Int.J. Behav. Med. DOI 10.1007/s12529-013-9375-1

Dimensional or Categorical Approach to Tinnitus Severity: an Item Response Mixture Modeling Analysis of Tinnitus Handicap Hugo Hesser & Gerhard Andersson

# International Society of Behavioral Medicine 2013

Abstract Background Whether handicap due to tinnitus—sound(s) in the ears and/or in the head in the absence of an external auditory source—is best conceived as dimensional or categorical remains an unanswered empirical question. Purpose The objective was to investigate whether tinnitus severity was best conceptualized as qualitatively distinct subtypes, quantitative differences varying along a single continuum, or as severity differences within subtypes. Methods Various forms of item response mixture models (latent class models, factor analysis models, and hybrid models) that corresponded to the competing hypotheses were fitted to item responses on the Tinnitus Handicap Inventory in a Swedish sample of individuals with tinnitus (N =362). Results A latent class model could be fitted to the data with a high probability of correctly classifying individuals into three different classes: high-, moderate-, and low-severity classes. However, a comparison of models showed that a unidimensional factor analysis model with a single class provided the best fit to the data. Conclusions The analysis provided evidence that tinnitus severity varies along a single severity continuum from mild to moderate to severe tinnitus-related handicap. The result that tinnitus severity exists on a continuum rather than as discrete categories has important implications for clinical research. Keywords Tinnitus . Tinnitus handicap . Item mixture analysis . Tinnitus Handicap Inventory H. Hesser (*) : G. Andersson Department of Behavioural Sciences and Learning, Linköping University, 581 83 Linköping, Sweden e-mail: [email protected] G. Andersson Department of Clinical Neuroscience, Psychiatry Section, Karolinska Institutet, Stockholm, Sweden

Introduction Tinnitus—the experience of sound(s) in the ears and/or in the head without any external auditory source—is a highly prevalent symptom, affecting about 10 to 15 % in the general adult population [1, 2]. Not all who are afflicted by the condition perceive tinnitus as a debilitating symptom; rather, most people will go on to live healthy and productive lives despite of experiencing these sounds. Still, for a significant proportion of those affected (approximately 10 %), the symptom is associated with negative affect, sleep and concentration problems that severely impair their ability to function in everyday situations [3]. Unfortunately for those severely troubled by tinnitus, in the majority of cases, the sounds cannot be eliminated. Numerous therapies, however, have been developed that target severity and distress associated with the symptom [4]. Among these therapies, cognitive and behavioral treatment approaches (CBT) are the most researched, and an increasing body of evidence supports their use in the management of tinnitus distress [5, 6]. It is clear that individuals with tinnitus vary in the degree of tinnitus severity. Yet, whether the population consists of discrete clinical subpopulations or varies with respect to the degree of severity along dimensions of functioning remains an unanswered empirical research question. The current report addresses this important, but up to now, neglected, research target, namely whether tinnitus severity should be conceived as dimensional or categorical. This feature of various phenomena has been investigated and debated extensively in psychiatry and psychology [7, 8] but has received far less attention within audiology. Given significant differences in terms of severity between individuals with tinnitus [3, 9], it seems likely that this heterogeneous population consists of qualitatively different subtypes. To identify such subpopulations—if they truly do exist in the population—is of interest in order to match individuals to subtype-specific treatments or to

Int.J. Behav. Med.

