Looking for bipolar spectrum psychopathology: identification and expression in daily life

6.1. Goals and hypotheses

Study 1 examined the relation of bipolar spectrum psychopathology (as measured by the HPS) with interview and questionnaire measures of psychopathology, personality, and functioning. It was hypothesized that HPS scores would be associated with lifetime diagnoses of DSM-IV-TR bipolar episodes and disorders and the bipolar spectrum disorders of Akiskal [4] as well as global ratings of impaired functioning. Bipolar spectrum psychopathology was hypothesized to be associated with interview-based ratings of borderline personality disorder symptoms, increased substance use and abuse, history of treatment of bipolar psychopathology, and family history of bipolar disorders. On the questionnaire measures, it was hypothesized that bipolar spectrum psychopathology would be associated with affective temperaments of hyperthymia, cyclothymia, and irritability; with personality domains of extraversion and openness to experience; and with impulsivity. We expected these findings to remain when participants with DSM-IV-TR bipolar disorders were removed from the analyses, consistent with a spectrum model. Note that we hypothesized that bipolar spectrum psychopathology would be associated with depressive symptoms but that this relation would not remain significant after omitting participants with DSM-IV-TR bipolar disorders.

7.1. Participants

Approximately 1200 students enrolled in general psychology courses completed the HPS in mass-screening sessions during 3 consecutive semesters. College students are an appropriate sample for examining bipolar characteristics, given that bipolar psychopathology frequently first emerges in late adolescence and early adulthood [6] (the age of the sample in the present study), rates of bipolar disorder are relatively comparable among college-educated and non– college-educated samples [34], and numerous studies have examined bipolar characteristics in student samples (eg, Meyer and Hautzinger [24]). A total of 191 students were invited to participate in study 1. Specifically, all of the mass-screening participants who scored at least 1.5 SD above the mean on the HPS and a comparable number of randomly selected participants who scored less than 1.5 SD above the mean were invited to participate. This recruitment strategy was designed to ensure that a sufficient number of individuals with bipolar spectrum psychopathology were included in the study, while maintaining a continuous distribution. A total of 147 participants took part in study 1 (and, subsequently, in study 2). Two participants were dropped because of invalid questionnaire measures. Neither age nor sex was significantly correlated with HPS scores (r = −0.09 and −0.02, respectively). The final sample included 100 women and 45 men. Mean age was 19.5 years (SD, 2.3 years). The sample was 65% white, 16% African American, 4% Hispanic, 4% Asian/Pacific Islander, 4% other, and 7% unspecified.

7.2. Materials and procedures

8.1. Hypomanic Personality Scale data

Participants completed the HPS at mass screening and at the time of the interview (2–12 weeks later; mean, 5.5 weeks). Hypomanic Personality Scale scores were examined at both time points (mass screening: mean, 22.6; SD, 11.0; range, 3–42; interview: mean, 17.5; SD, 10.0; range, 0–41). The lower mean HPS score at the second assessment likely reflected regression to the mean (given the selection procedure). Hypomanic Personality Scale scores were strongly correlated across the 2 time points (intraclass correlation, 0.85; P < .001), consistent with the test-retest reliability of Eckblad and Chapman [22] of 0.81 across 15 weeks. Therefore, participants were assigned an average HPS score that was used for all subsequent analyses. Note that the results were unchanged when using the average or individual scores.

8.2. Relation of bipolar spectrum psychopathology with interview and questionnaire measures

Thirty interviews (21%) were independently rated by both interviewers to assess interrater reliability. Intraclass correlations using 2-way mixed models for absolute agreement for single ratings were computed for continuous measures including: global functioning, 0.82; borderline symptoms, 0.84; alcohol use, 0.97; and drug use, 0.99. κ Was computed for dichotomous measures including: DSM-IV-TR bipolar diagnoses, 1.00; bipolar spectrum diagnoses, 1.00; interview-rated hyperthymia, 0.83; and major depressive episodes, 0.92.

