DOI 10.3233/NRE-2011-0726 IOS Press
Effectiveness of holistic neuropsychological
rehabilitation for Spanish population with
acquired brain injury measured using Rasch
analysis
A. Caracuel
a,b,c,∗, G. Cuberos-Urbano
a, S. Santiago-Ramajo
a, R. Vilar-Lopez
a,b, M.A. Co´
ı
n-Megias
a,
A. Verdejo-Garc´
ı
a
a,band M. P´erez-Garc´
ı
a
a,baDepartment Personalidad, Evaluaci´on y Tratamiento Psicol´ogico, Granada, Spain
bInstituto de Neurociencias F. Ol´oriz, Universidad de Granada, Granada, Spain
cHospital Universitario Virgen de las Nieves, Granada, Spain
Abstract.Introduction: The Rasch model is increasingly used in thefield of rehabilitation because it improves the accuracy of measurements of patient status and their changes after therapy.
Objective:To determine the long-term effectiveness of a holistic neuropsychological rehabilitation program for Spanish
outpa-tients with acquired brain injury (ABI) using Rasch analysis.
Methods: Eighteen patients (ten with long evolution – patients who started the program>6 months after ABI- and eight with
short evolution) and their relatives attended the program for 6 months. Patients’ and relatives’ answers to the European Brain Injury Questionnaire and the Frontal Systems Behavior Scale at 3 time points (pre-intervention. post-intervention and 12 month follow-up) were transformed into linear measures called logits.
Results:The linear measures revealed significant improvements with large effects at the follow-up assessment on cognitive and
executive functioning, social and emotional self-regulation, apathy and mood. At follow-up, the short evolution group achieved greater improvements in mood and cognitive functioning than the long evolution patients.
Conclusions: The program showed long-term effectiveness for most of the variables, and it was more effective for mood and
cognitive functioning when patients were treated early. Relatives played a key role in the effectiveness of the rehabilitation program.
Keywords: Acquired brain injury, stroke, traumatic brain injury, holistic neuropsychological rehabilitation, outcome measures, Rasch analysis
1. Introduction
Holistic Neuropsychological Rehabilitation Pro-grams (HNRP) are multimodal treatments that train compensatory abilities to face residual deficits and dis-abilities after acquired brain injury (ABI) [11]. HNRP increase patients’ independence and the rate of return
∗Corresponding author: Alfonso Caracuel, Facultad de
Psi-colog´ıa, Campus de Cartuja, 18071, Granada, Spain. Tel.: +34 958 242948; Fax: +34 958 243749; E-mail: [email protected].
to independent work [16,28]. These programs have several therapeutic modules for the intensive and pro-longed treatment of cognitive disorders, both emotion-al and behavioremotion-al [17]. For people with moderate or severe ABI these interventions provide the best long-term outcome [18,19,22]. Nevertheless, those reha-bilitation outcomes remain largely unexplored in mi-nority populations, and specifically in Hispanic peo-ple [11], constituting one of the main caveats in neu-ropsychological rehabilitation research. Despite many interesting studies have focused on theoretical guide-lines for the neuropsychological rehabilitation of
turally diverse populations (i.e. [21,32]), the very few studies that have reported the effectiveness of HNRP with minorities have shown lower functional outcomes compared to white Americans [4].
Taxonomies of rehabilitation outcome measures have proposed several unidimensional domains such as physical, neuropsychological (perceptual/cognitive), psychological (personal-behavioral), social (individu-al, family/other), vocational (work), and avocational (life activity) measures [30]. Current health policies emphasize the importance of using patient-centered measures to evaluate the latent constructs that under-lie several of these domains [26]. For this purpose, thousands of questionnaires and scales are used around the world. In these tools, numbers assigned to cate-gory responses involve a certain degree of increase in the amount of the variable under study [20]. However, analysis based on raw ordinal scores raises drawbacks because numbers assigned to item response categories say nothing about distances between categories [51]. Magnitude between pairs of adjacent item response cat-egories, for instance, between “seldom” and “some-times” versus “some“some-times” and “frequently” might be quite different. In addition, these distances are not the same across all items in a scale. Usually, by adding the individual’s responses, researchers get an overall raw score that is often subjected to parametric opera-tions such as mean and standard deviation. However, even though these raw scores may seem to correspond to numbers, they do not [43], and then they are not suitable for such analyses [40]. Using linear scales is a prerequisite to unequivocal statistical analyses, and then results of arithmetic from raw ordinal data could
lead to misinference [29]. However, in clinicalfields
such as rehabilitation, where many questionnaires and scales are needed, raw ordinal scores have come to play a major role in decisions about health care [25].
To address the above mentioned shortcomings of using raw ordinal scores, George Rasch developed a stochastic model that measures latent variables while meeting the fundamental measurement requirements used in the physical sciences [39]. By constructing a Guttmann matrix of the data, Rasch analysis provides statistics capable of estimating two relevant parame-ters that meet the requirements of linear measurement: item difficulty and personal ability. Both can be lo-cated on a common scale or ruler with a linear unit of
measurement, the so calledlogit[1].