find underlying subtype-specific mechanisms. Indeed, scholars within the field have since long worked under the assumption that identifying subgroups is necessary to find appropriate treatments (e.g., [10, 11]). For example, in a recent proposal for international methodological standards for clinical trials in tinnitus, some of the leading experts in the field stated the following: “Based on the accumulated evidence of results from numerous clinical studies in tinnitus underlining its heterogeneity, it seems more and more likely, that specific sub-forms of tinnitus will benefit from specific treatment interventions.” ([12], p. 113). Accordingly, from this perspective, it is imperative to identify subpopulations within the tinnitus population. Given that most treatments target the impact of tinnitus rather than the sounds, per se, and considering the heterogeneity in terms of tinnitus impairment within the population, one important way of identifying such subgroups is by examining item response patterns on established measures of tinnitus severity. Indeed, to analyze correlations among item responses on severity measures is one way to determine whether a given phenomenon is dimensional or categorical [13, 14]. To date, however, we are only aware of one study that has empirically examined individuals' response patterns on tinnitus severity measures to identify meaningful and qualitatively distinct subpopulations. That study used cluster techniques to group individuals, and the authors concluded that distinct subgroups of individuals with tinnitus existed in the population [11]. Yet, the study does not provide strong support for the existence of subgroups, nor does it demonstrate that tinnitus severity is, in fact, best conceptualized as categorical, rather than dimensional. To make such an inference, one needs to directly compare models with different ways of accounting for the correlations among item responses. In fact, most of the statistical clustering techniques (e.g., cluster analysis, latent class analysis) assume that all of the variance observed among the items should be explained by the cluster/classes. In other words, there should be no variance within clusters. This may not be a reasonable assumption in many situations and may create spurious subgroups [15]. In other words, applying statistical models that assume that observed item responses are in fact due to the existence of homogenous subgroups does not provide evidence for or against the existence of such subgroups. Recent advances in statistics have made it possible to estimate various models with different statistical ways to account for responses on items/outcomes and compare them within a general statistical framework, factor mixture modeling framework [13, 15]. Factor mixture models are hybrid models that combine factor analysis and latent class analysis. In factor analysis, the observed items are clustered and covariance between items is modeled using continuous latent variables, the so-called factors. Individuals are allowed to vary in degree with respect to these factors (e.g., severity), but the

factor model assumes that the observed data is obtained from a single homogenous population. Latent class analysis, in contrast, assumes that the population consists of different subpopulations and uses categorical latent variables to identify homogeneous groups of individuals (similar to cluster analysis). Finally, the factor mixture model is a combination of these models. The model serves to account for unknown population heterogeneity as well as covariance between observed variables. In other words, the factor mixture model allows for variation within class. Thus, these different statistical models correspond directly to the very research question at hand, namely the distinction between categorical and dimensional views on severity [13]. Furthermore, by estimating various models in one general statistical framework, different assumptions can be explicitly tested. Indeed, factor mixture models have been applied in psychiatry [14], developmental psychology [16], addiction [17], among other areas, to shed light on this central issue. In the present study, we adopted this general modeling approach to answer the question of whether tinnitus severity as a phenomenon is best conceptualized as categorical versus dimensional. By examining item responses on the Tinnitus Handicap Inventory (THI [18]), which is one of the most widely used questionnaires to assess the impact of tinnitus on daily functioning, various factor mixture models (i.e., factor analysis models, latent class analysis models, or hybrid models) were applied in exploratory fashion and compared to provide the first empirical evidence on how tinnitus severity is best conceptualized.

Method Participants Participants used in the current analysis were 362 individuals who had visited the audiological clinic at Linköping University Hospital, Linköping, Sweden for tinnitus between 2004 and 2011 and who had completed the THI [18] as part of a larger postal survey conducted in spring 2012. All participants had a confirmed diagnosis of tinnitus. The reasons for visiting the clinic varied between individuals. Some were merely interested in receiving more information about tinnitus, whereas others were actively seeking help for the condition. The clinic uses a step-cared approach to the management of tinnitus. This meant that participants received different treatments. However, at the very minimum, all participants received information about tinnitus in group format. More pertinently, this also meant that participants varied in their degree of severity. The demographic and clinical characteristics of the sample are reported in Table 1. The experiment followed the ethical principles as outlined in the Declaration of Helsinki for human studies. Signed

Int.J. Behav. Med. Table 1 Sample characteristics Characteristics

Statistics

Age in years, mean (SD) Gender female, n (%) Higher education (completed university degree), n (%) Occupational status, n (%) Employed Retired Long-term sick leave Student/unemployed/other Tinnitus duration in years, mean (SD) Tinnitus primary location, n (%)

59.6 (11.64) 173 (47.8) 151 (43.1)

