Mild to severe social fears: Ranking types of feared social situations using item response theory

Journal of Anxiety Disorders 28 (2014) 471–479

Contents lists available at ScienceDirect

Journal of Anxiety Disorders

Mild to severe social fears: Ranking types of feared social situations using item response theory Erica Crome *, Andrew Baillie NHMRC Centre of Research Excellence in Mental Health and Substance Use/Centre of Emotional Health, Macquarie University, Australia

A R T I C L E I N F O

A B S T R A C T

Article history: Received 1 January 2014 Received in revised form 2 May 2014 Accepted 7 May 2014

Social anxiety disorder is one of the most common mental disorders, and is associated with long term impairment, distress and vulnerability to secondary disorders. Certain types of social fears are more common than others, with public speaking fears typically the most prevalent in epidemiological surveys. The distinction between performance- and interaction-based fears has been the focus of long-standing debate in the literature, with evidence performance-based fears may reflect more mild presentations of social anxiety. This study aims to explicitly test whether different types of social fears differ in underlying social anxiety severity using item response theory techniques. Different types of social fears were assessed using items from three different structured diagnostic interviews in four different epidemiological surveys in the United States (n = 2261, n = 5411) and Australia (n = 1845, n = 1497); and ranked using 2-parameter logistic item response theory models. Overall, patterns of underlying severity indicated by different fears were consistent across the four samples with items functioning across a range of social anxiety. Public performance fears and speaking at meetings/classes indicated the lowest levels of social anxiety, with increasing severity indicated by situations such as being assertive or attending parties. Fears of using public bathrooms or eating, drinking or writing in public reflected the highest levels of social anxiety. Understanding differences in the underlying severity of different types of social fears has important implications for the underlying structure of social anxiety, and may also enhance the delivery of social anxiety treatment at a population level. ã 2014 Elsevier Ltd. All rights reserved.

Keywords: Social anxiety disorder Performance Interaction Item response theory Exposure therapy

Social anxiety disorder (SAD), previously social phobia (American Psychiatric Association (APA), 2000, 2013), is one of the most prevalent mental disorders, with an estimated lifetime prevalence over 12% (Stein & Stein, 2008). As social anxiety typically emerges early in life, the effects of continued avoidance of feared social situations often accumulates with time to result in significant distress and impairment in social, financial, employment, educational and intimate relationship domains (Wittchen & Fehm, 2003). Social anxiety has also been demonstrated to frequently lead to the development of secondary disorders such as depression and substance use disorders (Ruscio et al., 2008). There is also evidence that, rather than being a discrete disorder which is present or absent, social anxiety disorder lies on a continuum from mild to severe social anxiety; potentially including avoidant personality disorder (Crome, Baillie, Slade, & Ruscio, 2010).

* Corresponding author at: Centre for Emotional Health, Department of Psychology, Building C3A, Macquarie University, NSW 2109, Australia. Tel.: +61 2 9850 8670/0 410 652 798; fax: +61 2 9850 8062. E-mail addresses: [email protected], [email protected] (E. Crome). http://dx.doi.org/10.1016/j.janxdis.2014.05.002 0887-6185/ ã 2014 Elsevier Ltd. All rights reserved.

Presentations of social anxiety below diagnostic thresholds are also essential to understand from clinical and public health perspectives, as they are also associated with increased distress, impairment and comorbidity (Fehm, Beesdo, Jacobi, & Fiedler, 2008); as well as vast personal and economic costs (Acarturk, Graaf, Van Straten, Have, & Cuijpers, 2008). Yet despite the burden of social anxiety above and below diagnostic thresholds, there is still relatively little understanding about the processes causing and maintaining this impairing disorder (Stein & Stein, 2008). There is also an urgent need to improve our ability to screen for social anxiety and engage people with social anxiety in treatment, as only a minority of people with social anxiety disorder seek treatment, and chronic social anxiety often goes unrecognised in primary health care settings (Katzelnick & Greist, 2001; Nordgreen et al., 2012). The types of social situations a person fears may provide important information not only about the underlying structure of social anxiety, but also enhance initiatives to extend social anxiety treatment beyond traditional clinical settings. There is longstanding debate about whether the types of social situations feared are unitary or are better conceptualised as distinct performance-

472

E. Crome, A. Baillie / Journal of Anxiety Disorders 28 (2014) 471–479

based and interaction-based fears. Some authors report differences in the somatic symptoms reported by people with performance compared with interaction fears (May et al., 2013); and there is evidence of different patterns of comorbidity or genetic vulnerabilities between performance and interaction subgroups (see Bögels et al., 2010). However, other authors have not found support of a performance and interaction distinction, with evidence both types of fears load onto a single underlying social anxiety dimension (Ruscio et al., 2008). This is consistent with several studies reporting linear relationships between an increase in the number of social situations feared and negative outcomes; with no additional support for subtyping fears either into the DSM-IV generalised and non-generalised subtypes, nor performance and interaction subtypes (e.g. Acarturk et al., 2008; Stein, Torgrud, & Walker, 2000). The Diagnostic and Statistical Manual of Mental Disorders (fifth edition; (DSM 5); 2013) has explicitly recognised a discrete subgroup of people experiencing social anxiety severe enough to meet diagnostic criteria, yet fearing only public performance situations such as public speaking. Yet the usefulness of this “performance-only” specifier is already being questioned, with preliminary estimates of DSM 5 social anxiety disorder highlighting extremely low prevalence or absence of these cases (Burstein et al., 2011; Kerns, Comer, Pincus, & Hofmann, 2013). However, one area of agreement throughout this ongoing debate is that some fears are consistently reported more frequently than others. Public speaking is amongst the most commonly reported social fears in large epidemiological samples, with up to 25% of the general population reporting a fear of public speaking (Ruscio et al., 2008; Wittchen & Fehm, 2003). This has been interpreted as evidence for performance-based fears (e.g. public speaking) reflecting lower levels of anxiety than interaction-based fears (e.g. initiating conversation; Bögels et al., 2010). However, the presence of differences in social anxiety severity across different types of social fears is yet to be formally tested. If formal testing showed a clear and consistent order of severity associated with each fear, this information would have far reaching effects. In particular, it would inform our understanding of the underlying structure of social anxiety and the ongoing debate about social anxiety subtypes. Understanding more about the underlying severity reflected by different social fears may also provide important information for the scaling up of existing social anxiety treatments beyond traditional treatment settings. The extensive burden of mental disorders and subsequent demand for effective mental health treatment has highlighted that existing treatment resources are insufficient to adequately treat highly prevalent disorders such as social anxiety disorder (Nordgreen et al., 2012). This has led to increasing support of, and investment in, stepped-care or internet interventions where initial treatment is provided with little or no trained clinician input (Andersson, 2009; Van Straten, Tiemens, Hakkaart, Nolen, & Donker, 2006). Whilst there is increasing evidence of the efficacy of self-guided treatments for social anxiety (Nordgreen et al., 2012), understanding more about the typical severity of different social fears would provide important information for developers of these programs. In particular, this information would be most relevant to the design of exposure therapy components, one of the most efficacious treatments of social anxiety (Feske & Chambless, 1995; Hofmann, 2004). Exposure-based therapies require a person to remain in a feared social situation, despite distress, with the aim of developing new experiences to challenge typically catastrophic beliefs about the likelihood or cost of potential negative social evaluation (Rodebaugh, Holaway, & Heimberg, 2004). Exposure protocols begin with the creation of a hierarchy of feared and avoided situations, ranked from least to most fear inducing. This ranking allows manageable progression through different feared

