). We did not find a substantial body of evidence testing varied approaches to inform several other clinical conditions. For musculoskeletal diseases, we found three trials that used interventions with no common features. Myocardial infarction, glaucoma, and multiple sclerosis had just one trial each. We found no eligible studies for cancer; reasons likely include the restrictions specified for this comparative effectiveness review to patient-administered medications and to outpatient settings. We found no eligible studies that explicitly focused on patients with adherence problems relating to polypharmacy, although a few studies included patients with two or more conditions and assessed adherence to more than one medication.

Table 75. Summary of results for patient, provider, and systems interventions (KQ 1).

Table 75

Summary of results for patient, provider, and systems interventions (KQ 1).

Collectively, the most consistent evidence was that various types of interventions improved medication adherence outcomes for hypertension, heart failure, depression, and asthma. These improvements were accompanied by improvements in systolic and diastolic blood pressure for case management and face-to-face education with pharmacists for hypertension; reduced emergency department (ED) visits and improved patient satisfaction for pharmacist-led multicomponent interventions for heart failure; improved symptoms, pulmonary function, health care utilization, and quality of life for shared decisionmaking; improved symptoms for case management for depression; and improved symptoms and patient satisfaction with medications and quality of care for collaborative care for depression. We generally graded these interventions as beneficial with low-to-moderate strength of evidence, depending on the specific type of intervention. Of note, three clinical conditions (hypertension, heart failure, and depression) included some interventions for which evidence was insufficient due to lack of consistency or precision in the evidence (Table 76).

Table 76. Summary of strength-of-evidence grades for medication adherence by type of intervention.

Table 76

Summary of strength-of-evidence grades for medication adherence by type of intervention.

For asthma and hypertension, because of several studies of low or moderate risk of bias that failed to find an effect, we judged that two interventions provided evidence of no benefit: these two interventions included collaborative care for hypertension and patient or provider access to patient adherence data for asthma.

Trials in diabetes, hyperlipidemia, and musculoskeletal diseases found a single intervention indicating benefit for medication adherence. These trials focused on care coordination and collaborative care approaches for diabetes, education and behavioral support for hyperlipidemia, and a virtual clinic for osteoporosis; all other approaches did not produce improvements and were judged to be insufficient for lack of consistency or lack of precision in the results.

The least consistent evidence of improvement in medication adherence pertained to patients with multiple chronic conditions: three trials, using pharmacist-based outreach, education, and problem-solving approaches, provided evidence of no benefit for medication adherence, and findings from another trial, using case management, were insufficient.

We found the least evidence for myocardial infarction, glaucoma, and multiple sclerosis. Single trials in each of these clinical areas suggested low strength of evidence of benefit for medication adherence.

Findings Specific to Interventions

We identified 20 intervention approaches (Table 76) across the clinical conditions included in this comparative effectiveness review. Intervention approaches tested in patient populations with different clinical conditions (either single diagnoses of chronic illnesses or, in some cases, two or more such ailments) included case management, collaborative care, decision aids, education, reminders, and pharmacist-led multicomponent approaches. Our findings suggest that educational interventions and case management approaches offer the most consistent and voluminous evidence of improvements in medication adherence across varied clinical conditions. We found moderate strength of evidence for self-management interventions for asthma, which generally include strong educational components. Trials showing improvement with case management and educational interventions provided some evidence of improvement for other health outcomes. We found low strength of evidence of benefit from educational interventions for medication adherence for hypertension, hyperlipidemia, and myocardial infarction, and insufficient evidence for diabetes. We found low or moderate strength of evidence of benefit from case management for diabetes, hypertension, heart failure, and depression, insufficient evidence for musculoskeletal diseases, and low strength of evidence of no benefit for persistence for multiple chronic conditions.

Other promising approaches tested and found to be effective in more than one clinical area include reminders and pharmacist-led multicomponent approaches. Interventions such as shared decisionmaking and blister packaging were tested in a single clinical area with a single trial; without additional evidence, their widespread applicability is difficult to judge but may well hold promise.

Some interventions may be most effective for a particular clinical condition. Collaborative care appeared to be effective primarily for patients with depression or with depression and diabetes; for other clinical conditions (hyperlipidemia and hypertension), the evidence was insufficient.

The categories noted above are shorthand for one or more key elements of very diverse interventions. As explained in earlier chapters, we opted not to try to impose any external taxonomy on these markedly different programs; none seemed suitable for capturing the underlying constructs or specific activities we encountered in this literature. For instance, of the two trials categorized as interventions that gave health care providers access to patient adherence data, one included a substantial pharmaceutical care program, whereas the other did not. Thus, the inductive approach we used to identify types of interventions allowed us to group them in ways that seemed to reflect key similarities, but doing so limited our ability to draw firm conclusions about the effectiveness of specific intervention features. In addition, the trials that tested multicomponent efforts did not have multiple intervention arms that would have provided information about particular (individual) elements of the intervention effort. Nevertheless, we attempted to address this limitation through analyses for KQ 3, and those findings offer further insights on some common elements across these interventions.

