Available online at www.sciencedirect.com

ScienceDirect

Comprehensive Psychiatry 55 (2014) 1050–1054www.elsevier.com/locate/comppsych

Translation into Arabic and validation of the Schedule for the Assessmentof Insight–Expanded Version (SAI-E) for use in Tunisia

Jaafar Nakhli⁎, Salem Mlika, Saoussen Bouhlel, Badii Amamou, Ines Chaieb,Selma Ben Nasr, Béchir Ben Hadj Ali

Farhat Hached University Hospital, Sousse, Tunisia

Abstract

Background: The Schedule for the Assessment of Insight–Expanded Version (SAI-E) consists of 11 items that encompass: awareness ofhaving a mental illness, ability to rename psychotic phenomena as abnormal, and compliance with treatment.Aims: To translate into Arabic and validate the Tunisian version of this instrument.Method: The Arabic translation of the SAI-E was obtained by the “forward/backward translation” method. Adaptations were made after apilot study involving 20 outpatients with schizophrenia and after taking account the opinions of 15 experts in psychiatry.

For validation, 150 outpatients suffering from schizophreniawere recruited by a randomdrawing in the psychiatric department in Sousse (Tunisia).For factor analysis, principal components analysis and Varimax rotation were adopted. Convergent validity was assessed by correlating the

translated scale with the G12 item (lack of judgment and awareness of the disease) of the positive and Negative Syndrome Scale (PANSS).Internal consistency was assessed by Cronbach alpha coefficient and inter-rater reliability was assessed by the use of intra-class

correlation coefficient (ICC).Results: Regarding construct validity, factor analysis revealed three factors that were responsible for 70.2% of the variance.

As for concurrent validity, we found a negative correlation between the score of the SAI-E and that of the G12 item of the PANSS (r = − 0.82and p b 10−3).

The study of internal consistency between the 11 items was found to be good (α = 0.82). The test–retest reliability was satisfactory(r = 0.8, p b 10−3), and so was inter-rater reliability (ICC = 0.84).Conclusion: In the Tunisian cultural context, the SAI-E presented three factors with good consistency and an inter-rater reliability compatiblewith the insight dimensions that are intended to be evaluated.© 2014 Elsevier Inc. All rights reserved.

1. Introduction

Unawareness of illness or lack of insight is an importantcharacteristic of psychotic disorders, especially schizophrenia.It is related to severity of illness, poor outcome, poormedication compliance and more hospitalizations [1,2].Different authors have developed standardized instrumentsassessing insight in its different dimensions [3–7]. Amongthese instruments the most well known and widely used areThe Insight and Treatment Attitudes Questionnaire (ITAQ)[8], the Schedule for the Assessment of Insight (SAI) [9] andits expanded version (SAI-E) [10], the Scale to Assess

⁎ Corresponding author.E-mail address: nakhlijaaf@yahoo.fr (J. Nakhli).

0010-440X/$ – see front matter © 2014 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.comppsych.2014.02.016

Unawareness of Mental Disorder (SUMD) [11], and theBirch-wood self-report Insight Scale (IS) [12].

Comparative studies have demonstrated that the scores ofthese scales are strongly correlated, which indicates that theyhave a good concurrent validity [8,13]. The SAI-E remainsthe most practical as it is brief and easy to administrate forschizophrenic patients who generally have cognitive impair-ments [9,10,14]. It was developed for the assessment ofinsight in psychotic patients and it is based on a concept ofinsight that includes three distinct dimensions. Thesedimensions are: 1) awareness of mental illness; 2) abilityto relabel unusual mental events (e.g. hallucinations) aspathological; and 3) treatment compliance, both expressedand observed [14]. They are closely corresponding to thedimensions of insight proposed by David et al. [14],

1051J. Nakhli et al. / Comprehensive Psychiatry 55 (2014) 1050–1054

suggesting that the SAI-E aptly captures the theoreticalconstruct put forward by the author.

Although Gigante and Castel [15] considered that insightis rather a construct that depends on the socio-culturalcontext, few studies have investigated this possible influenceof socio-cultural factors on insight in schizophrenia [16].

To our knowledge, in Tunisia, as well as in other Arabiccountries there is not yet any validated schedule for theassessment of insight.

The aim of the present study was to translate SAI-E intoTunisian Arabic, to adapt it cross-culturally and to assess itsinternal structure.

2. Methods

2.1. The Schedule for the Assessment ofInsight–Expanded version

The SAI-E is a semi-structured interview which consistsof 11 items, with a standardized mode of rating of the itemsby the interviewer. Items 1–6, 10, and 11 are rated from 0 to2, while items 7–9 are rated from 0 to 4, with higher scoresindicating better insight. All the items are summed to reach atotal score, ranging from 0 to 28 [14]. The scale has provenvalidity and reliability in patients with psychosis [9,14].