scenarios, building a sense of mastery, momentum and achievement with each successfully completed stage (Heimberg, 2002; Rodebaugh, Holaway, & Heimberg, 2004). The art of creating stepped tasks which are sufficiently challenging, yet not overwhelming, is often developed throughout training and clinical experience. If this wealth of clinical experience and training is not accessible, such as in self-directed treatments or interventions delivered by non-clinical healthcare workers, data from epidemiological surveys may provide a template for designing exposure hierarchies. If there was a typical order of increasing severity across different social fears on a population level, this could be readily translated into templates to personalise exposure hierarchies and normalise experiences for participants. In a similar manner, identifying particular fears typically indicative of more severe social anxiety on a population level could provide a “red-flag” screener during initial contacts with health professionals. Being able to briefly screen and detect people most likely to be experiencing severe social anxiety is essential given the high levels of attrition observed between the initial enquiries about social anxiety treatment and formal assessment (Coles, Turk, Jindra, & Heimberg, 2004). These “red-flag” situations would highlight that these people are more likely experience strong negative evaluation fears and social anxiety severity, and thus require additional support in therapy (Nordgreen et al., 2012). Item response theory (IRT) techniques provide a tool for formally assessing the underlying severity at which different social fears typically occur. Initially developed for educational testing, IRT techniques use mathematical relationships to infer information about unobservable states such as social anxiety, from observable variables such as self-reported anxiety or avoidance (Thomas, 2011). The two IRT parameters typically estimated in psychiatric research are the difficulty and discrimination parameters. The difficulty parameter (“b”) is the point on an underlying dimension at which a person has a 50% chance of endorsing that item. Larger difficulty estimates highlight items associated with higher, or more severe, levels of the trait. The discrimination parameter (“a” or slope) is analogous with precision, as it models the strength of the relationship between an individual item and the underlying trait. Larger discrimination values are desirable as they reflect items which separate people with different levels of the construct more efficiently (DeMars, 2010). Item response theory models have been used to rank the severity of diagnostic criteria for disorders such as conduct disorder (Gelhorn et al., 2009), panic attacks (Ietsugu, Sukigara, & Furukawa, 2007), nicotine dependence (Saha et al., 2010) and various alcohol and substance use disorders (Gillespie, Neale, Prescott, Aggen, & Kendler, 2007; Saha, Chou, & Grant, 2006). Thomas (2011) claims IRT techniques may revolutionise the practice of psychology by decreasing measurement error, as well as facilitating computer adaptive testing, objective calibration and greater ability to assess unobservable variables. This study aims to use IRT techniques to determine whether different types of social fears reflect different levels of underlying social anxiety severity. To enhance the validity of findings, analyses were replicated across four different samples using three different structured diagnostic interviews. 1. Materials and methods 1.1. Samples and measures Data was obtained from four large epidemiological surveys conducted using face-to-face structured diagnostic interviews. Interview skip structures designed to reduce respondent burden meant only subgroups of people reporting at least one social fear were asked further questions about their specific social fears. As

E. Crome, A. Baillie / Journal of Anxiety Disorders 28 (2014) 471–479

these populations were of interest from a measurement perspective, and to minimise errors associated with imputing data, analyses were restricted to these subgroups. A brief introduction to each survey, measure, subgroup of respondents and items used in analyses are outlined below, with key references provided containing further details. The investigation was carried out in accordance with the Declaration of Helsinki, with ethics approval and consent obtained within each of the four epidemiological survey, and ethical review of current analyses provided by relevant human research ethics committees.

473

International Diagnostic Interview (WMH-CIDI: Kessler & Ustun, 2004) developed as part of the World Health Organisation's (WHO) World Mental Health Initiative. The WMH-CIDI provides estimates of lifetime, 12 month and 30 days diagnoses for 22 disorders defined by the DSM-IV (4th ed.; APA, 2000). Overall, 8841 people participated in the NSMHWB 2007 (response rate 60%), and 1497 people were asked about specific types of social situations feared (Table 1). Items used in analysis were dichotomously coded ‘absent’ or ‘present’. 1.3. National Comorbidity Survey – Replication (NCS-R)

1.2. National Survey of Mental Health and Wellbeing (NSMHWB) 2007 The NSMHWB was conducted in 2007 to estimate the prevalence of common mental disorders and associated impairment in the Australian population aged 16 years and over (Slade, Johnston, Oakley-Browne, Andrews, & Whiteford, 2009). The NSMHWB 2007 used the World Mental Health Composite

The NCS-R was also conducted using the WMH-CIDI between 2001 and 2003 to estimate the prevalence and correlates of mental disorder in US residents aged 18 and over (Kessler et al., 2004). Of 9282 respondents, 2261 were asked about the presence of types of specific fears listed in Table 1. Items used in analysis were dichotomously coded ‘absent’ or ‘present’.

Table 1 Assessment items associated with types of social situations feared and rates of endorsement in the National Survey of Mental Health and Wellbeing (NSMHWB: 1997, n = 1845; 2007, n = 1497), National Comorbidity Survey – Replication (NCS-R, n = 2261) and National Epidemiologic Survey in Alcohol and Related Conditions (NESARC, n = 5411). Assessment items

Label

Endorsement rate (%) NSMHWB NCS-R 2007

WMH-CIDI Did you ever strongly fear any of the following situations? Meeting new people Talking to people in authority Speaking up in a meeting or class Going to parties or other social gatherings Acting, performing or giving a talk in front of an audience

Meet new Authority Meeting/class Party Public performance Exam/interview Taking an important exam or being interviewed for a job, even though prepared Working while someone watches Working watched Entering a room when others were already present Enter room Talking with people you didn't know very well Talk unfamiliar Expressing disagreement to people you don't know too well Disagree unfamiliar Write/eat/drink Writing or eating or drinking while someone watches Urinating in a public bathroom or using a bathroom away from home Bathroom Being in a dating situation Dating Any other social or performance situation where you could be the centre of attention or something embarrassing might happen. Other CIDI v2.1 In the past 12 months have you had an unusually strong fear or unusually strong avoidance of: Eating or drinking where someone could watch you Talking to people because you might have nothing to say or might sound foolish Writing while someone watches Taking part or speaking in a meeting or class Going to a party or other social outing Giving a speech or speaking in public Any other situation where you could be the centre of attention

