Professional Documents
Culture Documents
Caracuel 2012 BI Preliminary Validation FrSBe
Caracuel 2012 BI Preliminary Validation FrSBe
Granada, Spain
Abstract
For personal use only.
Primary objective: To explore the construct validity of the Spanish version of the Frontal Systems Behavioral Scale (FrSBe)
using Rasch modelling.
Methods: Item responses of 245 Spanish subjects were analysed using Rasch analysis: self-rating of 65 participants with
TBI or stroke (sample A), family-rating of the same 65 participants (sample B) and self-rating of 115 healthy individuals
(sample C).
Results: After removing or grouping several problematic items, the Apathy and the Executive Dysfunction sub-scales
were found to be valid measures for samples A and B and the Disinhibition sub-scale was valid for samples B and C. Person
Separation Index of reliability of sub-scales was greater than 0.83 for sample B and 0.72 for A and C. All items showed
disordered threshold categories in samples A and B and five items were ordered in sample C.
Conclusions: With a few modifications, the sub-scales of the FrSBe-Spanish version are adequate measures for the
assessment of the behavioural syndromes derived from frontal systems dysfunction in persons with brain injury. The family-
rating form is preferable to the self-rating form. Only the Disinhibition scale is a valid measure for the behavioural
assessment of the normal population. A reduction of response categories is suggested.
Keywords: FrSBe, Rasch analysis, prefrontal cortex systems, apathy, disinhibition, executive dysfunction, frontal behaviours,
traumatic brain injury, stroke
Correspondence: Alfonso Caracuel, PhD, Facultad de Psicologı́a, Campus de Cartuja, 18071, Granada, Spain. E-mail: acaracuel@ugr.es
ISSN 0269–9052 print/ISSN 1362–301X online ß 2012 Informa UK Ltd.
DOI: 10.3109/02699052.2012.655365
Validation of the FrSBe-Spanish version 845
associated with damage to particular prefrontal- substance abusers and healthy control participants.
subcortical (PFC) systems [6–8]. The scale is However, the scores of individuals with ABI and
composed of 46 items, which are grouped into healthy controls were moderately lower than those
the following three sub-scales: Apathy (14 items, collected for their corresponding groups in the
FrSBe-Ap), Disinhibition (15 items, FrSBe-Dis) and original norms. This was attributed to cultural
Executive dysfunction (17 items, FrSBe-Exe). There differences between Spanish and US samples [36].
is one self-rating form that is completed by the In spite of these positive findings regarding the
patient and one family form that is completed by scale’s reliability, further analysis that employs novel
an informant. Both forms assess behaviour before psychometric tools may improve the construct
(i.e. premorbid or baseline) and after frontal systems validity of the FrSBe-Spanish version. Since Rasch
damage occurred (i.e. morbid or current state). The analysis tests several measurement requirements
manual provides normative data (T scores) from a (construct validity, unidimensionality, appropriate-
sample of 436 healthy adults and their relatives [4]. ness of response categories) [37], it is a highly
The FrSBe has been used for the behavioural recommended method of refining measurement
assessment of a number of disorders involving tools [38]. Although the FrSBe is widely used
damage to PFC systems [9, 10], including frontal in both clinical and research settings, to date, no
and non-frontal ABI [5, 11, 12], Alzheimer’s studies have attempted to apply Rasch analysis to
Disease [13], Parkinson’s Disease [14, 15], multiple this scale. There is considerable scarcity of specific
Brain Inj Downloaded from informahealthcare.com by EBSCO on 05/16/12
sclerosis [16–18], amyotrophic lateral sclerosis [19] measures to assess executive dysfunction in Spanish
and substance use disorders [20–23]. Although it speakers. Some of the measures used in English-
was designed to be administered to neurological speaking countries have been translated, but they
samples, the scale has also been applied to obsessive mostly lack studies on psychometric properties and
compulsive disorder [24], schizophrenia [25, 26] validation. The aim of the current study was to
and secondary psychopathy [27]. Its use has also explore the construct validity of the Spanish version
spread among healthy populations for the assess- of the FrSBe using Rasch analysis in samples of
For personal use only.
ment of the risk of developing alcohol use disorders, participants with brain injury and the normal
maladaptative eating behaviours, financial decisions, population.
empathy and metacognitive awareness [28–35].