Binary logistic regression was used to examine the relation of the HPS with dichotomous measures, such as diagnoses. Pearson correlations were used to analyze the relation of the HPS with questionnaire and quantitative interview variables. To determine whether significant findings simply resulted from including participants with DSM-IV-TR bipolar disorders, analyses were recomputed with those participants omitted (with the obvious exception of analyses in which bipolar disorders were the criteria). Note that, in all but 3 analyses, the results remained significant when bipolar patients were omitted, suggesting that the findings were not simply due to the inclusion of a few diagnosed patients.

Table 1 presents the associations of HPS scores with dichotomous indicators of bipolar psychopathology, treatment, and family history. Fifteen (10%) of the participants met criteria for a DSM-IV-TR bipolar disorder (3 with bipolar I, 6 with bipolar II, 1 with cyclothymic, and 5 with bipolar not otherwise specified [NOS] disorders). Six additional participants qualified for bipolar IV; and 1, for bipolar II 1/2 disorders. Hypomanic Personality Scale scores were significantly associated with interview ratings of DSM-IV-TR bipolar disorders, the bipolar spectrum disorders of Akiskal [4], and history of hypomania or hyperthymia. Note that 20 of the 22 participants with diagnosable bipolar spectrum disorders scored at least 1.5 SD above the mean on the HPS. The relation of the HPS with history of major depressive episodes was not significant (despite the fact that major depressive episodes are part of many of the bipolar disorders). Likewise, relations with treatment or family history of mania or mood disorder were not significant.

Table 2 presents descriptive data for quantitative measures of personality, psychopathology, and functioning as well as correlations with bipolar spectrum psychopathology. As hypothesized, bipolar spectrum psychopathology was negatively associated with ratings of psychosocial functioning. The HPS was positively associated with hyperthymic, cyclothymic, and irritable temperament but was unrelated to dysthymic temperament (and only modestly to current depressive symptoms). Bipolar spectrum psychopathology was significantly associated with extraversion (large effect size) and openness (medium effect size) and inversely associated with agreeableness (medium effect size), with a modest direct association with neuroticism and inverse association with conscientiousness. Bipolar spectrum psychopathology was significantly associated with borderline traits (although only 1 participant met diagnostic criteria) as well as urgency, lack of premeditation, and sensation seeking. Surprisingly, bipolar spectrum psychopathology was generally unassociated with substance use, although there was a small association with alcohol impairment. All of the results remained significant after omitting the 15 participants with DSM-IV-TR bipolar disorders from the analyses, with the exception of the modest associations of the HPS with current depressive symptoms, neuroticism, and alcohol impairment.

8.3. Summary of study 1

As hypothesized, the HPS was associated with measures of psychopathology, personality, and functioning, including interview ratings of bipolar disorders. Note that the odds ratios were relatively modest, but this must be tempered by the fact that the study uses a nonclinically ascertained sample and the fact that the participants still have a considerable period of risk remaining for developing bipolar disorders. We also found predicted associations with measures of functioning, affective temperaments, and personality. Furthermore, the findings did not appear to be driven by a few markedly impaired participants with DSM-IV-TR bipolar disorders.

9.1. Goals and hypotheses

Study 2 examined the expression of bipolar spectrum psychopathology, as assessed by the HPS, in daily life using ESM. Table 3 shows hypothesized relations of bipolar spectrum psychopathology with experiences in daily life. The findings were expected to remain independent of DSM-IV-TR bipolar disorders.

In addition to the direct effects of bipolar spectrum psychopathology on experiences in daily life, it was expected that the HPS would moderate the effects of stress in daily life on behaviors. It was hypothesized that high scorers would be more reactive to the experience of stress in terms of affect, cognition, and behavior. We also examined whether changes in affect and cognition across the day differed across levels of bipolar spectrum psychopathology. Following Kwapil et al [33], it was hypothesized that exuberance would decline across the day in the low HPS scorers relative to the high HPS scorers, although this differential effect was not expected for other forms of affect. Also consistent with Kwapil et al [33], it was hypothesized that concentration difficulties would increase across the day for both high- and low-scoring subjects (with a main effect of higher concentration problems for the high HPS scorers) but that fullness of thought (which is more specific to bipolar spectrum psychopathology) would be elevated in the high HPS scorers but constant across the day in both high and low scorers.

10.1. Participants

Participants who completed study 1 took part in study 2. Seven participants were dropped because of failure to complete sufficient ESM protocols (final, n = 138).