69.0 60.9 80.5 55.0 87.9

60.1 49.4 57.5 58.2 61.1

58.7 48.4 49.7 53.9 51.9

32.3 21.0 40.9 70.3

32.8 23.6 46.9 64.6

NSMHWB 1997 Eat/drink Talk Write Meeting/class Party Public speech Other

Fear 8.8 29.5 9.7 36.3 22.5 55.1 17.2

Avoid 6.2 15.7 5.8 16.0 14.7 24.7 7.7 NESARC

AUDADIS Have you ever had a fear or avoidance of: Talking in front of others Engaging in conversation with people you don’t know too well Going to parties or other social gatherings Eating or drinking in public Writing whist being watched Dating Being in a small group situation Taking part of or speaking in class Being interviewed Taking part or speaking in a meeting Performing in front of others Taking an important exam Speaking to an authority figure such as a teacher or boss

72.8 64.7 83.0 68.7 83.8

Public speech Talk unfamiliar Party Eat/drink Write Dating Small group Class Interview Meeting Public performance Exam Authority

77.1 43.3 35.2 12.0 16.5 17.3 11.6 62.8 33.4 54.6 61.9 41.9 33.2

Measures are the Composite International Diagnostic Interview (CIDI) versions 2.1 and the World Mental Health (WMH) version, and the Alcohol Use Disorder and Associated Disabilities Interview Schedule-IV (AUDADIS-IV).

474

E. Crome, A. Baillie / Journal of Anxiety Disorders 28 (2014) 471–479

1.4. National Survey of Mental Health and Wellbeing (NSMHWB) 1997 The NSMHWB was the initial large scale mental health survey of the Australian population aged 18 and over (Andrews, Henderson, & Hall, 2001). It was based on an earlier version of the Composite International Diagnostic Interview (CIDI v2.1; WHO, 1997) assessing 12 month prevalence of up to 20 DSM-IV and ICD-10 diagnoses. Of 10,641 respondents (response rate 78.1%), 1845 were asked about of seven possible social scenarios feared and/or avoided (Table 1). 1.5. National Epidemiologic Survey in Alcohol and Related Conditions (NESARC)

created to minimise erroneous results (Steinberg & Thissen, 1996) and consisted of four levels: no fear/avoidance (0), fear only (1), avoidance only (2) or both fear and avoidance (3). Fear and avoidance were not combined into a single dichotomous indicator as there is evidence that avoidance may reflect a different level of anxiety than report of fear without avoidance (Karsten, Nolen, Penninx, & Hartman, 2011). As there were more than two possible responses, a graded response model was applied to estimate separate difficult parameters for each response category. An overall discrimination parameter was also estimated for each feared situation. 1.8. Item characteristic curves (ICC)

The NESARC survey was conducted between 2001 and 2002 to estimate the prevalence of alcohol use disorders and related concerns in the US population aged 18 and over (Grant, Moore, Shepard, & Kaplan, 2003). The NESARC was based on the Alcohol Use Disorder and Associated Disabilities Interview Schedule-IV (AUDADIS-IV). This is a structured diagnostic interview assessing the presence of alcohol/drug use and disorder, as well as several DSM-IV anxiety and mood disorders (Grant, Dawson, & Hasin, 2001). The NESARC survey was also designed to translate diagnostic criteria into terminology and language understandable to a lay person (Grant et al., 2003b). Of the 43,093 survey respondents (response rate 81.2%), 5411 were asked about 13 potential social fears (Table 1).

Outcomes from IRT analyses are best interpreted visually using ICC plots. These graphically reflect the probability of endorsing types of social fears at different levels of social anxiety severity. More difficult/severe items are located closer to the right of the x axis and highly discriminating items have steep vertical slopes. For ease of presentation, overall results from the graded response model used in the NSMHWB 1997 were summarised using item information curves (IIC). These IIC curves reflect the range of underlying social anxiety where these items are most discriminating, with more information denoting higher precision.

1.6. Statistical analyses

Factor scores for each model were derived using an empirical Bayesian estimator to form a dimensional social phobia severity measure. In the absence of a ‘gold-standard’ for comparison, dimensional severity measures should be highly related to existing categorical diagnoses (Kraemer, 2007). These relationships were tested by stratifying samples into 10 equal factor score percentiles, and testing for changes in the proportion of people meeting social phobia diagnostic criteria over each percentile. Consistent increases in percentages of people with social phobia over severity percentiles were taken as evidence of convergent validity.

Analyses followed the procedures for IRT with marginal maximum likelihood estimators outlined by Rizopoulos (2006) within the R statistical framework (version 2.13.2; R Development Core Team, 2011). Basic IRT assumptions of unidimensional structure and local independence were assessed by visual inspection of scree tests and exploratory factor analyses, as well as evidence of a single dimension underlying types of social fears previously reported (Ruscio et al., 2008). As universal guides for goodness of fit of IRT analyses are yet to be established, model fit was assessed by comparing different IRT models and assessing plotted residuals (DeMars, 2010). As IRT analyses create samplefree item parameter estimates, survey weights were not included in current analyses. 1.7. Item response theory models Two IRT models were compared; a one parameter (Rasch) model and a two parameter logistic (2PL) model. Rasch models estimate the difficulty or threshold of each assessment item and assume all items have the same discrimination. Rasch models most closely reflect DSM style nosologies, in which all diagnostic criteria, or types of situations feared, contribute equally to diagnosis of social anxiety disorder. The 2PL models extend Rasch models to separately estimate discrimination parameters for each type of social fear. As previously detailed, larger discrimination parameters are desirable, as they reflect items which efficiently separate people with different levels of underlying social anxiety. Discrimination parameter magnitudes can be interpreted as; <0.20 = very low discrimination; 0.21–0.40 = low discrimination; 0.41–0.80 = moderate discrimination; >0.80 = high discrimination (Baker, 2001). Differences between Rasch and 2PL models were formally tested using a likelihood ratio test (ANOVA) with smaller Bayesian Information Criterion (BIC) reflecting better model fit. Analyses in the NSMHWB 1997 sample were the exception to this method, as fear and avoidance were recorded separately for each situation. Testlets were

