Differences in beliefs about mood between people with and without bipolar disorder

Abstract Psychological models of bipolar disorder (BD), such as the self‐regulation model (SRM; Leventhal, Nerenz, & Steele, 1984), highlight the crucial role of beliefs about mood in relapse vulnerability. To date, no studies have directly compared these beliefs between people with and without BD. Based on the SRM, the current research examined beliefs about mood in people with and without BD and explored the impact of current affect on these beliefs. Fifty euthymic people with a diagnosis of BD and 50 controls were recruited through an online screening study, clinical services, and support organizations. Experience sampling methodology (ESM) was used to assess beliefs (according to the Brief Illness Perceptions Questionnaire; Broadbent, Petrie, Main, & Weinman, 2006) across a typical week of everyday life. Data were analysed using multilevel modelling. Forty‐two people with a diagnosis of BD and 50 controls were included in the analyses. Results indicated that the BD group reported less control over mood, a shorter duration of mood, and less understanding of mood and were more likely to report the cause of depressive symptoms as something internal, compared with controls. When controlling for current affect, the BD group also reported more positive consequences, made more internal attributions for hypomanic symptoms, and reported less concern about mood, compared with controls. Findings suggest important differences in beliefs about mood between people with and without BD that are not the result of current affect. These beliefs may be particularly important in understanding underlying vulnerability to future relapse into depression and/or mania.

modalities) can become disrupted through circadian rhythm changes (e.g., insomnia, jetlag, drug/alcohol use, and daylight changes) and that internal appraisals of this disruption (held at the schematic, propositional, and associative levels) may increase vulnerability to BD onset and recurrence (Healy & Williams, 1989;Jones, 2001). Based on this model, the Hypomanic Interpretations Questionnaire (HIQ; Jones, Mansell, & Waller, 2006) and the Interpretations of Depression Questionnaire (IDQ; Jones & Day, 2008) were developed to measure the tendency for people to believe that hypomanic or depressive symptoms respectively were linked to something internal/related to the self rather than external factors. For example, "If I felt in high spirits and full of energy, I would probably think it was because I am a talented person with lots to offer" (self-dispositional hypomanic appraisal) or "If I felt down on myself, I would probably think it was because I am a bad person" (self-dispositional depressive appraisal). Previous studies have found that people with BD score significantly higher on the HIQ and IDQ compared with controls (Banks, Lobban, Fanshawe, & Jones, 2016;Jones et al., 2006; and that this may be a vulnerability factor in BD (Jones & Day, 2008;Jones, Hare, & Evershed, 2005).
The importance of beliefs about changes in internal state (including mood state) is also highlighted in the Integrative Cognitive Model (ICM; Mansell, Morrison, Reid, Lowens, & Tai, 2007). The ICM suggests that beliefs that a change in mood has extreme personal meaning provoke exaggerated efforts to control that mood change. Such efforts exacerbate the initial mood change, resulting in a vicious cycle of mood intensification.
The Hypomanic Attitudes and Positive Predictions Inventory (HAPPI; Mansell, 2006) was developed to measure the extreme, personal positive and negative beliefs about internal states central to ICM. A number of studies have reported an association between scores on the HAPPI and BD-related experiences in nonclinical samples (Dodd, Mansell, Beck, & Tai, 2013;Dodd, Mansell, Bentall, & Tai, 2011;Mansell, Rigby, Tai, & Lowe, 2008). Preliminary evidence has been found for the predictive validity of the HAPPI for BD symptoms in a clinical sample (Dodd, Mansell, Morrison, & Tai, 2011), and people with BD have been found to score higher on the HAPPI than matched nonclinical and unipolar samples, while controlling for current symptoms (e.g., Alatiq, Crane, Williams, & Goodwin, 2010;Mansell, 2006;Mansell et al., 2011;. The self-regulation model (SRM; Leventhal et al., 1984) is a theoretical framework that attempts to explain how people use coping behaviours to try to bring their current health state in line with their desired health state. The beliefs people hold about their illness experiences guide how coping strategies are employed, which subsequently influences outcome. The SRM has proved to be helpful for understanding illness beliefs and their association with outcomes in physical health, providing useful information to inform treatment (Hagger & Orbell, 2003), and has since been applied to mental health problems (Baines & Wittkowski, 2012;Lobban, Barrowclough, & Jones, 2004;Peay, Rosenstein, & Biesecker, 2013). The Brief Illness Perception Questionnaire (BIPQ; Broadbent et al., 2006) is a measure of the illness beliefs proposed by the SRM. In a recent cross-sectional study (Dodd, Mezes, Lobban, & Jones, 2017), negative beliefs about mood swings were shown to be associated with poorer personal recovery in BD. However, only one study to date has used the BIPQ to explore the impact of beliefs about mood on outcomes over time in BD. In a study of 91 BD participants, Lobban, Solis-Trapala, Tyler, Chandler, and Morriss (2013) found that illness beliefs had important effects on symptomatic outcomes over a 24-week period. Specifically, higher perceived consequences, more symptoms experienced (identity), and greater concern about mood were predictive of time to relapse over 24 weeks. Additionally, higher perceived consequences and lower perceived effort to get well were associated with higher weekly depressive symptom scores after controlling for baseline depression, medication, and number of previous episodes. Although this study indicates that beliefs about mood may be a vulnerability factor for relapse, it is not clear whether these beliefs could be a vulnerability factor specifically for BD or mood fluctuations in general. Furthermore, despite beliefs about consequences being a strong predictor of outcome in this, and numerous other non-BD studies (Broadbent et al., 2015), the extent to which positive and negative consequences separately predicted outcome was not examined. Therefore, we cannot determine whether beliefs about positive and negative consequences lead to different outcomes.
All of these models hypothesize that beliefs about mood may be a vulnerability factor for BD. We aimed to explore whether such beliefs have a specific role in determining mood changes in people with BD compared with controls and therefore may be a vulnerability factor for extreme fluctuations in mood. We chose to use the SRM as our theoretical framework as this is inherently transdiagnostic in its development and application and therefore can be applied to people in both groups of the study.

