Bauer 

et al. Int J Bipolar Disord (2019) 7:19

https://doi.org/10.1186/s40345-019-0154-z

RESEARCH Open Access

Trajectories of adherence to mood stabilizers

in patients with bipolar disorder
M. Bauer1*  , T. Glenn2, M. Alda3, R. Bauer1, P. Grof4, W. Marsh5, S. Monteith6, R. Munoz7, N. Rasgon8,
K. Sagduyu9^ and P. C. Whybrow10

Abstract 
Background:  Nonadherence with mood stabilizers is a major problem that negatively impacts the course of bipolar
disorder. Medication adherence is a complex individual behavior, and adherence rates often change over time. This
study asked if distinct classes of adherence trajectories with mood stabilizers over time could be found, and if so,
which patient characteristics were associated with the classes.
Methods:  This analysis was based on 12 weeks of daily self-reported data from 273 patients with bipolar 1 or II
disorder using ChronoRecord computer software. All patients were taking at least one mood stabilizer. The latent class
mixed model was used to detect trajectories of adherence based on 12 weekly calculated adherence datapoints per
patient.
Results:  Two distinct trajectory classes were found: an adherent class (210 patients; 77%) and a less adherent class
(63 patients; 23%). The characteristics associated with the less adherent class were: more time not euthymic (p < 0.001)
and female gender (p = 0.016). No other demographic associations were found.
Conclusion:  In a sample of motivated patients who complete daily mood charting, about one quarter were in
the less adherent class. Even patients who actively participate in their care, such as by daily mood charting, may be
nonadherent. Demographic characteristics may not be useful in assessing individual adherence. Future research on
longitudinal adherence patterns in bipolar disorder is needed.

Introduction costs (Svarstad et  al. 2001; Gianfrancesco et  al. 2008;
Mood stabilizers are a fundamental treatment for both Hong et  al. 2011; Eaddy et  al. 2005; Hassan and Lage
acute episodes of bipolar disorder and the prevention of 2009; Schuepbach et  al. 2008), homelessness (Copeland
future episodes. Patient nonadherence is an important et  al. 2009), and involvement with the criminal justice
contributor to medication nonresponse (Osterberg and system (Robertson et  al. 2014). The harm from medica-
Blaschke 2005), and nonadherence with mood stabiliz- tion nonadherence in bipolar disorder is considerable to
ers may negatively impact every aspect of bipolar dis- both individual patients and society.
order. Poor medication adherence is associated with an Medication adherence refers to the extent to which
increase in relapses (Gutiérrez-Rojas et  al. 2010; Franks patients follow the medication regimen prescribed
et al. 2008), suicide and suicide attempts (Gonzalez-Pinto by their physician, often defined as taking ≥ 80% of
et  al. 2006; Pompili et  al. 2009), emergency room visits, the prescribed doses (Osterberg and Blaschke 2005).
hospitalizations, involuntary hospitalizations, healthcare Medication adherence involves three phases: initiation
of treatment, implementation or following the dosing
regimen, and persistence with treatment (Vrijens et al.
*Correspondence: michael.bauer@uniklinikum‑dresden.de 2012; Gellad et  al. 2017). Diverse factors and behav-
^
1
Deceased iors influence each phase including patient attitudes
Department of Psychiatry and Psychotherapy, Medical Faculty,
University Hospital Carl Gustav Carus, Technische Universität Dresden,
towards medication, clinical symptoms, adverse reac-
Fetscherstr. 74, 01307 Dresden, Germany tions, regimen complexity, health literacy, substance
Full list of author information is available at the end of the article

© The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License
(http://creat​iveco​mmons​.org/licen​ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium,
provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license,
and indicate if changes were made.
Bauer et al. Int J Bipolar Disord (2019) 7:19 Page 2 of 9