For several reasons, Rasch analysis has come to be widely used for the measurement of rehabilitation out-come. First, the Rasch model ensures that the items in
a tool meet basic requirements, such as (1) invariance, i.e., that the item works the same throughout the latent construct and for all individuals and subgroups [3]; (2) unidimensionality, i.e., that a single construct is being measured, and (3) local independence, i.e., that the an-swer to one item does not depend on the response to any other item [35]. For this reason, Rasch analysis is used in the development of new tools and in the valida-tion of those designed using tradivalida-tional approaches [8]. Second, using these validated scales, Rasch analysis can be used to transform raw ordinal scores into lin-ear measures of the person’s ability in the measured construct [45]. These linear measures enable more ac-curate assessments of improvements in patient ability and more accurate comparisons between patients [49]. Third, Rasch analysis detects abnormal response pat-terns [42] that signal invalid results. This is especially relevant in the rehabilitation of individuals with ABI, wherein certain factors can profoundly influence a pa-tient’s responses, such as deficits in awareness, impul-sivity or attention, and reading or understanding disor-ders. Finally, the Rasch model allows equivalent forms to be developed aimed at different raters but that do not need to be parallel. This is crucial for the assessment of anosognosia. Sometimes, different scores of patients and relatives on specific items indicate differential item functioning rather than an awareness disorder [14].
Considering the boundaries of Rasch analyses and the supposed suitability of its application to measure rehabilitation outcomes, the aims of the present study were (1) to determine the effectiveness at one year follow-up of a HNRP with Spanish patients using Rasch
analysis; (2) to use thelogitsobtained through Rasch
analyses to detect differences in outcomes between pa-tients who participated in the program 6 months or less after ABI versus patients with a longer time since injury; and (3) to compare outcomes using raw
ordi-nal scores versuslogits, the linear measures obtained
through Rasch analysis.
2. Methods
2.1. Participants and setting
A sample of 18 patients with traumatic brain injury (TBI) or stroke (and their 18 informal caregivers) was recruited in three waves to form three groups of six in-dividuals, one group every six months. The six patients in each group were selected using accidental sampling,
neuropsy-chology service of the trauma rehabilitation unit at the Virgen de las Nieves Hospital after a certain time and who met the inclusion criteria were included to start each wave. This hospital is a public center of the Na-tional Health Service for patients living in Granada re-gion (Spain). In this trauma rehabilitation unit the an-nual average of patients admitted due to moderate to severe ABI (both TBI and stroke) is 140.
The inclusion criteria for the patients of the present study were as follows: older than 16, a diagnosis of severe ABI due to TBI or stroke, a brain lesion by CT or MRI, scores of at least 1.5 standard deviations below the normative mean on the 3 subscales of the famil-iar version of the European Brain Injury Questionnaire (EBIQ) [41]. ABI of severe intensity was established
as8 points on the Glasgow Coma Scale at
admis-sion or an estimated period of posttraumatic amnesia greater than 7 days [7]. The exclusion criteria were: a psychotic disorder diagnosis, inability to move or a severe language disorder preventing group communi-cation. Also, time of evolution from the ABI was used to delineate three patients with six or fewer months af-ter brain damage and three patients with more than six months after brain damage. At the end of the interven-tion, a mistake was noticed in the date of damage of one patient; therefore, the number of short evolution individuals was only 8 (instead of 9) and the long evo-lution group had 10 people. All informal carers were first-degree relatives, so in the text both terms are used interchangeably. Descriptive data from the patients are in Table 1. The ethics committee of the hospital ap-proved the research, and written informed consent was obtained from each participant after a full explanation of the study.
2.2. Materials
Patients and relativesfilled out parallel forms of the
Spanish version of two instruments. The first, the
EBIQ [41], is an overall measure of outcome wide-ly used in Rehabilitation centers. Three factorial sub-scales labeled depressive mood, cognitive dysfunction, and poor social and emotional self-regulation [12] are used to assess the psychological, neuropsychological and social domains proposed by Mermis [30]. The EBIQ demonstrated a good construct validity to be used with Spaniards, and a Person Separation Index (an in-dex similar to the Cronbach’s alpha reliability coeffi-cient) of 0.90 for the Depressive Mood subscale, 0.88 for the Cognitive Dysfunction subscale, and 0.82 for
the Poor Social and Emotional Self-Regulation sub-scale [12].
Due to the high incidence of frontal lobe damage after ABI and its significant impact on the rehabili-tation process, a second tool was used to specifically assess behavioral symptoms derived from damage to prefrontal circuits: the Frontal Systems Behavior Scale (FrSBe) with three subscales for apathy, disinhibition and executive dysfunction assessment [23]. The FrSBe
showed good reliability (Cronbach’s alpha>0.91) and
proved its validity to assess Spanish patients with ABI and substance abusers [13,36,47].
These six subscales (EBIQ + FrSBe) adequately
cover the most frequent alterations after brain damage in multiple domains. Their level of difficulty is ade-quate for patients with moderate to severe brain dam-age and their relatives [13]. For both tools, the higher the score the worse the patient’s functioning.