Key Question 2. Effect of Policy Interventions on Medication Adherence and Other Outcomes

Five studies evaluated the effects of policy-level interventions on medication adherence, specifically for cardiovascular disease, diabetes, and respiratory conditions. One study was a randomized controlled trial (RCT). The other four studies used cohort designs. All of the studies assessed medication adherence using insurance claims data to measure either the medication possession ratio (MPR) or proportion of days covered (PDC). The use of similar adherence measures across the studies facilitates comparison of results.

All five studies evaluated policy-level interventions that reduced patient out-of-pocket expenses for prescription medications, either through reduced medication copayments or improved prescription drug coverage. The study by Zhang and colleagues evaluated the impact of Medicare Part D on medication adherence among groups of older adults who had different levels of prescription drug coverage prior to implementation of Medicare Part D.151 This study found a large improvement in adherence among individuals who had had no prescription drug coverage before Medicare Part D and smaller improvements among individuals with some prior coverage but whose out-of-pocket expenses were reduced following Medicare Part D implementation.

All five policy-level studies found statistically significant between-group differences in adherence to medications used to treat cardiovascular conditions, favoring the group that had out-of-pocket expenses reduced. However, we find these differences somewhat difficult to interpret because medication adherence decreased over time in all groups in two of the studies that used cohort designs. Nonetheless, the magnitude of effects observed in the cohort studies were similar to those reported in the RCT.153 Therefore, we concluded that evidence of moderate strength indicates that policy-level interventions that reduce patient out-of-pocket expenses can have a beneficial effect on adherence to medications used to treat cardiovascular conditions (Table 77).

Table 77. Summary of evidence for policy-level interventions (KQ 2).

Table 77

Summary of evidence for policy-level interventions (KQ 2).

Three policy-level studies found statistically significant between-group differences in adherence to medications used to treat diabetes, favoring the group that had out-of-pocket expenses reduced. As above, we find these differences somewhat difficult to interpret because all of these studies used cohort designs and medication adherence decreased over time in all groups in two of the studies. Nonetheless, the magnitude of effects observed in these two studies were similar to those in the Medicare Part D study among individuals who had had some prescription drug coverage before Medicare Part D but whose out-of-pocket medication expenses following its implementation dropped.151 Therefore, we concluded that evidence of moderate strength indicates that policy-level interventions that reduce patient out-of-pocket expenses can have a beneficial effect on adherence to medications used to treat diabetes (Table 77).

One study found no effect of a policy-level intervention on adherence to inhaled corticosteroids, usually used to treat reactive airway disease conditions. Therefore, we concluded that evidence is insufficient to draw conclusions for the effectiveness of policy-level interventions in this clinical area (Table 77).

One study examined the effect of policy-level interventions on clinical outcomes.153 This study found a 14 percent reduction in the rate of first vascular events following hospital discharge for a myocardial infarction. The same study found a 26 percent reduction in total patient spending, but no change in total insurer paying. We concluded that evidence is insufficient to draw conclusions regarding the effects of policy-level interventions on clinical and economic outcomes (Table 77).

Key Question 3a. Characteristics of Medication Adherence

Overall, the extreme heterogeneity of terminology used to describe medication adherence interventions in the studies reviewed hindered our ability to compare effects of different features of the interventions across studies and across diseases. In addition, the diversity of the interventions themselves made identification of “intervention type” clusters challenging.

Most, but not all, studies provided information (although not in a standardized manner) about six key intervention characteristics: the target(s), the agent(s), and the mode(s) of the intervention, as well as their intensity, duration, and components. The characteristics provided a framework by which we could describe the interventions. For example, for the intervention target, a little more than 50 percent of the interventions aimed at various combinations of multiple targets, whereas nearly 40 percent targeted only patients. Similarly, for the agent of intervention delivery, a pharmacist, physician, or nurse delivered about half of interventions. About half of interventions involved at least some face-to-face delivery of the program.

In addition to characterizing the interventions for these six key features, we identified some general patterns of combinations of the six features. For example, interventions varied in the number of contacts they entailed from 1 to 30, but those with more contacts tended to involve telephone contact. Similarly, certain intervention components, such as facilitation and knowledge-based components affecting the delivery of medical information, were commonly used across most interventions. In contrast, others, such as motivational interviewing and contingent rewards, were used less commonly. Similarly, we noted a greater frequency of combining awareness-raising activities with knowledge delivery among nurse-delivered programs than among either pharmacist- or physician-delivered interventions. The specific components of the interventions were the least well-characterized aspect of this literature, although often these components were the features that most meaningfully distinguished the interventions from one another. Some intervention types, such as decision aids, were not captured by existing taxonomies of adherence intervention components.