Concerning the reliability of the sub-scales,
the manual reports that the Cronbach’s alpha Methods
coefficients in the normal population range between
Participants and settings
0.78–0.87 for the family form and between 0.72–
0.79 for the self-rating form. Slightly greater values Data from the following three samples were ana-
were found for samples of neurological patients, with lysed: (A) Sixty-five participants (14 women) with
the FrSBe-Exe sub-scale showing the greatest ABI (45 TBI, 20 Stroke) aged 15–65 years old
reliability indexes [4, 9]. With regard to validity, (M ¼ 30.11; SD ¼ 12.12) with 8–16 years of educa-
a factor analysis in a normative sample yielded tion (M ¼ 10.23; SD ¼ 2.75). The average time
support to the validity of the three scales. However, incurred since their injury was 22.4 months
items 33 and 43 loaded onto FrSBe-Ap, which is not (SD ¼ 20.55 months). (B) Sixty-five relatives of the
the originally assigned sub-scale for these items [5]. sample A participants (10 men) aged 25–67 years
In addition, a factor analysis of the family-rating old (M ¼ 50.55; SD ¼ 22.34) with 4–16 years of
form in a large neurological patient sample found education (M ¼ 8.23; SD ¼ 3.55). (C) One hundred
that six items (items 6, 23, 33, 36, 40 and 43) loaded and fifteen healthy participants (17 women) aged
onto a factor other than the corresponding sub-scale 18–50 years (M ¼ 30.11; SD ¼ 8.48) with 8–17 years
[10]. Both series of items were maintained in their of education (M ¼ 12.25; SD ¼ 2.65).
original scales, but the authors suggested that some The selection criteria for sample A were:
revision or elimination of specific items may be (i) documented moderate–severe TBI or stroke
warranted to refine the scales and enhance their (i.e. initial Glasgow Coma Scale 13; post-traumatic
validity [9]. amnesia greater than 24 hours or a period of
A Spanish version of the FrSBe, which was unconsciousness longer than 6 hours); (ii) time
approved by the authors and the editors of the incurred since injury over 30 days; (iii) minimum age
scale, has been used for the assessment of persons of 15 years; (iv) the absence of severe language
with ABI and substance abusers in Spanish samples comprehension problems; (v) resides at home and
[22, 36]. The self-rating form of this Spanish version substantially self-reliant in daily life activities; and
had similar reliability values as those found in (vi) availability of a relative who was willing and
US populations. Regarding validity, this version capable of completing the scale. A consecutive series
discriminated between persons with frontal ABI, of 65 patients who met the selection criteria were
846 A. Caracuel et al.
recruited from the outpatient rehabilitation depart- in the category below by a factor of e1 ¼ 2.71 [42].
ment of the Virgen de las Nieves Hospital (Granada, Construct validity under the Rasch analysis is
Spain). They completed the scale during the first determined by examining the hierarchy of the
appointment as part of the typical assessment items in the latent construct as well as by evaluating
protocol. Healthy individuals were selected by its fit to the model [43]. Steps in conducting a Rasch
means of local advertisements and using snowball analysis and its interpretations have been explained
or chain sampling among adults from the commu- in detail by others [37, 44–46]. A brief explanation
nity. The selection criteria for the health sample of the key features of the Rasch analysis that
included the absence of a history of mental retarda- grounded the present study can be found in
tion, learning disability, psychiatric disorders, sub- Kersten et al. [47]. Rasch analysis was performed
stance abuse, neurological disorders or systemic using RUMM2020 software [48] and descriptive
diseases that might affect the central nervous statistics were calculated using SPSS V17.0 for
system. Individuals from samples A and C com- Windows.
pleted the scale by rating their own current
behaviour, whereas sample B participants rated the
current behaviour of their relatives. All participants Results
read, understood and signed an informed consent
prior to completing the scale in the presence of, and Rasch analysis of the whole scale
Brain Inj Downloaded from informahealthcare.com by EBSCO on 05/16/12
if necessary with the assistance of, a clinician. Inspection of the data revealed that only a few
missing data points were spread throughout the
Materials scale, with frequencies lower than 2% for all items.