Key Practitioner Message
• Specific beliefs about mood differ between people with and without bipolar disorder (BD), a subset of which are unaffected by changes in current affect and may therefore represent a core belief set in BD.
• When controlling for current affect, people with a diagnosis of BD reported beliefs that may affect escalation to hypomania.
• Psychoeducational and cognitive-behavioural approaches should include techniques to address internal appraisal mechanisms for mood changes. Recovery-focussed therapy could provide a way to do this. In order to examine differences in beliefs between people with and without BD in real time, during a typical week of everyday life and in the natural environment, we employed Experience Sampling Methodology (ESM). People with BD are more likely to be in lower affective state and show stronger fluctuations in affect, than those without BD, even when not in episode (Judd et al., 2002;Judd et al., 2003;Knowles et al., 2007). Therefore, to test whether differences in current affect impact on underlying beliefs about mood, we assessed beliefs with and without controlling for current affect. Our aims were 1. To identify differences in beliefs about mood between people with and without BD across a typical week of everyday life, using ESM.

| Sample
A convenience sample of 50 people with a diagnosis of BD and 50 controls were recruited through an online screening study advertised using online adverts on internet forums and media sites, and poster and newsletters circulated around universities across the North West of England (Banks et al., 2016). Additional participants were recruited directly through local clinical services and support organizations. BD I or II was confirmed using Structured Clinical Interview for DSM-IV (SCID; First, Spitzer, Gibbon, & Williams, 1997). Control participants were required to score < 0.5 standard deviations above the sample mean on the Hypomanic Personality Scale (HPS; Eckblad & Chapman, 1986), to exclude people at high behavioural risk for BD (Kwapil et al., 2000). Additional eligibility criteria for both groups included • Aged 18 years old or over.
• No current manic, hypomanic, mixed affective, or major depressive episode currently or within the 4 weeks prior to baseline assessment assessed by SCID-Life (First et al., 1997) scores.
• No physical brain injury.
• No current suicide plans or high suicide intent.
• Able and willing to give written informed consent to the study.
• Able to communicate in written and oral English to a sufficient level to allow the participant to complete the measures.
• Not a night shift worker.