Key Question 3b. Direct Comparisons of Medication Adherence Intervention Components

The vast majority of studies compared a multicomponent intervention with a usual-care control arm. Very few studies directly compared one feature of an intervention to another feature to determine which aspects of the intervention had the most effect on outcomes. A longstanding debate exists about the advantages and disadvantages of testing multicomponent interventions, which may increase the likelihood of having an impact versus those of testing each component in isolation to understand its individual effects. Researchers may first combine approaches to document an effect and in later studies “peel away the layers of the onion” to isolate relative effects of separate components. The paucity of this second type of study design may reflect the state of the field. As studies increasingly demonstrate efficacious combination interventions, in the future we may see more studies that attempt to isolate effects of intervention features. Among the four studies that did conduct this kind of comparison, each compared different aspects of different interventions.

As a result, we could not pool data across even these four studies. One demonstrated that shared decisionmaking (in which nonphysician clinicians and patients negotiated a treatment regimen that accommodated patient goals and preferences) had a greater effect on adherence to asthma medications than did a clinical decision-making approach (in which the physician prescribed the treatment without specifically eliciting patient goals or preferences). Both approaches were more efficacious than usual care. The effects of shared decisionmaking on adherence lasted up to 2 years, whereas those attributed to clinical decisionmaking had attenuated at that point. Another study, conducted among patients with heart failure, directly compared two different delivery modes of the same information (telephone vs. videophone). This study found no difference between the two delivery modes regarding improvement in adherence, but both were superior to usual care. Another study directly compared the agent of delivery (physician vs. research staff) using the same mode (face-to-face contact) to deliver a decision aid among patients with diabetes to try to help them decide whether to take statins to lower their risk of cardiovascular disease. Patients who were given the decision aid had better adherence than those receiving usual care, regardless of who delivered the aid.

Thus, we conclude that mode of delivery was an important feature only in certain settings. However, incorporation of patient preferences through shared decisionmaking about treatment seems more efficacious at improving and sustaining improvement in asthma medication adherence than traditional clinical decisionmaking that does not take into account patient preferences in selecting a recommended treatment. Shared decisionmaking appeared to improve pulmonary function tests when compared with clinical decisionmaking but this approach did not improve quality of life or health care utilization; we rated this evidence as having low strength (Table 78).

Table 78. Direct comparisons of medication adherence intervention components: strength of evidence summary table.

Table 78

Direct comparisons of medication adherence intervention components: strength of evidence summary table.

Key Question 4. Outcomes for Vulnerable Populations

We searched for evidence on a broad set of vulnerable populations. For certain vulnerable subgroups—specifically for patients with major depression, severe depression, or depression and coexisting hypertension; Black patients with depression and coexisting diabetes; elderly patients with diabetes, hyperlipidemia, heart failure, or hypertension—we determined that interventions with a positive impact on medication adherence had only low strength of evidence. Evidence was insufficient about benefit to adherence of interventions dealing with patients who had depression with coexisting HIV, patients who had diabetes and depression (except for African-American patients with diabetes and depression), patients with diabetes and hypertension, and patients from rural communities. The low number of studies and limited sample size of included studies curtailed our confidence in the strength of evidence. For some vulnerable subgroups, including low-income patients and populations with low health literacy, we did not find any evidence.

Key Question 5. Adverse Effects

Our review of studies that examined adverse events or harms associated with interventions aimed at improving adherence did not find any indication that these interventions resulted in any unintended negative consequences for patients. However, we found only three relevant studies, and the level of heterogeneity among these studies in terms of the intervention and outcomes was so great that we determined that the evidence was insufficient to reach definitive conclusions.

Findings in Relationship to What Is Already Known

This comparative effectiveness review contributes to the sizeable literature about medication adherence in several ways. A Cochrane review in 200828 of studies through 2007, demonstrated that medication adherence interventions can have moderate effects on medication adherence and health outcomes for several common chronic and acute medical conditions. Our review includes studies from 1994 through the present (2011).”In addition, patients’ observations of medical regimens for infectious diseases can differ from practices by patients with chronic illnesses. Because several reviews had been conducted on interventions to improve HIV medication adherence,80,157 we excluded studies on patients with HIV and other infectious. We also exclude studies of acute conditions to improve the ability to potentially pool findings-adherence to short-term, acute conditions is different than that for chronic medications; the Cochrane review included these. Hence, we are unable to comment on adherence interventions for those particular ailments.

We, like the Cochrane review, excluded substance abuse interventions also to improve ability to pool findings potentially since the involvement of physical and psychological addiction would make adherence to these treatments different than that of other treatment. We also excluded studies of adherence to medications for severe psychosis because these conditions require specific approaches that would not likely apply in other diseases.

