Medication Adherence in Pediatric Patients with Bipolar Disorder: A Systematic Review

Introduction

Bipolar disorder is a severe chronic mental illness characterized by discrete periods of elevated mood (mania or hypomania), typically along with episodes of major depression. Pediatric bipolar disorder is estimated to affect 1.8%–3.9% of children and adolescents worldwide, with bipolar I making up 0.6%–1.2% (Van Meter et al. 2011, 2019). Prior studies have found that 30%–60% of adult patients with bipolar disorder have mood symptoms before age 19 (Perlis et al. 2004; Post et al. 2008). Evidence also suggests that this diagnosis often persists into adulthood, especially with a bipolar I diagnosis (Geller et al. 2008). Earlier onset of bipolar disorder has been associated with a more severe disease course along with higher rates of comorbid psychiatric disorders, comorbid substance abuse, more recurrences, and a higher likelihood of suicide attempts and violence (Perlis et al. 2004). Pediatric bipolar disorder is also highly comorbid with other psychiatric and medical conditions, which affects disease severity and functional outcomes for these patients. Studies have found high rates of comorbid anxiety disorders, attention-deficit/hyperactivity disorder (ADHD), oppositional defiant disorder (ODD)/conduct disorder (CD), and substance use disorders (Sala et al. 2014; Frías et al. 2015).

Given the severity and chronicity of illness, patients with bipolar disorder suffer from significant morbidity and mortality. These patients have high rates of self-harm, suicide attempts, completed suicide, and inpatient psychiatric admission (Baldessarini and Tondo 2003; Hayes et al. 2015). This increased risk of suicide attempts and completed suicides also seems to hold for pediatric patients (Goldstein et al. 2005; Goldstein 2009). Pediatric bipolar disorder has been associated with varying levels of neurocognitive impairment, lower intelligence quotient, and functional impairment (Horn et al. 2011; Best et al. 2017). Pediatric patients often report associated reduction in quality of life compared with youth with other psychiatric illness, medical illnesses, and healthy controls (Freeman et al. 2009). Bipolar disorder overall also results in significant economic burden. The estimated direct and indirect costs of bipolar disorder in the United States were $151 billion in 2009 (Dilsaver 2011).

Timely and appropriate treatment is critical for treatment of symptoms and improvement of functioning in bipolar disorder. Treatment is similar for pediatric patients and adult patients, with atypical or second-generation antipsychotic medications and/or mood stabilizers as the primary recommended treatment, although psychotherapy and adjuvant medications are also often needed for appropriate and comprehensive treatment (Kowatch et al. 2005; McClellan et al. 2007). Patients often need to take medications consistently for extended periods of time to prevent recurrence of mood symptoms. Current recommendations are that a patient achieves remission of symptoms for a minimum of 12–24 months before considering medication tapering or discontinuation (Kowatch et al. 2005). There is extensive adult literature concerning medication adherence in bipolar disorder, which consistently shows that poor medication adherence is associated with worse patient outcomes (Berk et al. 2010). Unfortunately, reported mean rates of treatment adherence in adult bipolar disorder are as low as 41% (García et al. 2016). Adult studies have identified a variety of patient-level factors, including younger age, poor social support, and multimorbidity, and external factors such as poor access to care and high medication cost that are associated with poor adherence (Berk et al. 2010; Levin et al. 2016).

Although the total body of research concerning pediatric bipolar disorder has increased significantly over the past two decades, it is unclear how much of this research specifically addresses medication adherence for pediatric patients with bipolar disorder (Goldstein et al. 2017). More specifically, studies that not only measure adherence but also study patient characteristics and other factors that influence adherence rates could provide valuable insight into strategies for improving medication adherence. Although a recent previous study did review medication adherence among children and adolescents with severe mental illness, two of the studies addressing bipolar disorder in that review included patients >18 years old and the bipolar disorder articles were not separated out from other psychiatric conditions for the purpose of analysis (Edgcomb and Zima 2018). No prior reviews, to our knowledge, have focused on medication adherence and factors influencing medication adherence specifically in patients ≤18 years old with bipolar disorder. This age range is likely to have different factors, especially social factors, which could influence medication adherence rates in this group. The objective of this review was to examine the current state of the literature regarding medication adherence in pediatric patients with bipolar disorder.