The patients and their relatives completed both in-struments at three evaluation points: Pre-intervention (Pre, at the beginning of the program),Post-intervention (Post, after the 6 month program) and Follow-up (Follow-up, 12 months after program completion). All patients and relatives provided data for all three-time points, with no missed cases in the evaluation during both the post-intervention and the follow-up.
Table 1
Time after brain damage and demographics
Group Time after brain Age (years) Education Gender Diagnosis
damage(months) Mean (SD) (years) N (% of males) N (% of TBI)
Mean (SD) Mean (SD)
Short evolution 4.1 (1.2) 27.4 (11.2) 9.8 (2.57) 9 (90%) 5 (50%)
Long evolution 18.3 (5.3) 32.75 (16.18) 10.75 (3.15) 6 (75%) 5 (62.5%)
hour per week program designed to train caregivers as co-therapists was implemented for the whole 6 months. The intervention was delivered in an outpatient basis by a trained neuropsychologist. All the sessions with the patients had the same structure: 15 minutes devot-ed to increase activation and focus on the tion session; 45 minutes of group cognitive rehabilita-tion; a 10 minute break; 40 minutes of individual or by pairs cognitive rehabilitation (depending on the cog-nitive level of the patients); another 10 minute break; 45 minutes of group psychotherapy; and 15 minutes of therapeutic milieu. Towards the end of the 6-month program (the last 6 weeks), the vocational/academic module replaced the individual (or by pair) cognitive rehabilitation module.
2.3. Statistical analyses
The Rasch model produces scale-free person mea-sures [52]. A way to transform raw ordinal data
in-to logitsis to pool data of new participants into
data-banks of similar samples. Therefore, we turned to two databases previously generated in our neuropsychology unit with item responses for the FrSBe and the EBIQ.
Thefirst database contains item responses to self-rating
questions of both tools from 85 Spanish patients, and the second database contains item responses from their 85 relatives. For each of the three evaluation times of the study, data from the 18 patients and caregivers in the study were pooled with data from those 85 pa-tients and relatives. Rasch analyses were conducted with each data pool to obtain linear measures for the three assessment times. According to previous valida-tion studies of the Spanish versions of the EBIQ [12] and FrSBe [13], we had to remove or combine various items from the two forms of each subscale (see Ta-ble 2). Because there was no previous Rasch analysis of the EBIQ-Spanish version family rating form, we ran a factor analysis and then a Rasch analysis of our accumulated data from 85 relatives. The same factor structure as in the patient’s rating form was found [12], and nearly the same items had to be deleted (see Ta-ble 2). The rating scale model was applied for Rasch analysis performed using RUMM2020 software [2].
Variables were tested for normal distribution with the Kolmogorov-Smirnoff test and for homogeneity
of variance with Bartlett’s F test. Mixed analyses of
variance (ANOVAs) 2 (group of evolution, between groups: short vs. long evolution) * 3 (Time of evalu-ation, within-group: pre vs. post vs. follow-up) were
performed. When thep-value of the “group of
evolu-tion * time of evaluaevolu-tion” interacevolu-tion was less than 0.05, post-hoc analysis was conducted focused on comparing the two time of evolution groups at each of the three evaluation times. In these cases, t-tests for indepen-dent samples were used. The effect size was
calculat-ed using Cohen’sdcorrected for dependence between
means using equation no. 8 of Morris and DeShon [31] in cases of repeated measures. Statistical analyses were
conducted usinglogitsand raw ordinal scores. In order
to equate the two types of data, to compare them, those items removed and combined to obtain linear data (Ta-ble 2) were not included in the raw scores. Descrip-tive statistics, factor analysis and hypothesis tests were calculated using SPSS V17.0 for Windows.
3. Results
The short and long evolution groups were matched with respect to demographic characteristics (see Ta-ble 1). Normal distribution was confirmed and t-tests for independent samples revealed that the groups were matched in the study variables and the results of rou-tine cognitive tests before intervention. The average time since injury was four months for the patients in the short evolution group and 18 months for those in the long evolution group.