A Spanish version of the Frontal Systems Behavior Separate analysis for each sample were conducted
Scale (FrSBe) [4] that was adapted and reproduced for the 46 items of the whole scale. The Rating
by special permission of the publisher (Psychological Scale version of the Rasch model was adopted
because there were less than 10 observations in
For personal use only.
Table I. Summary of results of the Rasch analyses of the FrSBe for participants with ABI (sample A).
Correlated
Item-trait Deleted or item Item Fit Person Fit
interaction grouped residuals Residual Residual Reliability Unidimensionality
Sub-scale #Analysis 2 (Df) p items >0.30 Mean (SD) Mean (SD) PSI Sig t-test CI
Apathy
(n ¼ 65) 1st 21.01 (14) 0.09 11 (deleted) 11–39 0.176 (1.2) 0.170 (1.3) 0.70 2.31% (6.1–22.7)
(n ¼ 65) 2nd 13.33 (12) 0.34 41 þ 42 41–42 0.470 (0.88) 0.079 (1.1) 0.74 4.62% (1.5–12.7)
(grouped)
Disinhibition
(n ¼ 65) 1st 21.79 (15) 0.11 0.041 (1.2) 0.185 (1.3) 0.73 6.15% (1.9–15.2)
Executive dysfunction
(n ¼ 65) 1st 31.9 (17) 0.01 36 (deleted) 0.230 (1.6) 0.149 (1.5) 0.71 12.31% (6.1–22.7)
(n ¼ 65) 2nd 29.03 (16) 0.02* 0.224 (1.4) 0.178 (1.5) 0.74 4.62% (1.5–12.7)
# Analysis: number of each analysis. Item-trait interaction, 2, (Df), p: summary chi-square for all items in the sub-scale, indicating the
overall fit to the model; Df: Degree of freedom. Deleted or grouped items: deleted items for item misfit or grouped into a super-item.
Correlated item residuals >0.30: items showing residual correlations above 0.3. Item Fit Residual, Mean and SD: summary of item fit
statistics; Person Fit Residual, Mean and SD: summary of person fit statistics. Reliability PSI: Person Separation Index of reliability.
Brain Inj Downloaded from informahealthcare.com by EBSCO on 05/16/12
Unidimensionality, Sig t-test CI: % of significant t-test at 95% of confidence and Confidence Interval. * non-significant p after Bonferroni
adjustment.
Table II. Summary of results of the Rasch analyses of the FrSBe for relatives (sample B).
Correlated
For personal use only.
Apathy
(n ¼ 64) 1st 18.48 (14) 0.18 0.009 (1.4) 0.088 (1.1) 0.87 3.13% (0.8–10.7)
Disinhinbition
(n ¼ 65) 1st 30.89 (15) 0.009 43 0.213 (1.4) 0.147 (1.1) 0.79 6.15% (1.9–15.2)
(n ¼ 64) 2nd 25.94 (14) 0.2 0.170 (1.3) 0.151 (1.1) 0.83 3.13% (0.8–10.7)
Executive dysfunction
(n ¼ 65) 1st 22.71 (17) 0.15 0.51 (1.0) 0.136 (1.3) 0.86 4.62% (1.5–12.7)
2
# Analysis: number of each analysis. Item-trait interaction, , (Df), p: summary chi-square for all items in the sub-scale, indicating the
overall fit to the model; Df: Degree of freedom. Deleted items: deleted for item misfit. Correlated item residuals >0.30: items showing
residual correlations above 0.3. Item Fit Residual, Mean and SD: summary of item fit statistics; Person Fit Residual, Mean and SD:
summary of person fit statistics. Reliability PSI: Person Separation Index of Reliability. Unidimensionality, Sig t-test CI: % of significant
t-test at 95% of confidence and Confidence Interval.
