Adherence to Maintenance-Phase Antidepressant Medication as a Function of Patient Beliefs About Medication

INTRODUCTION

Major depressive disorder is associated with enormous personal suffering for affected patients, great distress for their family and friends, and major economic and societal costs.^1,2 Antidepressant medication reduces depressive symptoms,^3–,⁵ and its premature discontinuation increases the risk for relapse.⁶ Treatment guidelines recommend continuing antidepressant medication for at least 8 months after symptom remission and regimen stabilization, and this period is referred to as the maintenance phase of treatment.^3,7 Nevertheless, between 30% and 83% of patients who begin antidepressants discontinue treatment prematurely.^6,8–,¹⁰ Among primary care patients who filled a new prescription for a tricyclic antidepressant, 21% discontinued medication within 2 weeks of initially filling the prescription, an additional 3% to 10% discontinued every 2 weeks, until only about one half still took the medication at 4 months.⁹

Although most interventions to improve depression outcomes emphasize broad systems strategies, such as case identification and intensified follow-up, outcomes ultimately depend upon patients’ willingness to take antidepressants. In turn, willingness to take antidepressants is tightly linked to beliefs about medication. For example, patients report that treatment effectiveness and barriers are among the most critical aspects of depression care.¹¹ Data also indicate that pretreatment perceptions of the benefits of and barriers to antidepressants predict initial medication adherence, and that primary care patients frequently attribute their early discontinuation to their perception that they do not need an antidepressant.⁹

The brief timelines of these existing studies, however, make it difficult to extrapolate their findings to the maintenance phase of treatment. This research gap is important, because adherence during treatment initiation in the early acute phase is probably governed by mechanisms different from those that govern maintenance-phase adherence. Patients on maintenance antidepressants are by definition less depressed than those in the acute phase; therefore, their treatment perceptions are probably less negatively biased. Moreover, the existing studies (reviewed above) collectively suggest that early discontinuation is related to both side effects and perceiving medication as ineffective, factors that seem unlikely to influence established antidepressant users whose initial side effects have subsided or who have experienced a treatment response. Instead, we suspect that long-term adherence may gradually diminish as clinically improved patients begin to conclude that they no longer need medication or become less willing to continue tolerating previously acceptable medication problems, such as sexual side effects. Maintenance-phase patients may also be relatively more affected by fears of potential long-term cumulative or insidious adverse effects, such as personality change, addiction, or toxicity.

Existing studies of beliefs about antidepressants also have several methodologic limitations. All previous studies used single ad hoc items rather than measures of validated beliefs, none used multiple adherence indices, and none adjusted for self-report biases, such as social desirability. Some studies aggregated patient-initiated discontinuation together with physician-initiated discontinuation even though these events are different. Many studies disregard meaningful variance by reducing adherence into a binary construct (present vs absent), although adherence is clearly a continuous gradient. All previous studies include tricyclic antidepressants, which have been displaced as first-line approaches by more contemporary agents that tend to have fewer side effects. In the present study, therefore, we endeavored to overcome common methodologic limitations by using multiple validated measures of adherence and beliefs, treating adherence as a continuum and adjusting for social desirability bias.

Research on beliefs about antidepressants has been exploratory and generally atheoretical. The present study was guided by Horne’s theoretical model of medication representations,¹² which distinguishes between beliefs surrounding medication in general and beliefs surrounding the specific prescribed medication. This model breaks down medication-specific beliefs into 2 constructs—perceived need for the medication (necessity) and the perceived potential for the medication to cause problems (concerns)—and posits that adherence is determined by the balance between these 2 constructs. We thus hypothesized that adherence is highest for patients whose perceived need exceeds their concerns, and lowest under the reverse conditions. We did not expect beliefs about general medications to predict maintenance-phase adherence, because they seem more relevant to early adherence during treatment initiation rather than eventual adherence after patients are well established on the regimen.