misuse, medication costs and forgetfulness (Jawad Methods

et al. 2018; García et al. 2016; Levin et al. 2016; Fred- All data were from outpatients with a diagnosis of bipo-
ericksen et  al. 2018). An adherence rate is difficult to lar disorder by DSM-IV or DSM-5 criteria that was made
measure, and will vary with the definition, methodol- by the prescribing psychiatrist during a clinical interview.
ogy, adherence phase, study design and study popula- All participants volunteered, were age 18 years or older,
tion (Sajatovic et al. 2010; Levin et al. 2016). Less than and provided informed written consent. The participants
half of patients with bipolar disorder are estimated to were recruited from a university mood clinic or private
be fully adherent (García et al. 2016; Levin et al. 2016; practice and received treatment as usual throughout the
Scott and Pope 2002). For most patients with bipo- study. The participants agreed to record medications,
lar disorder, adherence is partial or intermittent, and mood and sleep daily using computer software in their
changes over time (Scott and Pope 2002; Jawad et  al. native language (ChronoRecord). Demographic vari-
2018). Sometimes patients take all the prescribed ables were obtained from the patient by the clinician at
doses, sometimes a partial dose, and sometimes none the time of enrollment. The demographic characteristics
for one or all drugs in the medication regimen. of patients who use ChronoRecord are similar to that for
The most frequent way that psychiatrists evalu- patients in other studies of bipolar disorder (Bauer et al.
ate medication adherence in bipolar disorder is to ask 2012). Patients were included in this analysis if they had
the patient (Vieta et  al. 2012). However, estimates are a diagnosis of bipolar I or bipolar II disorder, returned at
often incorrect and optimistic (Stephenson et al. 2012; least 12 weeks of data (84 days), and were taking at least
Baldessarini et al. 2008; Lopez et al. 2017; De las Cue- one mood stabilizer during the 12 weeks. Although arbi-
vas et al. 2013). Although physicians frequently adjust trary, 12 weeks of data provides sufficient time for adher-
medications at visits for bipolar disorder (Hodgkin ence analysis, and includes patients that are comfortable
et  al. 2018), it is challenging to differentiate non- or with and willing to use ChronoRecord. Data used in this
inadequate adherence from nonresponse (Velligan analysis were collected between 2000 and 2016. The
et al. 2009). A lack of recognition of patient nonadher- ChronoRecord database was previously used for a variety
ence may lead to higher dosages, medication switches of published analyses. Active collection of ChronoRecord
and increasingly complex medication regimens, which data is ongoing.
may further reduce adherence (Baldessarini et al. 2008;
Eaddy et  al. 2005; Velligan et  al. 2009; Colom et  al.
2005). Daily data entry
More information to help assess adherence in Patients entered mood, sleep, medications taken and
patients with bipolar disorder is needed. Adherence life events daily, and weight weekly into ChronoRecord
studies generally use conventional longitudinal mod- software (Bauer et  al. 2004, 2008). All patients received
eling approaches that assume the individuals in a about a half hour of training in person or by phone before
sample come from a homogeneous population, the entering data. During training, a medication list was cre-
outcome of interest has a single growth trajectory, and ated by selecting from a list of psychotropic medications
any defined covariates influence each individual in displayed by brand and generic name. The list includes
the same way (Jung and Wickrama 2008; Proust-Lima every medication taken for bipolar disorder. For each
et  al. 2017). In contrast, trajectory analysis accom- selected medication, the pill strength was chosen from
modates a heterogeneous population, and allows the a list of available strengths. Every day, for each medica-
detection of subgroups or latent classes within a sam- tion, the patient entered the total number of pills taken.
ple that have different trajectories of change over time Patients could enter partial pills (1/4, 1/2, or 3/4) for tab-
for an outcome of interest (Nagin and Odgers 2010; lets but not capsules. If a medication was not taken, the
Lennon et al. 2018; Jung and Wickrama 2008). Trajec- patient entered zero pills for that drug. The patient could
tory analysis was previously used to assess medication modify the drugs taken throughout the study as needed,
adherence for several chronic conditions (Greenley and a drug not included in the software list could be
et  al. 2015; de Vries McClintock et  al. 2016; Hommel added by the patient. Data not entered on one day could
et  al. 2017; Blalock et  al. 2019). Based on 12  weeks of be entered later. The software prevents entry for a future
daily self-reported data from patients with bipolar date, prevents modification of previously entered data,
disorder, the purpose of this study was twofold: (1) to and requires confirmation for entry of a large number of
determine whether distinct classes of adherence tra- pills for a drug.
jectories could be identified for patients taking mood To record mood, patients entered a single daily rating
stabilizers, and (2) if trajectory classes were present, to that best described the prior 24 h using a 100-unit visual