Methods

We conducted a comprehensive search of the Ovid Medline, PubMed Medline, Elsevier Embase, Wiley Cochrane Register of Controlled Trials, Wiley Cochrane Database of Systematic Reviews, EBSCO CINAHL, and Ovid PsycINFO from inception to June 2019. The search strategy included both free-text and controlled vocabulary for the following: bipolar disorder, bipolar, mania, manic, medication adherence, patient compliance, adherence, compliance, antipsychotic, neuroleptic, drug therapy, drug, drug prescriptions, prescription, mood stabilizer, medication, psychopharm, antipsychotic agents, adolescent, pediatric, boy, girl, teenage, adolescent, and child (Supplementary Appendix S1).

The initial search generated 822 titles, which was narrowed to 439 after duplicates between databases were removed. All 439 titles and abstracts were screened by one of the authors (M.M., S.L., M.S.) for further full-text review, with each approximately reviewing one-third of the titles and abstracts. One hundred thirty-three of these were selected for additional eligibility review. Each of these one hundred thirty-three was reviewed by two of the three above authors for inclusion. Sixteen texts were then selected for all three authors to review. Of these, six texts were selected for final inclusion in this article. The Preferred Reporting Items for Systematic Review and Meta-Analyses flow diagram in Figure 1 gives additional information about the evaluation process and reasons for exclusion.

OPEN IN VIEWER

FIG. 1.

Preferred Reporting Items for Systematic Review and Meta-Analyses flow diagram of studies through review process.

We included all articles that addressed bipolar I/II/not otherwise specified disorders, evaluated ≤18-year-old patients, included information about adherence to medications for treatment of bipolar disorder (at minimum antipsychotics or mood-stabilizers), and recorded characteristics that could influence medication adherence. Exclusion criteria included (i) inclusion of patients >18 years old or failure to separate out ≤18-year olds from adults in analysis; (ii) inclusion of only nonmedication treatment; (iii) no measure of medication adherence; (iv) no measure of characteristics that could impact adherence; (v) other psychiatric diagnoses besides bipolar disorder or failure to separate out patients with bipolar disorder; (vi) nonpsychiatric diagnoses; (vii) text not available in English; and (viii) not original research. The most common reason for exclusion was the age range restriction.

Results

Of the six articles identified above, two were retrospective database cohort studies (Evans-Lacko et al. 2010; Bhowmik et al. 2013), two were prospective cohort studies (Patel et al. 2005; DelBello et al. 2007), one was a single-institution retrospective cohort study (Dailey et al. 2005), and one evaluated a cohort during a lead-in phase for a randomized clinical trial (Drotar et al. 2007). Patients were followed from a minimum of 20 weeks to a maximum of 12 months.

Sample sizes ranged from 31 to 5460 (mean: 1021 ± 2180, median: 89, interquartile range: 304), with overall 6127 patients included within all studies. Ages in the studies ranged from 5 to 18 years (weighted mean: 14 years old ±1.62). Three studies included patients <12 years old. Medications studied included mood stabilizers for all six studies, antipsychotics for three, antidepressants for four, and stimulants for two (Fig. 2). Two studies evaluated adherence to only one medication class, which were mood stabilizers for both of these studies. All studies were carried out in the United States. The largest study, a database study, did not include any demographic information about patients (Bhowmik et al. 2013). Of the remaining 667 patients, 365 (55%) were male. The next largest study, another database study, did not include additional demographic information besides sex (Evans-Lacko et al. 2010). Of the remaining 241 patients, 151 (63%) were Caucasian. After age, sex, and race, all studies were variable in which demographic information they collected.