Finally, the Cochrane review included only adherence studies that also assessed health outcomes. To broaden understanding of the impact of interventions on adherence, we included adherence intervention trials even if they did not assess other health outcomes. This decision likely expanded the variety of medication interventions included in this comparative effectiveness review. On the other hand, it is possible that while statistical significance for improved medication adherence was not seen in some studies, this may still translate into improvement of clinical outcomes. Decisionmakers should consider this possibility when designing programs to improve adherence in their particular organizations.

We included studies that assessed the effects of policy-level interventions, although these changes are relevant chiefly to the United States. Our findings are fairly consistent with studies conducted of HIV adherence. Rueda and colleagues conducted a Cochrane Database review of 19 patient education and support interventions of 2,159 patients and found that methods were too heterogeneous to conduct a meta-analysis. They identified a broad range of intervention types, including cognitive behavioral therapy, motivational interviewing, medication management strategies, and interventions indirectly targeting adherence, such as programs directed to reduce risky sexual behaviors. Ten of the 19 studies indicated the invention was beneficial to adherence. Unlike our review, this HIV review showed some characteristics of interventions associated with improved adherence outcomes: targeting practical medication management skills, administering interventions to individuals rather than groups, and delivering over at least 12 weeks had a greater impact on adherence with improved adherence outcomes.157 In contrast, a meta-analysis by Simoni and colleagues showed that when data were pooled, participants in the intervention arms were more likely than controls to attain 95 percent adherence (OR = 1.50, 95% CI, 1.16 to 1.94), and this effect was stronger in studies that used recall periods of at least 2 weeks. They could not identify differences based on intervention features and concluded, as we have, that more research to identify the most efficacious intervention components is needed.80 Unlike other reviews, we analyzed intervention effects in relation to intervention type, to identify those programs with the strongest evidence. This information has the potential to offer actionable information for policymakers and practitioners working within clinical domains. The 20 intervention clusters we identified, which included categories like case management, coordinated care, shared decisionmaking, education with social support, and so forth, as listed in Table 74 provide a starting framework by which practitioners and researchers may develop, test, and report their adherence programs more explicitly and consistently.

In addition to identifying empirically derived clusters, this review has characterized interventions targeting medication adherence based on six intervention features: target, agent, mode, intensity, duration, and components. The information about variations in these six features has not been reported previously and provides a second approach to reporting adherence programs in a more standardized manner. Ultimately, if studies used this framework more consistently, future reviews might be able more easily to pool data and pursue syntheses that could provide more robust data and more precise estimates of effects. As with other active areas of research, ongoing trials have the potential to shift the weight of evidence: this systematic review will need to be updated frequently.

Finally, unlike other reviews of RCTs testing interventions for medication adherence, ours is the first attempt to understand the moderating effects of population characteristics on intervention effects. We did this by analyzing data from included studies that pertained to vulnerable populations (described in KQ 4 above). The paucity of evidence in this area highlights the need for future studies to include vulnerable populations.

Applicability

The interventions analyzed in this review were not highly selective; rather, they ranged from relatively minimalist to complex and intense, although evidence often came from small studies. Neither were these studies limited to narrow or unrepresentative disorders or disease severity; rather, they reflected studies done across a substantial variety of chronic conditions affecting adults. Thus, in one sense the evidence from this comparative effectiveness review might be regarded as relatively applicable across numerous different options for health care providers to pursue for their adult patients with major chronic diseases or multiple chronic conditions. Our findings are not generalizable to children or young adolescents because of our inclusion criteria.

As noted, many of our findings came from single, often small or short-term, trials, some with important questions about risk of bias. Findings from this diversity clinical conditions and interventions have not yet been replicated in trials in larger patient populations, in groups drawn from different settings and with different sociodemographic characteristics, or in investigations with longer observation and followup periods. These gaps in the evidence base constrain somewhat the applicability of our results.

Another limitation to the applicability of this evidence comes from the complexity of multicomponent interventions. Studies did not generally provide information on how researchers identified the separate active components in their interventions or how they had operationalized those components; generally, these complex programs lacked detailed instructions and users’ manuals by which other groups might try to replicate the original research.

Finally, the degree to which these interventions require fidelity to protocol when being implemented in other settings or through different study designs (e.g., nonexperimental studies) is unclear. The need for fidelity to protocol, or the allowable, appropriate adjustments for other patient populations (e.g., different illnesses; different sociodemographic characteristics) is likely a matter of some debate. These questions place some limits on the wide applicability of the evidence reported here.