detect associations with patient characteristics. analog scale. During training, the scale was calibrated to
Bauer et al. Int J Bipolar Disord (2019) 7:19 Page 3 of 9

the extremes of mania and depression the patient ever function. The individuals within each trajectory class will
experienced. Based upon the validation studies (Bauer be similar to each other, and different from individuals
et  al. 2004, 2008), a mood entry less than 40 was con- in other classes. The package “lcmm” in R software was
sidered depression, 40–60 euthymia, and greater than used for model estimation (Proust-Lima et al. 2017), and
60 hypomania/mania. The range of depression varied “LCMM toolkit” (Lennon et  al. 2018) package was used
between mild symptoms (an entry of 20–39) to moder- to verify results.
ate to severe symptoms (an entry of 0–19). The range The analysis models were checked for between 1 and
of mania varied from hypomania (an entry of 61–80) 5 trajectory classes based on a quadratic adherence tra-
to moderate to severe symptoms of mania (an entry of jectory function over the 12 weeks. Covariates were not
81–100). included in the trajectory function, and the variance–
covariance matrix was not constrained. The probability
Drugs analyzed that each patient was a member of a class was calculated
The mood stabilizers considered were lithium, valproate, and all patients were classified according to the highest
lamotrigine, carbamazepine, oxcarbazepine and second probability. Several measures were used to help select
generation antipsychotics: aripiprazole, olanzapine, ris- the optimal number of classes (van de Schoot et al. 2017;
peridone, quetiapine, ziprasidone, paliperidone, asenap- Lennon et al. 2018), including the lowest Bayesian Infor-
ine, lurasidone, and clozapine. For the analysis of total mation Criteria (BIC), class size, and probabilities of
psychotropic drugs taken and the daily pill burden, the class membership. The choice of preferred model also
other drugs considered were antidepressants, benzodiaz- included concerns for parsimony and interpretability of
epines, typical antipsychotics, insomnia medications, and results (Jung and Wickrama 2008). To verify the model
other anticonvulsants (topiramate, gabapentin, pregaba- estimates were not influenced by the initial values, a grid
lin, tiagabine, levetiracetam, zonisamide). search was performed to confirm the model estimates
reflected the best estimates.
Adherence For the selected number of classes, the patient char-
All medication data were self-reported, so the prescribed acteristics within the classes were compared using Chi-
dosage and medication changes were not known. For squared tests for categorical variables and ANOVA for
each patient, adherence was defined as taking at least continuous variables. Demographic characteristics for
one pill per day of each prescribed mood stabilizer. An the entire sample were also determined. SPSS version 25
entry of 0 pills or a missing day of data were treated as was used for all demographic calculations.
no pills taken. The adherence was calculated for each
day, and then a weekly adherence rate was created for a Results
total of 12 weekly datapoints per patient. For example to There were 560 patients with bipolar disorder in the data-
calculate a weekly rate, consider a patient taking 2 mood base of which 528 had a diagnosis of bipolar I or bipolar
stabilizers. If the patient took at least one pill for drug A II disorder. Of the 528 patients, 483 were taking a mood
on 7 days, and took at least one pill for drug B on 6 days, stabilizer, and 273 returned ≥ 84  days of data and were
adherence for that week would be 86% (adherent on included in the analysis. The 273 patients included in
6 days/7 days = 86%). A mood stabilizer that was discon- the analysis resided in the US (189, 69%), Germany (38,
tinued by tapering but remained on the medication list 14%), Canada (27, 10%), Australia (7, 3%), Chile (4, 1%),
was not included in the calculation. If patients returned Austria (3, 1%), Poland (3, 1%), and the UK (2, 1%). Of
more than 84 days of data, only the first 12 weeks of data the 273 patients, 151 (55%) were recruited at a university
were included to balance the contribution of each patient. mood center, and 122 (45%) from a private practice. Of
the 273 patients, 173 (63.4%) had bipolar I disorder and
Latent class trajectory modeling 100 (36.6%) had bipolar II disorder. Of the 273 patients,
The latent class mixed model (LCMM), an extension of a 192 (70.3%) were female and 81 (29.7%) were male. The
standard mixed model, was used to identify the trajectory demographic characteristics of 273 patients are shown in
classes for adherence with mood stabilizers (Proust-Lima Table 1. The 273 patients took a total of 2.8 ± 1.4 psychi-
et al. 2017; Lennon et al. 2018). Instead of one adherence atric drugs daily, with a mean pill burden of 6.3 ± 4.3 as
trajectory for the entire sample, the LCMM allows mul- shown in Table 2.
tiple trajectory classes. LCMM identifies the trajectory
classes without predefined assumptions as to the num- Latent classes
ber, size or trajectory pattern of the classes. With LCMM, A comparison of the models identified by LCMM for 1–5
each patient is included in only one class, and each class trajectory classes are shown in Table 3. Although the BIC
has a distinct trajectory based on a supplied trajectory for the 3-class model (− 2108) was slightly smaller than
Bauer et al. Int J Bipolar Disord (2019) 7:19 Page 4 of 9