METHODS

RESULTS

Hypothesis Testing

MANOVA was used to analyze general and recent adherence simultaneously as dependent variables (Table 2⇓). The additional variables considered for inclusion in the model were demographics (age, sex, and educational level), depression severity and duration, and social desirability bias. To preserve degrees of freedom, the 3 demographic covariates were dropped from the model because they had no significant multivariate effects (all P >.23); we retained depression characteristics and social desirability because of their theoretical relevance to adherence. As can be seen in Table 2⇓, only the medication-specific necessity-minus-concerns composite score had a significant multivariate association with adherence (P <.001), which univariate ANOVA on follow-up indicated was due to its association with both general adherence (ie, Morisky score, P <.001) and recent adherence (P = .001). There were no other significant multivariate associations, nor did the pattern of statistical significance change when social desirability and the depression variables were dropped from the model.

View this table:

• View inline
• View popup

Table 2.

Results of MANOVA Analysis of Adherence

To translate the necessity-concerns differential into a categorical perspective that might be more useful to clinicians, we split at the median both belief dimensions to create 4 antidepressant attitude groups: (1) skeptical (low necessity, high concerns; n = 24), (2) ambivalent (high necessity, high concerns; n = 18), (3) indifferent (low necessity, low concerns; n = 19), and (4) accepting (high necessity, high concerns; n = 20). Figure 1⇓ illustrates how these groups were derived. A series of Student t tests indicated that general adherence varied significantly by group (Figure 2⇓). Specifically, the antidepressant-accepting group had significantly lower Morisky scores than each of the other 3 groups (all P <.05). Second, patients who were skeptical about antidepressant medications were significantly less adherent to treatment than both ambivalent and accepting patients (both P <.05), although their difference from indifferent patients fell just short of statistical significance (P = .057). The pattern of findings was similar for recent percentage of adherence scores (Figure 1⇓), with the exception that ambivalent patients also did not differ significantly from accepting patients.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Antidepressant belief groups (recent percentage of adherence in parentheses).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Morisky adherence scores by beliefs group. Note: Higher scores indicate lesser adherence.

ns = nonsignificant.

* P = .057

† P = <.05

‡ P = <.01

§ P = <.001

DISCUSSION

To summarize, we found that adherence to maintenance antidepressant therapy averaged 85%, varied broadly, and varied as predicted according to patients’ beliefs about the medications. Of the many factors that we analyzed or adjusted; including demographics, social desirability, and depression characteristics; beliefs about the necessity for and concerns about antidepressant medications were the only variables that accounted for adherence. Specifically, both recent and general adherence were highest among patients whose perceived need for medication exceeded their concerns about taking medication, and lowest for those whose concerns about taking medication exceeded the perceived need. These 2 dimensions can be reduced into an intuitive composite score that reflects the balance between perceived necessity and concerns.

Interestingly, necessity and concerns were empirically orthogonal in our sample of patients on long-term antidepressant therapy. That is, antidepressant concerns were no more likely among patients who believed that they needed their medication than among those who saw it as unnecessary. Similarly, viewing one’s antidepressant as unnecessary was not associated with also viewing it as unsafe. Even more interesting, both beliefs were independent predictors of adherence. Taken together, this pattern of findings challenges the seemingly common notion that patients’ concerns about medication reflect their underlying disagreement with the treatment plan, and that a perceived need for medication mitigates medication-related concerns. Subsequent research ought to explore the empirical and pragmatic utility of comparing beliefs within patients, which has the advantage of operating at the patient level where decision making plays out. Another advantage of this approach is that it helps control for individual biases in how study participants use rating scales, ie, some patients tend to use extreme ratings whereas others tend to use more moderate or neutral ratings.

