Differentiating irritable mood and disruptive behavior in adults

Abstract Introduction Irritability has both mood and behavioral manifestations. These frequently co-occur, and it is unclear to what extent they are dissociable domains. We used confirmatory factor analysis and external validators to investigate the independence of mood and behavioral components of irritability. Methods The sample comprised 246 patients (mean age 45 years; 63% female) from four outpatient programs (depression, anxiety, bipolar, and schizophrenia) at a tertiary hospital. A clinical instrument rated by trained clinicians was specifically designed to capture irritable mood and disruptive behavior dimensionally, as well as current categorical diagnoses i.e., intermittent explosive disorder (IED); oppositional defiant disorder (ODD); and an adaptation to diagnose disruptive mood dysregulation disorder (DMDD) in adults. Confirmatory factor analysis (CFA) was used to test the best fitting irritability models and regression analyses were used to investigate associations with external validators. Results Irritable mood and disruptive behavior were both frequent, but diagnoses of disruptive syndromes were rare (IED, 8%; ODD, 2%; DMDD, 2%). A correlated model with two dimensions, and a bifactor model with one general dimension and two specific dimensions (mood and behavior) both had good fit indices. The correlated model had root mean square error of approximation (RMSEA) = 0.077, with 90% confidence interval (90%CI) = 0.071-0.083; comparative fit index (CFI) = 0.99; and Tucker-Lewis index (TLI) = 0.99, while the bifactor model had RMSEA = 0.041; CFI = 0.99; and TLI = 0.99 respectively). In the bifactor model, external validity for differentiation of the mood and behavioral components of irritability was also supported by associations between irritable mood and impairment and clinical measures of depression and mania, which were not associated with disruptive behavior. Conclusions Psychometric and external validity data suggest both overlapping and specific features of the mood vs. disruptive behavior dimensions of irritability.


Introduction
and impairment. 8 IED is characterized by presence of disruptive behaviors (e.g., extreme temper outbursts, aggression), whereas ODD and DMDD are characterized by both disruptive behaviors and irritable mood, i.e., persistent anger, including sullen nonverbal behaviors, and reports of being annoyed over many days. 1 Although irritable mood and behavior frequently co-occur, it is unclear whether these components of irritability can be measured separately and constitute distinct dimensions.
Beyond the implications for the conceptualization of irritability in the adult population, if irritability has distinct components, this raises the possibility of tailoring interventions to target specific components of irritability.
While research has begun to examine this question in youth, 9  for patients with irritability. Therefore, the purpose of this study is to investigate distinctions between the mood and behavioral components of irritability in clinically referred adults.
One way of investigating whether behavior and mood are distinct constructs is to use confirmatory factor analysis (CFA). CFA starts from the assumption that indicators (i.e., symptoms) are initiated by nonobservable latent constructs. By examining variance among those symptoms, theoretical models can be tested to decide which model is best suited to describing the symptom correlation patterns observed in a specific dataset. We tested three potential models. In the first, irritability is conceptualized as a single construct, i.e., all the variance of irritability symptoms is due to either a single latent factor or measurement error. The second model assumes that there are two correlated factors (mood and behavior), i.e., two separate, correlated sources of variance. Lastly, we tested a bifactor model, which assumes both shared and distinct sources of variance between irritable mood and disruptive behavior. The adult literature on irritability has mainly focused on behavioral manifestations of irritability, conceptualized as reactive aggression. 10,11 Aggression, defined as a behavior intended to harm another, is commonly divided into proactive and reactive aggression. 12 Proactive, or instrumental, aggression is designed to achieve a goal (e.g., gain social status or a job promotion), while reactive aggression (also called emotional or hostile aggression) occurs in response to frustrating or threatening events. However, previous studies in the adult irritability literature did not focus on psychometrics, such that neither specific associations between reactive aggression and irritable mood, nor the extent to which these constructs are distinct could be directly investigated.
In contrast to the adult psychiatric literature, child psychiatry research demonstrates the importance of studying both behavioral and mood manifestations of irritability in depth. 1,13,14 Research in children indicates that irritability is closely related to affective disorders such as anxiety and depression, in terms of longitudinal 15 and genetic 16 associations, as well as some shared cognitive mechanisms, such as biases towards threats. 17,18 Thus, the mood and behavioral components of irritability might be alternative manifestations of the same pathophysiological mechanism. Alternatively, these associations could be driven by a shared affective component that is common to anxiety, depression, and anger, but which excludes the behavioral components of irritability. The latter hypothesis is only plausible if the mood and behavioral components of irritability are indeed distinct constructs, a hypothesis that is examined in this paper.
Here, we investigate the independence of the mood and behavioral components of irritability in a sample of adults with severe mental disorders, using confirmatory factor analysis and correlations with external validators.
We hypothesize that a clinical interview specifically designed to probe mood and behavioral aspects of irritability will provide a means of investigating these two related dimensions. Consistent with this, measures of external validity will identify some specific associations with each of these two components of irritability.