10.2. Materials and procedures

The ESM protocol was designed to assess experiences relevant to bipolar spectrum psychopathology (eg, racing thoughts) and contextual factors (stressful situations). The Appendix A lists the ESM items and indices. After the completion of study 1, participants received a personal digital assistant (PDA) and were instructed about ESM procedures. The participants kept the PDAs for 7 days. The PDAs signaled the participants, administered the questionnaires, and time stamped and recorded responses. Participants were signaled to complete ESM questionnaires 8 times daily between noon and midnight. Participants had 3 minutes to initiate responses after the signal. After this time interval (or completion of the questionnaire), the PDA turned off and did not reactivate until the next signal, ensuring that participants did not skip questionnaires and complete them later.

10.3. Statistical method

The analyses examined the association of HPS scores with ESM responses. Experience sampling methodology data have a hierarchical structure in which ESM ratings (level 1) are nested within participants (level 2). Multilevel linear modeling provides a more appropriate method than conventional unilevel analyses for nested data [43,44]. The multilevel analyses examined 2 types of relations between the HPS score and experiences in daily life. The first was the intercept of the level 1 criterion, which assessed the independent effects of the level 2 predictor (HPS score) on level 1 dependent measures (ESM ratings in daily life). The intercept, β₀, was computed using the formula β₀ = γ₀₀ + γ₀₁ (HPS) + μ₀. In this model, γ₀₀ is the mean value of the level 1 dependent measure, γ₀₁ is the effect of the level 2 HPS predictor, and μ₀ is the residual variance term. The γ₀₁ coefficient provides information comparable with the unstandardized regression weight of the level 2 predictor with the level 1 measures.

The second set of analyses examined cross-level interactions of the relation of a level 1 predictor and criterion (eg, stress and irritability) with the level 2 HPS scores. Cross-level interactions [45] tested whether level 1 relations varied as a function of HPS scores. Cross-level interactions were evaluated using the equation β₁ = γ₁₀ + γ₁₁ (HPS) + μ₁ (in which γ₁₀ is the mean value of the level 1 slope, γ₁₁ is the effect of the level 2 HPS predictor, and μ₁ is the error term). If the HPS predictor was significant, then it explained variability in the within-person slopes. Note that the γ₁₀ coefficient evaluates the strength of the relations of the level 1 predictor and criterion, independent of the level 2 variable. These values provide an effective test of the validity of the ESM assessment, although they are not necessarily related to hypotheses regarding bipolar spectrum psychopathology.

Multilevel analyses were computed with HLM6 (Scientific Software International, Lincolnwood, IL, USA) [46]. Consistent with the recommendation of Luke [44], level 1 predictors were group mean centered, and HPS scores were grand mean centered. Parameter estimates were calculated using robust SEs, following Hox [43].

12.1. Reconsidering traditional views of bipolar psychopathology

Traditional psychiatric literature divides the world into those with bipolar disorders and those without. This categorical system simplifies the diagnosis of bipolar disorders but does not map onto the continuous expression of psychopathology and impairment existing in nature [3,47]. The DSM-IV-TR focuses on defining reliable diagnostic categories. As a result, it lacks guidelines for conceptualizing individuals who have bipolar symptoms that do not meet the diagnostic criteria [48]. The current diagnostic system also does not represent the various bipolar disorders as being on a continuum, separated by degree, not type. Therefore, DSM-IV-TR diagnoses offer only a glimpse of bipolar characteristics and may be better conceptualized as classification “short-cuts” than an accurate representation of the broader bipolar spectrum.

12.2. Hypomanic Personality Scale as a measure of bipolar spectrum psychopathology

Construct validation of the bipolar spectrum requires adequate measurement tools. The present studies supported the validity of the HPS as a measure of clinical and subclinical bipolar spectrum psychopathology in a nonclinically ascertained sample. In the first study, the HPS was positively associated with DSM-IV-TR bipolar disorders and the bipolar spectrum disorders of Akiskal as well as a range of subclinical bipolar characteristics and associated traits, including hypomania and hyperthymia, borderline symptoms, cyclothymic temperament, irritability, and depressive symptoms. The present results also suggested that the bipolar spectrum is a heterogeneous construct. Consistent with bipolar disorders, subclinical bipolar characteristics vary broadly across individuals.