3.1. Outcomes measured by logits(objective 1)
Usually, the linear logit scale ranges from −3 to
+3 where the more positive thelogitthe more intense
the symptom. The results of the patients and relatives
transformed tologitsare shown in Table 3. The mean
stan-Table 2
Deleted and modified items on the EBIQ and FrSBe subscales Spanish
ver-sion of tool Subscale Deleted items on patient’s form Deleted items on relative’s form Combined pars ofitems EBIQ
(Caracuel et al. 2010)
Depressive Mood 1: Headaches
6: Others do not understand your problems
16: Faintness or dizziness 63: Having problems in general
1: Headaches
16: Faintness or dizziness 63: Having problems in general
18 & 41: Feeling sad&Crying easily
Poor Social &
Emo-tional Self-regulation 27: Annoyance or irritation44: Getting into quarrels easily 3: Reacting too quickly to whatothers say or do 39: Thinking only of yourself
–
Cognitive
Dysfunction 35: Difwhat you wantficulty in communicating – –
FrSBe (Caracuel et al. in press)
Apathy 11: Neglects personal hygiene – 41&42: Gets involved
with activities sponta-neously & Does things without being request-ed to do so
Disinhibition – 43: Is sensitive to the needs of
other people –
Executive
Dysfunction 36: Uses strategies to remember im-portant things –
dard deviations indicated heterogeneity in symptom in-tensity. Social and emotional self-regulation was the
most deficient domains reported by relatives (−0.06)
and patients (−0.23), followed by cognitive
function-ing (relatives=0.12; patients=−0.28). Disinhibition
was the least affected facet for both informants (−0.74
logitsfor relatives and −0.50 for patients). Post-hoc
t-tests showed a significant main effect of time of as-sessment in the reports of both patients and their carers. Relatives reported improvement between baseline and follow-up in all three EBIQ subscales and in the apathy and executive dysfunction subscales of the FrSBe. In two of these subscales (the poor social and emotion-al self-regulation subscemotion-ale of the EBIQ and the apa-thy subscale of the FrSBe), patients’ self-ratings also showed significant changes (see Table 3). Because the
logitis a linear unit, the change inlogitscan be obtained
by subtracting the follow-up results from the baseline results. As reported by relatives, cognitive functioning was the most strongly affected domain in this group
of patients. Subtracting location on thelogitscale at
follow up (−1.27) from location at baseline (0.12) we
discovered that the biggest amount of improvement
measured inlogitshappened in cognitive dysfunction
(1.39logits). Post-hoc analyses showed that significant
improvement in cognitive functioning was reported by caregivers between the post-intervention and follow-up, but not immediately after program. Changes in so-cial and emotional self-regulation reported by patients also happened after the end of the intervention.
3.2. Differential outcomes between the short and long
evolution groups as measured by logits(objective
2)
Two interaction effects between group and time were significant in two EBIQ subscales for both types of informants. Because the results from the analysis of the patients’ ratings are quite similar, we only report the data from relatives. Post-hoc t-tests showed signif-icant differences between the groups at both the
post-intervention (p = 0.000) and follow-up (p =0.005)
evaluation times in the depressive mood subscale (p=
0.004), with the short evolution patients displaying bet-ter mood. The effect size of the difference at
follow-up was large (d =1.71) [34]. The same interaction
effect appeared in the cognitive dysfunction subscale
(p=0.037) at both the post (p=0.009) and follow-up
(p=0.007) times. Again, patients in the shorter
evo-lution group had a lower dysfunction level at follow-up
(d=1.54) (see Table 4 for groups means and standard
deviations).
3.3. Comparison between outcomes obtained using
Rasch analysis linear measures(logits)and those
obtained using raw ordinal scores(objective 3)
All significant improvements found usinglogitswere
also found when raw scores were analyzed. Howev-er, using raw scores revealed several significant
differ-ences that did not appear in the analyses using logits
Ta bl e 3 Res ults from the E B IQ and the Fr SB e of patients and relati ve s in a linear scale us ing the unit called lo git . 1. Pre: pre-interv ention; 2. Pos t: pos t-interv ention; 3. Fo llo w -up: 12 m onths af ter the inter vention. L inear m eas ur es w er e us ed to calcula te m eans and st andard de viations and the y are expres se d in lo gits ,t herefore the m ore pos iti ve the lo git ,t he m ore intens e the sy mp to m. Si gn i fi cant dif fe re nces betw een tim es ar e in bold. U nder lining indicates si gni fi cant dif ferences that appear only w hen us ing ordinal ra w data. d : C ohen’ s d . E ff ect si zes w ere calculated only betw een the pre-interv ention and follo w -up tim es . C ohen’ s d w as cor rected fo r dependence betw een m eans us ing Morris and DeShon’ s (2002) equation no. 8 T im es A m ount t -t es ts of m ain ef fect of tim e 1Pre 2Pos t 3F ollo w-up of change in
∗(p<
0. 001) M ean (S D ) M ean (S D ) M ean (S D ) lo gits : 3F ollo w -U sing lo gits Usi ng ra w sc ore s Lo gits Ra w Lo gits Ra w Lo gits Ra w up m inus 1Pre EB IQ De pr essi ve Mood Pa tients ( n = 18) − 0. 46 (0. 97) 35. 28 (1. 99) − 0. 74 (0. 88) 31. 70 (1. 34) − 1. 24 (0. 94) 28. 66 (1. 34) 0. 78 Pre = Post = Fo llo w -up ( d = 0. 54) Pre = Post = Fo llo w -up ( d = 1. 08) Relati ve s ( n = 18) − 0. 26 (0. 83) 36. 29 (1. 56) − 0. 71 (0. 87) 32. 43 (1. 57) − 1. 46 (0. 94) 27. 14 (1. 39) 1. 2 Pr e = F ollo w -u p ∗ ( d = 1. 36) Pr e = F ollo w -u p ∗ ( d = 1. 64) E B IQ Cogniti ve Dys function Pa tients ( n = 18) − 0. 28 (0. 79) 39. 72 (1. 78) − 0. 81 (1. 10) 34. 85 (1. 78) − 1. 32 (0. 96) 30. 93 (1. 34) 1. 04 Pre = Post = Fo llo w -up ( d = 0. 81) Pre = Post = Fo llo w -up ( d = 1. 14) Relati ve s ( n = 18) 0. 12 (0. 66) 43. 97 (1. 43) − 0. 62 (0. 94) 36. 94 (1. 89) − 1. 27 (0. 99) 31. 57 (1. 55) 1. 39 Pre = Po st = F ollo w -u p ∗ ( d = 1. 84) Pr e = Po st = F ollo w -u p ∗ ( d = 2. 0) E B IQ Poor So cial and E m otional Se lf-Re gulation Pa tients ( n = 18) − 0. 23 (0. 84) 26. 81 (1. 41) − 0. 65 (0. 87) 23. 82 (1. 22) − 1. 38 (0. 99) 19. 98 (1. 07) 1. 15 Pre = Po st = F ollo w -u p ∗ ( d = 0. 88) Pre = Po st = F ollo w -u p ∗ ( d = 1. 33) Relati ve s ( n = 18) − 0. 06 (1. 35) 28. 16 (1. 75) − 0. 76 (0. 82) 23. 27 (1. 19) − 1. 23 (0. 68) 20. 12 (0. 85) 1. 17 Pr e = F ollo w -u p ∗ ( d = 1. 35) Pr e = F ollo w -u p ∗ ( d = 1. 4) Fr SB e A pathy