Table III. Summary of results of the Rasch analyses of the FrSBe for control participants (sample C).
Correlated
Item-trait Deleted or item Item Fit Person Fit
interaction grouped residuals Residual Residual Reliability Unidimensionality
Sub-scale #Analysis 2 (Df) p items >0.30 Mean (SD) Mean (SD) PSI Sig t-test CI
Apathy
(n ¼ 114) 1st 37.2 (15) 0.0006 1 (deleted) 41–42 0.107 (0.97) 0.237 (0.97) 0.72 7.89% (4–14.5)
(n ¼ 114) 2nd 16 (12) 0.16 41 þ 42 0.170 (0.69) 0.210 (0.74) 0.71 6.14% (2.7–12.3)
(grouped)
Disinhinbition
(n ¼ 114) 1st 36.12 (15) 0.001 43 (deleted) 0.125 (0.94) 0.194 (0.87) 0.73 6.14% (2.7–12.3)
(n ¼ 114) 2nd 20.69 (14) 0.1 0.148 (0.92) 0.205 (0.85) 0.72 1.7% (1.6–4.8)
Executive dysfunction
(n ¼ 115) 1st 14.15 (17) 0.6 33 (deleted) 0.006 (1.7) 0.207 (1.1) 0.71 7.83% (4–14.3)
(n ¼ 115) 2nd 16.93 (16) 0.3 17 (deleted) 0.043 (1.3) 0.207 (1.1) 0.72 6.96% (3.3–13.3)
(n ¼ 114) 3rd 12.20 (15) 0.6 0.122 (1.1) 0.253 (1.1) 0.75 7.02% (3.4–13.4)
#Analysis: number of each analysis. Item-trait interaction, 2, (Df), p: summary chi-square for all items in the sub-scale, indicating the
overall fit to the model; Df: Degree of freedom. Deleted or grouped items: deleted items for item misfit or grouped into a super-item.
Brain Inj Downloaded from informahealthcare.com by EBSCO on 05/16/12
Correlated item residuals >0.30: items showing residual correlations above 0.3. Item Fit Residual, Mean and SD: summary of item fit
statistics; Person Fit Residual, Mean and SD: summary of person fit statistics. Reliability PSI: Person Separation Index of reliability.
Unidimensionality, Sig t-test CI: % of significant t-test at 95% of confidence and Confidence Interval. *non-significant p after Bonferroni
adjustment.
sub-scale. There was great concordance on item Differential Item Functioning (DIF). DIF indicates
location between the three samples; therefore, that different sub-groups within the sample respond
combined results from all samples are reported. in a different manner to an individual item, despite
For personal use only.
On the FrSBe-Ap, items 41, 42, 8, 14, 29, 1 and equal severity of the disorder being measured. Sub-
21 achieved the highest negative locations, whereas groups of participants were coded according to the
item 16 had the highest positive location. On the original normative data of the FrSBe [4]: gender,
FrSBe-Dis, the highest negative locations were for education (12 and >12 years), age (39 and >39
items 2, 4, 6, 45, 18, 12, 32 and 10, whereas the years) and diagnosis (only for the ABI sample: TBI
opposite was found for item 31. On the FrSBe-Exe, or Stroke). An ANOVA of person-item deviation
items 3, 25, 36, 7, 13, 37 and 35 were on the residuals with the above-mentioned characteristics
negative side of the latent construct and item 20 had and class intervals as factors was conducted [37]. No
the highest positive location. item showed DIF by the factors used.
Figure 1. Person-threshold map: Distribution of participants with ABI using self-rating form (upper part of the graph) and item thresholds
for the FrSBe-Exe subscale (lower part of the graph).
For personal use only.
Figure 2. Person-threshold map: Distribution of participants with ABI rated by their relatives (upper part of the graph) and item thresholds
for the FrSBe-Exe subscale (lower part of the graph).
using the Rasch model and, thus, the whole scale changes were needed within each sub-scale in order
cannot be considered a unidimensional instrument. to achieve both a perfect fit and unidimensionality.