Right Middle Left Tinnitus characteristics, n (%) More than one sound Change in pitch, location, and/or character Sensitive to external sounds, n (%) Hearing impairment, n (%) Problems with dizziness, n (%) Subjective tinnitus loudness, mean (SD) [0 = minimal level; 100 = highest imaginable] Tinnitus severity (Tinnitus Handicap Inventory), mean (SD)

178 (49.6) 151 (42.1) 14 (3.9) 16 (4.5) 12.5 (9.4) 72 (20.8) 157 (45.4) 117 (33.8) 159 (43.9) 327 (91.3) 299 (83.5) 280 (78.9) 185 (51.5) 54.7 (22.5) 39.15 (22.20), range 0−94

N =362. Descriptive statistics are based on completed responses on the particular item/questionnaire

informed consent was obtained from all participants. The regional ethics committee at Linköping University approved the study protocol. Measure THI is one of the most widely used self-report questionnaires for the assessment of the impact of tinnitus in daily life [12, 19, 20] and has documented robust psychometric properties, including high internal consistency [18], test–retest reliability [20], and convergent validity with other established measures of tinnitus severity [21]. The questionnaire has also been evaluated for postal use [21]. Although the measure was developed primarily for classification of tinnitus severity and for diagnostic purposes, the THI has also been frequently employed to document outcomes in treatment trials ([19]; see also, e.g., [22, 23]) The THI consists of 25 items, each is rated on a three-point graded response category scale—0 (“no”), 2 (“sometimes”), and 4 (“yes”)—yielding a total score range from 0 to 100 points. A high score is indicative of a high tinnitus-related handicap. While originally proposed to measure three aspects of tinnitus handicap—emotional, functional, and catastrophic

factor—exploratory factor analysis of the THI has pointed to a unifactorial solution [24–26]. Indeed, the total score is often used in clinical research and has been suggested as a means of distinguishing between individuals with no handicap (0–16 points), mild handicap (18–36 points), moderate handicap (28–56 points), and severe handicap (58–100 points) [20]. A more recent study [27] has advocated a five-category solution for severity grading that also distinguishes between severe (58–76 points) and catastrophic handicap (78–100 points).

Data Analysis To address the aim of the study, we examined item response patterns on the THI—ordered categorical responses on the 25 items—by fitting several different models to the data in the general statistical framework of factor mixture modeling [15, 28, 29]. By estimating different models to the same data within this general framework, we could directly compare the models in terms of relative fit to make an empirically informed decision on whether tinnitus severity should be best conceptualized as continuous or categorical (or a combination of the two). The different models that were estimated corresponded to different assumptions of the underlying structure of item responses. Specifically, factor analysis (FA) (also known as item response theory modeling when the indicators are categorical) uses continuous latent variables (factors), which reflect gradual severity differences. Latent class analysis (LCA), in contrast, uses a categorical latent variable that captures unobserved subgroups in a heterogonous population, by two or more categories, the so-called latent classes. A strong assumption in LCA is that the specified latent classes should explain all of the covariation between observed items. In other words, once indicators are conditioned on the latent class variable, the within-class correlation between items should be zero (i.e., the conditional independence assumption). Factor mixture models (FMM), however, relax this assumption of zero variance within subgroups using a combination of categorical latent and continuous variables, corresponding to subgroups and variation within the subgroups (i.e., covariation between items), respectively. Of note is that all models (LCA and FA) can be generally conceived as FMM [13]. Specifically, FMM with a zero factor variance reduces to LCA, and FMM with only one class reduces to FA [13]. It is also important to note that fitting one of these types of models does not provide support for one over the other. Rather, only a direct comparison of model fit within the same statistical framework can provide evidence for which model actually best captures the item responses. In this particular scenario, a better fit of the FA models would suggest that tinnitus severity is best conceived as continuous, whereas a better fit of LCA models would suggest that it is best conceived as categorical.

Int.J. Behav. Med.

