Faurholt‑Jepsen et al.

Int J Bipolar Disord (2019) 7:5

https://doi.org/10.1186/s40345-019-0141-4

RESEARCH Open Access

Differences in mood instability

in patients with bipolar disorder type I and II:
a smartphone‑based study
Maria Faurholt‑Jepsen1*  , Mads Frost2, Jonas Busk3, Ellen Margrethe Christensen1, Jakob E. Bardram3,
Maj Vinberg1 and Lars Vedel Kessing1

Abstract 
Background:  Mood instability in bipolar disorder is associated with a risk of relapse. This study investigated differ‑
ences in mood instability between patients with bipolar disorder type I and type II, which previously has been spar‑
ingly investigated.
Methods:  Patients with bipolar disorder type I (n = 53) and type II (n = 31) used a daily smartphone-based self-moni‑
toring system for 9 months. Data in the present reflect 15.975 observations of daily collected smartphone-based data
on patient-evaluated mood.
Results:  In models adjusted for age, gender, illness duration and psychopharmacological treatment, patients with
bipolar disorder type II experienced more mood instability during depression compared with patients with bipolar
disorder type I (B: 0.27, 95% CI 0.007; 0.53, p = 0.044), but lower intensity of manic symptoms. Patients with bipolar disor‑
der type II did not experience lower mean mood or higher intensity of depressive symptoms compared with patients
with bipolar disorder type I.
Conclusions:  Compared to bipolar disorder type I, patients with bipolar disorder type II had higher mood instability
for depression. Clinically it is of importance to identify these inter-episodic symptoms. Future studies investigating the
effect of treatment on mood instability measures are warranted.
Trial registration NCT02221336
Keywords:  Bipolar disorder, Hypomania, Depression

Background between affective episodes is associated with sub-

Bipolar disorder is characterized by recurrent episodes stantial disability and poor prognostic factors such as
of depression and (hypo)mania and in the form of sub- increased risk of hospitalization, high risk of relapse,
syndromal mood swings. A substantial proportion of and impaired functioning (Bopp et al. 2010; Judd et al.
patients with bipolar disorder does not experience full 2003; Joffe et  al. 2004; Kupka et  al. 2007; MacQueen
remission, remains symptomatic during inter-episode et  al. 2003; Strejilevich et  al. 2013; Patel et  al. 2015).
periods and experience daily mood swings above what This emphasizes the serious nature of bipolar disorder.
is experienced by healthy individuals (Bopp et al. 2010; Noticing the clinically and functionally impact of mood
Judd et  al. 2003; Joffe et  al. 2004). Mood instability instability, mood instability has been suggested as a
target for treatment and may be a more sensitive out-
come measure in randomized controlled maintenance
*Correspondence: maria@faurholt‑jepsen.dk trials (RCT) than for example relapse or recurrence
1
Copenhagen Affective Disorder Research Center (CADIC), Psychiatric of affective episodes (Bopp et  al. 2010; Bonsall et  al.
Center Copenhagen, Rigshospitalet, Blegdamsvej 9, 2100 Copenhagen,
Denmark
2012; Saunders et al. 2016). However, few studies have
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.
Faurholt‑Jepsen et al. Int J Bipolar Disord (2019) 7:5 Page 2 of 8