OPEN IN VIEWER

FIG. 2.

Medication categories measured in included articles.

Inclusion criteria, rigor of bipolar diagnoses, methods of determining adherence and factors influencing adherence varied widely between studies (Table 1). The database studies relied on claim submission data to establish a diagnosis and both required at least two different service dates noting a diagnosis of bipolar disorder or an inpatient admission diagnosis of bipolar disorder based on this claims data. In contrast, the other studies evaluated smaller populations with a more detailed process for evaluation. The two database studies also relied on prescription database records to evaluate adherence, while the four other studies at a minimum relied on clinical interviews along with blood levels, pill counts, or prescription refill data for some studies. All studies evaluated different factors associated with or possibly influencing adherence, will little to no overlap between these evaluated categories between studies.

Discussion

The literature concerning medication adherence and factors influencing adherence in pediatric patients with bipolar disorder appears to be very limited. After a systematic literature search and thorough review, only six studies met our inclusion criteria. Overall there was significant clinical heterogeneity (study populations), methodological heterogeneity (study design), and statistical heterogeneity (associations measured) between studies. Notably, all studies did measure adherence to mood stabilizers, but only three studies measured adherence to antipsychotics. Studies used varying ways to categorize medication adherence or continuation and one study did not directly categorize patients at all in terms of adherence. Of the studies that did categorize adherence, full adherence rates ranged from 34% to 68%. All studies did note possible factors associated with medication adherence. Many of these factors noted above were similar to what has been found in the adult literature (comorbidities, lower SE status, male sex, worse social support, worse subjective rating of drug/clinician helpfulness, and less consistent follow-up associated with worse adherence). Dailey et al. (2005) was unique in that it evaluated and found an association between medication noncompliance and criminal offenses in juveniles discharged from a corrections facility, which may help to inform policy toward juveniles with mental illness who commit a crime. No studies evaluated any intervention to improve medication adherence in pediatric bipolar disorder patients.

One crucial aspect to consider is the strictness of the diagnostic criteria for bipolar disorder used in studies. Some prior articles have noted a significantly higher rate of pediatric bipolar disorder diagnosis in the United States when compared with other countries. One recent study noted at 12.5 times, a higher rate of pediatric bipolar disorder discharge diagnoses in the United States versus England (James et al. 2014). Larger epidemiologic articles suggest that the true prevalence of pediatric bipolar disorder in the United States is no higher than in other countries (Van Meter et al. 2019). Recognition of this likely differential application in diagnostic criteria, even within the same country, is essential when reviewing research. All of our included studies were conducted in the United States, which may mean that a wider phenotype was included in these studies than would be included in other countries. Two studies used insurance claims information to include patients (Evans-Lacko et al. 2010; Bhowmik et al. 2013). Due to the nature of insurance claims data, including the possibility for incomplete or incorrect information, loss of patients who switched insurance plans or lost insurance and reimbursement policy incentives, these articles likely include a significantly different population of patients than the other included articles (Monteith et al. 2016). In contrast, two of the cohort studies included in-depth evaluation with multiple diagnostic instruments and the one lead-in cohort study was part of a larger randomized control trial that had stringent requirements for inclusion (Patel et al. 2005; DelBello et al. 2007; Drotar et al. 2007). It is not clear that adherence rates or associated factors in the larger claims-based studies generalize to a more rigorously diagnosed population. For instance, DelBello et al. (2007) found that medication polytherapy groups had higher rates of medication continuation than antipsychotic monotherapy groups. This could be explained by the polytherapy group meeting more stringent criteria for pediatric bipolar disorder, and so, greater emphasis being placed on medication continuation for this group. In addition, even among the more rigorously diagnosed groups, different studies used different criteria, with one using DSM-IIIR and three others using DSM-IV. No studies were recent enough to utilize the Diagnostic and Statistical Manual of Mental Disorders, 5th ed. (DSM-5) criteria (American Psychiatric Association 2013).