1.9. Factor scores

2. Results Percentages of participants in each sample endorsing each situation are detailed in Table 1. Initial dimensional structure of each dataset was established using exploratory factor analysis following guidelines recommended by Norris and Lecavalier (2010) in MPlus version 6.11 (Muthén & Muthén, 2010). Tetrachoric correlations and a robust weighted least squares estimator for categorical indicators were used to estimate models, with oblique rotation allowing correlation between extracted factors. The number of factors to retain was guided by Catell's scree test (the last sharp drop in values representing the number of factors to retain), given other methods such as Kaiser criterion are less reliable (Norris & Lecavalier, 2010). In addition, Root Mean Square Error of Approximation (RMSEA) values were used as a guide for overall model fit, with values between .05 and .08 reflecting reasonable fit, .08 and 1.0 as marginal fit and over 1.0 as unacceptable fit. In all instances, RMSEA values were within acceptable ranges. Scree plots for models of up to four extracted factors indicated a very steep drop after the first factor, creating an elbow between the first and second factors. Whilst there was a slight improvement in overall model fit for two-factor models compared with single factor models in all four samples, a majority of items had cross loadings greater than .30 on both factors suggesting poor differentiation between multiple factors. Therefore, overall outcomes from the scree plots and high cross loadings

E. Crome, A. Baillie / Journal of Anxiety Disorders 28 (2014) 471–479

475

Table 2 Outcomes from one parameter (1PL: Rasch) and two parameter (2PL) item response theory analyses of items assessing types of social situations feared in the National Surveys of Mental Health and Wellbeing (NSMHWB: 1997, n = 1845; 2007, n = 1497), National Comorbidity Survey – Replication (NCS-R, n = 2261) and National Epidemiologic Survey in Alcohol and Related Conditions (NESARC, n = 5411). NSMHWB 07

Rasch (1PL) model 2PL model ANOVA

Public performancea Public speecha Classa Meetinga Meet new Exama Interviewa Authority Party Talk unfamiliar Disagree unfamiliar Enter room Working watched Dating Writea Eat/drinka Bathroom Small group Other fear

NSMHWB 97

NESARC

Log likelihood

BIC

Log likelihood

BIC

Log likelihood

BIC

Log likelihood

BIC

11,334.07 11,204.12 LRT: 259.90

22,777.80 22,612.95 (df = 13, p < .001)

17229.40 17023.32 LRT: 412.16

34,574.65 34,262.90 (df = 13, p < .001)

9282.68 9244.78 LRT: 75.8

18,730.81 18700.13 (df = 6, p < .001)

34,740.54 34,620 LRT: 240.57

34,574.65 34,262.90 (df = 12, P <.001)

b (SE)

Type of fear

NCS-R

b rank

2.49 (.30)

1

a (SE) 0.63 (.08)

b (SE)

b rank

4.16 (.58)

1

a (SE)

b (SE) 1

2

3

a (SE)

0.50 (.07)

–

– 5.68 – 2.34 – – – – 1.86 1.78 – – – – 4.55 2.51 – – 2.494

– 6.10 – 2.49 – – – – 1.98 1.85 – – – – 4.74 2.60 – – 2.604

– 0.20 – 0.78 – – – – 1.16 1.15 – – – – 0.65 1.40 – – 1.223

1.42 1.53 (.12)

2

1.05 (.10)

1.73 (.12)

2

0.97 (.08)

0.66 (.04) 0.62 (.07)

3 4

2.91 (.24) 1.11 (.09)

0.57 (.04) 0.35 (.05)

4 6

2.81 (.18) 1.34 (.08)

0.44 (.05) 0.44 (.04) 0.17 (.04) 0.21 (.05) 0.13 (.04) 0.01 (.05) 0.24 (.05) 0.76 (.06)

7 6 9 8 10 11 12 13

1.60 (.11) 2.05 (.14) 2.67 (.19) 1.63 (.11) 2.15 (.15) 1.43 (.10) 1.48 (.10) 1.39 (.10)

0.41(.04) 0.17 (.04) 0.11 (.03) 0.07 (.04) 0.02 (.04) 0.06 (.04) 0.12 (.04) 0.66 (.05)

5 7 8 9 10 11 12 13

1.50 (.09) 1.82 (.10) 2.41 (.14) 1.44 (.08) 2.00 (.11) 1.50 (.09) 1.55 (.09) 1.55 (.09)

1.48 (.12) – 0.62 (.06)

14 – 5

1.06 (.09) – 1.46 (.11)

1.52 (.11) – 0.59 (.05)

14 – 3

0.90 (.07) – 1.34 (.08)

– 0.69 – – – – 1.23 0.87 – – – – 3.54 2.11 – – 1.543

b (SE) .42 (.03) 1.09 (.04) 0.43 (.02) .15 (.02) – .34 (.03) .56 (.03) .60 (.03) .58 (.03) .25 (.03) – – – 1.53 (.06) 1.70 (.07) 1.78 (.06) – 1.72 (.06) –

b rank

a (SE)

3 1 2 4 – 6 7 9 8 5 – – – 10 11 13 – 12 –

1.81 (.07) 1.58 (.07) 1.99 (.08) 2.12 (.08) – 1.20 (.05) 1.95 (.08) 1.74 (.07) 1.42 (.06) 1.48 (.06) – – 1.33 (.06) 1.18 (.06) 1.52 (.07) – 1.67 (.08) –

Measures are the Composite International Diagnostic Interview (CIDI) versions 2.1 and the World Mental Health (WMH) version, and the Alcohol Use Disorder and Associated Disabilities Interview Schedule-IV (AUDADIS-IV). Bayesian Information Criteria (BIC), Likelihood Ratio Test (LRT), degrees of freedom (df) difficulty parameter (b), discrimination parameter (a), standard error (S.E.). a These items were assessed by one item in the WMH CIDI and two separate items in AUDADIS.

and using a public bathroom the highest. Different types of social fears functioned across a wide range of social anxiety severity, with a majority of items within one standard deviation above or below mean social anxiety levels (62.7% of total information, see Fig. 1). Fears of public performance were only moderately discriminating, with slightly lower discrimination also associated with using a public bathroom or participating in a class/meeting. Distributions of factor scores from this model were approximately normal (skew = 0.02, kurtosis 3 = 0.66), and percentages of people reporting lifetime social phobia incrementally increased from 1.6% to 8.4% across percentiles.

were interpreted as support for underlying assumptions of unidimensionality. In comparison of the Rasch (1PL models) and graded response models (2PL), AVOVAs and inspection of residuals supported the interpretation of 2PL models in all surveys. These results for 2PL models are thus summarised in Table 2 and interpreted below. 2.1. NSMHWB 2007 Consistent with high endorsement rates, public speaking/ performance was associated with the lowest level of social anxiety

1 0.9 0.8

Public Performance Meeting/Class Other Meet New Authority Exam/Interview Party Talk Unfamiliar Disagree Unfamiliar Enter Room Working Watched Dating Write/Eat/Drink Bathroom

0.7

Probability

0.6 0.5 0.4 0.3 0.2 0.1 0 -4

-3

-2

-1

0

1

2

3

4

Situational Social Anxiety Fig. 1. Item characteristic curve for types of social situations feared in the National Survey of Mental Health and Wellbeing 2007.