Patients (n=
cognitive dysfunction subscale of the EBIQ and the ap-athy subscale of the FrSBe indicated that improvements attributable to the program were detected at different assessment times (see Table 3). Second, significant improvements in the disinhibition (relative’s ratings) and executive dysfunction (patient’s ratings) subscales of the FrSBe only appeared when using raw scores. Third, the above-mentioned significant interaction ef-fects indicating that the short evolution group achieved better mood and cognitive outcomes at follow-up were not found using raw scores. However, a similar
signifi-cant interaction effect (p=0.017) was found using raw
scores from patients’ apathy FrSBe subscale ratings at
both post-intervention (p=0.043) and follow-up (p=
0.016). Finally, most of effect sizes are greater when raw scores are used.
4. Discussion
The main objective was to determine the effective-ness of holistic neuropsychological rehabilitation as public assistance for Spanish patients with a brain
in-jury acquired from TBI or stroke. Using thelogitscale
for assessing the amount of change between baseline and other (post or follow-up) evaluation times, accord-ing to the relatives’ rataccord-ings, cognition changed more than any other domain, followed by depressive mood, social and emotional self-regulation, apathy, executive function and disinhibition. Those changes can be eas-ily related to each therapeutic module of the holis-tic program. Cognitive rehabilitation was the largest module, and the greatest improvements in the patients were in cognitive status. Because rehabilitation of ex-ecutive functions was also included on the cognitive module, changes in executive dysfunction are also pri-marily connected to this module. Another relationship might exist between the psychotherapy module and re-ductions in depressive mood in the patients. We tried to prompt changes in apathy through the whole pro-gram but especially with the vocational module. Be-havioral improvements associated with social and emo-tional self-regulation can derive from the group format of most therapeutic activities. Significant changes in disinhibition status did not occur, possibly because this area was a less deteriorated facet of the patients.
Regarding the effectiveness of the caregiver inter-vention module, it may be connected to the changes in cognitive functioning (reported by relatives) and so-cial and emotional self-regulation (reported by patients) that occurred between the post-intervention and
follow-up evaluations. During these 12 months, trained rela-tives are the patient’s only support. This module aimed at emotional support and trained relatives in areas such as behavior modification techniques, assertiveness and relaxation. All this learning may have contributed to the effectiveness of the program because it helped rel-atives to act as co-therapists during the program and later as expert caregivers. Although studies support the cost-effectiveness of different neuropsychological rehabilitation formats [15,37,50], there are still many countries that do not include this type of assistance in their national health system. The present study sup-ports the implementation of holistic neuropsycholog-ical rehabilitation in Spain and highlights the critneuropsycholog-ical role that the family plays on it. Nevertheless, the gener-alizability of our results to other groups of Hispanics is unknown. Considering the numerous differences found between Hispanic groups [9], more research is
guar-anteed before extending ourfindings to other Latinos,
including those living in the US (i.e., the importance of family support in the Hispanic communities could be diminished in immigrants whose families reside in their origin country).
The second aim was to determine differences in out-comes obtained by patients who were included in the program within 6 months of their ABI versus patients with a longer evolution time. The authors support the hypothesis that it is better to start neuropsychological rehabilitation in the acute phase of ABI [48]. We found that the long-term cognitive functioning and mood sta-tus outcomes, outside of any spontaneous recovery ef-fects, were better for those patients who were involved in a holistic program early. It is known that the severity of behavioral and emotional problems increases after the seventh month following severe ABI [27]. Starting
a holistic intervention within thefirst six months might
associ-ated with severe damage for many months. A follow-up evaluation at 12 months allows the maintenance of improvements after the intervention to be assessed and eliminates the effects of spontaneous recovery at the post-intervention assessment in the short evolution group. More important than the statistical significance is the magnitude of the differences between the short and long evolution groups found at follow-up. Such findings are relevant given their impact on the func-tioning and quality of life of patients and their families. Just as early occupational therapy, speech and language pathology, and physical therapy are clearly beneficial to patients [5], this study supports the early implemen-tation of holistic neuropsychological rehabiliimplemen-tation.