Although our cross-sectional data cannot support causal conclusions, several major theoretical models (eg, Social-Cognitive Theory, Health Beliefs Model, Commonsense/Self-Regulation Model) emphasize that beliefs determine behavior. Of these traditions, the Health Beliefs Model makes the most specific statements about treatment beliefs through its effectiveness and barriers constructs. The extension of Leventhal’s Commonsense Model²¹ by Horne et al,¹² however, is the most specific model in regard to medication, postulating that adherence is governed by patients’ mental representations of whether they require medication and whether it will cause problems. Our findings specifically support the medication-specific dimensions of the medication representations model and complement a growing literature on medication beliefs across several other conditions, including hypertension,²² human immunodeficiency virus infection and acquired immunodeficiency syndrome,²³ asthma, renal failure, heart disease, and cancer.¹⁶ In contrast, we did not find that general beliefs about medications did not predict maintenance adherence, nor did we expect to find this association. We are currently analyzing medication beliefs in 2 additional study groups in which we expect general beliefs to play a more central role: patients initiating antidepressants, and persons taking over-the-counter St John’s wort for depression.

According to most cognitive models, beliefs usually remain stable in the face of repeated experiences that specifically disconfirm them. Thus, data such as ours could lead to the development of new belief-based adherence promotion strategies for clinicians as well as for those conducting clinical trials. For example, patients who are indifferent about antidepressant medications (Figure 1⇑, lower left quadrant) might be most influenced by pragmatic factors, such as ease of administration and cost; or by social factors, such as patient-physician relationship, life stress, stigma, and social support. Perhaps these patients’ adherence behavior would be enhanced by an aggressive “built for speed” regimen that is specifically tailored to achieve rapid and full effects. These patients might also benefit from education about symptom course, medication response lag, the rationale of maintenance treatment, and the link between early discontinuation and subsequent decline.

In contrast, patients who are ambivalent about antidepressant medications might be more motivated to minimize medication problems than to achieve rapid and complete symptom relief. Thus, their adherence might be enhanced by a more conservative “built for comfort” regimen that minimizes side effects. They might also benefit if clinicians proactively identify and correct any patient misunderstandings about medications, emphasize the transient or reversible nature of most side effects, and respond rapidly to any patient-expressed medication concerns.

Finally, patients who are skeptical of antidepressant medications will probably not be motivated to take antidepressants until either their perceived need increases, concerns diminish, or symptoms worsen. Behavioral strategies might initially affect their beliefs but not their adherence behavior, and sustained effort will be needed to achieve adequate medication trials. It may be beneficial to explore whether they view themselves as affected by nonbiological depression or consider their condition as otherwise misdiagnosed, whether they have a history of insufficient or overaggressive treatment or they have cultural beliefs about medication. Although sometimes their multiple treatment barriers can be resolved, to do so within the constraints of the primary care setting is particularly challenging.

Given certain study limitations, our findings should be taken as preliminary and not necessarily generalizable to all patient populations. Our study does not apply to patients who discontinue drug therapy during treatment initiation (a group that is already well-researched). Even though we did recruit some patients who did not adhere to their treatment regimen, self-selection probably biased respondents toward a more adherent study sample. We did not control the study for the length of depression or its treatment, although each proved unrelated to adherence. A more ethnically diverse sample would have increased the sensitivity of our study to cultural effects upon beliefs, and the findings might not apply to patients who seek specialized mental health treatment.

Our cross-sectional data cannot prove that beliefs play a causal role. Although we were able to rule out confounding by depression variables, medication side effects, social desirability bias, medical comorbidity, and demographic features, adherence or beliefs may actually be governed by some unmeasured factor, such as a patient-physician relationship, stigma, or memory problems. Finally, we could not confirm that patients met criteria for a depressive syndrome when their treatment episode began. Even so, patients with physician-recognized depression tend to have greater symptom severity and recurrence than do patients whose criteria-defined depression is not recognized by their physician.²⁴

In closing, our findings raise a variety of testable implications for predicting treatment adherence and developing strategies tailored to enhance it. Future research ought to examine whether attitudes about antidepressant medications prospectively predict treatment adherence and outcome. If so, subsequent studies ought to determine whether depression outcomes improve when regimens are tailored to patients’ beliefs, and explore the generality of our proposed attitudinal prototypes apart from antidepressant medication regimens.

Disclaimer

The views expressed in this article are those of the authors and do not necessarily represent the views of the University of Michigan or the Blue Cross Blue Shield Foundation of Michigan.