Table 1  Patient demographics (N = 273) for the 2-class model (− 2091), one of the classes in the
Category N Percent 3-class model had only 20 members (7%), the probability
of class membership was lower, and trajectories for two
Gender of the three classes overlapped, making the interpretation
 Female 192 70.3 of results more difficult. Therefore the 2-class model was
 Male 81 29.7 selected as the preferred model.
Diagnosis For the preferred model, the actual and predicted
 BP I 173 63.4 weekly adherence trajectory over the 12  weeks for the
 BP II 100 36.6 two classes is shown in Fig.  1. One class has consist-
Disabled ently high adherence over the 12 weeks and is referred to
 No 175 73.5 as the adherent class, while the other class has consist-
 Yes 63 26.5 ently lower adherence and is referred to as the less adher-
Working full time ent class. The adherent class includes 210 patients or
 No 131 55.0 77% of the group, while the less adherent class includes
 Yes 107 45.0 63 patients or 23% of the group. The trajectory of the
Any college less adherent class was also relatively stable over the
 No 32 12.5 12 weeks.
 Yes 224 87.5 Only two patient characteristics differed between the
College graduate classes. The less adherent class spent a smaller percentage
 No 115 44.9 of days euthymic, and included more females, as shown
 Yes 141 55.1 in Table 4. There were no significant differences between
Married the two classes in other demographics including age,
 No 133 52.2 age of onset, diagnosis, college graduate, marital status,
 Yes 122 47.8 employed full time, number of hospitalizations for bipo-
Mean SD lar disorder, years of illness, or if receiving government
disability payment due to bipolar disorder. There were no
Hospitalizations (N = 248) 2.8 4.51
significant differences in the total number of psychiatric
Age of onset (N = 253) 22.3 10.34
medications, daily pill burden, specific mood stabilizer
Age (N = 273) 40.8 11.07
taken, or the use of any antidepressant, benzodiazepine,
Years of illness (N = 253) 18.8 12.12
stimulant or insomnia medication. There were no signifi-
Percent days depressed (N = 273) 20.5 22.3
cant differences between the classes in the percentage of
Percent days manic/hypomanic 8.4 11.8
(N = 273)
days with depression, severe depression, mania/hypoma-
Percent days euthymic (N = 273) 64.3 27.5
nia or mania. The smaller percentage of days euthymic in
the less adherent class was due to a larger percentage of
days with depressive symptoms in some patients, and to
a larger percentage of days with manic/hypomanic symp-
toms in others.
The mean percent of days missing medication data
Table 2  Patient medications (N = 273)
for all 273 patients was 7.0% ± 13.5. The mean percent
Medication N Percent of days missing medication data was 2.2% ± 3.4 for the
Lamotrigine 117 42.9
adherent class and 22.9% ± 20.5 for the less adherent
Lithium 97 35.5
class (p < 0.001).
Valproate 58 21.2
Carbamazepine/oxcarbazepine 34 12.5 Discussion
Any antipsychotic 123 45.1 This analysis detected two distinct patterns of adherence
Any antidepressant 135 49.5 with taking mood stabilizers over 12  weeks—an adher-
Any benzodiazepine 58 21.2 ent class and a less adherent class. About one quarter of
Any insomnia medication 18 6.6 the patients were in the less adherent class. Patients in
Mean SD the less adherent class spent significantly more time with
symptoms (not euthymic). In prior research in bipolar
­ edicationsa
Total number of m 2.8 1.4
disorder, affective morbidity including frequent or severe
Total pill ­burdena 6.3 4.3
episodes, psychosis, rapid cycling, and longer standing
a
  Only psychiatric drugs
Bauer et al. Int J Bipolar Disord (2019) 7:19 Page 5 of 9