476

E. Crome, A. Baillie / Journal of Anxiety Disorders 28 (2014) 471–479

1 0.9 0.8

Public Performance Meeting/Class Other Meet New Authority Exam/Interview Party Talk Unfamiliar Disagree Unfamiliar Enter Room Working Watched Dating Write/Eat/Drink Bathroom

0.7 0.6

Probability

0.5 0.4 0.3 0.2 0.1 0 -4

-3

-2

-1

0

1

2

3

4

Situational Social Anxiety Fig. 2. Item characteristic curve for types of social situations feared in the National Comorbidity Survey – Replication.

2.2. NCS-R Results from the NCS-R were similar to those in the NSMHWB 2007, with public speaking/performance fears reflecting lowest levels of social anxiety and using a public bathroom the highest. The range of social anxiety captured assessment by items was larger in the NCS-R; however, this was largely due to public speaking fears performing at much lower levels of social anxiety relative to average social anxiety. A majority of item information was contained within one standard deviation above or below mean social anxiety levels (62.9%, see Fig. 2). Discrimination values in this sample were slightly lower than in the NSMHWB 2007; however, similar patterns of least and most discriminating items remained. Factor scores from this model approximated a normal distribution (skew = 0.08; kurtosis 3 = 0.61), and again, percentages of people reporting lifetime social phobia consistently increased with each factor score percentile (2.5 to 7.8%).

endorsement of both fear and avoidance of public speaking occurred six standard deviations above the sample social anxiety mean (Fig. 3). This relationship was similar to the range of social anxiety captured by fear/avoidance of talking in meetings. Most other items, however, functioned in more discrete ranges, with minimal difference between reporting fear only, avoidance only or both. Item testlets differed in the amount of information captured by each item. Those items assessing fear and avoidance of eating or drinking in public were the most informative, along with attending parties and talking to new people. In contrast, fear of speaking at meetings or in public contained minimal information. Factor scores from this solution were not normally distributed (skew = 1.37, kurtosis 3 = 2.22), and were poor predictors of social phobia diagnosis. This was highlighted by more people in the lowest percentile of this factor met criteria for social phobia in the past 12 months (2.1%) than the highest percentile (0.4%). This indicates that this solution was not too stable, possibly due to a smaller number of items or larger differences between fear and avoidance in some testlets.

2.3. NSMHWB 1997 2.4. NESARC Interpreting item rankings along a social anxiety dimension was more complex within the graded response model. For example, whilst movement between reporting no fear to endorsing a fear of public speaking occurred at relatively low levels of social anxiety;

A majority of items within this survey functioned optimally above average social anxiety, with just over half (52.35%) of the total information occurring within one standard deviation above

0.4 0.3 0.0

0.1

0.2

Information

0.5

0.6

0.7

EAT TALK WRITE MEETINGS PARTY PUBLIC OTHER

-4

-2

0

2

4

Situational Social Anxiety Fig. 3. Item information curve for types of social situations feared in the National Survey of Mental Health and Wellbeing 1997.

E. Crome, A. Baillie / Journal of Anxiety Disorders 28 (2014) 471–479

477

1 0.9 Public Speech

0.8

Class Public Performance

Probability

0.7

Meeting

0.6

Talk Unfamiliar Exam

0.5

Interview

0.4

Party

0.3

Authority Dating

0.2

Write

0.1

Small Group Eat/Drink

0 -4

-3

-2

-1

0

1

2

3

4

Situational Social Anxiety Fig. 4. Item characteristic curve for types of social situations feared in the National Epidemiologic Survey in Alcohol and Related Conditions.

and below average social anxiety. The low endorsement rate of ‘other’ meant minimal information was contained in this item and was thus excluded from analyses. As seen in Fig. 4, items in the NESARC study were fairly evenly spread across the underlying social anxiety dimension. In addition, all items in this sample were highly discriminating, with roughly equivalent curve slopes. Consistent with this, factor scores from this model approximated a normal distribution (skew = 0.24, kurtosis 3 = 0.32). Whilst there was minimal change in the proportion of people meeting lifetime criteria for social phobia within each percentile group, there was a decline in the number of people not meeting lifetime social phobia diagnosis from lowest to highest percentile groups (10.5 to 7.1%). 3. Discussion Overall, the severity ranking of different types of social situations feared was largely consistent across samples and measures. Patterns of items functioning at lower levels (e.g. public speaking, participating in meetings and classes, being the centre of attention) up to higher levels (e.g. eating, drinking or writing in public or using public bathrooms) of social anxiety were comparable between surveys. Where severity rankings varied between surveys, items often functioned within a very narrow range and had overlapping discrimination parameters. This complicates interpretation as the precise order of item difficulties actually changes as a function of the level of underlying social anxiety. However, overall, items such as taking exams or being interviewed typically reflected lower social anxiety compared with other items such as disagreeing with unfamiliar people. Based on these results, items appeared to reflect a range of social anxiety with fears in lower or “mild” ranges including public speaking, participating in meetings or classes, being interviewed or examined and other fears related to being the centre of attention. As social anxiety severity increases to more moderate levels, people are more likely to report fears regarding talking to people in authority, being assertive, talking to unfamiliar people or attending parties. Other items such as anxiety associated with entering a room when others are present, working in small groups, writing, eating or drinking in public or having to use public bathrooms appear to reflect the more severe end of the social anxiety spectrum. As presented earlier, there are numerous practical uses for this spectrum of social anxiety from optimising the development of exposure-therapy modules for population treatments to using fears of situations such as eating or drinking in public as “red-

flags” for people likely to be experiencing high levels of social anxiety. There are also important theoretical implications for these results. Firstly, this dimension appears to suggest a linear relationship between the severity of social anxiety and irrationality of fears. Fears typically falling on the less severe end of the social anxiety spectrum may actually be more realistic, as evaluation in these scenarios is likely to be both more likely and more costly. For example, participating in a meeting or class often involves explicit assessment of an individual's personal ideas or ability. Poor performance in these situations may impact career progression or academic grades. On the other hand, the potential likelihood or cost of negative evaluation in situations such as eating or writing in public is much lower, and thus more unrealistic. More severe social fears, such as eating, drinking or writing in public or using public bathrooms are also more likely to be activities which routinely occur in daily life, and therefore fear and avoidance of these situations is also likely to lead to much higher levels of functional impairment than fear or avoidance of public speaking. Secondly, there are measurement implications for these results, with a large variety in discrimination parameters between surveys despite the consistency in difficulty parameters suggesting vast differences in precision associated with how items have been phrased. As expected, items from the two surveys using the WMH-CIDI had the most similar discrimination parameters, with these different to discrimination parameters of AUDADIS items. Whilst the WMHCIDI covers the highest number of potential social fears, differences in discrimination parameters between the WMH-CIDI and AUDADIS suggest combining situations, as in the WMH-CIDI, occurs at the expense of accuracy. Fears of public speaking and using public bathrooms were consistently less informative than other items, which may be due to various measurement factors (e.g. ceiling effects from very high or low item endorsement) or incomplete theoretical distinction from other concepts (e.g. a fear of contamination associated with obsessive compulsive traits may also have been assessed in fears of using public bathrooms). Results from the NSMHWB 1997 highlight the loss of precision in items such as public speaking fears may also arise from combining fear and avoidance into a single assessment item. When fear and avoidance are assessed separately, a large number of people appear to fear public speaking, yet those who actually go on to avoid these situations appear to have much higher levels of social anxiety. These results contribute to the ongoing debate about potential performance and interaction subtypes in several ways, Firstly, it suggests that differences in performance and interaction fears may