suggested methods and measures on how to investigate and intensity of depressive symptoms compared with
mood instability and its possible predictors. patients with bipolar disorder type I.
Patients with bipolar disorder type II seem to spend
more time depressed, less time euthymic and experi- Methods
ence greater depression instability than patients with Design, settings and participants
bipolar disorder type I (Joffe et al. 2004; Faurholt-Jepsen The patients were recruited as part of a randomized
et al. 2015a; O’Donnell et al. 2018; Vinberg et al. 2017). controlled single-blind parallel-group trial investigat-
Previous studies investigating differences in mood ing the effect of smartphone-based self-monitoring, the
between bipolar disorder type I and II have collected MONARCA II trial (Faurholt-Jepsen et al. 2014) includ-
data on mood on a weekly and not daily basis and fur- ing adult patients with bipolar disorder for a 9  months
thermore used paper based charts, increasing the risk follow-up period. All patients with a diagnosis of bipolar
of recall bias and low compliance (Judd et al. 2003; Joffe disorder who previously had been treated at The Copen-
et  al. 2004; Kupka et  al. 2007; MacQueen et  al. 2003; hagen Clinic for Affective Disorders, Denmark in the
Strejilevich et al. 2013). One previous pilot study by the period 2004 to January 2016, currently treated at com-
authors used smartphone-based daily electronic self- munity psychiatric centers, private psychiatrists and gen-
monitoring of mood (Faurholt-Jepsen et  al. 2015a). In eral practitioners were invited to participate in the trial,
this study (the MONARCA I trial) including 33 patients corresponding to approximately 735 potential partici-
with bipolar disorder, the patients with bipolar disor- pants. Treatment at the Copenhagen Clinic for Affective
der type II spend less time euthymic, a higher propor- Disorders comprised combined psychopharmacological
tion of time with depressive symptoms and experienced treatment as according to international guidelines, group
higher mood instability for depression compared with based psychoeducation and supporting therapy for a
patients with bipolar disorder type I (Faurholt-Jepsen total of 2-year (Kessing et al. 2013).
et al. 2015a). Apart from this previous pilot study by the Inclusion criteria: Bipolar disorder diagnosis according
authors, no study has investigated differences in mood to ICD-10 using the schedules for clinical assessments in
instability between patients with bipolar disorder type neuropsychiatry (SCAN) (Wing et  al. 1990) and previ-
I and bipolar disorder type II using fine-grained daily ously treatment at the Copenhagen Clinic for Affective
electronic data. A recent paper discussed differences Disorders. Exclusion criteria: schizophrenia, schizotypal
in momentary and retrospective trait self-report tech- or delusional disorders; previous use of the smartphone-
niques pointing out that retrospective self-monitoring based monitoring system; pregnancy and lack of Danish
is influenced by peak moments with greater salience of language skills.
moments that occur closest in time to the assessment Patients were diagnostically categorized into bipolar
(Conner and Barrett 2012), which may be a particu- disorder type I or II according to the SCAN interview.
lar problem when investigating mood instability given During the study the patients’ severity of depressive
its dynamic structure in both polarity, variation and and manic were evaluated at baseline, after 4  weeks,
intensity. Ecological momentary assessments (EMA) 3  months, 6  months and 9  months using the Hamilton
reflect the methods used to collect assessments of indi- Depression Rating Scale 17-items (HDRS) (Hamilton
vidual’s real-time states repeatedly over time and dur- 1967) and the Young Mania Rating Scale (YMRS) (Young
ing naturalistic settings, and may reduce recall bias et al. 1978).
(Shiffman et  al. 2008). Smartphones extends the use
of EMA beyond its classical use for self-reports and Smartphone‑based monitoring
offer the opportunity to collect fine-grained data unob- Patients were randomized to using a smartphone-based
trusively and outside the clinical settings (Ebner-Prie- monitoring system (the MONARCA II system) for daily
mer and Trull 2009), and enable collection of data on self-monitoring (intervention group) or treatment as
daily subsyndromal mood fluctuations. Based on daily usual (control group). Only patients included in the
smartphone-based self-monitoring data, the present intervention group collecting daily smartphone-based
report aimed to investigate, and replicate findings from self-monitoring data were included and presented in the
the before mentioned study by the authors (Faurholt- present report.
Jepsen et al. 2015a). We hypothesized that patients with Patients in the intervention group used a smartphone
bipolar disorder type II had lower mean mood level, with the MONARCA II app installed and were instructed
spend less time euthymic, a higher proportion of time to use the system for daily evaluation for 9  months.
with depressive symptoms, experienced higher instabil- The MONARCA II system allowed for evaluation of
ity for depression, i.e. more fluctuations in self-moni- the following measures on a daily basis: mood (scored
tored depressive mood, and experienced higher levels on a scale from − 3 to + 3); sleep duration (measured
Faurholt‑Jepsen et al. Int J Bipolar Disord (2019) 7:5 Page 3 of 8

