General Behaviour Inventory (GBI)

The items on the GBI are distributed across two subscales:

• Depression Subscale: Focuses on capturing symptoms associated with depressive episodes, such as fatigue, sleep disturbances, and feelings of hopelessness.
• Hypomania/Biphasic Subscale: Evaluates the presence of hypomanic and manic symptoms, including elevated mood, increased energy, and impulsivity, along with shifts between depressive and hypomanic states. This subscale effectively captures the fluctuating nature of bipolar.

The GBI is useful for assessing bipolar disorder risk and monitoring symptoms in clinical practice. It has been validated across diverse demographics, including adolescents (13+), adults, in community health and psychiatric settings (Danielson et al., 2003).

The inventory is useful for screening for subtypes of bipolar disorders:

• Bipolar I Disorder: Characterised by one or more manic episodes, which are often severe and can be accompanied by depressive episodes. A manic episode typically involves elevated or irritable mood, increased energy, and risky behavior. Depressive episodes can also occur, though not necessarily.
• Bipolar II Disorder: Involves one or more major depressive episodes and at least one hypomanic episode. Hypomania is a milder form of mania with less severe symptoms that do not cause significant impairment. Individuals with Bipolar II do not experience full manic episodes.
• Cyclothymic Disorder (Cyclothymia): Features periods of hypomanic symptoms and periods of depressive symptoms lasting for at least two years (one year in children and adolescents). These symptoms do not meet the criteria for hypomanic or major depressive episodes but do cause distress or impairment.

The GBI total score ranges from 0 to 219. The GBI total score reflects the frequency and intensity of symptoms related to bipolar disorder.

Score is also expressed as the “percent endorsed”, which uses a binary scoring where “Never or Hardly Ever” and “Sometimes” are not endorsed, while responses “Often” and “Very Often or Almost Constantly” are considered endorsed. Percent endorsed convert scores to 0% to 100%.

Total scores are classified into three risk categories:

• Unlikely to have Bipolar Disorder (Percent endorsed between 0 and 26%): Individuals are likely to not have bipolar classified as likely non-cases, meaning they are unlikely to have bipolar. About 80% of the population typically falls within this category, indicating that their mood variability falls within the normal range.
• Indeterminate (Percent endorsed between 27% and 36%): Scores in this range indicate respondents may exhibit mild or atypical mood disturbances that are consistent with bipolar symptoms, but not to the extent that they clearly indicate bipolar disorder. This score range represents 12% of individuals among members of the community.
• Likely to have Bipolar Disorder (Percent endorsed 37% and above): Individuals within this range are classified as likely cases of bipolar disorder. Scores in this range reflect meaningful syndromal behavior consistent with bipolar types I, II, or Cyclothymia (Klein et al., 1986).

The subscale scores are also computed using the Binary scoring system, where scores are expressed as the percent of items endorsed as occurring “Often” and “Very Often or Almost Constantly”:

• Depression Subscale
  • (items 3, 23, 45, 63, 47, 56, 62, 73, 9, 10, 13, 70, 21, 33, 49, 59, 1, 12, 41, 5, 25, 37, 52, 14, 39, 55, 29, 36, 50, 71, 16, 60, 65, 67, 20, 32, 34, 72, 18, 26, 58, 68, 6, 28, 69)
  • A higher score on the Depression subscale indicates higher depressive symptomatology. The percent of symptoms endorsed are classified into three categories (Pendergast et al., 2014):
    • Low – 0 to 4% – Unlikely to have any mood disorder
    • Moderate – 5% to 31% – May have a mood disorder of some kind
    • High – 32% to 100% – Likely to have a mood disorder of some kind (4.75 times more likely than the general public)
• Hypomania/Biphasic Subscale
  • (items 2, 24, 35, 48, 19, 40, 53, 4, 7, 15, 22, 30, 31, 66, 11, 17, 42, 51, 27, 44, 54, 8, 57, 64, 38, 43, 46, 61)
  • A higher score on the Hypomania/Biphasic subscale suggests mood variability, impulsivity, or energy dysregulation consistent with hypomanic or manic episodes. The percent of symptoms endorsed are classified into three categories (Pendergast et al., 2014):
    • Low – 0 to 18% – Unlikely to have bipolar disorder
    • Moderate – 19% to 68% – 4.21 times more likely to have bipolar disorder than the general public.
    • High – 69% to 100% – 6.49 times more likely to have bipolar disorder than the general public.

A percentile is used to contextualise a respondent’s score compared to the normative community sample. The community sample represents the typical level of mood-related symptoms found in the general population (Bullock et al., 2011; Pendergast et al., 2014). A percentile of 50 suggests typical (and relatively healthy) experience of psychomotor energy, whereas a percentile of 80 indicates that the respondent scores higher than 80 percent of individuals, indicating some symptoms consistent with bipolar disorder.