Implications for Clinical and Policy Decisionmaking

We found evidence of effective interventions to improve medication adherence for many chronic conditions. These analyses suggest that patients’ adherence to chronic-disease medications can be improved through programs targeting patients, providers, health systems, or policy. They demonstrated that a broad range of approaches can work.

Adherence is typically the result of a combination of patient, provider, and policy factors. Indeed, most of the interventions we identified were multifactorial; over half were aimed at multiple targets and most had multiple components, including several with multiple delivery modes. In other words, no single “silver bullet” exists for medication adherence.

We found the strongest evidence for enhancing adherence with reduced copays across clinical conditions, self-management of asthma (for short-term outcomes), and collaborative care or case management for depression. Within clinical conditions, we found the strongest evidence with depression case management for depression symptom improvement and pharmacist-led hypertension approaches for systolic blood pressure improvement. We found consistent evidence or evidence from more than one clinical area supporting medication adherence interventions such as education, reminders, and pharmacist-led multicomponent interventions.

Clinicians and policymakers should keep in mind that we found very little evidence of any relationship between medication adherence and adverse events, although what we found suggests that improving adherence did not increase the incidence of adverse events. However, many of the conditions studied did not involve medications typically associated with very severe common side effects. This review is the first we are aware of that systematically reviewed information on adverse events. It thus provides information that should be confirmed in future studies and reviews.

The lack of studies evaluating potential mechanisms that link improved adherence with other health-related or health services outcomes somewhat constrains policymakers’ and clinicians’ options. We did not find evidence of studies among patients with chronic illnesses who tend to have more intermittent disease trajectories, such as certain types of arthritis, diverticulitis, and other gastrointestinal conditions. In particular, decisionmakers should exercise caution in trying to use any a la carte approach to implementing components of complex interventions to enhance patients’ medication adherence. We do not think that sufficient information is yet available to guide choices among the considerable array of program components, especially to pick and choose only some parts of multicomponent approaches. Therefore, future studies must do a better job not only of clearly describing each component of their intervention but also of designing studies and conducting analyses that can identify which components are driving the effects of the intervention. Meanwhile, however, if studies have not been done in their specific clinical patient population, clinicians and health system administrators may want to give more thought to how they might be able to extrapolate existing results to their specific patient populations—that is take apparently successful programs and apply them to groups with diagnoses and other characteristics similar to those in the successful program. For example, interventions similar to those that were successful at improving adherence to medication for hypertension and hyperlipidemia may help in other settings in which the illness is asymptomatic and medication is taken primarily to prevent long-term complications.

Poor medication adherence is known to result in large downstream health care costs. An important finding for policymakers contemplating changes in health policy is our assessment of moderate-strength evidence, from five consistent studies, that reducing patients’ out-of-pocket costs or improving prescription drug coverage can improve their medication-taking behavior. Policies that enhance patient adherence by easing patient copayments or other patient-paid medication expenses may prove highly cost-effective. Cost-effectiveness studies that assess the long-term effects of such policies could be beneficial to policymakers.

Limitations of the Comparative Effectiveness Review Process

The constraints for population and setting we imposed on the systematic review limit the applicability of this review, as discussed above. We did not review the evidence on populations with human immunodeficiency virus/acquired immune deficiency syndrome (HIV/AIDS), mania, bipolar disorder, or substance abuse. We excluded studies among patients with HIV/AIDS because existing comprehensive reviews of these interventions had been conducted recently. We also excluded studies of acute conditions, severe mental illness and substance abuse to improve our ability to potentially pool findings since adherence to short-term, acute conditions, those involving addictions or cognitive limitations are different than that for chronic medications. However, interventions for these excluded clinical conditions may have applicability to the conditions that we included in our review. We limited this review to adults and cannot, therefore, address important adherence concerns for children and adolescents with chronic conditions such as type 2 diabetes. Another limitation is geographic location: we excluded non-English and non-U.S. studies. This criterion may well have decreased the pool of eligible studies we might have examined, but their applicability to the United States is unclear. Our approach to categorizing interventions for KQ 1 relied essentially on the short descriptions in published manuscripts; their similarities or differences substituted for any overarching taxonomy, as none that we considered seemed fit our purpose. Thus, we have introduced intervention labels that, admittedly, do not fully describe or account for heterogeneity within and across clinical conditions or patient populations. This approach limits our ability to make definitive statements about the effectiveness of interventions across clinical areas; we believe the clusters and categorizations we used are useful heuristics, but they may be regarded more as hypothesis generating than reflecting settled principles of classification. Finally, our pool of included interventions is limited to those that were designed specifically to address medication adherence as a primary or secondary outcome. We did not include clinical trials of drugs that considered adherence as a component of safety and efficacy; as a result, we do not address the effectiveness of specific drug formulations that may improve adherence by limiting adverse effects.