Measuring medication adherence is a challenging endeavor and there is no one standard for adherence measurement. Studies rely on measures ranging in invasiveness and objectivity from patient self-report to pill counts or electronic monitoring of pillboxes to serum drug levels. These different methods of adherence have been found in some cases to over- or underestimate adherence significantly when compared with each other (Busby and Sajatovic 2010). Due to these discrepancies, guidelines on adherence have suggested using at least two measures of adherence (Velligan et al. 2009). Defined cutoffs for levels of adherence (e.g., “adherent,” “partially adherent,” “nonadherent”) tend to be mostly based off of expert consensus, with this consensus defining medication nonadherence as missing ≥20%–30% of medications (Levin et al. 2016).

This challenge in consistently measuring adherence both within and between studies was also seen in this review. The larger database studies followed linked prescription fill data to patients, which means these data had more to do with prescription continuity rather than measurable adherence to prescription instructions. At the other extreme, one study used multiple measures of adherence for each patient, including serum levels, and compared these measures to each other for consistency (Drotar et al. 2007). Other studies used patient self-report at weekly interview or used an undefined cutoff of patient “unreliability” for medication count (Patel et al. 2005; DelBello et al. 2007). Each study used its own definitions to divide patients into variably defined categories that corresponded to adherent, partially adherent, or nonadherent. Definitions used by Delbello et al. (2007) and Patel et al. (2005) are most consistent with expert consensus (adherent >75%, partial 25%–75%, and nonadherent <25% medication taken as prescribed). Erring on the side of the most precise and consistent adherence measures possible is likely to be the most helpful when attempting to determine factors that could influence medication adherence in bipolar disorder.

Investigators tracking medication adherence must also decide which medications they will track. Atypical/second-generation antipsychotics and mood stabilizers are the mainstay of medical treatment for pediatric bipolar disorder with additional medication classes generally thought of as augmenting or used to treat other comorbid conditions. However, these additional classes (antidepressants, anxiolytics, psychostimulants) of psychotropic medications can be critical for proper treatment of comorbid conditions, and adherence to medication for comorbidities can be equally as important. In this review, although all six studies did track mood stabilizer use, only three studies evaluated antipsychotic use. This was easily explained in one study that was specifically designed to prospectively evaluate adherence to mood stabilizing medication (Drotar et al. 2007). It is unclear why the other two studies did not attempt to evaluate antipsychotic use. Studies were also variable in terms of the additional medications they evaluated, with four studies evaluating antidepressants and two studies evaluating psychostimulants. As these various medication classes have drastically different side effect profiles, the class of medication tracked has the potential to significantly influence the observed adherence rate in studies.

This review did have some limitations. First, we intentionally chose to restrict our age range to ≤18 years old. However, other studies include young adults (typically up to age 21) at times as “adolescents” when studying adolescent populations. The main reason most studies were excluded in our analysis was due to this age range and there were multiple studies we excluded that included young adults >18 years old. Notably, two studies in a previous review and meta-analysis evaluating medication adherence for pediatric patients with severe mental illness (including pediatric bipolar disorder patients) were excluded in our analysis primarily due to age (Edgcomb and Zima 2018). Coletti et al. (2005) evaluated primarily adherence rates among patients up to 19 years old with bipolar disorder but did include several factors (age, duration of illness) that they compared with adherence rates. Goldstein et al. (2016) evaluated adherence rates using both subjective and objective measurements and found that greater illness severity, male sex, greater self-reported barriers to adherence, day of the week, and number of doses per day were associated with worse adherence, however, this study included patients up to 22 years old. This review was also primarily concerned with studies relating medication adherence to factors that could influence adherence, which led to exclusion of additional studies. For instance, Findling et al. (2015) performed a randomized control trial evaluating lithium versus placebo, which recorded rates of adherence using a dosing diary and pill counts and found an adherence rate of 92 ± 11.8% for the lithium group but did not relate this to any other patient or environmental characteristics. This review only included English language studies or studies that had been translated into English. Expansion of our inclusion criteria or age range may have allowed us to include additional studies. However, we note that both the patient-level (i.e., comorbid conditions, age/maturity level) and external/systems-level factors (i.e., stability of family unit, school, family socioeconomic status) that could influence medication adherence are unique to pediatric patients compared with adult patients (El-Rachidi et al. 2017).