Sample
The sample consisted of 246 patients, recruited from four outpatient programs at the Hospital de

The Mood Disruptive Scale (MOODS)
The MOODS is a self-rated questionnaire with 56

Electronic Chart Review Instrument
The chart review instrument 19

Statistical analysis
We conducted CFA to evaluate a unidimensional model (with all items), a correlated model (with two dimensions, mood and behavior), and a bifactor model (with one general and two specific dimensions, mood and behavior). Correlations between CFA items were calculated with the mean and variance adjusted weighted least squares (WLSMV) estimator, implemented in the R (version 3.3.2) lavaan package (version 0.5-23). 24 The model was considered to have a good fit to the data if the comparative fit index (CFI) and Tucker-Lewis index (TLI) were both ≥ 0.95 and the root mean square error of approximation (RMSEA) was ≤ 0.06. An acceptable fit to the data was defined as when fit indices CFI and TLI were ≥ 0.90, and RMSEA was ≤ 0.80. 25 We also used Reykov omega coefficients as reliability indices. 26 Item information curves were estimated from CFA using the R psych package. 27 Path analysis was used to investigate associations with irritability-related impairment. Spearman correlation coefficients were used to investigate associations with clinical scales.
Missing data were treated with imputation by chained equations using the mice package. 28,29 Imputation accounted for 4% of the data.

Sample description
The majority of our sample was composed of adult females from low to middle-income classes. One third of the sample had an anxiety disorder, two thirds had a mood disorder, and one-fourth met criteria for a psychotic disorder. Complete information on the sample can be found in Table 1.

Prevalence of irritability levels and disruptive disorders
Item frequencies for each component of the MOODS questionnaire are shown in Table 2. Most of the participants showed low to moderate frequency of irritable mood and behavior symptoms. The most prevalent disruptive disorder diagnosis was IED with 8% prevalence, followed by DMDD with 2% and ODD with 2% prevalence.

Confirmatory factor analysis
The CFA indices showed that the unidimensional   Panel A illustrates the unidimensional model, Panel B shows the correlated model, and Panel C depicts the bifactorial model. an_ = disruptive behavior dimension; ang = disruptive behavior dimension; ds = general factor; dsr = general factor; md_ = irritable mood dimension; mod = irritable mood. Full descriptions of the abbreviations for each question (m3a-t2q) are given in Tables 2 and 3.   Thresholds represent the amount of factor loading z-score needed for a person be more likely to endorse the next response category (from never to rarely, for example) than the previous one. Sample recruited from outpatient psychiatry services at the Hospital de Clínicas de Porto Alegre (n = 246).  Table 3.

Associations with irritability related impairment
In the bifactor model, only the common factor (b

Associations with depression and mania
The bifactor model revealed that most of the associations were driven by the common factor (ρ = Although research discriminating mood and behavior components of irritability in adults is lacking, preliminary research has begun to discriminate between distinct aspects of irritability in children. 9 The pediatric literature has described irritability as having a tonic aspect, characterized by angry mood between temper outbursts, which does not necessarily have behavior manifestations; and a phasic aspect, characterized by temper outbursts or tantrums, which is not necessarily associated with persistent angry mood. Although the main distinction between phasic and tonic irritability in children relies on the duration of the manifestation (tonic lasting longer than phasic); operational methods for classifying tonic and phasic irritability are very reliant on the mood and behavior distinction; with mood representing proneness to persistently experience anger as an emotion, and behavior representing proneness to respond frequently with anger to frustration. Consistent with our findings of a shared variance between mood and behavior, this study also found that these two aspects of irritability are closely related. 9 Second, another important finding from item level CFA analysis showed mood items were more frequently endorsed than behavioral items in this sample of adults.

Conclusions
This study concludes that there are both shared and potentially distinct aspects of irritability in adults that can be differentiated by a clinical interview, thus potentially facilitating specifically targeted treatments.
Irritable mood is very common in individuals with severe mental disorders, and behavioral symptoms are present in more severe cases of irritability. These data provide some psychometric support for studying the shared and specific aspects of irritability in adults.