Limitations of the Evidence Base

Methodological Limitations

Our review identified several gaps in the literature that may be filled by future research efforts. In many disease areas for KQ 1, interventions and adherence measures were heterogeneous, which limited our ability to pool results from studies. If investigators could use more standardized, objective adherence outcomes in future research, their results might be more easily analyzed and interpreted in the context of other adherence studies.

In addition, a lack of focus on mediating relationships through which the interventions acted on medication adherence limited the conclusions that we could safely draw about the efficacy of specific intervention features. Although some studies showed that interventions improved adherence, only a few had large effects on adherence. Hence, future studies could be designed to identify how to enhance the effects of efficacious interventions, such as by using a factorial design that combines efficacious interventions and can assess both additive and multiplicative effects.

Most trials were not placed in a larger context of improving the quality of health care delivered; only a minority examined issues such as quality of life and patient-reported outcomes or patient satisfaction. This limitation interacts with the issues noted above about understanding the effectiveness of these programs, not simply their efficacy, which is especially important for providing information suitable for broadly based clinical and policy decisionmaking. At a minimum, using guidelines from the Standards for Quality Improvement Reporting Excellence (SQUIRE) group (http://squire-statement.org/guidelines) will improve the quality of reporting so that future studies of complex interventions routinely clarify the mechanisms by which intervention components are expected to cause change, the course of the implementation, and the success of tests of the mechanism of action.158

Finally, although many studies did assess some health outcomes, these often were not reported by patients themselves, and many were relatively short term (at least in the context of lifelong chronic ailments). Including long-term health outcomes and mounting efforts to solicit information directly from patients in future trials or observational studies of adherence would enhance the nation’s capacity to assess the overall significance of adherence interventions. While the minimum length of followup indicated may vary by condition, for lifelong chronic ailments, medication adherence often decays over at least the first year. Hence, studies that follow patients longer than one year could provide information about adherence levels once they have reached a plateau. Collecting information about costs will be crucial, because no health systems or facilities can afford to try all approaches across the diverse patient populations they serve. Economic information is essential in and of itself, but it will facilitate cost-effectiveness analyses of such interventions.

Research Gaps

We found numerous gaps in the literature, described in the sections below. The following key research gaps have emerged across key questions and clinical conditions:

  • Some clinical areas revealed a paucity of evidence. Among the conditions that we reviewed, we found limited evidence for myocardial infarction, multiple sclerosis, glaucoma, and multiple chronic conditions.
  • The evidence focuses on clinical conditions with relatively stable or increasing levels of morbidity; effective adherence interventions for these conditions may not be effective for conditions with episodic symptomatology.
  • Information on subgroup analysis was limited; despite our relatively wide search for evidence on vulnerable populations, we found very little evidence.
  • Information on adverse events, health outcomes, quality of life, costs, and healthcare utilization was limited.
  • Information on long-term outcomes was limited.
  • Information was limited or not available on the effectiveness of components or mechanisms of action of complex or practice-driven interventions.
  • The wide heterogeneity of measures and outcomes made synthesis challenging. Future efforts to pool evidence would benefit from the use of standard and valid measures.

Key Question 1. Patient, Provider, and Systems Interventions

Diabetes

The body of evidence for diabetes was relatively sparse and provided low strength of evidence. The evidence did not clarify which aspects of the various models were important. Future studies would benefit from factorial designs that identify which aspects of interventions are most important, which are working together, and which have an independent influence. Additional research to assess such models in a wide range of settings, on a larger scale, and over a longer term would be particularly valuable. Studies that seek to advance understanding whether the impact of interventions for diabetes medications varies for different subgroups (such as groups with low health literacy, very poorly controlled diabetes, or other vulnerable populations) may be beneficial. This analysis can be accomplished by assessing the moderating effects of such characteristics as literacy level on the effects of the intervention on adherence. Most but not all studies included HbA1C assessments. It is important that future studies include such important biomarkers as outcome measures. One trial that found an effect of a decision aid on medication adherence assessed the effects of the intervention on patient satisfaction. No trials assess costs or health care utilization. Inclusion of assessments of intervention effects on patient satisfaction and other outcomes, costs, quality of care, utilization, or quality of life in future studies will be important.

Cardiovascular Disease and Hyperlipidemia

We found that interventions and measures of adherence were heterogeneous among included trials evaluating interventions to improve adherence in patients with cardiovascular disease and hyperlipidemia. This heterogeneity limited our ability to pool results within respective disease categories. Among studies in cardiovascular disease and hyperlipidemia, reporting of additional outcomes beyond medication adherence varied by disease. For example, all three heart failure trials that found improved medication adherence also reported additional outcomes, including health care utilization in two of them. Among the 17 trials conducted in patients with hypertension, seven found improved adherence or persistence and six of the seven reported systolic and diastolic blood pressure outcomes, but only two reported health care utilization outcomes. Among the nine trials in hyperlipidemia, four found improvements in either medication adherence or persistence; only two of the four reported additional outcomes, including low-density lipoprotein cholesterol (LDL-C) levels and patient satisfaction. Thus, while a majority of trials in the heart failure section evaluated health care utilization outcomes, among the trials with improved adherence, few in the hypertension group and none in the hyperlipidemia group reported such outcomes. Future research could help to fill this gap.