in half-hour intervals); medicine intake; activity level 2. Mood symptomatic factors: A mood symptomatic
(scored on a scale from − 3 to + 3); mixed mood (scored factor for depression was calculated as the number
on a scale from 1 to 3); irritability (scored on a scale from of days with depressive symptoms < − 0.5 on a mood
1 to 3); anxiety (scored on a scale from 1 to 3); cognitive scale from − 3 to + 3 divided by the number of weeks
problems (scored on a scale from 1 to 3); alcohol con- followed. A mood symptomatic factor for mania was
sumption; and stress (scored on a scale from 1 to 3). The calculated as the number of days with manic symp-
patients were prompted to evaluate by an alarm set at a toms > + 0.5 on a mood scale from − 3 to + 3 divided
self-chosen time during the day. Further details regard- by the number of weeks followed.
ing the MONARCA II system are described elsewhere 3. Mood intensity: A mood intensity factor for depres-
(Faurholt-Jepsen et al. 2014). sion was calculated as the summary score of depres-
sive symptoms divided by the number of weeks
followed. A mood intensity factor for mania was cal-
Statistical methods culated as the summary score of manic symptoms
The hypotheses and statistical analyses were defined á divided by the number of weeks followed.
priori. The validity of self-monitored mood compared
with clinically rated depressive and manic reflected by In addition to these measures and as suggested by oth-
the HDRS and the YMRS, respectively was investigated. ers, mood instability scores reflecting the extent to which
Since the HDRS and the YMRS reflects clinically rated consecutively measured mood differ from one another
severity of depressive and manic symptoms for the past during the follow-up period were computed for each user
4  days, the mean self-monitored mood for the days the by applying the root mean square successive difference
scales were reflecting were used in the analyses of the (rMSSD) method, taking the square root of the sum of
validity of self-monitored mood. A two-level linear mixed the squared differences between daily and previous day
effect model, which accommodates both variation of the mood scores (O’Donnell et  al. 2018; Ebner-Priemer and
variables of interest within patients (intra-individual vari- Trull 2009). The rMSSD reflects both size and temporal
ation) and between individuals (inter-individuals varia- order of changes in mood. Higher rMSSD scores indicate
tion) was employed. The models including a fixed effect more instability. Previous reports have demonstrated the
of visit number (baseline, 4 weeks, 3 months, 6 months, construct validity of this method (Ebner-Priemer and
9 months) and a patients-specific random effect allowing Trull 2009; Jahng et al. 2008).
for individual intercept and a slope for each participant Regression analyses were used to evaluate the rela-
were employed. Level one represented repeated meas- tionship between the different calculated mood indexes
ures of clinically rated symptoms (HDRS and YMRS) and and bipolar disorder type I or bipolar disorder type II,
level two represented inter-individual variation. respectively. First, we considered an unadjusted model
Differences in mean sleep duration, mood level and (model 1). Secondly, we considered a model adjusted
proportion of time with mixed symptoms were investi- for age and gender (model 2). Thirdly, we considered a
gated by means of analysis of covariance (ANCOVA). To model adjusted for age, gender, illness duration and psy-
summarize the patient profiles over time summary meas- chopharmacological treatment (anticonvulsant treatment
ures or indexes covering three areas (instability, symptom yes/no, antipsychotic treatment yes/no, antidepressant
load and intensity) as indicators of illness activity were treatment yes/no, lithium treatment yes/no) (model 3) as
calculated for each patient in the same way as in our pre- these variables could affect the differences in mood insta-
vious paper based on daily self-monitored data on mood bility measures between the two groups. Model assump-
scores (Faurholt-Jepsen et al. 2015a) and as suggested and tions were checked for each of the statistical analyses. As
used by others (Strejilevich et al. 2013) (data presented in few prior studies have investigated differences in mood
Table 2): instability indexes between bipolar disorder type I and II
we were not able to make statistical power analyses prior
1. Mood instability factors: A mood instability fac- to the study since potential effects were unknown. The
tor measure was calculated as the number of mood overall hypotheses were made a priori based on our prior
changes divided by the number of weeks followed. A findings within patients with bipolar disorder. Neverthe-
mood instability factor for depression was calculated less, as these hypotheses included several measures on
as the number of mood changes of depressive polar- mood instability, the present study should be considered
ity divided by the number of weeks followed. A mood as hypotheses generating and the results needs further
instability factor for mania was calculated as the replication. Consequently, we did not account for multi-
number of mood changes of manic polarity divided ple testing in the statistical models. p-values below 0.05
be the number of weeks followed. were considered statistical significant. Data were entered
Faurholt‑Jepsen et al. Int J Bipolar Disord (2019) 7:5 Page 4 of 8

­ pidata®, STATA (StataCorp LP, Collega

using Excel and E the days. Data in the present report are based on 15.975
Station, TX, USA) version 12.1 was used for statistical daily smartphone-based self-monitored observations.
analyses. Regarding the validity of self-monitored mood com-
pared with clinically rated depressive and manic symp-
toms, there was a statistically significant negative
Ethical considerations
association between self-monitored mood and scores
Ethical permission was obtained from the Regional Eth-
on the HDRS in both unadjusted models and in mod-
ics Committee in The Capital Region of Denmark and
els adjusted for age, gender and psychopharmacological
The Danish Data Protection Agency (H-2-2014-059).
treatment (adjusted model B: − 0.033, 95% CI − 0.046;
− 0.20, p < 0.0001). Thus, for every increase of 10 points
Results on the HDRS there was a decrease of 0.33 point on the
Background characteristics self-monitoring mood scale (range − 3 to + 3). Also, there
From October 2014 to January 2018, a total of 85 patients was a statistically significant positive association between
collected daily smartphone-based self-monitoring data self-monitored mood and scores on the YMRS in both
(the intervention group of the MONARCA II trial) for unadjusted models and in models adjusted for age, gen-
a study period of 9 months. In total 84 patient provided der and psychopharmacological treatment (adjusted
self-monitoring data during the study and were included model B: 0.044, 95% CI 0.023; 0.066, p < 0.0001). Thus,
in the present study (bipolar disorder type I n = 53, bipo- for every increase of 10 points on the YMRS there was an
lar disorder type II n = 31). As presented in Table  1, increase of 0.44 point on the self-monitoring mood scale
patients had a mean age of 43.0  years (SD 12.3) and 51 (range − 3 to + 3). Analyses including BD type (BD type I
(61.2%) were of female gender. Patients with bipolar dis- or II) did not alter the estimates.
order type I were more often in prescribed treatment
with antipsychotics compared with patients with bipolar Bipolar disorder subtype and self‑monitored smartphone
disorder type II (bipolar disorder type I: 59.3%; bipolar data
disorder type II: 32.3%) (Table 1). Overall, regardless the bipolar subtype, the patients had a
At baseline a total of 52.9% (n = 44) of the patients had mean mood level below zero and reported mixed symp-
a HDRS score ≤ 7 and 87.1% (n = 73) had a HDRS score toms a large proportion of time when experiencing levels
≤ 14. A total of 85.3% (n = 72) had a YMRS score ≤ 7 and of depression or mania during the study, but there were
97.6% (n = 82) had a YMRS score ≤ 14. During the entire no statistically significant differences (all p-values > 0.05)
9 months follow-up period 85.6% (n = 72) of the patients between patients with bipolar disorder type I and bipolar
had a HDRS score ≤ 14 and 97.5% (n = 82) had a YMRS disorder type II in the self-monitored mean mood level
score ≤ 14. During the study patients adhered to self- [patients with bipolar disorder type I: − 0.15 (95% CI
monitoring with a mean of 72.6% (a mean of 190 days) of − 0.26; − 0.043) vs. patients with bipolar disorder type II:

Table 1  Background characteristics of patients with bipolar disorder I and II, N = 84

All patients (n = 84) Bipolar disorder type I Bipolar disorder type II p
(n = 53) (n = 31)

Age (years) 43.0 (12.3) 42.8 (11.9) 43.4 (13.3) 0.85

Female gender, % (n) 61.2 (51) 64.8 (34) 54.8 (17) 0.36
Full time employment, % (n) 17.7 (15) 16.7 (9) 19.4 (6) 0.73
Illness duration (years) 18.0 (10.1) 17.7 (10.6) 18.6 (9.5) 0.71
Number of previous depressive episodes 4 [2–10] 3.5 [2–9] 5 [3–10] 0.21
Number of previous (hypo)manic episodes 3 [2–7] 3 [2–6] 5 [2–8] 0.33
Hamilton depression rating scale score, baseline 7.6 (5.0) 7.9 (5.0) 7.2 (6.0) 0.57
Young Mania Rating Scale score, baseline 3.2 (4.6) 4.1 (5.6) 2.6 (3.0) 0.16
Psychopharmacological treatment
 Anticonvulsants, % (n) 55.9 (47) 57.4 (30) 54.8 (17) 0.82
 Antipsychotics, % (n) 48.8 (41) 59.3 (31) 32.3 (10) 0.017
 Antidepressants, % (n) 22.6 (19) 26.4 (14) 16.1 (5) 0.28
 Lithium, % (n) 63.1 (53) 68.5 (36) 54.8 (17) 0.21
Data are mean (SD), median [IQR] or proportions (n) unless otherwise stated
Faurholt‑Jepsen et al. Int J Bipolar Disord (2019) 7:5 Page 5 of 8

− 0.20 (95% CI − 0.33; − 0.065)], the proportion of over- duration (B: − 0.01, 95% CI − 0.040; 0.02, p = 0.44). Fur-
all time with mixed symptoms [patients with bipolar dis- thermore, patients with bipolar disorder type II had sta-
order type I: 8.77% (95% CI 4.75; 15.63) vs. patients with tistically significantly lower Mood symptomatic factor for
bipolar disorder type II: 3.51% (95% CI 0.86; 13.23)], the mania and mood intensity factor for mania (mood inten-
proportion of time with mixed symptoms when scoring sity factor for mania: models adjusted for age, gender,
< − 0.5 on a mood scale [patients with bipolar disorder illness duration and psychopharmacological treatment
type I: 16.67% (95% CI 8.07; 41.40) vs. patients with bipo- (anticonvulsant treatment yes/no, antipsychotic treat-
lar disorder type II: 11.11% (95% CI 1.45; 37.73)] and the ment yes/no, antidepressant treatment yes/no, lithium
proportion of time with mixed symptoms when scoring treatment yes/no): B: − 0.074, 95% CI − 0.15; − 0.0022,
> + 0.5 om a mood scale [patients with bipolar disorder p = 0.043) compared with patients with bipolar disorder
type I: 25.00% (95% CI 4.56; 54.56) vs. patients with bipo- type I. There were no differences in mood instability fac-
lar disorder type II: 16.66% (95% CI 5.95; 26.18)]. tor, mood instability factor for mania, mood symptomatic
Patients with bipolar disorder type II had a statistically factor for depression, mood intensity factor depression
significantly lower mean sleep duration per night (hours) and mood instability score between patients with bipolar
compared with patients with bipolar disorder type I disorder type I and bipolar disorder type II.
(6.54 (95% CI 6.18; 6.90) vs. 7.27 (95% CI 6.98; 7.57) vs.,
p = 0.036). Discussion
Apart from one previous study by the authors (Faurholt-
Bipolar disorder subtype and mood instability Jepsen et al. 2015a), this is the first study investigating dif-
Unadjusted and adjusted regression models on indexes ferences in daily smartphone-based self-monitored data
reflecting mood instability measures are presented in collected over 9  months on illness activity in patients
Table 2. As hypothesized, patients with bipolar disorder with bipolar disorder type I and bipolar disorder type
type II had statically significantly higher mood instabil- II. As hypothesized we found that patients with bipo-
ity factor for depression (models adjusted for age, gender, lar disorder type II experienced higher mood instability
illness duration and psychopharmacological treatment for depression compared with patients with bipolar dis-
(anticonvulsant treatment yes/no, antipsychotic treat- order type I, but lower intensity of manic symptoms. In
ment yes/no, antidepressant treatment yes/no, lithium addition, we found that in the entire sample regardless
treatment yes/no): B: 0.11, 95% CI 0.0085; 0.21, p = 0.034) the bipolar disorder subtype, the patients had a mean
compared with patient with bipolar disorder type I. mood level below zero during the 9-month study period
There was no significant correlation between mood and reported mixed symptoms a large proportion of
instability factor for depression and self-monitored sleep time when experiencing levels of depression or mania.