| Experience sampling methodology (ESM)
ESM refers to a set of empirical methods that allow participants to respond to questions within the context of their daily lives, following a specific event or at a specific time (often signalled by a watch or mobile device; Christensen, Barrett, Bliss-Moreau, Lebo, & Kaschub, 2003). Thoughts, feelings, and experiences are captured in detail (through repeated assessments often using a paper diary or mobile device), as they occur (momentary assessments of current state), and in the context in which they occur (increasing ecological validity).
ESM is a well-established approach for investigating processes in mental health research. ESM has been employed to investigate crosssectional (Myin-Germeys, Delespaul, & van Os, 2003) and temporal (Pavlickova et al., 2013) associations between variables in BD, as well as biological markers (Havermans, Nicolson, Berkhof, & deVries, 2011) and as a technique for intervention (Depp et al., 2010). However, ESM has not yet been used to examine the types of beliefs measured by the BIPQ, which has generally relied on cross sectional self-report and therefore does not address the dynamic nature of BD and neglects contextually and temporally sensitive information such as changes in current affect. By using multiple, frequent assessments of momentary phenomena, ESM can provide an ecologically valid, real-time method for examining the inherently dynamic nature of BD that traditional assessments fail to capture. As with the majority of ESM studies using BD samples (e.g., Fulford, Johnson, Llabre, & Carver, 2010;Havermans et al., 2011;Havermans, Nicolson, Berkhof, & deVries, 2010;Havermans, Nicolson, & Devries, 2007;Knowles et al., 2007;Myin-Germeys et al., 2003;Pavlickova et al., 2013), we used momentary paper diary assessments in real time triggered by a watch/mobile phone signal. To ensure accurate completion of diaries, participants recorded the time at which they completed each diary entry and were instructed to complete the diary as soon as possible following an alert (within no more than 10 min). Where a response was missed, participants were asked to provide a reason and to wait until the next alert to respond. The number of diary items selected was consistent with good practice in ESM research (Palmier-Claus et al., 2011). See Section 2.4 for more detail.

| Measures
Screening measures (see Figure 1) The HPS (Eckblad & Chapman, 1986) was used to screen controls (participants who scored < 0.5 standard deviations above the sample mean on the HPS).
The SCID (First et al., 1997) was used to confirm that the clinical sample met research diagnostic criteria for BD I or II to verify any additional DSM-IV Axis I psychological disorders and to screen control participants. SCID interviews were administered over the telephone by a trained researcher to allow participants across the North of England to participate.
The SCID Life (First et al., 1997) was administered at baseline to assess participants' mood during the previous 4 weeks. It incorporates the Hamilton Depression Rating Grid (Williams et al., 2008) and the Bech-Rafaelsen Mania Scale (Bech, Rafaelsen, Kramp, & Bolwig, 1978), which also provided measures of depressive or manic symptoms at baseline.

ESM Diary (available on request)
Affect assessment: Momentary affect assessment diary items were drawn from items used in previous ESM research assessing positive and negative affect in clinical samples (e.g., Havermans et al., 2010) and validated measures of affect (e.g., Positive and Negative Affect Schedule (Watson, Clark, & Tellegen, 1988) and the Internal State Scale (Bauer et al., 1991)). Item ratings for "cheerful," "energetic," and "confident" were averaged to assess positive affect, and "bad about myself," "down," and "guilty" were averaged to form a negative affect scale. All were rated on a 7-point Likert scale (1 to 7 [very]).
The BIPQ (Broadbent et al., 2006) was used to measure beliefs about mood. The BIPQ was adapted for use with a control, as well as clinical, sample by omitting the treatment control dimension (as this is not relevant to the former group) and emotional representation (due to its overlap with our focus of interest on mood; see Appendix A). The following dimensions were included: identity (current affect), beliefs about consequences (positive and negative), personal control, concern, comprehensibility, and duration (time line).
Beliefs about the cause of current mood were assessed using single items from the HIQ  Item "If I felt in high spirits and full of energy, I would probably think it was because I'm a talented person with lots to offer") and the IDQ (Jones & Day, 2008; "If I felt down on myself, I would probably think it was because I am a bad person, even towards myself") to allow multilevel modelling analysis. These were selected based on the items with the highest loading in principal components analyses of these positive and negative appraisal measures.