Figure 1 shows schematic presentations of the three overall models that were applied. All analyses were carried out using the Mplus software (vs. 5.2 [30]) with maximum likelihood robust estimation. To ensure that the best solution corresponded to the global rather than to a local maximum likelihood solution, we increased the number of random starts (100 at minimum) in the estimation process [30]. To determine the relative fit of estimated models to observed responses, we made use of the Bayesian information criterion (BIC) and the log likelihood values, where lower BIC values and higher log likelihood values indicate better fit. The BIC is commonly used to select the best-fitted parsimonious model and has been shown to outperform other information criteria in deciding on the number of classes in mixture modeling, even in relatively small sample sizes (N =200) [31]. In comparison with the log likelihood that favors more complex models with more estimated parameters (e.g., models with more classes), the BIC assigns penalties to models with more parameters and is thus the preferred choice, particularly in the case of categorical indicators [15, 31]. We added classes to the model until BIC did not decrease further or until the model failed to converge1, which may indicate that the absolute number of classes that can be extracted from the data has been reached (for practical recommendations, see [32]). In addition to using BIC for deciding on the number of classes, we also employed the entropy value, which ranges between 0 and 1 and describes the uncertainty of classification of subjects into latent classes. Higher probability values (close to 1) indicate less uncertainty, i.e., greater precision of classification. Finally, we determined how readily the response patterns from the estimated classes were conceived as meaningful and interpretable.

Results Preliminary Analyses of the Factor Structure of THI Scores on the THI varied substantially between individuals in the sample, with an actual total score response range from 0 to 94 (M =39.15, SD=22.20). Demographic and clinical variables in the sample were unrelated to the total score on the THI, with the expectation for age that had a small negative correlation with the THI (r =−0.12, p =0.03). An initial exploratory FA (carried out within the structural equation framework) provided evidence for the previous demonstrated unifactorial solution of the THI [24, 25]. In the current sample, all factor loadings were statistically significant 1 Convergence problems can occur because there is insufficient data to estimate some of the parameters in the model. See, for example, the Mplus user's guide [30] for more information about convergence problems, specifically regarding mixtures models.

THI1

THI2

THI2

THI25

THI3

THI4

THI25

THI4

THI25

C

b THI1

THI4

F

a THI1

THI3

THI2

THI3

C

c

F

Fig. 1 Three overall models that were estimated to the observed item responses on the Tinnitus Handicap Inventory (THI). a A factor analysis model (FA) in which individual differences on item responses are captured by an underlying continuous latent variable (F) that allows individuals to vary in degree of severity along one continuum. b A latent class model (LCA) in which individual differences on item responses are captured by a categorical latent variable (C) that classifies individuals into homogenous severity classes. c A factor mixture model (FMM) in which item responses are captured by both a categorical (C) and a continuous latent variable (F). In contrast to LCA that forms classes based on the assumption of independence of items within class, the FMM allows for variation in degree of severity within each class (i.e., correlated items)

(p 0.4). Furthermore, the 2- and 3-factor solution did not have a clear and interpretable loading pattern (with most items ending up on the first factor and several double loadings). Thus, the one-factor structure was retained in the current sample. The internal consistency of the THI was Cronbach's alpha=0.93. Comparison of Models For the main analysis, we estimated several different LCAs and FMMs and compared model fit. Table 2 shows model fitting results for estimated models. First, we fitted LCA models to examine how many classes we could successfully extract from the data. As can be seen in Table 2, both the log likelihood and the BIC improved from 1 class to 2 and from 2 to 3; yet, the 4-class solution failed to converge. Thus, we retained the 3-class LCA. Indeed, the 3-class solution was readily interpretable. Figure 2 shows the 3-class LCA estimates of the item profiles on the THI. As can be seen in the Figure 2, the 3-class LCA contains high-, moderate-, and low-

Int.J. Behav. Med. Table 2 Results from estimated models on item responses on the Tinnitus Handicap Inventory Estimated model

No. of Log parameters likelihood

Latent class analysis 1 class 2 class 3 class 4 class

50 101 152 203

−8685.994 −7409.918 −7095.659 Failed to convergea

Factor mixture model 1 factor, 1 class 75 1 factor, 2 class

Bayesian Entropy information criterion

102

17,666.570 15,414.892 15,086.848 −

− 0.958 0.932 −

−6996.057

14,433.987 –

−6916.759

14,434.465 0.55

N =362. The 1 factor with 1-class model reduces to a simple one-factor analysis model (i.e., item response model). Higher log likelihood values and lower Bayesian information criterion values indicate better fit. Entropy refers to the average probability of correctly classifying individuals into their most likely class membership a