Table 2  Differences in mood instability indexes reflecting self-monitored illness activity in patients with bipolar disorder
type I and type II using smartphones (bipolar disorder type I serve as reference), N = 84
Model ­1a Model ­2a Model ­3a
B 95% CI p B 95% CI p B 95% CI p

A: Mood instability factor (MIF) − 0.19 − 0.50; 0.025 0.51 − 0.25 − 0.81; 0.32 0.38 − 0.33 − 0.91; 0.26 0.27
B: Mood instability factor for depression (MIFD) 0.24 − 0.025; 0.50 0.076 0.25 0.011; 0.51 0.061 0.27 0.007; 0.53 0.044
C: Mood instability factor for mania (MIFM) − 0.11 − 0.46; 0.24 0.53 − 0.15 − 0.50; 0.20 0.40 − 0.20 − 0.60; 0.20 0.32
D: Mood symptomatic factor for depression (MSFD) 0.30 − 0.37; 0.98 0.37 0.26 − 0.39; 0.92 0.43 0.26 − 0.39; 0.92 0.26
E: Mood symptomatic factor for mania (MSFM) − 0.24 − 0.49; 0.025 0.076 − 0.24 − 0.52; 0.029 0.080 − 0.33 − 0.64; − 0.021 0.037
F: Mood intensity factor for depression (MIntFD) 0.36 − 0.72; 1.43 0.51 0.29 0.75; 1.34 0.58 0.29 0.75; 1.34 0.58
G: Mood intensity factor for mania (MIntFM) − 0.24 0.50; 0.025 0.076 − 0.24 − 0.50; 0.023 0.073 − 0.29 − 0.56; − 0.10 0.042
H: Mood instability score (MIS) 0.037 − 0.53; 0.61 0.90 − 0.064 − 0.66; 0.53 0.83 − 0.064 − 0.66; 0.53 0.83
Italic values indicate statistical significance level (p value)
A: MIF = number of mood changes/number of weeks followed; B: MIFD = number of mood changes of depressive polarity/number of weeks followed; C:
MIFM = number of mood changes of manic polarity/number of weeks followed; D: MSFD = number of days with depressive symptoms < − 0.5 on a mood scale from
− 3 to + 3/number of weeks followed; E: MSFM = number of days with manic symptoms > 0.5 on a mood scale from − 3 to + 3/number of weeks followed; F: MIntFD:
the summaric score of depressive symptoms/number of weeks followed; G: MintFM: the summaric score of manic symptoms/number of weeks followed; H: MIS: scores
were computed for each user by applying the root mean square successive difference (rMSSD) method, taking the square root of the sum of the squared differences
between daily and previous day mood scores
a
  Model 1: unadjusted. Model 2: adjusted for age and gender. Model 3: adjusted for age, gender, illness duration and psychopharmacological treatment
(anticonvulsant treatment yes/no, antipsychotics treatment yes/no, antidepressant treatment yes/no, lithium treatment yes/no)
Faurholt‑Jepsen et al. Int J Bipolar Disord (2019) 7:5 Page 6 of 8