2.2. Translation–Adaptation

We applied the original SAI-E index forward/backwardtranslation procedure. For Translation–Adaptation, weconsidered it necessary to obtain the consent of the designerof the SAI-E scale (D. Antony) about validation. He gave uspermission and answered all our questions and suggestions.We used the manual published by the linguistic validation“Mapi Research Institute” (MAPI) [17].

The translation was made from English to TunisianArabic dialect by two bilingual translators. The choice ofTunisian Arabic dialect and not of Standard Arabic aims fora better understanding by patients of the 11 items.

The stage of the translation was completed by assemblingmeeting translators, an experienced psychiatrist and apsychologist in addition to the authors of this study. Theitems were tested one by one to find the most appropriateterms. Each term was chosen after the approval of theinvolved parties.

The second stage was the back-translation of thetranslated text from Tunisian Arabic dialect to English andthe verification of compliance with the original schedule.

This back-translation was done by a bilingual translatorand followed by a meeting of all parties involved. The 11back-translated items were compared with the originalEnglish version of the scale SAI-E.

Fifteen psychiatrists had agreed to judge our translation.Their suggestions for items that could pose a problem wereencouraged. Their responses and comments were taken

into consideration in the revision of the Arabic version ofthe schedule.

2.3. Pre-testing

For pre-testing, a sample of 20 outpatients withschizophrenia answered the translated questionnaire inorder to test for the misunderstanding and acceptability ofquestions. Patients were randomly selected and not part ofthe sample statistical validation.

2.4. Procedure

A study protocol was written, evaluated and approved bythe ethics committee of the Faculty of Medicine in Sousse.

Symptoms of schizophrenia were rated using the Positiveand Negative Syndrome Scale (PANSS) [18].

We recruited 150 outpatients, 107 (73.3%) males and 43(26.7%) females, with a diagnosis of schizophrenia from theoutpatient service of Farhat Hached hospital in Sousse.Participants were recruited by a random draw from apopulation of 623 patients based on the following criteria:fulfilling the diagnosis of schizophrenia according to theDSM-IV-TR criteria [19], having a duration of illness of atleast three years and meeting the remission criterion offeredby “The Working Group Remission in schizophrenia” with ascore less than or equal to 90 at PANSS and no history ofhospitalization during the last 6 months [20]. Exclusioncriteria were mental retardation, personal or family history ofany neurological disorder, history of head injury, alcohol orsubstance abuse in the preceding 6 months. Basic demo-graphic and clinical data were obtained from the participants'medical records. All patients were taking an antipsychoticmedication at the time of assessment.

2.5. Reliability

Reliability in terms of internal consistency of the SAI-Ewas evaluated with Cronbach's alpha coefficient. This canbe considered as adequate when values greater than 0.7are obtained.

For the test–retest reliability evaluation, the “r” Pearson'scoefficient was used between the SAI-E scores obtained in theinitial evaluation and those obtained at 15 days. It hadconcerned 41 outpatients from the same sample of 150 patients.

To determine inter-rater reliability, the SAI-E wasadministered twice during a period between 10- and 15-dayinterval by two investigators to 34 schizophrenic patients. Wechose this interval to avoid variations in clinical status andpatient's remembering previous answers. The intra-classcorrelation coefficient (ICC), with two-way random effectmodel and absolute agreement definition, was adopted for theanalysis of inter-rater reliability according to the recommen-dations of Shrout and Fleiss [21]. It concerned 34 schizo-phrenic outpatients.

able 2ternal consistency of the Tunisian version of the SAI-E.

ems of the SAI-E Cronbach's alpha

. Awareness of emotional/psychological changes 0.46

. Awareness of having “something wrong,” a condition 0.51

. Attribution of one's condition to mental disorders 0.59

. Patient's account for his own condition/illness/disorder 0.47

. Awareness of consequences of one's condition 0.48

. Recognition of the need of treatment 0.54

. Awareness of mental symptoms 0.68

. Attribution of symptoms to a mental disorder 0.72

. Hypothetical contradiction item 0.730. Treatment acceptance 0.851. Spontaneous request for treatment 0.87

1052 J. Nakhli et al. / Comprehensive Psychiatry 55 (2014) 1050–1054

2.6. Validity

Content validity was determined through the applicationof a factorial analysis in order to check if the Tunisian Arabicversion of SAI-E had a similar factorial structure to theoriginal version. The extraction method of the principalcomponents was used and extraction of the number of factorswas predetermined based on auto values equal to or superiorto one.

Convergent validity was assessed by correlating thetranslated scale with the score of item 12 in the generalsubscale of the PANSS. This included item is one of the mostscales used in the evaluation of insight. The scores of bothmeasures should correlate positively and significantly.

able 3ter-rater reliability for the SAI-E items and total score.

ems 95% CI ICC

. Awareness of emotional/psychological changes 0.17–0.55 0.42

. Awareness of having “something wrong,” a condition 0.18–0.90 0.60

3. Results

3.1. Translation

Forward translation was carried out by three translatorsand followed by a meeting of all parties involved. Synthesisof the translations led to a unique version. The two back-ward translations of this version were comparable to theoriginal scale.