The model did not converge due to a non-positive definite first-order derivative product matrix

severity classes with class percentages of 25, 37, and 38 %, respectively. In addition, the entropy value is high (i.e., small error in misclassification of individuals) for the 3-class LCA. Taken together, the results indicate that individuals can be classified into different classes based on their item responses on the THI and that a 3-class solution provided the best fit among LCA models. However, comparing the best-fitted LCA solution (3 class) with the estimated factor mixture models (1 factor and 1 or 2 classes) revealed that more parsimonious mixture models outperformed the LCA. In fact, the best-fitted model overall was the simple 1 factor with 1-class solution, as the BIC did not decrease further when an additional class was added to the 1-factor 1-class model (see Table 2). In addition, entropy was "high severity class" (25%) "moderate severity class" (37%) "low severity class" (38%)

Item probability (4="YES")

1 0.8

low for the 2-class FMM, indicating a high probability of misclassifying individuals into their most likely class. This FMM with a single class is statistically equivalent to a simple unidimensional FA (i.e., item response theory model). In conclusion, the best fitted-model to the data was a unidimensional FA model and thus provided empirical evidence for tinnitus severity as dimensional rather than categorical.2

Discussion An important, but to date neglected, topic in tinnitus research is the issue of whether tinnitus severity is best conceptualized as categorical or dimensional. Although this has been investigated and debated extensively in other areas, such as in psychology and psychiatry [7, 8], this is, to our knowledge, the first study to address this issue with regard to tinnitus severity. Examining individual differences on item responses on the THI and fitting various models within a general statistical framework that allowed models to be directly compared provided evidence that tinnitus severity could best described in terms of severity differences along a continuum. Importantly, we were able to successfully estimate a 3-class LCA model that characterized the sample in terms of a high-, moderate-, and low-severity class and that classified individuals with a low probability of misclassification (i.e., high entropy). The LCA model, however, assumes that individuals are homogenous within each class, which may not be a reasonable assumption. A more plausible assumption is that individuals vary in degree of severity within a class. Thus, we also examined the data with FMM analyses that allow for variation within a class. Indeed, the best-fitted model to the current sample data was a unidimensional factor solution, that is, a FMM analysis with a single class. We also explored a 2-class FMM, but that model did not improve the model fit beyond the simple factor solution. Thus, the findings provide clear support for the existence of a tinnitus severity continuum. This is the first study to address this issue with regard to tinnitus severity. Given this, it is important to see the results as tentative at this stage. Obviously, our results and inferences are restricted to tinnitus severity, specifically item responses on the THI. It would therefore be valuable to apply a similar

0.6 2

0.4 0.2 0 1

4

7

10

13

16

19

22

25

Item on Tinnitus Handicap Inventory Fig. 2 Response patterns of the three classes from the latent class analysis

Our data analytic approach for identifying the correct number of classes was to add classes in an exploratory fashion using BIC as relative fit measure (see e.g., [32]). Another approach would be to compare several different LCA and FMM analyses based on prior assumptions about the number of latent classes that would be theoretically justified. Given that we made no such assumptions, we used the exploratory approach here. However, it is important to note that both of these approaches yielded results that provided evidence for the same overall conclusion. That is, FMM analyses that allowed for variation within class with 3, 4, and 5 classes provided a worse relative fit than the simple FMM with 1 class or failed to converge.