The horizontal graphs presents the total score (0 to 219, scored using the likert method) and percent endorsed score for the subscales (from 0 to 100%). These graphs shows scores in comparison to the normative community and clinical distributions, with shaded areas around the two middle quartiles (between the 25th and 75th percentile). The clinical distribution represents individuals diagnosed with bipolar disorder (Bullock et al., 2011; Pendergast et al., 2014). This graph helps contextualise patterns of responding in comparison to the distribution of responses among community samples and those with a diagnosis of bipolar disorder (Pendergast et al., 2014).

For the total score, when 37% of symptoms are endorsed the scale has a sensitivity of .90, meaning it correctly identifies 90% of true cases (Klein et al., 1986). At this level, its specificity is .98, indicating it correctly identifies 98% of non-cases.

Higher scores on the GBI may suggest the presence of bipolar disorder, which frequently coexists with other psychiatric conditions. For instance, research indicates that one in six adults with bipolar disorder also has ADHD (Schiweck et al., 2021). People with ADHD also score higher on the GBI than the general public. Other commonly associated disorders include autism spectrum disorder, borderline personality disorder, and substance use disorders (Menezes et al., 2019; Skokauskas & Frodl, 2015). Recent evidence highlights the GBI’s effectiveness in differentiating bipolar disorder from conditions such as unipolar depression and ADHD, making it a valuable tool for addressing comorbidity challenges in psychiatric assessments (Pendergast et al., 2014).

The GBI’s psychometric properties have been evaluated extensively, demonstrating it as a reliable and valid tool for assessing symptoms of bipolar disorder. The GBI captures both the core behavioural symptoms and dimensions like intensity, duration, frequency, and rapid shifts in mood. The items on the GBI were developed based on several sources:

1. symptom literature related to adult and adolescent depression, hypomania, and mania;
2. symptoms observed in bipolar patients during depressive phases;
3. behavioural characteristics associated with cyclothymia; and
4. symptoms outlined in the Research Diagnostic Criteria (RDC) manual (Depue et al., 1981).

During its development, the GBI was evaluated against a binomial classification of bipolar “cases” and “non-cases,” making it effective for identifying individuals at risk of bipolar disorder versus those with subclinical mood variations or other mood disorders. Recent research suggests that the GBI’s effectiveness could be improved by reorganising its items into more homogeneous subscales (Youngstrom et al., 2001). While earlier studies pointed to a single general factor accounting for most variance (Depue et al., 1981), newer analyses (Danielson et al., 2003; Depue et al., 1989; Pendergast et al., 2015; Youngstrom et al., 2001, 2013) support a two-factor structure, comprising of the following subscales:

• Depression and
• Hypomania/Biphasic

In the original study (Depue et al., 1981), analysis was performed on a non-clinical sample of 850 undergraduate students (51.5% female) with an age range of 18 to 32 years old. The majority of participants (90%) were 22 years of age or younger. The GBI exhibited strong construct validity, as evidenced by its significant correlation with clinician-rated assessments of mood disturbance (Depue et al., 1981). The scale showed strong overall internal consistency (alpha = .94) (Depue et al., 1981; 1989) and test-retest reliability over 15 weeks was reported to be 0.73 (Depue et al., 1981).

The Hypomania and Depressive scale scores demonstrate strong psychometric properties across various samples, including internal reliability alphas above .90, test-retest reliabilities over .70, robust predictive validity, and satisfactory convergent and discriminant validity (Youngstrom, 2008; Youngstrom et al., 2013).

The GBI uses two methods of scoring: Binary and Likert.

Using the binary scoring method the responses “Never or Hardly Ever” and “Sometimes” are scored as 0, and responses “Often” and “Very Often or Almost Constantly” are scored as 1. In NovoPsych, subscales are scored using the binary method, yielding a “percent endorsed” metric.

Alternatively, using the Likert scoring method, the response “Never or Hardly Ever” is scored as 0, “Sometimes” is scored as 1, “Often” is scored as 2, and “Very Often or Almost Constantly” is scored as 3. NovoPsych uses both the likert method (score range = 0 to 219) and binary method (percent endorsed) for the total score.

When using the Binary scoring, the GBI uses a two-threshold system, with one threshold for likely cases and another for non-cases, to classify individuals based on their total scores. This two-cut-off score system divides scores into three categories:

• Total scores of 19 or lower (0% to 26% endorsement) are classified as likely non-cases.
• Total scores between 20 and 26 (27% to 36% endorsement) are classified as indeterminate.
• Total scores of 27 to 73 (more than 37%+ endorsement) indicate likely cases of bipolar disorder. (Depue et al., 1981; Klein et al., 1986).