To achieve the third objective,findings obtained
us-ing raw scores andlogits were compared. The tools
used to obtain the data have been validated methodolog-ically with both Classical Test Theory and the Rasch model of the Item Response Theory. Notably, the effect of early intervention on mood and cognitive domain
appeared usinglogitsbut not with raw scores. This
dif-ference may result from a logistic relationship between
the measures expressed inlogitsversus raw scores. The
two measures are related by an S-shaped logistic curve, indicating that there is high correspondence between the two data sets in the center of the continuum. As an example, when a treatment produces “x” amount of improvement, patients who are in the middle range of
the variable get the same value with bothlogitsand raw
score. However, for patients at the extremes of the vari-able, the same “x” amount of improvement corresponds
to a lower value in raw scores than inlogitsdue to the
S-shape of the function [24]. Using tools validated by the Rasch model and Rasch analysis of data, all patients are measured with same standard across the whole con-struct [44]. The differences detected between groups may have resulted from more accurate measurement of mood and cognitive function of subjects in different ranges of the variable. This possibility is supported by the heterogeneity of symptom severity in the study sample. There are large standard deviations for all the variables, indicating that patients are at scattered po-sitions throughout the constructs. As this diversity is very prevalent in the clinical practice of ABI, it is im-portant to have methods to increase the accuracy and equality of the assessment of changes in individuals in a treatment group, regardless of the severity of their symptoms or the position they occupy in the construct. Having the same guarantees of quality for the raw
scores and for the data transformed intologits, several
reasons prompt us to think that the results usinglogits
are more accurate. First, because Rasch model studies have repeatedly indicated this accuracy [20,43].
Sec-ond, becausefindings fromlogitsclearly support
rel-evant clinical hypotheses that have been demonstrat-ed. One such hypothesis is that intervention target-ing relatives is one of the main keys to long-term im-provements in neuropsychological rehabilitation pro-grams [46]. Another hypothesis is that early interven-tion provides better long-term outcomes [5,33]. Sup-porting this second hypothesis, evidence that patients’ assessments of apathy are reduced through early inter-vention was only obtained using raw scores. We have no clear explanation for this. However, as apathy is a
subjective concept and thisfinding was found only in
family reports, it might be questionable.
The main limitations of this study concern its small percentage of women, its small sample size and its lack of control over potentially relevant variables of
patients and relatives. Despite thefirst limitation, it is
still possible to compare the results with other studies because the imbalance between men and women is the same as shown in most research [22]. In future studies, we aim to increase the number of subjects and the percentage of women and to control for variables such as patient personality traits, age, emotional state and coping strategies of relatives.
The authors declare no conflicts of interest.
5. Conclusions
This study supports the implementation of holis-tic neuropsychological rehabilitation programs for pa-tients with acquired brain injury. Long-term outcomes rated by relatives indicate patient improvements in cog-nitive and executive functioning, mood, apathy and so-cial and emotional self-regulation. The results high-light the critical role that a trained relative plays in ob-taining and mainob-taining long-term improvements. This study also supports the early implementation of neu-ropsychological rehabilitation. Mood and cognition were the domains that benefited most from early holis-tic intervention. Research on effectiveness has the po-tential to impact decision-making in health services; therefore it is worth applying Rasch analysis to improve
the accuracy offindings.
References
[2] Andrich, D., Lyne, A., Sheridan, B., & Luo, G. (2003). RUMM2020. Perth: RUMM Laboratory.
[3] Andrich, D. (1988).Rasch Models for Measurement. London: Sage Publications.
[4] Arango-Lasprilla, J. C., Rosenthal, M., DeLuca, J., Cifu, D. X., Hanks, R., & Komaroff, E. (2007). Functional out-comes from inpatient rehabilitation after Traumatic Brain In-jury: How do Hispanics fare?Archives of Physical Medicine and Rehabilitation, 88,11-18.
[5] Bennett, T. L. (2001). Neuropsychological evaluation in reha-bilitation planning and evaluation of functional skills.Archives of Clinical Neuropsychology, 16(3), 237-253.
[6] Ben-Yishay, Y. (2000). Postacute neuropsychological rehabil-itation. A holistic perspective. En A.-L. Christensen & B. P. Uzzell (Eds.).International handbook of neuropsychological rehabilitation(pp. 127-135). New York: Kluwer Academ-ic/Plenum Publishers.
[7] Bond, M. R. (1990). Standardized methods of assessing and predicting outcome. En Rosenthal, M., Bond, M. R., Griffith, E. R., & Miller, J. D. (Eds.).Rehabilitation of the adult and child with traumatic barin injury(2nd. ed.). Philadelphia: F.A.
Davis.
[8] Bond, T. G., & Fox, C. M. (2007).Applying the Rasch Model: Fundamental Measurement in the Human Sciences(2nd. ed.).
Mahwah. NJ: Lawrence Erlbaum.