Regardless the bipolar subtype, the patients were able Throughout the present study period all patients had
to validly evaluate their level of depressive and manic continuous contact to a study nurse through a bi-direc-
symptoms. tional feedback loop based on changes in the smart-
The finding that patients with bipolar disorder type II phone-based self-monitored data incorporated in the
experienced more instability for depressive symptoms MONARCA II system (Faurholt-Jepsen et  al. 2014).
compared with patients with bipolar disorder type I is However, despite this rather intensive intervention they
in line with findings from other studies including a study presented with subsyndromal levels of symptoms dur-
from our group suggesting that bipolar disorder type ing large proportions of time, reflecting the serious and
II is not simply a less serious subtype of bipolar disor- chronic nature of bipolar disorder.
der (Bopp et  al. 2010; Judd et  al. 2003; Joffe et  al. 2004; In contrast to our hypotheses and in contrast to find-
Kupka et  al. 2007; Vinberg et  al. 2017). During the last ings from our previous study patients with bipolar disor-
decades there has been an increased shift from a focus der type II had a lower mean mood level or experienced
on awareness of the impact of inter-episodic mood insta- higher levels and intensity of depressive symptoms com-
bility (MacQueen et  al. 2003; Harrison et  al. 2016), and pared with patient with bipolar disorder type I (Faurholt-
a substantial proportion of patients with bipolar disorder Jepsen et al. 2015a). The findings on mood instability for
experience subsyndromal mood swings on a daily basis depression and mood symptomatic factor for mania from
even though they overall seem to be in remission. Mood the present study are in line with findings from our pre-
instability is associated with poor prognostic factors vious study. However, in contrast to our previous study,
including impaired functioning, increased risk of hospi- the present patients with bipolar disorder type II did not
talization, high risk of relapse, and comorbid substance experience higher mood symptomatic factor for depres-
use disorder (Bopp et al. 2010; Judd et al. 2003; Joffe et al. sion and mood intensity factor for depression factor. Dif-
2004; Kupka et  al. 2007; Strejilevich et  al. 2013; Patel ferences in findings between the two studies may be due
et al. 2015; O’Donnell et al. 2018; Gershon and Eidelman to differences in the included samples in the two studies.
2015). The patients in the present study had a mean ill- Patients in our previous study were recruited early during
ness duration of 18.0  years (SD 10.1) and several prior their course of treatment, whereas patients in the present
depressive and manic episodes. Studies investigating report had received two full years of outpatient treatment
whether differences in mood instability between bipolar comprising combined evidence-based psychopharma-
disorder type I and bipolar disorder type II are intrinsic cological treatment and supporting therapy, including
to bipolar disorder or a consequence of illness progres- psychoeducation at the specialized outpatient clinic. In
sion have not been conducted. However, illness duration our previous study a total of 27.2% of the patients were
was included as a possible confounder in the statisti- prescribed with antidepressants and 57.6% were pre-
cal analyses in the present study and did not alter the scribed with lithium, whereas in the present study a total
estimates. Smartphone-based monitoring of mood and of 22.6% were prescribed with antidepressants and 63.1%
mood instability in high-risk groups such as the adoles- were prescribed with lithium. In addition, a larger pro-
cent and emerging adult offspring of parents with bipolar portion of the patients in the present study were in full
disorder may be a fruitful new direction of early identifi- time employment compared with the patients from the
cation and characterization of the prospective course of previous study (17.7% vs. 8.6%).
mood instability (Duffy et al. 2018; Kessing et al. 2017).
Mood instability may predict worse long-term clini- Limitations
cal course and functioning beyond what is predicted by Some limitations to the present report should be men-
prior and current depressive and/or (hypo)manic epi- tioned. First, the patients included in the study were
sodes (Strejilevich et al. 2013; McKnight et al. 2017) sup- recruited as part of a larger RCT and thus not recruited
porting the idea that different measures of subsyndromal according the bipolar disorder subtype. The unequal
illness activity may be a more sensitive outcome meas- group size between the two disorders was therefore not
ure in maintenance RCTs than for example time to first intended. The statistical power for the study was calcu-
relapse or recurrence of affective episodes (Bopp et  al. lated for the original RCT and consequently power cal-
2010; Bonsall et  al. 2012; Saunders et  al. 2016; Broome culations were not conducted for the present report
et al. 2015). Thus, there is a need to identify and monitor (Faurholt-Jepsen et  al. 2015a). We did not correct for
inter-episodic symptoms giving the opportunity to inter- multiple testing since the study is considered as hypoth-
vene. Innovative RCT’s investigating the effect of treat- eses generating and the results needs further replication.
ment interventions on mood instability are ongoing and However, it should be noted that if Bonferroni correc-
will hopefully provide more insight to this area (Saunders tions were applied none of the models would be statis-
et al. 2016). tically significant. Second, regarding the generalizability