Table 3  LCMM parameter estimates for 1 to 5 classes (N = 273) using a Quadratic Trajectory Function
N classes Maximum log AICa BICb Relative entropy Class parameter Class 1 Class 2 Class 3 Class 4 Class 5
likelihood

1 716.85 − 1419.69 − 1394.43 N 273

% 100%
APPAc 1.000
2d 1076.39 − 2130.79 − 2091.09 0.878 N 210 63
% 77% 23%
APPA 0.971 0.968
3 1095.99 − 2161.98 − 2107.84 0.889 N 20 202 51
% 7% 74% 19%
APPA 0.965 0.968 0.905
4 1095.99 − 2153.98 − 2085.40 0.549 N 20 198 0 55
% 7% 73% 0% 20%
APPA 0.965 0.507 0.000 0.870
5 1095.99 − 2145.98 − 2062.97 0.428 N 20 0 0 193 60
% 7% 0% 0% 71% 22%
APPA 0.965 0.000 0.000 0.344 0.824
Trajectory function was weekly adherence = a0 + a1 * week + a2 * week2. No covariates were included
a
  Akaike Information Criterion
b
  Bayesian Information Criterion
c
  Average posterior probability of assignment
d
  Preferred model

illness were associated with nonadherence (Leclerc et al. Other demographic characteristics associated with
2013; Greene et  al. 2018; Baldessarini et  al. 2008; Levin nonadherence in bipolar disorder in prior research were
et  al. 2016; García et  al. 2016). Symptoms that may be not significant in this study, including less education,
most severe during episodes, such as impulsivity (Swann younger age, younger age of onset, and a marital status of
et al. 2008) or lack of insight (Dell’Osso et al. 2002), were single (Leclerc et al. 2013; Levin et al. 2016). Adherence
also associated with poor adherence (Leclerc et al. 2013, involves a wide range of factors related to the individual,
García et  al. 2016; Pompili et  al. 2009; Belzeaux et  al. culture, language and communication, disease, drugs,
2015). Other symptoms such as chaotic lifestyles, loss of physician, and healthcare system (Leclerc et  al. 2013,
daily routine, circadian disruption of sleep–wake cycles, Pompili et  al. 2009; McQuaid and Landier 2018). The
and a preference for nighttime activities may make it routinely collected demographic characteristics may not
harder to adhere with medication regimens (Frank et al. be associated with adherence, as found in non-psychiat-
2000; Melo et al. 2017). ric conditions (Franklin et  al. 2016; Steiner et  al. 2009)
Female gender was also associated with the less adher- including studies using latent trajectory models (Blalock
ent class in this study. Although poorer adherence in et  al. 2019; de Vries McClintock et  al. 2016). Addition-
females with bipolar disorder was reported previously ally, when significant, the relation between adherence
(Kessing et al. 2007; Gianfrancesco et al. 2006; Belzeaux and demographic characteristics is usually weak, provid-
et al. 2013; Murru et al. 2013), review articles show con- ing little practical assistance in discriminating between
tradictory results for a link between gender and adher- adherent and nonadherent patients (Steiner et  al. 2009;
ence in bipolar disorder: no difference (Colom et  al. Osterberg and Blaschke 2005). The difficulty in inferring
2005; Jawad et  al. 2018; Greene et  al. 2018), inconsist- individual adherence based on demographics suggests
ent (Levin et  al. 2016), and males more associated with that self-reported measures may be helpful in clini-
nonadherence (Pompili et al. 2009; Leclerc et al. 2013). In cal settings, and a variety of paper and automated tools
the European Social Survey of 45,700 participants from are available (Steiner 2012; Stirratt et  al. 2015). Review
24 countries, females were more likely to be nonadher- articles report moderate to high correlation between
ent than males (Stavropoulou 2011). In a study in Japan, self-reported adherence questionnaires and diaries and
females with depression were more likely to hide infor- electronic monitoring (Monnette et al. 2018; Garber et al.
mation related to medication adherence (Sawada et  al. 2004; Shi et  al. 2010). While all approaches to measure
2012). However, the reasons for females being associated adherence have strengths and weaknesses (Sajatovic et al.
with the poor adherence class in this study are unknown. 2010; Lehmann et al. 2014; Levin et al. 2015; Di Matteo
2004), good agreement was found between self-reported
Bauer et al. Int J Bipolar Disord (2019) 7:19 Page 6 of 9

Fig. 1  Mood stabilizer adherence by week for each class. pred predicted adherence, obs observed adherence, CI 91% confidence interval

measures and serum levels of various psychiatric drugs Since taking one pill of a mood stabilizer was consid-
(Jónsdóttir et al. 2010), and with antidepressants (Loayza ered adherent, patients taking a lower dose than pre-
et al. 2012). scribed were included as adherent. The dosage timing
Given the wide range of behaviors that may impact the and drug administration instructions were also not con-
initiation, implementation and persistence with treat- sidered during the calculation. Drugs taken for general
ment, understanding the specific reasons for nonadher- medical conditions and over the counter drugs were not
ence is needed to determine the appropriate remedy included in the analysis. Complex medication regimens
(Stirratt et al. 2018; Gellad et al. 2017). The detection of and requirements for dosing more than once daily are
two distinct classes in the current study suggests that associated with decreased adherence in a wide range of
future research using trajectory analysis, designed to chronic illness (Ingersoll and Cohen 2008; Coleman et al.
evaluate nonadherence and define membership in trajec- 2012). In our prior research, a larger number of psychiat-
tory classes, may increase understanding of patterns of ric medications and greater pill burden were associated
adherence in bipolar disorder. with irregularity in daily dosage of mood stabilizers and
single day omissions (Pilhatsch et  al. 2018; Bauer et  al.
Limitations 2013a), but not with adherence when defined as at least
This analysis overestimates adherence with mood stabi- one pill of a mood stabilizer a day (Bauer et  al. 2010).
lizers for several reasons. By requiring 12 weeks of data, The 12 week time period would not provide information
the sample analyzed was pre-selected for adherence. about long-term persistence with treatment.
Bauer et al. Int J Bipolar Disord (2019) 7:19 Page 7 of 9