This study built on previous cross-sectional (eg, Meyer and Hautzinger [24]) and longitudinal (eg, Kwapil et al [23]) investigations that reported that the HPS predicts bipolar disorders. However, previous validation studies have not fully examined the relation of the HPS with bipolar spectrum characteristics. The present research indicated that the HPS provides a useful “foot in the door” for identifying the bipolar spectrum. The present studies highlighted that the scale measures a wide range of bipolar psychopathology, well beyond initial assumptions that it only measured hypomanic functioning. Furthermore, previous studies of the validity of the HPS were often hampered by the fact that the HPS was administered in screening sessions that were weeks or months before the validity studies. This raised concerns regarding both the stability of the measure and the stability of the construct being assessed (given the cyclical nature of bipolar psychopathology). The present studies addressed this by administering the measure at the initial screening and the day of the interview and ESM information session. The stability coefficient was high, suggesting that the HPS is a reliable instrument across a 3-month interval and that the construct that it assesses is relatively stable, despite fluctuations in bipolar characteristics.

Hypomanic Personality Scale scores were associated with a history of bipolar disorders. As noted, more than 10% of the sample had a history of DSM-IV-TR bipolar disorders, and the rate increased to 15% when the broader diagnostic categories of Akiskal were considered. However, the association of the HPS with bipolar spectrum psychopathology remained independent of participants with DSM-IV-TR bipolar disorders. Thus, the results were not driven by a subset of disordered participants and support the construct of a broader bipolar spectrum.

The positive relation of HPS scores with borderline traits is consistent with the notion that bipolar and borderline characteristics represent overlapping constructs [21]. Both bipolar and borderline characteristics include labile affect, impulsive behavior, and suicidal gestures. However, bipolar spectrum psychopathology is also associated with changes in cognitions (eg, racing thoughts, fullness of thought) and energy (eg, hypomania, hyperthymia) that fall outside the range of borderline personality disorder.

The finding that HPS scores were not associated with family history of bipolar disorders was contrary to the hypotheses. It is unclear whether this reflects that our assessment method identifies participants with nonfamilial bipolar characteristics or that participants lack knowledge of or are unwilling to divulge family psychopathology. Future studies should include in-depth assessment of family psychopathology including additional informants.

12.3. Expression of the bipolar spectrum in daily life

Study 1 used diagnostic interviews and self-report questionnaires to assess clinical and subclinical bipolar spectrum psychopathology in the laboratory. Study 2 built upon this by examining the expression of the bipolar spectrum in real-world settings—independent of the effects of DSM-IV-TR bipolar disorders. The continuity of findings across the 2 methodologies further supports construct validity of the bipolar spectrum.

Disturbances in thought are a hallmark of bipolar spectrum psychopathology. As expected, bipolar spectrum psychopathology was positively associated with racing thoughts, thinking about many things, daydreaming, and difficulty concentrating. Disruptive behaviors are also associated with bipolar spectrum psychopathology—and are especially problematic during episodes of hypomania and mania. These mood states are often characterized by an increase in goal-directed activity and psychomotor agitation as well as reckless behavior. Bipolar spectrum psychopathology was positively associated with risky behavior, restlessness, exciting activities, and with doing many things in daily life. Bipolar spectrum psychopathology involves marked disruptions in affect, and we found associations with negative affect as well as energy and enthusiasm. Note that the HPS was positively associated with grandiosity and optimism as well as with uncertainty and boredom. These findings highlight that bipolar-like disruptions in cognition, behavior, affect, and sense of self are not limited to clinical episodes, supporting the notion that the bipolar spectrum extends beyond clinical boundaries.

Study 2 also explored the moderating effects of bipolar spectrum psychopathology on the relations of experiences in daily life. In general, bipolar spectrum psychopathology did not appear to moderate stress reactivity. This may reflect that the main effects on affect, cognition, and behavior in daily life were so robust that they were not differentially impacted by stress. However, the present study replicated findings of Kwapil et al [33] that bipolar spectrum psychopathology moderated the expression of exuberance during the day. Furthermore, it replicated the findings differentiating the experience of ubiquitous concentration difficulties from fullness of thought, which appears to be specific to bipolar spectrum psychopathology.