Faurholt‑Jepsen et al. Int J Bipolar Disord (2019) 7:5 Page 7 of 8

of the results, the patients included in the present study using the same smartphone-based self-monitoring sys-
had previously been treated at The Copenhagen Clinic tem have showed neither positive nor negative effects
for Affective Disorders, Denmark which is a specialized of smartphone-based self-monitoring, and that the self-
outpatient clinic in the period 2004 to January 2016. monitored mood levels correlated statistically significant
Along this line, it cannot be excluded that the patients with clinically rated depressive and manic symptoms
who agree to participate in the present study may repre- measured using the Hamilton Depression Rating Scale
sent a subgroup of patients with bipolar disorder. Third, and the Young Mania Rating Scale (Faurholt-Jepsen et al.
the patients presented with rather low levels of depres- 2015b, 2015c).
sive and manic symptoms, and received different types,
doses and combinations of psychopharmacological Perspectives
treatment during the study. Since a higher proportion Smartphones are ubiquitous and reflect an unobtrusive
of patients with bipolar disorder type I received antip- way of collecting fine-grained real-time data during natu-
sychotics compared with patients with bipolar disorder ralistic settings on the course of illness and detailed char-
type II, the statistical analyses included the different psy- acterization of prodromal and subsyndromal symptoms
chopharmacological treatments prescribed (anticonvul- of depression and (hypo)mania in patients with bipolar
sant treatment yes/no, antipsychotic treatment yes/no, disorder. Measures of mood instability may be a prom-
antidepressant treatment yes/no, lithium treatment yes/ ising measure of inter-episodic illness activity in bipolar
no) as covariates. However, residual confounding due to disorder and may be a more sensitive outcome meas-
differences in psychopharmacological treatment between ure in maintenance RCTs than for example time to first
the two groups cannot be excluded. It is likely that differ- relapse or recurrence of affective episodes since patients
ences in psychopharmacological treatment (antipsychot- continue to experience subsyndromal mood swings.
ics) may affect differences in mood instability and should
be investigated further in larger studies. Fourth, to date
there are no clear consensus on the definition, measure- Conclusions
ment, use and reporting of mood instability in patients Patients with bipolar disorder type II experienced higher
with bipolar disorder, and thus the measures used in the mood instability for depression compared with patients
present report were a combination of measures used with bipolar disorder type I. Patients with bipolar disor-
by others and measures used by the authors in a previ- der type I and II experienced mixed symptoms a large
ous study (Strejilevich et  al. 2013; Faurholt-Jepsen et  al. proportion of time. These findings emphasize the need to
2015a; O’Donnell et  al. 2018; Ebner-Priemer and Trull monitor, identify and treat subsyndromal inter-episodic
2009). Using, defining, measuring and reporting mood symptoms. Future studies investigating the effect of treat-
instability in a standardized way across studies could ment on mood instability measures are needed.
improve the quality, reproducibility and comparison of Authors’ contributions
results between studies. Fifth, we did not use last-obser- MFJ and LVK performed the statistical analyses and wrote the first draft of the
manuscript. MFJ, MF, JB, EMC, MV and LVK contributed to the final version of
vation-carried-forward method or imputation techniques the manuscript. All authors read and approved the final manuscript.
to handle missing data since there was a high adherence
to self-monitoring during the study. The patients pre- Author details
1
 Copenhagen Affective Disorder Research Center (CADIC), Psychiatric Center
sented with rather low level of depressive and manic Copenhagen, Rigshospitalet, Blegdamsvej 9, 2100 Copenhagen, Denmark. 2 IT
symptoms and therefore data is estimated to be missing University of Copenhagen, Rued Langgaards Vej 7, 2300 Copenhagen, Den‑
at random. It cannot be excluded that patients did not mark. 3 Department of Applied Mathematics and Computer Science, Technical
University of Denmark, Kongens Lyngby, Denmark.
conduct the self-monitoring during the most severe time
point. However, the patients were able to validly evalu- Acknowledgements
ated their level of depressive and manic symptoms. Inves- The authors would like to thank the patients for participating in the study and
nurse Bente Støyer for assisting the patients in using the MONARCA II system.
tigating the validity of self-monitored symptoms during
more severe affective episodes could be a target for future Competing interests
studies. MFJ and JB have no competing interests. MF and JEB are co-founders and
shareholders in Monsenso ApS. EMC has been a consultant for Eli Lilly, Astra
Sixth, it may be that systematic self-monitoring and Zeneca, Servier, Bristol-Myers Squibb, Lundbeck, and Medilink. MV has been a
reporting of mood could potentially in itself affect mood consultant for Lundbeck within the last 3 years. LVK has been a consultant for
and mood awareness. Along this line, the risk of patients Sunovion.
using the smartphone-based self-monitoring system sys- Availability of data and materials
tematically reported lower mood, to gain attention from Data are managed by the authors.
the study nurse leading to non-random measurement
Consent for publication
error, must be considered. However, our previous studies Not applicable.
Faurholt‑Jepsen et al. Int J Bipolar Disord (2019) 7:5 Page 8 of 8