Table 4  Significant differences in the classes by LCMM class (N = 273)

Class 1 (adherent) Class 2 (less adherent) Total Test
N Percent N Percent N Percent Χ2 Sig. DF

Gender
 Male 70 33.3 11 17.5 81 29.7 0.016 1
 Female 140 66.7 52 82.5 192 70.3
Diagnosis
 BP I 135 64.3 38 60.3 173 63.4 0.566 1
 BP II 75 35.7 25 39.7 100 36.6
College graduate
 Yes 108 54.8 33 55.9 141 55.1 0.880 1
 No 89 45.2 26 44.1 115 44.9
Married
 Yes 89 45.9 33 54.1 122 47.8 0.262 1
 No 105 54.1 28 45.9 133 52.2
Mean SD Mean SD Mean SD F Sig. DF

Percent days euthymic 68.7 26.0 49.7 27.4 64.3 27.5 < 0.001 1271
Age of onset 22.5 10.3 21.6 10.5 22.3 10.3 0.547 1251
Age 40.9 11.0 40.5 11.5 40.8 11.1 0.834 1271
­ edicationsa
Total number of m 2.7 1.4 3.1 1.5 2.8 1.4 0.097 1271
Total pill ­burdena 6.1 4.2 6.9 4.3 6.3 4.3 0.145 1271
a
  Only psychiatric drugs

There are other limitations to this study. All data were Conclusion
self-reported and there was no objective confirmation In conclusion, in a sample of motivated patients who
of the medication data. There were more females than complete daily mood charting, there were two trajec-
males. Only oral drugs were included. The optimal rates tories of adherence with taking at least one pill of each
to define adherence are not uniform, but vary with the prescribed mood stabilizer, adherent and less adherent.
pharmacokinetic and pharmacodynamic properties of About one quarter of the patients were in the less adher-
the drug (Morrison et  al. 2017). Adherence to specific ent class. Characteristics associated with being in the
mood stabilizers was not investigated, as drug regimens less adherent class were more time with symptoms (not
for bipolar disorder are highly customized in clinical euthymic), and female gender. Motivated patients may be
practice (Bauer et  al. 2013b). Adherence rates may be nonadherent, and demographic characteristics may not
larger than one if patients take more than the prescribed be useful to assess individual adherence. Future research
dose, as noted in some newly admitted inpatients (Ger- to identify longitudinal adherence trajectory patterns is
etsegger et  al. 2019). The least adherent patients who needed in bipolar disorder.
would not use mood charting, and those who never fill
Acknowledgements
the initial prescription for a mood stabilizer were not None.
included. In a year-long study of 195,930 new electronic
prescriptions, the rate of non-filling for adults was 30.2% Authors’ contributions
MB and TG worked on the conception and design of the study, interpreted
overall, with 27.7% for drugs classified as neuropsychi- the results and drafted the manuscript. TG performed the analyses. MA, RB, PG,
atric, and 29.5% for antidepressants (Fischer et al. 2010). WM, SM, NR, KS and PW performed data collection, and interpretation of the
ChronoRecord was not designed to evaluate the reasons findings. All authors read and approved the final manuscript.
for nonadherence. Many important variables were not Funding
available for analysis relating to patient attitudes, cultural None.
issues, financial concerns, and the use of adherence tools
Availability of data and materials
such as pill boxes. Finally, it was not known if patients The data will not be shared or made publicly available. Informed consent for
were receiving any type of psychosocial interventions for this was not sought from the study participants prior to the collection of data.
bipolar disorder.
Bauer et al. Int J Bipolar Disord (2019) 7:19 Page 8 of 9