Int.J. Behav. Med.

data analytic approach in other samples and settings using other measures of tinnitus distress. Replication of the results could strengthen the argument that tinnitus severity is best conceptualized as continuous rather than categorical. It is however important to state that our results do not provide any evidence for or against the contention that subtypes of tinnitus can be identified. Thus, it is important to note that even if tinnitus severity may be viewed as dimensional rather than categorical, the tinnitus population may still consist of discrete homogenous subpopulations, for example, in relation to underlying neural mechanisms involved in sound perception [33]. Still, the finding runs counter to the widespread characterization that dichotomizes individuals with tinnitus into compensated and decompensated tinnitus, that is, distressed and non-distressed individuals (e.g., [34]). In other words, this discrete distinction between individuals may not be an accurate description of the variation in severity observed in the population. The results may have implications for clinical research and practice. For example, the findings may explain why we, to date, despite numerous research efforts, have been fairly unsuccessful in matching individuals to subtype-specific treatments, at least when it comes to treating the adverse effects of tinnitus. If tinnitus severity exists on a continuum, the pursuit of identifying latent (unobserved) subgroups of individuals with tinnitus in terms of severity in order to tailor the treatment to the individual may not be fruitful. Instead, these findings can be seen as a clear invitation to develop evidence-based interventions grounded on principles and processes that vary along dimensions of functioning. Thus, in addition to examining how individuals who are severely troubled by the symptom differ qualitatively from those less troubled, we may need to consider the degree to which individuals differ on various processes. For example, the degree to which individuals use avoidance strategies to cope with the sounds and associated emotional experience has more recently been proposed as a core underlining psychological dimension explaining variation in tinnitus distress and functioning [22, 35, 36]. As such, the tinnitus severity dimension may cut across even wider boundaries and is potentially related to human functioning more broadly. Indeed, other scholars have argued that tinnitus severity shares pathophysiology with other chronic health conditions [37, 38] and psychiatric disorders [39]. Irrespective of the mechanisms underlying tinnitus handicap, given the evidence provided here, it is important to acknowledge the dimensional aspect of tinnitus severity. In addition to the clinical implications of the findings, there are methodological considerations relevant for those aiming to identify subpopulations by statistical means (e.g., LCA) in the area. The findings point at the importance of considering the variation within clusters when using statistical approaches that aim to identify subpopulations among individuals with tinnitus (c.f., [13]). Otherwise, the researcher may end up with spurious

subgroups that may misdirect further research efforts. Furthermore, we believe it is essential that different models may be estimated to the same data so that they may be directly compared and that the aim of such a comparison should be to identify the most parsimonious model that can account for variation in responses. To adopt such a methodological approach will most likely enable the identification of substantive subgroups of individuals with tinnitus and will benefit the research aiming to match individuals to subtype-specific treatments [12]. The study has several limitations. First, despite using a fairly large sample of individuals with tinnitus, it is still in the lower range for the type of analysis conducted (i.e., FMM). Second, although the THI is probably the most widely used questionnaire to assess tinnitus handicap, it has a restricted number of response options. Other measures with a greater number of response options may produce another pattern of results. Third, there are other statistical approaches that have been specifically developed for the purpose of the study, most significantly the taxometric procedure developed by Meehl [40]. However, the model is fairly restrictive as compared with the more general statistical framework of FMM [13]. Fourth, the sample consisted of individuals who had sought help/ information due to their tinnitus, so additional analyses with individuals who have not sought help for the condition are warranted. Of note, however, is that there was no restriction in range on the THI. Still, as previously stated, our findings are in need of replication in other tinnitus samples, in particular using individuals in a younger age range, with early onset tinnitus, and more diverse educational background. Bearing these limitations in mind, this is, to our knowledge, the first study to provide evidence for or against a categorical versus a dimensional approach to tinnitus severity. We conclude that tinnitus severity as a phenomenon is best viewed as dimensional rather than categorical. As the quest to find better and more suitable treatments for those who are suffering from tinnitus moves on, we argue that it is time to consider the dimensional aspect of tinnitus, not just the categorical, which has up to now received considerably more attention in the field. Acknowledgments We wish to thank Ellinor Bånkestad for assisting in collecting the data. The preparation of this manuscript was sponsored in part by a grant from the Swedish Council for Working and Life Research. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Conflict of Interest The authors report no conflict of interest.

References 1. Shargorodsky J, Curhan GC, Farwell WR. Prevalence and characteristics of tinnitus among US adults. Am J Med. 2010;123(8):711–8. 2. Hasson D, Theorell T, Westerlund H, Canlon B. Prevalence and characteristics of hearing problems in a working and non-working

Int.J. Behav. Med.