| Procedure
Participants were screened (using the SCID overview leading to relevant sections of the SCID) for primary BD diagnosis (or lack of) and completed the HPS online. Following further screening questions over the phone to verify eligibility, participants were invited to take part in a baseline interview in person. Participants gave written informed consent and were screened using the SCID Life. Participants with BD who could not be included due to scores above the threshold (SCID Life score > 4 indicating more than normal mood symptoms) were offered the opportunity to take part once their symptoms had reduced.
Participants then completed sociodemographic and clinical measures (listed in Table 1) prior to a briefing session to explain use of the diaries. Participants were provided with seven A5 ESM diaries (one for each day) and an information book detailing study procedure.
Participants were alerted to the need to fill in the ESM diary by either text alerts from a mobile phone or beeps from an Ironman Data Link watch (five BD and five control participants chose to use a watch and the remaining participants used mobile phones). Text or beep alerts prompted participants to complete the same set of diary questions (see above) 10 times a day for 7 days. All alerts were programmed to occur between 7:45 a.m. and 10:15 p.m., at pseudorandom intervals. The minimum gap between alerts was 24 min and the maximum was 159 min, with an average gap of 90 min.
Participants who completed fewer than 30% of alerts within 10 min of the alert were excluded, consistent with a minimum threshold of 20 completed reports (Palmier-Claus et al., 2011). Again, consistent with good practice for ESM, participants were contacted by phone during the study week to provide support, increase motivation, and arrange a final appointment (Palmier-Claus et al., 2011). During this appointment, the researcher collected the study equipment, debriefed the participant, and gave a £10 "thank you" for their involvement.  time-since-first-response correlation term was modelled. Autocorrelation was also modelled because observations from a participant that were closer in time were more likely to be more similar than observations further apart. To estimate effects of current affect on group differences in beliefs about mood (Aim 2), the general MLM framework was adjusted for positive affect (PA) and negative affect (NA) simultaneously using two separate parameters for each affect type. Statistical significance was interpreted as p ≤ .05. The Bonferroni correction for multiple tests has been criticized for being overly conservative (Hochberg & Benjamini, 1990); to ensure a low false positive rate (< 5%), we used the more powerful false discovery rate (FDR; Glickman, Rao, & Schultz, 2014).
As age and gender did not improve the goodness-of-fit according to the Akaike information criterion value, they were not included in the final MLM. Only variables directly associated with the current hypotheses and those necessary for appropriate analysis of ESM data were included.

| Participant sample
The final sample was 92 adults with either BD (n = 42) or no history of mental health problems (control; n = 50). Eight participants with BD were excluded because they completed fewer than 30% of ESM alerts within 10 min of the signal.
The demographic and clinical characteristics of the sample are displayed inTable 1. Both groups were predominantly female and White British. BD participants were older, had less experience of higher education, were less likely to be working, were more likely to be single or divorced, and were less likely to be living with a partner than controls.
The majority of the BD group had a diagnosis of BDI with an established course of BD: Over half had experienced at least seven episodes of mania and/or depression and between 1 and 29 hospitalizations. The majority of the BD group had never received psychological treatment for mood (38%) or had received past treatment (36%).

| ESM results
There were 4,197 eligible responses (within 10 min of the alert) in total. Sixty eight per cent of alerts sent to controls and 55% of alerts sent to BD were eligible. Fifteen per cent of ineligible control alerts were completed after 10 min (23% for BD alerts); the remainder were missed completely. Explanations for missing data were provided in 66% of cases for controls and 43% for BD. Controls missed alerts most often due to work or study, whereas those with BD reported most commonly being asleep or resting.
Aim 1. Identify any differences in beliefs about mood between people with and without BD across a typical week of everyday life The BD group reported significantly less personal control, more concern, a significantly shorter expected duration of current mood, less understanding of mood, and more likely to attribute the cause of depressive symptoms to something internal compared with controls (see Table 2). These group differences remained significant after the FDR correction.

Aim 2. Examine the effect of current affect on differences in beliefs
There were statistically significant differences between groups in both positive (MD = −0.52, p = .01) and negative (MD = 0.79, p ≤ .001) affect. PA and NA were moderately negatively correlated with each other (r = −0.43, p < .001). Table 3 displays the effect of group ("Group difference," Columns 1 and 2), PA ("Positive affect," Columns 3 and 4), and NA ("Negative affect," Columns 5 and 6) on each type of belief. When current affect was controlled for, the BD group still reported less control over mood, a shorter expected duration of mood, less understanding of mood and were more likely to attribute the cause of depressive symptoms to something internal compared with controls. In addition, controlling for current affect impacted on group differences in beliefs about consequences and cause: BD participants reported significantly more positive consequences associated with mood and were more likely to attribute the cause of hypomanic symptoms to something internal compared with controls. These group differences remained significant after the FDR correction. There were no longer statistically significant group differences in concern.
In both groups, higher PA and lower NA were associated with reports of more personal control, understanding, positive consequences, and internal attributions for hypomanic symptoms. Lower PA and higher NA were associated with more concern and internal attributions for depressive symptoms. Both higher PA and higher NA were associated with beliefs about longer duration of mood.