The main implication of this is that a sum-score of Some of these proposed changes were previously
the 46 items included in the FrSBe does not suggested in the literature. First, during the process
constitute a valid and meaningful measure. Rather of development of the original scale, an independent
than an overall score, Rasch analysis supports the expert rater sorted most of the items into the same
use of a behavioural profile [51] of frontal systems sub-scales proposed by the authors, with the excep-
dysfunction formed by the scores of the three sub- tion of items 11, 20, 31, 33, 34, 35 and 43. Then, a
scales. subsequent factor analysis in a normative sample
The construct validity of the sub-scales for persons showed that five of these items were correctly
with brain injury is supported by the results from assigned by the authors, but items 33 and 43 loaded
Rasch analysis, with the exception of the self-rating onto a sub-scale other than the originally assigned
form of the Disinhibition sub-scale. However, some sub-scale [5]. The current Rasch analysis also found
850 A. Caracuel et al.
Brain Inj Downloaded from informahealthcare.com by EBSCO on 05/16/12
Figure 3. Person-threshold map: Distribution of healthy participants (upper part of the graph) and item thresholds for the FrSBe-Exe
subscale (lower part of the graph).
that several items (items 1, 11, 17, 33, 36 or 43) fail to these items reflect pathognomonic signs of frontal
For personal use only.
fit to the model. These results raise concern about systems disorder, including items 31 (find that food
potential problems with the coherence between has no taste or smell), 20 (make up fantastic stories when
the content of the item and the latent trait [52]. unable to remember something) and 16 (lose control of
The deletion of these items improved unidimension- my urine or bowels and it doesn’t seem to bother me).
ality and reliability of the three sub-scales, supporting This excessive specificity might be problematic
their poor contribution to the latent traits. However, for targeting the sub-scales to healthy or non-
several hypotheses might explain the misfits found neurological populations (see Figure 3).
across samples. For example, problems derived from Regarding the index of reliability, only the
unclear and subjective content, negative wording or relatives’ ratings had a PSI greater than 0.80 to
the use of more than one statement might impact on adequately classify syndrome severity into the three
the items’ behaviour. Furthermore, since grouping meaningful groups of mild, moderate and severe.
items with lack of local independency improved the fit Taking into account construct validity, reliability
of the Apathy sub-scale, it was suggested to reword and targeting of the authorised FrSBe-Spanish
items 41 (get involved with activities spontaneously) version, the family-rating form achieved all require-
and 42 (do things without being requested to do so) into ments of the Rasch model of measurement
a single item. after deleting item 43 (s/he is sensitive to the needs
The construct validity of the sub-scales for normal of other people). There was no DIF by gender,
populations was supported only for the Disinhibition age, education or diagnostic, further supporting the
sub-scale. Unidimensionality for the Apathy and psychometric properties of the scale.
Executive dysfunction sub-scales was not achieved. However, with respect to response categories,
In these healthy participants, some behaviours 41 items from the healthy participants’ sample
contained in these sub-scales may be connected showed disordered thresholds. This lack of order
to dimensions other than the functioning of the indicates that participants failed to use the 5-point
prefrontal cortex systems. scale in a manner that is consistent with the metric
The construct validity of the three sub-scales was estimate of the underlying construct [37]. The
also supported by the hierarchy of items within each fact that, similar to persons with brain injury,
sub-scale. All items on the negative side of the logit healthy individuals’ responses showed disordered
scales were the most representative and common in responses indicates that this limitation of the scale
clinical practice of each latent construct. As further is not linked to the degree of brain damage-
support, several infrequent items fell on the positive related disturbance. Threshold disorder may indi-
extreme of the logit scale and these are the typical cate that more categories exist in the scale than
symptoms of the greatest dysfunctions. Some of are needed to describe the construct [53]. As
Validation of the FrSBe-Spanish version 851
recommended by some authors [52], simplifying the Declaration of Interest: The authors report no
response format can be a useful strategy for conflicts of interest. The authors alone are respon-
improving the precise assessment of behaviour. sible for the content and writing of the paper.