[9] Bur´e-Reyes, A., Puente, A. E., Gontier. J., S´anchez, L., Hi-dalgo, N., Vilar-L´opez, R,. & P´erez-Garc´ıa, M. (2008). Neu-ropsychological test performance of four different Spanish speaking groups.Archives of Clinical Neuropsychology, 23, 681.
[10] Caetano, C., & Christensen, A.-L. (1999). Outpatient/Day Pa-tient Rehabilitation at the Centre for Rehabilitation of Brain Injury, Copenhagen Denmark.Neuropsychological Rehabili-tation: An International Journal, 9(3), 447.
[11] Caracuel, A., Perez-Garcia, M., Salinas-S´anchez, I., Asensio-Avil´es, M. M., S´anchez-Casta˜no, J. M., & P´erez-Ure˜na, M. B. (2005). Datos preliminares de la adaptaci´on a un servicio de rehabilitaci´on p´ublico de un programa hol´ıstico de reha-bilitaci´on neuropsicol´ogica para pacientes con da˜no cerebral adquirido.Rehabilitaci´on, 39(3), 95-102,
[12] Caracuel, A., Bateman, A., Teasdale, T. W., Verdejo-Garc´ıa, A., & P´erez-Garc´ıa, M. (2011). Spanish, French, and British Cross-cultural Validation of the European Brain Injury Ques-tionnaire,The Journal of Head Trauma Rehabilitation, 26(6), 478-488.
[13] Caracuel, A., Verdejo-Garc´ıa, A., Fern´andez-Serrano, M. J., Moreno-L´opez, L., Santiago-Ramajo, S., Salinas-S´anchez, I., & P´erez-Garc´ıa, M. (in press). Preliminary validation of the Spanish version of the Frontal Systems Behavior Scale (FrS-Be) using Rasch analysis.Brain Injury.
[14] Chan, R. C. K., & Bode, R. K. (2008). Analysis of patient and proxy ratings on the Dysexecutive Questionnaire: an applica-tion of Rasch analysis.Journal of Neurology, Neurosurgery & Psychiatry, 79(1), 86-88.
[15] Christensen, A.-L. (2005). Neuropsychological experiences in neurotraumatology.Acta Neurochirurgica Supplementum, 93, 195-198.
[16] Christensen, A.-L., Caetano, C., & Rasmussen, G. (1996). Psychosocial outcome after an intensive, neuropsychological-ly oriented day program: contributing program variables. In Uzzell B.P. & Stonnington H.H. (Eds.).Recovery after trau-matic brain injury(pp. 235-255). Mahwah: Lawrence Erl-baum Associates.
[17] Christensen, A.-L., & Uzzell, B. P. (2000).International hand-book of neuropsychological rehabilitation. New York: Kluwer Academic/Plenum Publishers.
[18] Cicerone, K. D., Dahlberg, C., Kalmar, K., Langenbahn, D. M., Malec, J. F., Bergquist, T. F., Felicetti, T., et al. (2000). Evidence-based cognitive rehabilitation: Recommendations for clinical practice.Archives of Physical Medicine and Reha-bilitation, 81(12), 1596-1615.
[19] Cicerone, K. D., Dahlberg, C., Malec, J. F., Langenbahn, D. M., Felicetti, T., Kneipp, S., Ellmo, W., et al. (2005). Evidence-based cognitive rehabilitation: Updated review of the literature from 1998 through 2002.Archives of Physical Medicine and Rehabilitation, 86(8), 1681-1692.
[20] Fisher, W. J. (2004). Ordinal vs. Ratio revisited again.Rasch Measurement Transactions, 18(2), 980-982.
[21] Fletcher-Janzen, E., Strickland, T., & Reynolds, C. R. (2000). Handbook of Cross-Cultural Neuropsychology. New York: Springer.
[22] Geurtsen, G., van Heugten, C., Martina, J., & Geurts, A. (2010). Comprehensive rehabilitation programmes in the chronic phase after severe brain injury: A systematic review. Journal of Rehabilitation Medicine, 42(2), 97-110.
[23] Grace, J., & Malloy, P. F. (2001).Frontal Systems Behavior Scale(FrSBe). Professional Manual.Lutz, FL: Psychological Assessment Resources, Inc.
[24] Hagquist, C., Bruce, M., & Gustavsson, J. P. (2009). Using the Rasch model in nursing research: An introduction and illustrative example.International Journal of Nursing Studies, 46(3), 380-393.
[25] Hobart, J., & Cano, S. (2009). Improving the evaluation of therapeutic interventions in multiple sclerosis: the role of new psychometric methods.Health Technology Assessment (Winchester, England), 13(12), iii, ix-x, 1-177.
[26] Hobart, J. C., Lamping, D. L., Freeman, J. A., Langdon, D. W., McLellan, D. L., Greenwood, R. J., & Thompson, A. J. (2001). Evidence-based measurement: Which disability scale for neurologic rehabilitation? Neurology, 57(4), 639-644. [27] Johnson, R., & Balleny, H. (1996). Behaviour problems
af-ter brain injury: incidence and need for treatment.Clinical Rehabilitation, 10(2), 173 -180.