3.

4.

5.

6.

7. 8.

9.

10. 11.

12.

13. 14.

15. 16.

17.

18.

19.

20.

21.

Swedish population. J Epidemiol Community Health. 2010;64(5): 453–60. Henry J, Dennis K, Schechter M. General review of tinnitus: prevalence, mechanisms, effects, and management. J Speech Lang Hear Res. 2005;48(5):1204–35. Hoare DJ, Kowalkowski VL, Kang S, Hall DA. Systematic review and meta-analyses of randomized controlled trials examining tinnitus management. Laryngoscope. 2011;121(7):1555–64. Hesser H, Weise C, Westin VZ, Andersson G. A systematic review and meta-analysis of randomized controlled trials of cognitivebehavioral therapy for tinnitus distress. Clin Psychol Rev. 2011;31(4):545–53. Martinez-Devesa P, Perera R, Theodoulou M, Waddell A. Cognitive behavioural therapy for tinnitus. Cochrane Database of Systematic Reviews. 2010. doi:10.1002/14651858.CD005233.pub3 Widiger TA, Clark LA. Toward DSM-V and the classification of psychopathology. Psychol Bull. 2000;126(6):946–63. De Boeck P, Wilson M, Acton GS. A conceptual and psychometric framework for distinguishing categories and dimensions. Psychol Rev. 2005;112(1):129–58. Scott B, Lindberg P. Psychological profile and somatic complaints between help-seeking and non-help-seeking tinnitus subjects. Psychosomatics. 2000;41(4):347–52. Langguth B, Salvi R, Elgoyhen AB. Emerging pharmacotherapy of tinnitus. Expert opinion on emerging drugs. 2009;14(4):687–702. Tyler R, Coelho C, Tao P, Ji H, Noble W, Gehringer A, et al. Identifying tinnitus subgroups with cluster analysis. Am J Audiol. 2008;17(2):S176–S84. Landgrebe M, Azevedo A, Baguley D, Bauer C, Cacace A, Coelho C, et al. Methodological aspects of clinical trials in tinnitus: a proposal for an international standard. J Psychosom Res. 2012;73(2):112–21. Lubke GH, Muthen B. Investigating population heterogeneity with factor mixture models. Psychol Methods. 2005;10(1):21–39. Lubke GH, Muthen B, Moilanen IK, McGough JJ, Loo SK, Swanson JM, et al. Subtypes versus severity differences in attention-deficit/ hyperactivity disorder in the Northern Finnish Birth Cohort. J Am Acad Child Adolesc Psychiatry. 2007;46(12):1584–93. Muthen B, Asparouhov T. Item response mixture modeling: application to tobacco dependence criteria. Addict Behav. 2006;31(6):1050–66. Nylund K, Bellmore A, Nishina A, Graham S. Subtypes, severity, and structural stability of peer victimization: what does latent class analysis say? Child Dev. 2007;78(6):1706–22. Neale MC, Aggen SH, Maes HH, Kubarych TS, Schmitt JE. Methodological issues in the assessment of substance use phenotypes. Addict Behav. 2006;31(6):1010–34. Newman C, Jacobson G, Spitzer J. Development of the tinnitus handicap inventory. Arch Otolaryngol Head Neck Surg. 1996;122(2):143–8. Zeman F, Koller M, Figueiredo R, Aazevedo A, Rates M, Coelho C, et al. Tinnitus handicap inventory for evaluating treatment effects: which changes are clinically relevant? Otolaryngol Head Neck Surg. 2011;145(2):282–7. Newman C, Sandridge S, Jacobson G. Psychometric adequacy of the Tinnitus Handicap Inventory (THI) for evaluating treatment outcome. J Am Acad Audiol. 1998;9(2):153–60. Baguley D, Humphriss R, Hodgson C. Convergent validity of the tinnitus handicap inventory and the tinnitus questionnaire. J Laryngol Otol. 2000;114(11):840–3.