The identification of only one trial of medication adherence in patients with myocardial infarction suggests significant research gaps in this area. Studies need to evaluate clinical outcomes in addition to adherence outcomes for patients after myocardial infarction. We only included trials in the myocardial infarction section that aimed to improve adherence to medications to treat myocardial infarction. We discussed trials that aimed to improve adherence to medications to treat diseases that are risk factors for myocardial infarction (hypertension, diabetes, hyperlipidemia) or that may have been related to a myocardial infarction (heart failure) elsewhere as independent clinical categories.

We noted that quality of life and patient satisfaction were evaluated in few trials and that cost was evaluated in only one trial, conducted in patients with heart failure. Quality of care was not evaluated in any of the included cardiovascular disease or hyperlipidemia trials. Future research could enhance our understanding of how medication adherence interventions could affect these outcomes as well.

Asthma

Among included asthma trials, we found that no long-term outcomes were reported for short-term interventions; this finding was true for many of the trials included in this review for other clinical conditions as well. For asthma, interventions lasting 4 to 6 weeks generally only reported outcomes within the intervention period or a month thereafter. Six of eight interventions for asthma-related medication adherence reported improvement in medication adherence; unlike other clinical conditions, all of these studies reported health outcomes. Our review of the evidence for asthma did not find any information on patient satisfaction, costs, or quality of care. We found a single trial on a potentially promising approach, shared decisionmaking. Further research on this intervention will help to clarify its applicability to other settings.

Depression

Seven out of 11 depression interventions reported improvements in medication adherence, with seven of these trials reporting on health outcomes. However, these trials provided limited information on patient satisfaction, costs, and quality of care. We found one trial that met our criteria on the use of reminder letters to nonadherent patients and lists of nonadherent patients to their health care providers. An added limitation of the evidence base was the lack of information on the clinical utility of medication adherence improvements. For example, one trial found a 1 to 3 percent statistically significant difference between the intervention and control arms of the study. A better understanding of the clinical implications of this difference in medication adherence requires that future research evaluate the effects of the intervention on clinical outcomes in addition to medication adherence outcomes.

Other Chronic Conditions

For interventions in the areas of unspecified or multiple chronic conditions, glaucoma, multiple sclerosis, and musculoskeletal diseases, we found only a few trials overall that met our inclusion criteria. In many cases we only identified one trial per disease area that met our inclusion criteria, indicating significant research gaps in these disease areas. For example, among included studies dealing with unspecified or multiple chronic conditions, we found four trials that varied in the intervention used and outcomes reported. One of the trials showed no effect of the intervention on adherence and mentioned that a post-hoc study showed the intervention may actually be inferior to usual care in improving medication adherence. In the other three trials, the variation among studies was too significant to meaningfully assess the evidence. More studies focused on multiple chronic conditions are required to fill this gap. For glaucoma and multiple sclerosis, where we found only one trial each, more studies with larger sample size and lower risk of bias are required to reach meaningful conclusions regarding interventions to improve adherence to medication. We found three trials dealing with musculoskeletal diseases, but again, were unable to reach conclusions due to a lack of precision in the results and significant differences in the nature of the interventions and the outcomes measured.

Key Question 2. Policy-Level Interventions

The five studies investigating policy-level interventions yielded important evidence that reducing patient out-of-pocket expenses for prescription medications can improve medication adherence. However, only one of these studies examined the effect of these policy changes on any patient-centered or health-related outcomes. Thus, future studies on policy interventions should focus more on how such interventions can improve actual management of these chronic conditions. Of particular interest are measures of blood pressure, lipid levels, and other intermediate outcomes and biomarkers; long-term health outcomes, such as rates of myocardial infarctions or strokes and measures of patient-reported quality of life and health status; and use of health care services.

In addition, none of the studies examined whether the impact of these interventions varied across different population subgroups. For example, policy-level interventions designed to reduce out-of-pocket costs most likely have the greatest effect among individuals with limited incomes and those using several medications. This type of question remains to be answered by future research. Finally, because the studies investigating the effect of copayment reductions found that adherence decreased in all study groups over time, research using new-user designs is needed to clarify how policy-level interventions may change the trajectory of adherence over time, beginning at the initiation of therapy.