[28] Malec, J. F., Smigielski, J. S., DePompolo, R. W., & Thomp-son, J. M. (1993). Outcome evaluation and prediction in a comprehensive-integrated postacute outpatient brain injury re-habilitation programme.Brain Injury, 7, 15-29.
[29] Merbitz, C., Morris, J., & Grip, J. C. (1989). Ordinal scales and foundations of misinference.Archives of Physical Medicine and Rehabilitation, 70(4), 308-312.
[30] Mermis, B. J. (2005). Developing a Taxonomy for Reha-bilitation Outcome Measurement.Rehabilitation Psychology, 50(1), 15.
[31] Morris, S. B., & DeShon, R. P. (2002). Combining effect size estimates in meta-analysis with repeated measures and independent-groups designs. Psychological Methods, 7(1), 105-125.
[32] Niemeier, J. P., Burnett, D. M., Whitaker, D. (2003). Cul-tural competence in the rehabilitation setting: Are we falling short of meeting needs? Archives of Physical Medicine and Rehabilitation, 84, 1240-1245.
[33] Niemeier, J. P., Kreutzer, J. S., & Taylor, L. A. (2005). Acute cognitive and neurobehavioural intervention for individuals with acquired brain injury: Preliminary outcome data. Neu-ropsychological Rehabilitation, 15(2), 129-146.
Journal of Consulting and Clinical Psychology.Journal of Consulting and Clinical Psychology, 78(3), 287-297. [35] Pallant, J. F., & Tennant, A. (2007). An introduction to the
Rasch measurement model: an example using the Hospital Anxiety and Depression Scale (HADS).The British Journal of Clinical Psychology, 46(Pt 1), 1-18.
[36] Pedrero-P´erez, E. J., Ruiz-S´anchez de Le´on, J. M., Llanero-Luque, M., Rojo-Mota, G., Olivar-Arrollo, A., & Puerta-Garc´ıa, C. (2009). Sintomatolog´ıa frontal en adictos a sus-tancias en tratamiento mediante la versi´on espa˜nola de la es-cala de comportamiento frontal.Revista de Neurolog´ıa, 48, 624-631.
[37] Prigatano, G. P., & Pliskin, N. H. (2002).Clinical neuropsy-chology and cost outcome research: A beginning. New York: Psychology Press.
[38] Prigatano, G. P., Fordyce, D. J., Zeiner, H. R. Roueche, J. R., Pepping, M. & Wood. B. C. (1986).Neuropsychological Re-habilitation after Brain Injury. Baltimore: The Johns Hopkins University Press.
[39] Rasch, G. (1980).Probabilistic Models for Some Intelligence and Attainment Tests. Univ of Chicago Pr (Tx).
[40] Siegal, S., & Castellan, N. J. (1988).Nonparametric Statis-tics for the Behavioral Sciences(Secon edition.). New York: McGraw Hill.
[41] Teasdale, T. W., Christensen, A.-L., Willmes, K., Deloche, G., Braga, L., Stachowiak, F., Vendrell, J. M., et al. (1997). Subjective experience in brain-injured patients and their close relatives: A European Brain Injury Questionnaire study.Brain Injury, 11(8), 543-563.
[42] Tennant, A., & Conaghan, P. G. (2007). The Rasch measure-ment model in rheumatology: What is it and why use it? When should it be applied, and what should one look for in a Rasch paper?Arthritis Care and Research, 57(8), 1358-1362. [43] Tesio, L, (2007a), Functional assessment in rehabilitative
medicine: Principles and methods.Europa Medicophysica, 43(4), 515-523.
[44] Tesio, L. (2007b). Rehabilitation and outcome measurement: where is Rasch analysis-going? Europa Medicophysica, 43(3), 417-426.
[45] Tesio, L. (2003). Measuring behaviours and perceptions: Rasch analysis as a tool for rehabilitation research.Journal of Rehabilitation Medicine: Official Journal of the UEMS Euro-pean Board of Physical and Rehabilitation Medicine, 35(3), 105-115.
[46] Vangel, S. J., Rapport, L. J., & Hanks, R. A. (2011). Effects of family and caregiver psychosocial functioning on outcomes in persons with traumatic brain injury.Journal of Head Trauma Rehabilitation, 26(1), 20-29.
[47] Verdejo-Garc´ıa A., Rivas-P´erez, C., Vilar-L´opez, R., & P´erez-Garc´ıa, M. (2007). Strategic self-regulation, decision-making and emotion processing in poly-substance abusers in theirfirst year of abstinence.Drug and Alcohol Dependence, 86, 139-146.
[48] Wilson, B. A. (2008). Neuropsychological rehabilitation. An-nual Review of Clinical Psychology, 4, 141-162.
[49] Wilson, M. (2005).Constructing measures: An item response modelling approach. London: Lawrence Erlbaum Associates. [50] Worthington, A. D., Matthews, S., Melia, Y., & Oddy, M. (2006). Cost-benefits associated with social outcome from neurobehavioural rehabilitation.Brain Injury, 20(9), 947-957. [51] Wright, B. D., & Linacre, J. M. (1989). Observations are always ordinal; measurements, however, must be interval. Archives of Physical Medicine and Rehabilitation, 70(12), 857-860.