Ethics approval and consent to participate Colom F, Vieta E, Tacchi MJ, Sánchez-Moreno J, Scott J. Identifying and improv-
All participating centers obtained approval from their Institutional Review ing non-adherence in bipolar disorders. Bipolar Disord. 2005;7(Suppl
Boards and all patients gave a written informed consent to participate in this 5):24–31.
naturalistic observational non-interventional study. The data used for this Copeland LA, Miller AL, Welsh DE, McCarthy JF, Zeber JE, Kilbourne AM. Clinical
paper were derived from a naturalistic observational non-interventional study, and demographic factors associated with homelessness and incar-
which at the time of inclusion was not registered. ceration among VA patients with bipolar disorder. Am J Public Health.
2009;99:871–7.
Consent for publication De las Cuevas C, Peñate W, Sanz EJ. Psychiatric outpatients’ self-reported
Consent to publish has been obtained from the participants. adherence versus psychiatrists’ impressions on adherence in affective
disorders. Hum Psychopharmacol. 2013;28:142–50.
Competing interests de Vries McClintock HF, Morales KH, Small DS, Bogner HR. Patterns of adher-
The authors declare that they have no competing interests. ence to oral hypoglycemic agents and glucose control among primary
care patients with type 2 diabetes. Behav Med. 2016;42:63–71.
Author details Dell’Osso L, Pini S, Cassano GB, Mastrocinque C, Seckinger RA, Saettoni M,
1
 Department of Psychiatry and Psychotherapy, Medical Faculty, University et al. Insight into illness in patients with mania, mixed mania, bipolar
Hospital Carl Gustav Carus, Technische Universität Dresden, Fetscherstr. 74, depression and major depression with psychotic features. Bipolar Disord.
01307 Dresden, Germany. 2 ChronoRecord Association Inc., Fullerton, CA, 2002;4:315–22.
USA. 3 Department of Psychiatry, Dalhousie University, Halifax, NS, Canada. Di Matteo MR. Variations in patients’ adherence to medical recommendations:
4
 Mood Disorders Center of Ottawa, University of Toronto, Toronto, Canada. a quantitative review of 50 years of research. Med Care. 2004;42:200–9.
5
 Department of Psychiatry, University of Massachusetts, Worcester, MA, USA. Eaddy M, Grogg A, Locklear J. Assessment of compliance with antipsychotic
6
 Michigan State University College of Human Medicine, Traverse City Campus, treatment and resource utilization in a Medicaid population. Clin Ther.
Traverse City, MI, USA. 7 Department of Psychiatry, University of California 2005;27:263–72.
San Diego, San Diego, CA, USA. 8 Department of Psychiatry and Behavioral Fischer MA, Stedman MR, Lii J, Vogeli C, Shrank WH, Brookhart MA, et al. Pri-
Sciences, Stanford School of Medicine, Palo Alto, CA, USA. 9 Department mary medication non-adherence: analysis of 195,930 electronic prescrip-
of Psychiatry, University of Missouri Kansas City School of Medicine, Kansas tions. J Gen Intern Med. 2010;25:284–90.
City, MO, USA. 10 Department of Psychiatry and Biobehavioral Sciences, Semel Frank E, Swartz HA, Kupfer DJ. Interpersonal and social rhythm therapy: man-
Institute for Neuroscience and Human Behavior University of California Los aging the chaos of bipolar disorder. Biol Psychiatry. 2000;48:593–604.
Angeles (UCLA), Los Angeles, CA, USA. Franklin JM, Shrank WH, Lii J, Krumme AK, Matlin OS, Brennan TA, et al.
Observing versus predicting: initial patterns of filling predict long-term
Received: 8 May 2019 Accepted: 24 July 2019 adherence more accurately than high-dimensional modeling techniques.
Health Serv Res. 2016;51:220–39.
Franks M, Macritchie KA, Mahmood T, Young AH. Bouncing back: is the bipolar
rebound phenomenon peculiar to lithium? A retrospective naturalistic
study. J Psychopharmacol. 2008;22:452–6.
References Fredericksen RJ, Gibbons L, Brown S, Edwards TC, Yang FM, Fitzsimmons E, et al.
Baldessarini RJ, Perry R, Pike J. Factors associated with treatment nonadher- Medication understanding among patients living with multiple chronic
ence among US bipolar disorder patients. Hum Psychopharmacol. conditions: implications for patient-reported measures of adherence. Res
2008;23:95–105. Soc Adm Pharm. 2018;14:540–4.
Bauer M, Grof P, Gyulai L, Rasgon N, Glenn T, Whybrow PC. Using technology to Garber MC, Nau DP, Erickson SR, Aikens JE, Lawrence JB. The concordance of
improve longitudinal studies: self-reporting with ChronoRecord in bipolar self-report with other measures of medication adherence: a summary of
disorder. Bipolar Disord. 2004;6:67–74. the literature. Med Care. 2004;42:649–52.
Bauer M, Wilson T, Neuhaus K, Sasse J, Pfennig A, Lewitzka U, et al. Self-report- García S, Martínez-Cengotitabengoa M, López-Zurbano S, Zorrilla I, López P,
ing software for bipolar disorder: validation of ChronoRecord by patients Vieta E, et al. Adherence to antipsychotic medication in bipolar disorder
with mania. Psychiatry Res. 2008;159:359–66. and schizophrenic patients: a systematic review. J Clin Psychopharmacol.
Bauer M, Glenn T, Grof P, Marsh W, Sagduyu K, Alda M, et al. The association 2016;36:355–71.
between concurrent psychotropic medications and self-reported adher- Gellad WF, Thorpe CT, Steiner JF, Voils CI. The myths of medication adherence.
ence with taking a mood stabilizer in bipolar disorder. Hum Psychophar- Pharmacoepidemiol Drug Saf. 2017;26:1437–41.
macol. 2010;25:47–54. Geretsegger C, Pichler EM, Gimpl K, Aichhorn W, Stelzig R, Grabher-Stoeffler
Bauer M, Glenn T, Keil M, Bauer R, Marsh W, Grof P, et al. Brief depressive symp- G, et al. Non-adherence to psychotropic medication assessed by plasma
toms in patients with bipolar disorder: analysis of long-term self-reported level in newly admitted psychiatric patients: prevalence before acute
data. Aust N Z J Psychiatry. 2012;46:1068–78. admission. Psychiatry Clin Neurosci. 2019;73:175–8.
Bauer M, Glenn T, Alda M, Sagduyu K, Marsh W, Grof P, et al. Regularity in daily Gianfrancesco FD, Rajagopalan K, Sajatovic M, Wang RH. Treatment adherence
mood stabilizer dosage taken by patients with bipolar disorder. Pharma- among patients with bipolar or manic disorder taking atypical and typical
copsychiatry. 2013a;46:163–8. antipsychotics. J Clin Psychiatry. 2006;67:222–32.
Bauer M, Glenn T, Alda M, Sagduyu K, Marsh W, Grof P, et al. Drug treatment Gianfrancesco FD, Sajatovic M, Rajagopalan K, Wang RH. Antipsychotic treat-
patterns in bipolar disorder: analysis of long-term self-reported data. Int J ment adherence and associated mental health care use among individu-
Bipolar Disord. 2013b;1:5. als with bipolar disorder. Clin Ther. 2008;30:1358–74.
Belzeaux R, Correard N, Boyer L, Etain B, Loftus J, Bellivier F, et al. Depressive Gonzalez-Pinto A, Mosquera F, Alonso M, López P, Ramírez F, Vieta E, et al.
residual symptoms are associated with lower adherence to medication in Suicidal risk in bipolar I disorder patients and adherence to long-term
bipolar patients without substance use disorder: results from the FACE- lithium treatment. Bipolar Disord. 2006;8:618–24.
BD cohort. J Affect Disord. 2013;151:1009–15. Greene M, Paladini L, Lemmer T, Piedade A, Touya M, Clark O. Systematic litera-
Belzeaux R, Boyer L, Mazzola-Pomietto P, Michel P, Correard N, Aubin V, et al. ture review on patterns of pharmacological treatment and adherence
Adherence to medication is associated with non-planning impulsivity in among patients with bipolar disorder type I in the USA. Neuropsychiatr
euthymic bipolar disorder patients. J Affect Disord. 2015;184:60–6. Dis Treat. 2018;14:1545–59.
Blalock DV, Bosworth HB, Reeve BB, Voils CI. Co-occurring reasons for medica- Greenley RN, Karazsia B, Schurman JV, Gumidyala AP, Nguyen EU, Thomason
tion nonadherence within subgroups of patients with hyperlipidemia. J MM, et al. Trajectories of oral medication adherence in youth with inflam-
Behav Med. 2019;42:291–9. matory bowel disease. Health Psychol. 2015;34:514–21.
Coleman CI, Limone B, Sobieraj DM, Lee S, Roberts MS, Kaur R, et al. Dosing Gutiérrez-Rojas L, Jurado D, Martínez-Ortega JM, Gurpegui M. Poor adherence
frequency and medication adherence in chronic disease. J Manag Care to treatment associated with a high recurrence in a bipolar disorder
Pharm. 2012;18:527–39. outpatient sample. J Affect Disord. 2010;127:77–83.
Bauer et al. Int J Bipolar Disord (2019) 7:19 Page 9 of 9