478

E. Crome, A. Baillie / Journal of Anxiety Disorders 28 (2014) 471–479

largely be due to the degree of social anxiety these different fears represent. This is consistent with previous research highlighting that there is no additional predictive value of performance versus other subtypes over a continuum model of social anxiety (Acarturk, de Graaf, Van Straten, Have, & Cuijpers, 2008; Vriends, Becker, Meyer, Michael, & Margraf, 2006). Secondly, that typical performance fears such as public speaking, or speaking at meetings or classes, typically reflect situations at the lowest end of the social anxiety spectrum, highlights the need to examine the usefulness of the “performance-only” subtype in DSM 5 further. The absence (Kerns et al., 2013) or very low prevalence rates (0.8%: Burstein et al., 2011) of people within this subgroup may be due to the fact that people fearing only performance situations are likely to be experiencing a level of social anxiety which is not sufficient to reach diagnostic thresholds. In this case, the “performance-only” specifier provides little information for either clinicians or researchers. Alternately, there may be a small, yet distinct population, who experience severe levels of social anxiety only in public performance situations, with this population qualitatively different to people experiencing public speaking fears below diagnostic thresholds. However, taxometric analyses using indicators based on types of situations feared, including public speaking fears, dating, social interaction, being assertive and interacting with authority figures found no evidence for qualitative differences between these situations (Kollman, Brown, Liverant, & Hofmann, 2006). The presence of a discrete “performance-only” subtype also raises questions about how such severe social anxiety and fear of negative evaluation in one situation does not then generalise to others. Some authors suggest that there are differences in the types of core fears a person may experience which are more complex than a central fear of negative evaluation. Authors such as Moscovitch (2009) theorise that the presentation of social anxiety differs with numerous combinations of potential perceived flaws in social skills or behaviours, controlling and concealing internal feelings of anxiety, physical appearance or characterlogical flaws. Alternately, individuals with performance fears severe enough to meet diagnostic criteria for social anxiety disorder may differ only on lifestyle factors, such as the type of career or work environment, which exacerbate the associated distress and impairment, and thus anxiety severity in a cyclical manner. This highlights the urgent need to understand who and what is captured by the new “performance-only” specifier now included in the international DSM 5 nosology. Whilst the utility of using IRT techniques has been demonstrated in these analyses, these analyses have also highlighted important considerations about applying IRT techniques to psychological constructs. In traditional educational applications, such as math test development, “easy” items are those which are most likely to be answered correctly by the largest number of participants from low to high math ability. This same relationship is observed in IRT for psychological constructs, with “easy” items or criteria being markers for lower levels of disorder severity (Krueger & Piasecki, 2002). In this study, an example of an “easy” item may be public speaking fears. Respondent may interpret public speaking as presenting to an audience of hundreds, which is something likely to be anxiety provoking for many people who would generally not experience anxiety in other social situations. Yet, using this same example of public speaking fears, this “easy” item is also likely to be experienced as one of the most anxiety provoking or “severe” fears for most people more than one social fear. It is hard to imagine many people experiencing a range of social fears would rate a public speech to hundreds as less challenging in a graded exposure hierarchy than filling in a form in a waiting room (writing in public). In this way, the items which appear to are the “easiest” because they reflect lower levels of

severity, may actually the items which will be the most challenging and enduring in treatment. It is also important to highlight that, when using unidimensional IRT techniques such as in this study, often any scale derived from the IRT analyses provides little additional information to the simple summation of items (Anderson, 1999; Xu & Stone, 2012). Therefore, unless being combined in large scale item banks such as in the PROMIS initiative (Cella et al., 2007), once the severity ranking of items is confirmed, the simple summation of number of social situations feared is likely to provide the optimal balance between utility and simplicity. This study has several strengths associated with secondary analysis of existing epidemiological data, including accessing large sample sizes to increase power, using community samples to increase generalisation and ability to compare outcomes over varied measures, samples and timeframes (Ruscio, 2009). Whilst the results from each sample cannot be directly compared, the consistency of these results over different samples adds to the validity of these results. Limitations include potentially unstable results of analyses in the NSMHWB 1997, however, previous research demonstrates as few as five indicators can produce stable IRT models (e.g. McGlinchey & Zimmerman, 2007). The current focus on a single aspect of social anxiety may be considered a strength by some and a limitation by others. On one hand, focusing solely on the types of situations feared overcomes barriers created by constant revisions to diagnostic criteria meaning outcomes from older studies quickly become obsolete (Furmark, 2002). However, focusing exclusively on the types of social fears also excludes items such as the fear of negative evaluation considered to be a central aspect of social anxiety disorder (APA, 2013). It is also unclear whether the wording of situational items would have a profound impact on the rates each of these items were endorsed. Ultimately, whilst there is likely to be individual variation in the perceived severity of different types of social fears, this research highlights a typical ordering to these fears over a range of underlying social anxiety severity. This range of fears from mild public speaking fears through to severe fears such as eating or drinking in public provide an important template for scaling social anxiety treatments in stepped-care or internet formats. It also highlights that people reporting fears at more severe ends of the social anxiety spectrum such as eating, drinking or writing in public, or using public bathrooms may be more likely to experience severe and impairing social anxiety. This range also highlights that the performance and interaction subtypes may be most reflective of different social anxiety severity underlying these feared scenarios. As participants in each survey did not provide individual severity rankings for each fear, rankings of the types of social fears inferred by IRT analyses should be validated in clinical and community samples. An additional focus on understanding differences in severity between fear and avoidance may also provide further information about the severity of different types of social fears. References Acarturk, C., de Graaf, R., Van Straten, A., Have, M., & Cuijpers, P. (2008). Social phobia and number of social fears, and their association with comorbidity, health-related quality of life and help seeking. Social Psychiatry and Psychiatric Epidemiology, 43(4), 273–279. http://dx.doi.org/10.1007/s00127-008-0309-1. American Psychiatric Association ((2000)). Diagnostic and statistical manual of mental disorders: DSM-IV-TR, 4th ed. Washington, D.C: American Psychiatric Association. American Psychiatric Association ((2013)). Diagnostic and statistical manual of mental disorders, 5th ed. Arlington, VA: American Psychiatric Publishing. Anderson, J. O. (1999). Does complex analysis (IRT) pay any dividends in achievement testing? Alberta Journal of Educational Research, 45(4), 344–352. Andersson, G. (2009). Using the internet to provide cognitive behaviour therapy. Behaviour Research and Therapy, 47(3), 175–180. http://dx.doi.org/10.1016/j. brat.2009.01.010.