Experience sampling methodology appears to be an effective method of capturing disruptions in affect, thoughts, behaviors, and sense of self associated with bipolar spectrum psychopathology. Furthermore, ESM data should ultimately allow for more complex relations, such as examining ESM outcomes over longer periods. This technique may be especially useful for assessing bipolar spectrum psychopathology because individuals may cycle in and out of various affective states. In addition, ESM may also be valuable in translational research. For example, ESM may be useful in clinical settings for monitoring bipolar symptoms between appointments.

12.4. Conceptualizing a spectrum of bipolar psychopathology

Conjecturing a bipolar spectrum carries the burden of operationalizing the construct and its boundaries. The bipolar spectrum includes clinical and subclinical expressions of bipolar symptoms, ranging from mild hypomanic episodes and mild hyperthymic or cyclothymic temperament at the low end to severe expressions of bipolar I disorder at the high end. It may best be conceptualized as a constellation of personality and psychopathology characteristics (the extremes of which are represented in DSM-IV-TR diagnoses). Specifically, the bipolar spectrum involves dysregulation in affect, energy, cognition, and behavior that can be expressed at clinical or subclinical levels. By identifying individuals who experience these subclinical characteristics, we will be better able to understand the nature of the bipolar spectrum as well as factors associated with the etiology and phenomenology of bipolar disorders. Furthermore, studying bipolar spectrum psychopathology will provide insight to the risk and resilience factors associated with the development of bipolar disorders as well as the ways to prevent, assess, and treat bipolar psychopathology.

This model of a bipolar spectrum does not suggest that there is a normal personality dimension of bipolarity— contrasting it from personality dimensions such as extraversion-introversion. Thus, it is not suggesting that everyone falls somewhere along a bipolar continuum. Rather, it suggests that the spectrum represents a confluence of personality and psychopathology factors that will presumably have discernable etiologic pathways. This formulation begs the questions of what are the etiologic pathways and what factors (and to what degree) must be present to be considered as representative of the bipolar spectrum. These questions fall outside the current construct validity study. However, the examination of both clinical features (study 1) and daily life experiences (study 2) provided a unique approach for examining bipolar characteristics, and the current findings are consistent with the idea of a broader bipolar spectrum and support further investigations of etiology.

Bipolar spectrum characteristics and psychopathology are multifaceted. This is apparent in the temperaments of Akiskal, in the cyclical nature of the disorders, and the domains of expression and impairment (elevated and dysphoric affect, form and content of thought, impulsive and erratic behavior, etc). These findings are suggestive of possible multidimensional formulations. However, such formulations should be based upon a priori theorizing and consider possible etiologic mechanisms and should avoid overinterpreting post hoc groupings of phenomenological characteristics.

Dimensional models of psychopathology are being considered for many other disorders, including schizophrenia (eg, Meehl [49]) and depression (eg, Hankin et al [50]). These models suggest that the clinical disorders are part of a larger spectrum of symptoms and impairment. Patients often exhibit subclinical manifestations before the onset of the disorder, and people with subclinical expressions are presumed to be at heightened risk for transitioning into clinical disorders. However, the bipolar spectrum is relatively unique compared with these other models in that mild manifestations of the bipolar spectrum can be advantageous (although Claridge [51] suggested advantageous aspects of preschizophrenic conditions). Numerous studies [52–55] indicated that sporadic hypomania or trait-like hyperthymia can enhance functioning in many domains (albeit conveying heightened risk for mania and depression). Nevertheless, this creates challenges for conceptualizing the spectrum and identifying people who fall on this continuum. Assessment must not simply rely on impairment or dysfunction but also has to identify spectrum characteristics that can be adaptive.

Following Cronbach and Meehl [56], the construct of the bipolar spectrum is at the present time relatively loosely defined. However, the findings provide preliminary support for the conceptualization. Investigation of the issues presented above provides a road map toward operationalizing and testing a more systemized construct. Furthermore, the bipolar spectrum model appears to provide a conceptually richer basis for understanding and ultimately treating bipolar psychopathology than current diagnostic formulations.