| DISCUSSION
This study explored differences in beliefs about mood between people with and without BD across 7 days.
During a typical week, people with BD reported significantly less personal control, more concern, a shorter duration of current mood, and less understanding of mood and were more likely to attribute the cause of depressive symptoms to something internal, compared with people without BD. All of these differences, apart from greater  concern, remained when current affect was controlled for and are therefore unlikely to be due to reporting biases associated with current affect. These beliefs may be particularly important in BD because they may distinguish those who go on to experience worse clinical outcomes. Indeed, the importance of beliefs about control has been highlighted previously (e.g., Crowe et al., 2012;Mansell et al., 2007;Proudfoot et al., 2012). In a recent review and meta-analysis of the BIPQ, Broadbent et al. (2015) highlighted the importance of beliefs about personal control for outcomes in a range of psychological disorders, including BD.
The current findings regarding control have important implications both for psychological therapy and for the wider treatment of people with BD. Although perceptions of a short duration of current mood and lack of control over current mood in BD may be accurate for some people given that BD is associated with increased mood variability even in remission (Knowles et al., 2007) Whereas beliefs about control, duration, understanding, and selfdispositional depressive appraisals were not affect dependent, beliefs about positive consequences and self-dispositional hypomanic appraisals in BD were associated with PA and therefore were only revealed when low affect in BD was controlled for in analyses. Similarly, beliefs about concern were influenced by current affect. When current affect was not controlled, consistent with previous research (Lobban et al., 2013), we found increased concern in people with BD. However, concern was associated with higher NA, which was elevated in the BD group. When this was controlled for, group differences in concern disappeared. As mood increases, these less catastrophizing, more positive beliefs may become more apparent and could influence escalation into hypomania. The challenge for interventions is to acknowledge positive aspects of BD and possible risks associated with mood escalation in order to genuinely encompass both aspects into therapy. Positive aspects of BD are highly valued by some people, and the chance to talk about them is welcomed (Lobban, Taylor, Murray, & Jones, 2012). Ambivalence towards treatment is likely where the focus is only on risk, ignoring the positive aspects of BD. Indeed, there is evidence that positive illness models enhanced recovery (Dodd et al., 2017).
Taken together, current results have important clinical implications.
Structured psychological interventions are effective in reducing relapse risk in BD (Oud et al., 2016). These generally share common elements of providing information about BD, improving cognitive and behavioural coping strategies and developing action plans to address relapse signs (Miklowitz, Goodwin, Bauer, & Geddes, 2008).
Although it is clear that such approaches should be helpful in providing coherent illness models and improving sense of control, most of these approaches do not explicitly address internal appraisal mechanisms for mood changes or address the positive as well as negative aspects of mood. Enhancement of psychoeducational and cognitive behavioural approaches to include techniques to address these factors should, based on SRM theory, promote increased therapeutic effectiveness.
Indeed, recovery-focussed therapy includes these elements and does show promise in both reducing relapse risk and improving functional and recovery outcomes in BD (Jones et al., 2015).

| Limitations and future directions
This study benefits from being theoretically driven by the SRM and diagnostically valid due to the completion of SCID diagnostic interviews. ESM allowed for multiple, momentary assessments during a typical week of everyday life, thus we collected a rich amount of data per participant and ensured high ecological validity. However, there are also limitations that should be considered in interpreting the findings.
First, as with most research studies, this study recruited a specific group of people with BD, generally through NHS secondary care.
Therefore, caution should be paid to the generalizability of the current findings to the general population of people with BD. In addition to generalizability, the specificity of current findings should be tested in future research using a clinical control group matched on symptomatology to allow examination of beliefs specific to BD.
Second, groups differed on a number of demographic variables and clinical comorbidities. However, there is no evidence that these variables would explain differences in beliefs between people with and without BD. The majority of the BD group were taking some form of psychiatric medication, whereas the control group were not. It could be argued that medication could confound any group differences found by altering affect in the BD sample only; however, this was controlled for in the second set of analyses and so is very unlikely.
Finally, this research did not establish any causal links between beliefs and affect. Future research should explore whether changing these beliefs can impact on outcomes in BD, such as relapse.
In conclusion, these findings are consistent with previous literature and psychological models such as the SRM, which suggests that underlying beliefs about mood play a crucial role in mood fluctuations in BD. The current study builds on existing research literature by highlighting important differences in beliefs about mood between people with and without BD, which are not the result of current affect and which therefore may explain variation in outcome. Prospective studies are needed to examine the impact beliefs about mood have on important outcomes in BD, such as relapse.