E. Crome, A. Baillie / Journal of Anxiety Disorders 28 (2014) 471–479 Andrews, G., Henderson, S., & Hall, W. (2001). Prevalence, comorbidity, disability and service utalisation: Overview of the Australian National Mental Health Survey. British Journal of Psychiatry, 178, 145–153. http://dx.doi.org/10.1192/ bjp.178.2.145. Baker, F. B. (2001). The basics of item response theory, 2nd ed. Madison Wisconsin: ERIC Clearinghouse on Assessment and Evaluation. Bögels, S. M., Alden, L., Beidel, D. C., Clark, L. A., Pine, D. S., Stein, M. B., & Voncken, M. (2010). Social anxiety disorder: Questions and answers for the DSM-V. Depression and Anxiety, 27(2), 168–189. http://dx.doi.org/10.1002/da.20670. Burstein, M., He, J.-P., Kattan, G., Albano, A. M., Avenevoli, S., & Merikangas, K. R. (2011). Social phobia and subtypes in the national comorbidity surveyadolescent supplement: prevalence, correlates, and comorbidity. Journal of the American Academy of Child & Adolescent Psychiatry, 50(9), 870–880. http://dx. doi.org/10.1016/j.jaac.2011.06.005. Cella, D., Yount, S., Rothrock, N., Gershon, R., Cook, K., Reeve, B., & Rose, M. (2007). The patient-reported outcomes measurement information system (PROMIS): Progress of an NIH road map cooperative group during its first two years. Medical Care, 45(5 Suppl. 1), S3. http://dx.doi.org/10.1097/01. mlr.0000258615.42478.55. Coles, M. E., Turk, C. L., Jindra, L., & Heimberg, R. G. (2004). The path from initial inquiry to initiation of treatment for social anxiety disorder in an anxiety disorders specialty clinic. Journal of Anxiety Disorders, 18, 371–383. http://dx.doi. org/10.1016/S0887-6185(02)00259-1. Crome, E., Baillie, A., Slade, T., & Ruscio, A. M. (2010). Social phobia: Further evidence of dimensional structure. Australian and New Zealand Journal of Psychiatry, 44 (11), 1012–1020. http://dx.doi.org/10.3109/00048674.2010.507544. DeMars, C. (2010). Item response theory. Oxford: Oxford University Press. Fehm, L., Beesdo, K., Jacobi, F., & Fiedler, A. (2008). Social anxiety disorder above and below the diagnostic threshold: Prevalence, comorbidity and impairment in the general population. Social Psychiatry and Psychiatric Epidemiology, 43, 257–265. http://dx.doi.org/10.1007/s00127-007-0299-4. Feske, U., & Chambless, D. L. (1995). Cognitive behavioral versus exposure only treatment for social phobia: A meta-analysis. Behavior Therapy, 26(4), 695–720. http://dx.doi.org/10.1016/S0005-7894(05)80040-1. Furmark, T. (2002). Social phobia: Overview of community surveys. Acta Psychiatrica Scandinavica, 105(2), 84–93. http://dx.doi.org/10.1034/j.1600-0447.2002.1r103. x. Gelhorn, H., Hartman, C., Sakai, J., Mikulich-Gilbertson, S., Stallings, M., Young, S., & Hopfer, C. (2009). An item response theory analysis of DSM-IV conduct disorder. Journal of the American Academy of Child & Adolescent Psychiatry, 48(1), 42–50. http://dx.doi.org/10.1037/0021-843X.115.4.807. Gillespie, N. A., Neale, M. C., Prescott, C. A., Aggen, S. H., & Kendler, K. S. (2007). Factor and item-response analysis DSM-IV criteria for abuse of and dependence on cannabis, cocaine, hallucinogens, sedatives, stimulants and opioids. Addiction, 102(6), 920–930. http://dx.doi.org/10.1111/j.1360-0443.2007.01804.x. Grant, B., Dawson, D., & Hasin, D. (2001). The alcohol use disorder and associated disabilities interview schedule-DSM-IV version. Bethesda, MD: National Institute on Alcohol Abuse and Alcoholism. Grant, B., Moore, T., Shepard, J., & Kaplan, K. (2003a). Source and accuracy statement: wave 1 national epidemiologic survey on alcohol and related conditions (NESARC). Bethesda, MD: National Institute on Alcohol Abuse and Alcoholism. Grant, B. F., Dawson, D. A., Stinson, F. S., Chou, P. S., Kay, W., & Pickering, R. (2003b). The alcohol use disorder and associated disabilities interview schedule-IV (AUDADIS-IV): Reliability of alcohol consumption, tobacco use, family history of depression and psychiatric diagnostic modules in a general population sample. Drug and Alcohol Dependence, 71, 7–16. http://dx.doi.org/10.1016/S0376-8716 (03)00070-X. Heimberg, R. G. (2002). Cognitive-behavioral therapy for social anxiety disorder: Current status and future directions. Biological Psychiatry, 51(1), 101–108. http:// dx.doi.org/10.1016/S0006-3223(01)01183-0. Hofmann, S. G. (2004). Cognitive mediation of treatment change in social phobia. Journal of Consulting and Clinical Psychology, 72(3), 392–399. http://dx.doi.org/ 10.1037/0022-006X.72.3.392. Ietsugu, T., Sukigara, M., & Furukawa, T. (2007). Evaluation of diagnostic criteria for panic attack using item response theory: Findings from the national comorbidity survey in USA. Journal of Affective Disorders, 104(1–3), 197–201. http://dx.doi.org/10.1016/j.jad.2007.03.005. Karsten, J., Nolen, W. A., Penninx, B. W. J. H., & Hartman, C. A. (2011). Subthreshold anxiety better defined by symptom self-report than by diagnostic interview. Journal of Affective Disorders, 129, 236–243. http://dx.doi.org/10.1016/j. jad.2010.09.006. Katzelnick, D. J., & Greist, J. H. (2001). Social anxiety disorder: An unrecognised problem in primary care. Journal of Clinical Psychiatry, 62(S1), 11–16. Kerns, C. E., Comer, J. S., Pincus, D. B., & Hofmann, S. G. (2013). Evaluation of the proposed social anxiety disorder specifier change for DSM-5 in a treatmentseeking sample of anxious youth. Depression and Anxiety, 30(8), 709–715. http:// dx.doi.org/10.1002/da.22067. Kessler, R. C., Berglund, P., Chiu, W. T., Demler, O., Heeringa, S., Hiripi, E., & Jin, R. (2004). The US national comorbidity survey replication (NCS-R): Design and field procedures. International Journal of Methods in Psychiatric Research, 13(2), 69–92. http://dx.doi.org/10.1002/mpr.167. Kessler, R. C., & Ustun, B. (2004). The world mental health (WMH) survey initiative version of the world health organization (WHO) composite international diagnostic interview (CIDI). International Journal of Methods in Psychiatric Research, 13(2), 93–121. http://dx.doi.org/10.1002/mpr.168.