Going forward, prospective studies that include well-defined populations of pediatric bipolar disorder patients and objective measures of medication adherence would be the most helpful for identifying particular patient-level and systems-level characteristics associated with lower adherence. Although it is difficult to state one best measure of adherence, studies including both patient/family-reported adherence using standardized scales and some objective prospective measure (e.g., pill checks or electronic pill cap monitoring) would strike a balance between invasiveness and objectivity. The first-line treatments for bipolar disorder (antipsychotics and mood stabilizers) would be the most important medications to track. Pilot studies assessing patient perspectives and self-report on barriers to medication adherence would also be useful to narrow down particular domains to target. We anticipate that factors affecting adherence in this unique population will represent the intersection between nonadherence in adult bipolar disorder patients and characteristics associated with nonadherence in pediatric patients with chronic medical conditions overall. None of the characteristics associated with nonadherence in the studies included above were replicated across studies. However, there are likely unique nuances in pediatric bipolar disorder in comparison with adult bipolar disorder or other pediatric severe mental illness that contribute to nonadherence. For instance, due to mania/hypomania suppression, patients may overall have a more negative attitude toward treatment than in other pediatric psychiatric disorders. Due to the intermittent (yet severe) nature of bipolar illness, families may have a more negative attitude toward chronic medication use when comparing risk/benefit as opposed to more clinically persistent disorders (e.g., severe major depressive disorder, schizophrenia). Conversely, patients/families who feel that medication is not improving severe symptomatology may not find benefit in prioritizing adherence to medication. Bipolar disorder is also highly heritable and parental illness may itself partially mediate nonadherence, as seen in one study included in this article. Prospective randomized studies using measures targeted at improving adherence in this population will then be the ideal final step in attempting to improve adherence rates. The existing pediatric literature about adherence in chronic medical conditions gives insight into potential intervention targets. For instance, the pediatric self-management model envisions patients as operating within individual, family, community, and health care system domains and identifies modifiable domain-specific influences that can be targeted to improve self-management (Modi et al. 2012). Because there are no studies yet evaluating adherence promoting interventions in this population, identifying ways to help promote adherence for these patients at both the individual clinician level and health care system level remains a challenge. Many interventions used in studies for other patient populations would require some level of resource expenditure (e.g.,. additional education/therapy for families, targeted psychotherapy for patients, text/phone reminders and check-ins for patients, mobile applications to track medication pickup and administration), and so, demonstrating value for these interventions will be critical to garner support for deployment.

Conclusion

Overall, data about medication adherence and factors affecting medication adherence in the pediatric population appear to be sparse. The data that are available vary widely in multiple important domains, including methodology, patient population, diagnostic criteria, measures of medication adherence, outcomes measurement, and the possible influencing patient or environmental factors. However, based on the available data, pediatric bipolar disorder patients have similar, if not worse, rates of full medication adherence as adult bipolar disorder patients. No consistent factors associated with adherence or nonadherence were found between studies, although some studies found factors similar to what has been seen in the adult literature.

Clinical Significance

Few studies have evaluated medication adherence specifically in bipolar disorder patients ≤18 years old. We could not identify any studies evaluating methods to improve adherence among patients with pediatric bipolar disorder, highlighting an important gap in the current state of research. This review highlights critical areas in the existing pediatric bipolar disorder literature that would benefit from further study. Prospective studies utilizing both subjective and objective measures of adherence to evaluate characteristics associated with nonadherence and identify effective interventions to improve adherence are needed.