Key Question 3. Intervention Characteristics

We identified six main properties of medication adherence interventions, which we called their target, agent, mode, intensity, duration, and components. Our capacity to describe fully the variation in these features was limited in two ways: by the sheer diversity of the programs and the measures used to assess outcomes, and by language that the various investigator teams used to describe their interventions’ features.

We suggest that future studies in this field adopt a standardized manner for describing interventions. It should include a clear report of the intended targets of the intervention, all agents, and modes of delivery using the categories we have identified here. We believe that investigators would find describing the intensity and duration of all interventions in a similarly standardized manner relatively simple; such descriptions should include the total number and type of contacts, the total amount of time for each contact, the frequency of the contacts, and the duration of calendar time over which the contacts are delivered. For interventions that do not involve contacts per se, such as policy changes, these variables would be categorized as “not applicable.” Much as specifications of CONSORT statement almost 15 years ago159 enabled systematic reviewers to do a much better job than previously of comparing and pooling clinical trial results, such a simple step as standardizing reporting descriptions of interventions might similarly enhance capacity to understand the effects of different aspects of these intervention. Similarly, researchers in this field might consider using deBruin’s taxonomy,74 which consists of specific definitions of each of several components to report their intervention components. Others could then have a better basis for cataloguing these features as a first step in comparing their utility across studies.

Finally, we found only four studies that directly compared specific components or approaches of interventions. More standardized descriptions of interventions, as advocated above, will enhance the capacity of systematic reviewers to pool data across studies and efficiently compare effects of specific features. Nevertheless, as we gain insight into what features are most critical, more studies will be needed that directly compare elements of interventions. Given that some coordinated care and other multicomponent interventions appear to be effective, study designs, such as factorial or step-wedged approaches that may help to delineate both the additive and synergistic aspects of multicomponent interventions will be particularly beneficial. Observational studies (not included in this review) may generate hypotheses regarding the mechanisms by which complex or practice-driven interventions work.

While not the goal of this review, there appeared to be a paucity of post-trial qualitative studies to understand from the patients’ perspective the aspects of the interventions that they found most useful. Use of such mixed methods may inform the refinement of efficacious interventions to make them most effective in real-world settings.

Key Question 4. Vulnerable Populations

We encourage health systems, insurers, and others to mount studies for the considerable range of population groups that we had intended to examine but on whom we found little to no literature. These include most racial and ethnic minorities, although African-American populations were reasonably well covered in this evidence base. People with a variety of characteristics putting them at risk of disparities in health care and health outcomes warrant more attention, especially those for whom English is a second language, those with low levels of literacy or health literacy, and those of low income or poor or no health insurance. As to the latter, more studies of children covered by state Medicaid programs or the Child Health Insurance Program might be warranted.

We believe that the evidence base for mainstream patient populations with common chronic conditions points toward a variety of medication adherence programs suitable for these groups. Other clinical populations facing substantial health challenges remain understudied. These include persons with dual mental health diagnoses (e.g., depression and a substance abuse problem) and persons with complex medical histories (e.g., multiple chronic conditions).

Key Question 5. Adverse Events

Interventions designed to improve medication adherence did not, in our very small evidence base, appear to increase adverse events, harms, or unintended consequences. However, routine tracking of adverse events related to attempts to improve adherence has apparently not received much (certainly not sufficient) attention in the literature. The fact that all pharmacotherapies for chronic conditions pose some risks to at least some patients—and in some cases (such as depression) the choice of drug may turn on the adverse events profile, not efficacy or effectiveness data—makes clear the need to improve and expand evaluation of harms, particularly over the long run. We advocate that investigators build into their trials or effectiveness studies more routine measurement of possible harms or unintended effects, in addition to benefits of greater medication adherence per se.

Conclusions

Despite the heterogeneity of adherence measurement, interventions tested, and characterization of interventions, we found the most consistent evidence of improvement in medication adherence for policy-level interventions to reduce out-of-pocket expenses or improve prescription drug coverage, case management, and educational interventions across clinical conditions. Within clinical conditions, we found the strongest support for self-management of medications for short-term improvement in adherence for asthma patients; collaborative care or case management programs for short-term improvement in adherence and symptom improvement for patients taking depression medications; and pharmacist-led approaches in hypertensive patients for improvement of systolic blood pressure.

We found low strength of evidence for many other interventions; these diverse groups of approaches offer promise but require more research to establish their value (or lack of it). Far less evidence was available to show whether most of these interventions improved patients’ health outcomes, given better adherence to their medication regimens. Several reviews that researchers have conducted over the past two decades—now complemented by our comparative effectiveness review—confirm that medication adherence can be improved via formal programs of various sorts. At this stage, new studies need to be asking “What specific intervention element or elements work best for improving medication adherence?” and “How can we further enhance medication adherence interventions to improve health outcomes?”

PubReader format: click here to try

Views

  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (4.2M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...