479

Kollman, D. M., Brown, T. A., Liverant, G. I., & Hofmann, S. G. (2006). A taxometric investigation of the latent structure of social anxiety disorder in outpatients with anxiety and mood disorders. Depression and Anxiety, 23, 190–199. http:// dx.doi.org/10.1002/da.20158. Kraemer, H. C. (2007). DSM categories and dimensions in clinical and research contexts. International Journal of Methods in Psychiatric Research, 16(S1), S8–S15. http://dx.doi.org/10.1002/mpr.211. Krueger, R. F., & Piasecki, T. M. (2002). Toward a dimensional and psychometricallyinformed approach to conceptualizing psychopathology. Behaviour Research and Therapy, 40(5), 485–499. http://dx.doi.org/10.1016/S0005-7967(02)000165. May, A. C., Rudy, B. M., Davis, T. E., Jenkins, W. S., Reuther, E. T., & Whiting, S. E. (2013). Somatic symptoms in those with performance and interaction anxiety. Journal of Health Psychology (July) . http://dx.doi.org/10.1177/1359105313490773 Published online ahead of print. McGlinchey, J., & Zimmerman, M. (2007). Examining a dimensional representation of depression and anxiety disorders’ comorbidity in psychiatric outpatients with item response modeling. Journal of Abnormal Psychology, 116(3), 464–474. http://dx.doi.org/10.1037/0021-843X.116.3.464. Moscovitch, D. A. (2009). What is the core fear in social phobia? A new model to facilitate individualized case conceptualization and treatment. Cognitive and Behavioral Practice, 16(2), 123–134. http://dx.doi.org/10.1016/j. cbpra.2008.04.002. Muthén L.K., & Muthén B.O. (2010). Mplus 6.11 [Computer Software] Los Angeles: Authors. Nordgreen, T., Havik, O. E., Ost, L. G., Furmark, T., Carlbring, P., & Andersson, G. (2012). Outcome predictors in guided and unguided self-help for social anxiety disorder. Behaviour Research and Therapy, 50(1), 13–21. http://dx.doi.org/ 10.1016/j.brat.2011.10.009. Norris, M., & Lecavalier, L. (2010). Evaluating the use of exploratory factor analysis in developmental disability psychological research. Journal of Autism and Development Disorders, 40, 8–20. http://dx.doi.org/10.1007/s10803-009-0816-2. R Development Core Team (2011). R: A language and environment for statistical computing, reference index version 2.13.2. Vienna, Austria: R Foundation for Statistical Computing. http://www.R-project.org. Rizopoulos, D. (2006). ltm: An R package for latent variable modeling and item response theory analyses. Journal of Statistical Software, 17(5), 1–25. Rodebaugh, T. L., Holaway, R. M., & Heimberg, R. G. (2004). The treatment of social anxiety disorder. Clinical Psychology Review, 24(7), 883–908. http://dx.doi.org/ 10.1016/j.cpr.2004.07.007. Ruscio, A. M. (2009). Integrating structural and epidemiological research to inform the classification of psychopathology. International Journal of Methods in Psychiatric Research, 18(4), 240–250. http://dx.doi.org/10.1002/mpr.295. Ruscio, A. M., Brown, T. A., Chiu, W. T., Sareen, J., Stein, M. B., & Kessler, R. C. (2008). Social fears and social phobia in the USA: Results from the national comorbidity survey replication. Psychological Medicine, 38(01), 15–28. http://dx.doi.org/ 10.1017/S0033291707001699. Saha, T. D., Chou, S. P., & Grant, B. F. (2006). Toward an alcohol use disorder continuum using item response theory: Results from the national epidemiologic survey on alcohol and related conditions. Psychological Medicine, 36(07), 931–941. http://dx.doi.org/10.1017/S003329170600746X. Saha, T. D., Compton, W. M., Pulay, A. J., Stinson, F. S., Ruan, W., Smith, S. M., & Grant, B. F. (2010). Dimensionality of DSM-IV nicotine dependence in a national sample: An item response theory application. Drug and Alcohol Dependence, 108 (1–2), 21–28. http://dx.doi.org/10.1016/j.drugalcdep.2009.11.012. Slade, T., Johnston, A., Oakley-Browne, M. A., Andrews, G., & Whiteford, H. (2009). 2007 National survey of mental health and wellbeing: Methods and key findings. Australian and New Zealand Journal of Psychiatry, 43(7), 594–605. http://dx.doi.org/10.1080/00048670902970882. Stein, M. B., & Stein, D. J. (2008). Social anxiety disorder. The Lancet, 371(9618), 1115– 1125. http://dx.doi.org/10.1016/S0140-6736(08)60488-2. Stein, M. B., Torgrud, L. J., & Walker, J. R. (2000). Social phobia symptoms, subtypes, and severity: findings from a community survey. Archives of General Psychiatry, 57(11), 1046–1052. http://dx.doi.org/10.1001/archpsyc.57.11.1046. Steinberg, L., & Thissen, D. (1996). Uses of item response theory and the testlet concept in the measurement of psychopathology. Psychological Methods, 1(1), 81–97. http://dx.doi.org/10.1037/1082-989X.1.1.81. Thomas, M. L. (2011). The value of item response theory in clinical assessment: A review. Assessment, 18(3), 291–307. http://dx.doi.org/10.1177/ 1073191110374797. Van Straten, A., Tiemens, B., Hakkaart, L., Nolen, W. A., & Donker, M. C. H. (2006). Stepped care vs. matched care for mood and anxiety disorders: A randomized trial in routine practice. Acta Psychiatrica Scandinavica, 113(6), 468–476. http:// dx.doi.org/10.1111/j.1600-0447.2005.00731.x. Vriends, N., Becker, E. S., Meyer, A., Michael, T., & Margraf, J. (2006). Subtypes of social phobia: Are they of any use? Journal of Anxiety Disorders, 21, 59–75. http:// dx.doi.org/10.1016/j.janxdis.2006.05.002. Wittchen, H. U., & Fehm, L. (2003). Epidemiology and natural course of social fears and social phobia. Acta Psychiatrica Scandinavica, 108(s417), 4–18. http://dx.doi. org/10.1034/j.1600-0447.108s417.1.x. Composite international diagnositic interview version 2.1. Geneva: World Health Organisation. Xu, T., & Stone, C. A. (2012). Using IRT trait estimates versus summated scores in predicting outcomes. Educational and Psychological Measurement, 72(3), 453– 468. http://dx.doi.org/10.1177/0013164411419846.