Ethics approval and consent to participate Gershon A, Eidelman P. Affective intensity and instability: predictors of depres‑
Ethical permission was obtained from the Regional Ethics Committee in The sion and functional impairment in bipolar disorder. J Behav Ther Exp
Capital Region of Denmark and The Danish Data Protection Agency (H-2- Psychiatry. 2015;46:14–8.
2014-059). The authors assert that all procedures contributing to this work Hamilton M. Development of a rating scale for primary depressive illness. Br J
comply with the ethical standards of the relevant national and institutional Soc Clin Psychol. 1967;6:278–96.
committees on human experimentation and with the Helsinki Declaration of Harrison PJ, Cipriani A, Harmer CJ, Nobre AC, Saunders K, Goodwin GM, et al.
1975, as revised in 2008. All participants consented to participate. Innovative approaches to bipolar disorder and its treatment. Ann NY
Acad Sci. 2016;1366:76–89.
Funding Jahng S, Wood PK, Trull TJ. Analysis of affective instability in ecological
The MONARCA II trial received financial support from The Mental Health Ser‑ momentary assessment: indices using successive difference and group
vices in the Capital Region of Denmark including the Centre for Telemedicine comparison via multilevel modeling. Psychol Methods. 2008;13:354–75.
in the Capital Region of Denmark and the EU 7th Frame Programme, the IT Joffe RT, MacQueen GM, Marriott M, Trevor Young L. A prospective, longitu‑
University of Copenhagen and the Lundbeck Foundation. The research team dinal study of percentage of time spent ill in patients with bipolar I or
has no financial connection to any of the contributors. The funding organiza‑ bipolar II disorders. Bipolar Disord. 2004;6:62–6.
tions had no role in the design and conduct of the study; collection, manage‑ Judd LL, Schettler PJ, Akiskal HS, Maser J, Coryell W, Solomon D, et al. Long-
ment, analysis, and interpretation of the data; preparation, review, or approval term symptomatic status of bipolar I vs. bipolar II disorders. Int J Neu‑
of the manuscript; and decision to submit the manuscript for publication. ropsychopharmacol. 2003;6:127–37.
Kessing LV, Hansen HV, Hvenegaard A, Christensen EM, Dam H, Gluud C, et al.
Treatment in a specialised out-patient mood disorder clinic v. standard
Publisher’s Note out-patient treatment in the early course of bipolar disorder: randomised
Springer Nature remains neutral with regard to jurisdictional claims in pub‑ clinical trial. Br J Psychiatry. 2013;202:212–9.
lished maps and institutional affiliations. Kessing LV, Munkholm K, Faurholt-Jepsen M, Miskowiak KW, Nielsen LB, Frikke-
Schmidt R, et al. The bipolar illness onset study: research protocol for the
Received: 7 November 2018 Accepted: 8 January 2019 BIO cohort study. BMJ Open. 2017;7:e015462.
Kupka RW, Altshuler LL, Nolen WA, Suppes T, Luckenbaugh DA, Leverich GS,
et al. Three times more days depressed than manic or hypomanic in both
bipolar I and bipolar II disorder. Bipolar Disord. 2007;9:531–5.
MacQueen GM, Marriott M, Begin H, Robb J, Joffe RT, Young LT. Subsyndromal
References symptoms assessed in longitudinal, prospective follow-up of a cohort of
Bonsall MB, Wallace-Hadrill SMA, Geddes JR, Goodwin GM, Holmes EA. Non‑ patients with bipolar disorder. Bipolar Disord. 2003;5:349–55.
linear time-series approaches in characterizing mood stability and mood McKnight RF, Bilderbeck AC, Miklowitz DJ, Hinds C, Goodwin GM, Ged‑
instability in bipolar disorder. Proc Biol Sci. 2012;279:916–24. des JR. Longitudinal mood monitoring in bipolar disorder: course of
Bopp JM, Miklowitz DJ, Goodwin GM, Stevens W, Rendell JM, Geddes JR. The illness as revealed through a short messaging service. J Affect Disord.
longitudinal course of bipolar disorder as revealed through weekly text 2017;223:139–45.
messaging: a feasibility study. Bipolar Disord. 2010;12:327–34. O’Donnell LA, Ellis AJ, de Loo MMV, Stange JP, Axelson DA, Kowatch RA, et al.
Broome MR, Saunders KEA, Harrison PJ, Marwaha S. Mood instability: signifi‑ Mood instability as a predictor of clinical and functional outcomes
cance, definition and measurement. Br J Psychiatry. 2015;207:283–5. in adolescents with bipolar I and bipolar II disorder. J Affect Disord.
Conner TS, Barrett LF. Trends in ambulatory self-report: the role of momen‑ 2018;236:199–206.
tary experience in psychosomatic medicine. Psychosom Med. Patel R, Lloyd T, Jackson R, Ball M, Shetty H, Broadbent M, et al. Mood instability
2012;74:327–37. is a common feature of mental health disorders and is associated with
Duffy A, Keown-Stoneman C, Goodday SM, Saunders K, Horrocks J, Grof P, et al. poor clinical outcomes. BMJ Open. 2015;5:e007504.
Daily and weekly mood ratings using a remote capture method in high- Saunders KEA, Cipriani A, Rendell J, Attenburrow M-J, Nelissen N, Bilderbeck
risk offspring of bipolar parents: compliance and symptom monitoring. AC, et al. Oxford Lithium Trial (OxLith) of the early affective, cognitive,
Bipolar Disord. 2018. https​://doi.org/10.1111/bdi.12721​. neural and biochemical effects of lithium carbonate in bipolar disorder:
Ebner-Priemer UW, Trull TJ. Ecological momentary assessment of mood disor‑ study protocol for a randomised controlled trial. Trials. 2016;17:16.
ders and mood dysregulation. Psychol Assess. 2009;21:463–75. Shiffman S, Stone AA, Hufford MR. Ecological momentary assessment. Annu
Faurholt-Jepsen M, Vinberg M, Frost M, Christensen EM, Bardram J, Kessing Rev Clin Psychol. 2008;4:1–32.
LV. Daily electronic monitoring of subjective and objective measures of Strejilevich SA, Martino DJ, Murru A, Teitelbaum J, Fassi G, Marengo E, et al.
illness activity in bipolar disorder using smartphones—the MONARCA II Mood instability and functional recovery in bipolar disorders. Acta Psychi‑
trial protocol: a randomized controlled single-blind parallel-group trial. atr Scand. 2013;128:194–202.
BMC Psychiatry. 2014;14:309. Vinberg M, Mikkelsen RL, Kirkegaard T, Christensen EM, Kessing LV. Differ‑
Faurholt-Jepsen M, Ritz C, Frost M, Mikkelsen RL, Margrethe Christensen E, ences in clinical presentation between bipolar I and II disorders in the
Bardram J, et al. Mood instability in bipolar disorder type I versus type early stages of bipolar disorder: a naturalistic study. J Affect Disord.
II-continuous daily electronic self-monitoring of illness activity using 2017;208:521–7.
smartphones. J Affect Disord. 2015a;186:342–9. Wing JK, Babor T, Brugha T, Burke J, Cooper JE, Giel R, et al. SCAN: sched‑
Faurholt-Jepsen M, Frost M, Ritz C, Christensen EM, Jacoby AS, Mikkelsen RL, ules for clinical assessment in neuropsychiatry. Arch Gen Psychiatry.
et al. Daily electronic self-monitoring in bipolar disorder using smart‑ 1990;47:589–93.
phones—the MONARCA I trial: a randomized, placebo-controlled, single- Young RC, Biggs JT, Ziegler VE, Meyer DA. A rating scale for mania: reliability,
blind, parallel group trial. Psychol Med. 2015b;45:2691–704. validity and sensitivity. Br J Psychiatry. 1978;133:429–35.
Faurholt-Jepsen M, Vinberg M, Frost M, Christensen EM, Bardram JE, Kessing LV.
Smartphone data as an electronic biomarker of illness activity in bipolar
disorder. Bipolar Disord. 2015c;17(7):715–28. https​://doi.org/10.1111/
bdi.12332​