These preliminary findings may yield relevant
clinical and psychometric implications. The FrSBe-
SA is the only tool for the assessment of executive
References
functioning in Spanish individuals with ABI that
has demonstrated fulfilment of the stringent valida- 1. Tagliaferri F, Compagnone C, Korsic M, Servadei F,
tion criteria used by the Rasch analysis. Because Kraus J. A systematic review of brain injury epidemiology
in Europe. Acta Neurochirurgica 2006;148:255–268.
executive dysfunction is one of the most common
2. Stuss DT, Levine B. Adult clinical neuropsychology: Lessons
problems experienced by ABI sufferers, this scale from studies of the frontal lobes. Annual Review of
guarantees an appropriate clinical assessment for Psychology 2002;53:401–433.
a substantial amount of potential Spanish-speaking 3. Nys GMS, van Zandvoort MJE, de Kort PLM, Jansen BPW,
clients. The FrSBe-SA is also a valid measure for the de Haan EHF, Kappelle LJ. Cognitive disorders in acute
stroke: Prevalence and clinical determinants. Cerebrovascular
assessment of rehabilitation outcomes, such that
Disease 2007;23:408–416.
it may allow carrying out comparative studies of 4. Grace J, Malloy PF. Frontal Systems Behavior Scale (FrSBe).
results of rehabilitation programmes implemented in Professional manual. Lutz, FL: Psychological Assessment
Brain Inj Downloaded from informahealthcare.com by EBSCO on 05/16/12
18. Goverover Y, Chiaravalloti N, DeLuca J. The relationship 36. Caracuel A, Verdejo-Garcı́a A, Vilar-Lopez R, Perez-
between self-awareness of neurobehavioral symptoms, cog- Garcia M, Salinas I, Cuberos G, et al. Frontal behavioral
nitive functioning, and emotional symptoms in multiple and emotional symptoms in Spanish individuals with
sclerosis. Multiple Sclerosis 2005;11:203–212. acquired brain injury and substance use disorders. Archives
19. Grossman A, Woolley-Levine S, Bradley W, Miller R. of Clinical Neuropsychology 2008;23:447–454.
Detecting neurobehavioral changes in amyotrophic lateral 37. Tennant A, Conaghan PG. The Rasch measurement model
sclerosis. Amyotrophic Lateral Sclerosis 2007;8:56–61. in rheumatology: What is it and why use it? When should it be
20. Spinella M. Relationship between drug use and prefrontal- applied, and what should one look for in a Rasch paper?
associated traits. Addiction Biology 2003;8:67–74. Arthritis Care and Research 2007;57:1358–1362.
21. Verdejo-Garcı́a A, Bechara A, Recknor EC, Pérez-Garcı́a M. 38. Bond TG, Fox CM. Applying the Rasch Model:
Executive dysfunction in substance dependent individuals Fundamental measurement in the human sciences. 2nd ed.
during drug use and abstinence: An examination of the Mahwah, NJ: Lawrence Erlbaum; 2007.
behavioral, cognitive and emotional correlates of addiction. 39. Rasch G. Probabilistic models for some intelligence and
Journal of the International Neuropsychological Society attainment tests. Chicago, IL: University of Chicago Pr (Tx);
2006;12:405–415. 1980.
22. Verdejo-Garcı́a A, Pérez-Garcı́a M. Substance abusers’ self- 40. Lawton G, Lundgren-Nilsson A, Biering-Sorensen F,
awareness of the neurobehavioral consequences of addiction. Tesio L, Slade A, Penta M, et al. Cross-cultural validity of
Psychiatry Research 2008;158:172–180. FIM in spinal cord injury. Spinal Cord 2006;44:746–752.
23. Verdejo-Garcı́a A, Rivas-Pérez C, López-Torrecillas F, 41. Tesio L. Measuring behaviours and perceptions: Rasch
Pérez-Garcı́a M. Differential impact of severity of drug use analysis as a tool for rehabilitation research. Journal of
on frontal behavioral symptoms. Addictive Behaviors Rehabilitation Medicine 2003;35:105–115.
Brain Inj Downloaded from informahealthcare.com by EBSCO on 05/16/12