Hassan M, Lage MJ. Risk of rehospitalization among bipolar disorder patients Pompili M, Serafini G, Del Casale A, Rigucci S, Innamorati M, Girardi P, et al.
who are nonadherent to antipsychotic therapy after hospital discharge. Improving adherence in mood disorders: the struggle against relapse,
Am J Health Syst Pharm. 2009;66:358–65. recurrence and suicide risk. Expert Rev Neurother. 2009;9:985–1004.
Hodgkin D, Stewart MT, Merrick EL, Pogue YZ, Reilly-Harrington NA, Sylvia Proust-Lima C, Philipps V, Liquet B. Estimation of extended mixed models
LG, et al. Prevalence and predictors of physician recommendations for using latent classes and latent processes: the R Package lcmm. J Stat
medication adjustment in bipolar disorder treatment. J Affect Disord. Softw. 2017;78:1–56.
2018;238:666–73. Robertson AG, Swanson JW, Van Dorn RA, Swartz MS. Treatment participation
Hommel KA, McGrady ME, Peugh J, Zacur G, Loreaux K, Saeed S, et al. Longitu- and medication adherence: effects on criminal justice costs of persons
dinal patterns of medication nonadherence and associated health care with mental illness. Psychiatr Serv. 2014;65:1189–91.
costs. Inflamm Bowel Dis. 2017;23:1577–83. Sajatovic M, Velligan DI, Weiden PJ, Valenstein MA, Ogedegbe G. Measurement
Hong J, Reed C, Novick D, Haro JM, Aguado J. Clinical and economic conse- of psychiatric treatment adherence. J Psychosom Res. 2010;69:591–9.
quences of medication non-adherence in the treatment of patients with Sawada N, Uchida H, Watanabe K, Kikuchi T, Suzuki T, Kashima H, et al. How
a manic/mixed episode of bipolar disorder: results from the European successful are physicians in eliciting the truth from their patients? A
Mania in Bipolar Longitudinal Evaluation of Medication (EMBLEM) study. large-scale Internet survey from patients’ perspectives. J Clin Psychiatry.
Psychiatry Res. 2011;190:110–4. 2012;73:311–7.
Ingersoll KS, Cohen J. The impact of medication regimen factors on adherence Schuepbach D, Novick D, Haro JM, Reed C, Boeker H, Noda S, et al. Determi-
to chronic treatment: a review of literature. J Behav Med. 2008;31:213–24. nants of voluntary vs. involuntary admission in bipolar disorder and the
Jawad I, Watson S, Haddad PM, Talbot PS, McAllister-Williams RH. Medication impact of adherence. Pharmacopsychiatry. 2008;41:29–36.
nonadherence in bipolar disorder: a narrative review. Ther Adv Psychop- Scott J, Pope M. Nonadherence with mood stabilizers: prevalence and predic-
harmacol. 2018;8:349–63. tors. J Clin Psychiatry. 2002;63:384–90.
Jónsdóttir H, Opjordsmoen S, Birkenaes AB, Engh JA, Ringen PA, Vaskinn A, Shi L, Liu J, Koleva Y, Fonseca V, Kalsekar A, Pawaskar M. Concordance of adher-
et al. Medication adherence in outpatients with severe mental disorders: ence measurement using self-reported adherence questionnaires and
relation between self-reports and serum level. J Clin Psychopharmacol. medication monitoring devices. Pharmacoeconomics. 2010;28:1097–107.
2010;30:169–75. Stavropoulou C. Non-adherence to medication and doctor-patient
Jung T, Wickrama KAS. An introduction to latent class growth analysis and relationship: evidence from a European survey. Patient Educ Couns.
growth mixture modeling. Soc Personal Psychol Compass. 2008;2:302–17. 2011;83:7–13.
Kessing LV, Søndergård L, Kvist K, Andersen PK. Adherence to lithium in Steiner JF. Rethinking adherence. Ann Intern Med. 2012;157:580–5.
naturalistic settings: results from a nationwide pharmacoepidemiological Steiner JF, Ho PM, Beaty BL, Dickinson LM, Hanratty R, Zeng C, et al. Sociode-
study. Bipolar Disord. 2007;9:730–6. mographic and clinical characteristics are not clinically useful predictors
Leclerc E, Mansur RB, Brietzke E. Determinants of adherence to treat- of refill adherence in patients with hypertension. Circ Cardiovasc Qual
ment in bipolar disorder: a comprehensive review. J Affect Disord. Outcomes. 2009;2:451–7.
2013;149:247–52. Stephenson JJ, Tunceli O, Gu T, Eisenberg D, Panish J, Crivera C, et al. Adher-
Lehmann A, Aslani P, Ahmed R, Celio J, Gauchet A, Bedouch P, et al. Assess- ence to oral second-generation antipsychotic medications in patients
ing medication adherence: options to consider. Int J Clin Pharm. with schizophrenia and bipolar disorder: physicians’ perceptions of adher-
2014;36:55–69. ence vs. pharmacy claims. Int J Clin Pract. 2012;66:565–73.
Lennon H, Kelly S, Sperrin M, Buchan I, Cross AJ, Leitzmann M, et al. Framework Stirratt MJ, Dunbar-Jacob J, Crane HM, Simoni JM, Czajkowski S, Hilliard ME,
to construct and interpret latent class trajectory modelling. BMJ Open. et al. Self-report measures of medication adherence behavior: recom-
2018;8:e020683. mendations on optimal use. Transl Behav Med. 2015;5:470–82.
Levin JB, Sams J, Tatsuoka C, Cassidy KA, Sajatovic M. Use of automated Stirratt MJ, Curtis JR, Danila MI, Hansen R, Miller MJ, Gakumo CA. Advancing
medication adherence monitoring in bipolar disorder research: pitfalls, the science and practice of medication adherence. J Gen Intern Med.
pragmatics, and possibilities. Ther Adv Psychopharmacol. 2015;5:76–87. 2018;33:216–22.
Levin JB, Krivenko A, Howland M, Schlachet R, Sajatovic M. Medication adher- Svarstad BL, Shireman TI, Sweeney JK. Using drug claims data to assess the
ence in patients with bipolar disorder: a comprehensive review. CNS relationship of medication adherence with hospitalization and costs.
Drugs. 2016;30:819–35. Psychiatr Serv. 2001;52:805–11.
Loayza N, Crettol S, Riquier F, Eap CB. Adherence to antidepressant treatment: Swann AC, Steinberg JL, Lijffijt M, Moeller FG. Impulsivity: differential rela-
what the doctor thinks and what the patient says. Pharmacopsychiatry. tionship to depression and mania in bipolar disorder. J Affect Disord.
2012;45:204–7. 2008;106:241–8.
Lopez LV, Shaikh A, Merson J, Greenberg J, Suckow RF, Kane JM. Accuracy of Van De Schoot R, Sijbrandij M, Winter SD, Depaoli S, Vermunt JK. The GRoLTS-
clinician assessments of medication status in the emergency setting: a checklist: guidelines for reporting on latent trajectory studies. Struct Equ
comparison of clinician assessment of antipsychotic usage and plasma Model. 2017;24:451–67.
level determination. J Clin Psychopharmacol. 2017;37:310–4. Velligan DI, Weiden PJ, Sajatovic M, Scott J, Carpenter D, Ross R, et al. The
McQuaid EL, Landier W. Cultural issues in medication adherence: disparities expert consensus guideline series: adherence problems in patients with
and directions. J Gen Intern Med. 2018;33:200–6. serious and persistent mental illness. J Clin Psychiatry. 2009;70(Suppl
Melo MCA, Abreu RLC, Linhares Neto VB, de Bruin PFC, de Bruin VMS. Chrono- 4):1–46.
type and circadian rhythm in bipolar disorder: a systematic review. Sleep Vieta E, Azorin JM, Bauer M, Frangou S, Perugi G, Martinez G, et al. Psychia-
Med Rev. 2017;34:46–58. trists’ perceptions of potential reasons for non- and partial adherence to
Monnette A, Zhang Y, Shao H, Shi L. Concordance of adherence measurement medication: results of a survey in bipolar disorder from eight European
using self-reported adherence questionnaires and medication monitor- countries. J Affect Disord. 2012;143:125–30.
ing devices: an updated review. Pharmacoeconomics. 2018;36:17–27. Vrijens B, De Geest S, Hughes DA, Przemyslaw K, Demonceau J, Ruppar T, et al.
Morrison A, Stauffer ME, Kaufman AS. Relationship between adherence rate A new taxonomy for describing and defining adherence to medications.
threshold and drug ‘forgiveness’. Clin Pharmacokinet. 2017;56:1435–40. Br J Clin Pharmacol. 2012;73:691–705.
Murru A, Pacchiarotti I, Amann BL, Nivoli AM, Vieta E, Colom F. Treatment
adherence in bipolar I and schizoaffective disorder, bipolar type. J Affect
Disord. 2013;151:1003–8. Publisher’s Note
Nagin DS, Odgers CL. Group-based trajectory modeling in clinical research. Springer Nature remains neutral with regard to jurisdictional claims in pub-
Annu Rev Clin Psychol. 2010;6:109–38. lished maps and institutional affiliations.
Osterberg L, Blaschke T. Adherence to medication. N Engl J Med.
2005;353:487–97.
Pilhatsch M, Glenn T, Rasgon N, Alda M, Sagduyu K, Grof P, et al. Regularity
of self-reported daily dosage of mood stabilizers and antipsychotics in
patients with bipolar disorder. Int J Bipolar Disord. 2018;6:10.