With the primary cut-off of 27 (37% endorsement), for bipolar conditions, the scale has a sensitivity of .90, meaning it correctly identifies 90% of true cases. Its specificity was .98 indicating it correctly identifies 98% of non-cases (Klein et al., 1986). In the study by Depue et al. (1981), a cut score of 27 secured a Positive Predictive Value (PPV) of 0.98, indicating that 98% of individuals who receive a positive result on the scale (indicating the presence of a condition) truly have bipolar disorder. Finally, the Negative Predictive Value (NPV) of the scale is .87, indicating that 87% of the individuals who receive a negative test result (indicating they do not have the condition) are correctly identified as not having bipolar disorder (Waugh et al., 2014). However, the false negative rate, where actual cases are missed, was 9% for scores below 19 and increased for scores in the 20-26 range (Depue et al., 1981).

For the depression subscale (range 0 to 46 as per Binary scoring method), a range of scores from 0 to 2 (0% to 4%) are categorised as “Low”, 3 to 14 (5% to 31%) as “Moderate”, and 15 to 46 (32% to 100%) as “High”.

For the hypomania/biphasia subscale (range 0 to 28 as per Binary scoring method), scores from 0 to 5 (0% to 18%) are categorised as “Low”, 6 to 19 (19% to 68%) are categorised as “Moderate”, and 20 to 28 (69% to 100%) are categorised as “High” (Pendergast et al., 2014).

Although the study by Pendergast et al., (2014) does not provide sensitivity and specificity data directly, it measured related constructs through Diagnostic Likelihood Ratios (DLRs). Diagnostic Likelihood Ratios (DLRs) are tools that examine how much a test result changes the chances that someone has a specific condition (Straus et al., 2011). A DLR of 1 is neutral, DLR < 1 decreases the likelihood of the condition, and DLR > 1 increases the likelihood (Straus et al., 2011). For the hypomania/biphasia scale, the following DLRs were reported for the severity categories when comparing individuals with and without bipolar disorder:

• Low (0–5): DLR = 0.09. This means that if a person scores in the low range, they have a very low probability of having bipolar disorder.
• Moderate (6–19): DLR = 4.21. A score in this range makes bipolar disorder 4.21 times more likely compared to the general public.
• High (20+): DLR = 6.49. A score in this high range increases the odds of bipolar disorder by 6.49 times compared to the general public, indicating a strong likelihood of BD when scores are 20 or above.

A similar analysis was conducted on the depression subscale, however, the comparison was made between individuals with and without any mood disorder:

• Low (0–2): DLR = 0.14. A score in the low range means it is unlikely the person has any mood disorder. A DLR of 0.14 greatly decreases the likelihood of a mood disorder.
• Moderate (3–14): DLR = 0.78. This value is close to 1, which means it doesn’t significantly change the odds of having or not having a mood disorder.
• High (15+): DLR = 4.75. A score in this high range increases the likelihood of having a mood disorder by about 4.75 times, making it a strong indicator of the presence of a mood disorder.

The sample used to compute normative percentiles for the total score consisted of 87 Australian participants (68% female) with a mean age of 26.3. (Bullock et al., 2011). The study used the Likert scoring method. The GBI scores were divided between low and high scorers, so a pooled average was calculated to represent the typical score of an individual in the normative sample. For the total score,

• Normative sample: M = 59.16, SD = 19.18
• Individuals with Bipolar Disorder, M = 112.67, SD = 52.85

Two other samples were used for comparison purposes at the subscale level. In Sample 1, a total of 359 adolescents (69.6% female), aged 14 to 19 years (M = 18.35, SD = 1.52), participated in the Teen Emotion and Motivation (TEAM) project (Alloy et al., 2012a). Sample 2 consisted of 614 university students (33.2% female), aged 18 to 24 years (M = 19.63, SD = 1.76), who were part of the Longitudinal Investigation of Bipolar Spectrum (LIBS) project (Alloy et al., 2008; 2012b). The above samples were used in the study by Pendergast et al., (2014), which analysed both individuals from the general public sample and those with bipolar disorder. This study used the binary scoring method, meaning that responses “Never or Hardly Ever” and “Sometimes” were scored as 0, and responses “Often” and “Very Often or Almost Constantly” were scored as 1 (Pendergast et al., 2014). The means and standard deviations for both the subscales in Pendergast et al. (2014) are as follows:

• For Depression subscale:
  • Normative Sample: M = 7.24, SD = 9.43
  • Individuals with Bipolar Disorder: M = 20.76, SD = 12.01 (Pendergast et al., 2014)
• For Hypomania/ Biphasic subscale:
  • Normative Sample: M = 5.66, SD = 9.43
  • Individuals with Bipolar Disorder: M = 14.22, SD = 6.65 (Pendergast et al., 2014)

Assess your patients with the General Behaviour Inventory (GBI) using a NovoPsych account

• For Psychologists & Mental Health Clinicians
• Send Assessments to Patient’s Phone
• Receive Comprehensive Reports
• Access Over 100 Validated Assessments​
• Instant Psychometric Scoring
• Track Symptoms​
• Inform Treatment​