22. Hesser H, Gustafsson T, Lunden C, Henrikson O, Fattahi K, Johnsson E, et al. A randomized controlled trial of internetdelivered cognitive behavior therapy and acceptance and commitment therapy in the treatment of tinnitus. J Consult Clin Psychol. 2012;80(4):649–61. 23. Landgrebe M, Binder H, Koller M, Eberl Y, Kleinjung T, Eichhammer P, et al. Design of a placebo-controlled, randomized study of the efficacy of repetitive transcranial magnetic stimulation for the treatment of chronic tinnitus. BMC Psychiatry. 2008;8:23. 24. Baguley D, Andersson G. Factor analysis of the tinnitus handicap inventory. Am J Audiol. 2003;12(1):31–4. 25. Zachariae R, Mirz F, Johansen LV, Andersen SE, Bjerring P, Pedersen CB. Reliability and validity of a Danish adaptation of the Tinnitus Handicap Inventory. Scand Audiol. 2000;29(10718675):37–43. 26. Monzani D, Genovese E, Marrara A, Gherpelli C, Pingani L, Forghieri M, et al. Validity of the Italian adaptation of the Tinnitus Handicap Inventory: focus on quality of life and psychological distress in tinnitus-sufferers. Acta otorhinolaryngologica Italica : organo ufficiale della Società italiana di otorinolaringologia e chirurgia cervico-facciale. 2008;28(3):126–34. 27. McCombe A, Baguley D, Coles R, McKenna L, McKinney C, Windle-Taylor P, et al. Guidelines for the grading of tinnitus severity: the results of a working group commissioned by the British Association of Otolaryngologists, Head and Neck Surgeons, 1999. Clin Otolaryngol Allied Sci. 2001;26(5):388–93. 28. Muthen B, Shedden K. Finite mixture modeling with mixture outcomes using the EM algorithm. Biometrics. 1999;55(2):463–9. 29. Dolan CV, Van der Maas HLJ. Fitting multivariage normal finite mixtures subject to structural equation modeling. Psychometrika. 1998;63(3):227–53. 30. Muthén L, Muthén B. Mplus user’s guide. Los Angeles, CA: Muthen & Muthen; 2008. 31. Nylund KL, Asparoutiov T, Muthen BO. Deciding on the number of classes in latent class analysis and growth mixture modeling: a Monte Carlo simulation study. Struct Equ Modeling. 2007;14(4):535–69. 32. Tofighi D, Enders CK. Identifying the correct number of classes in growth mixture models. In: Hancock GR, Samuelson KM, editors. Advances in Latent Variable Mixture Models. Charlotte, NC: Information Age Publishing; 2008. p. 317–41. 33. Baguley D. Mechanisms of tinnitus. Br Med Bull. 2002;63:195–212. 34. Jastreboff PJ. Phantom auditory perception (tinnitus): mechanisms of generation and perception. Neurosci Res. 1990;8(4):221–54. 35. Hesser H, Molander P, Jungermann M, Andersson G. Costs of suppressing emotional sound and countereffects of a mindfulness induction: an experimental analog of tinnitus impact. PLoS One. 2013;8(5):e64540. 36. Hesser H, Pereswetoff-Morath CE, Andersson G. Consequences of controlling background sounds: the effect of experiential avoidance on tinnitus interference. Rehabil Psychol. 2009;54(4):381–9. 37. De Ridder D, Elgoyhen AB, Romo R, Langguth B. Phantom percepts: tinnitus and pain as persisting aversive memory networks. Proc Natl Acad Sci U S A. 2011;108(20):8075–80. 38. Moller AR. Similarities between severe tinnitus and chronic pain. J Am Acad Audiol. 2000;11(3):115–24. 39. Langguth B, Landgrebe M, Kleinjung T, Sand GP, Hajak G. Tinnitus and depression. World J Biol Psychiatry. 2011;12(7):489–500. 40. Meehl PE. Factors and taxa, traits and types, differences of degree and differences in kind. J Pers. 1992;60(1):117–74.

Dimensional or categorical approach to tinnitus severity: an item response mixture modeling analysis of tinnitus handicap.

Whether handicap due to tinnitus-sound(s) in the ears and/or in the head in the absence of an external auditory source-is best conceived as dimensiona...
228KB Sizes 0 Downloads 0 Views