3.2. Pre-testing

Twenty schizophrenic outpatients participated in thisstep. The pilot study resulted in few linguistic changes in theinstrument. Incomprehensible terms were changed. A finaltranslated version was obtained.

3.3. SAI-E reliability

Table 1 shows the demographic and clinical characteris-tics of the sample. The mean SAI-E total score was 21.3(SD = 6.7). In regard to internal consistency of the scale,Cronbach's alpha was 0.82 for the entire sample. Table 2shows the mean of each item.

The test–retest reliability for 2-week interval betweenadministrations was calculated based on a subsample of 41patients. The total “r” Pearson's coefficient between bothadministrations was r = 0.8 (p b 10−3). It was between 0.36(item 1) and 0.94 (item 11).

The inter-rater reliability was calculated in the sample of 34patients. The ICC for the 11 items was 0.84. Items 4, 10 and 11had the highest ICC scores (respectively: 0.82, 0.9 and 0.95).

Table 1Demographic and clinical characteristics of patients with schizophrenia(n = 150).

Age (years) 40.3 ± 12.6Education (%) 93Professional activity (%) 40Duration of illness (years) 14.2 ± 9.8PANSS total score 47.6 ± 9.8SAI-E total score 21.3 ± 6.7

TIn

It

12345678911

Items 1 and 6 had the lower ICC scores (respectively: 0.42 and0.48) (Table 3).

3.4. SAI-E validity

The factorial analysis showed three factors that explained70.2% of the variance of the scores. Table 4 shows thefactorial weights of each item.

In relationship to the validity of convergent construct,Pearson's correlation coefficient between the total score ofthe G12 item of the PANSS and the DAI-E score wasnegative and significant (r = −0.82; p b 10−3). High scoresof the G12 indicated worse insight in schizophrenic patients.

4. Discussion

4.1. Translation

In our study, we chose to use the linguistic validationmethodology described by “The Mapi Research Institute”(MAPI) [17] which consists of three steps: translation, back-translation and pre-testing. In fact, this book seems to offerseveral advantages over other methods of validation. Triandis

. Attribution of one's condition to mental disorders 0.33–0.83 0.67

. Patient's account for his own condition/illness/disorder 0.65–0.91 0.82

. Awareness of consequences of one's condition 0.49–0.87 0.74

. Recognition of the need of treatment 0.13–0.67 0.48

. Awareness of mental symptoms 0.26–0.82 0.63

. Attribution of symptoms to a mental disorder 0.41–0.86 0.71

. Hypothetical contradiction item 0.51–0.88 0.750. Treatment acceptance 0.88–0.95 0.901. Spontaneous request for treatment 0.91–0.97 0.95otal 0.7–0.92 0.84

I, confidence interval (95%); ICC, intra-class correlation (two-way randomffect mode, absolute agreement definition).

TIn

It

12345678911T

Ce

Table 4Factor analysis of the 11 items of the Tunisian version of SAI-E.

Items of the SAI-E Factor1

Factor2

Factor3

1. Awareness of emotional/psychological changes 0.712. Awareness of having “something wrong,”a condition

0.74

3. Attribution of one's condition to mental disorders 0.784. Explanation of mental condition 0.44 0.425. Awareness of consequences of one's condition 0.636. Recognition of the need of treatment 0.717. Awareness of mental symptoms 0.898. Attribution of symptoms to a mental disorder 0.879. Hypothetical contradiction 0.8010. Treatment acceptance 0.9611. Spontaneous request for treatment 0.96Percentage of the variance explained by the factors 41.7 16.6 11.9Accumulated variance 70.2%

Extraction method: Principal components analysis. Rotation method: Varimaxwith Kaiser normalization. Values ≤0.40 were eliminated from the table.

1053J. Nakhli et al. / Comprehensive Psychiatry 55 (2014) 1050–1054

and Brislin [22] suggested that pre-testing is necessary after acareful translation.

For the pre-test, in contrast to the procedure described bySchuman [23], we decided to administer all items to 20patients. All the remarks made by patients concerning thecomprehension of certain terms and expressions were takeninto account.

4.2. Reliability

In our study, we noted a good internal consistency betweenthe 11 items of the scale SAI-E in the Tunisian version with aCronbach's α coefficient of 0.82. The 11 items appear to beconsistent which helps tomaintain a homogeneous structure ofthe Tunisian version of the SAI-E scale, and to measure theconcept of insight [14,24].

Konstantakopoulos et al. [25] had found Cronbach's alphabetween 0.35 (item 10) and 0.88 (item 4). Cronbach's alpha ofitem 11 (request for treatment) was higher than item 10(acceptance of treatment). For these authors, acceptance ofproposed treatment along with awareness of clinician-explained need for treatment and awareness of psychosocialconsequences of the illness lie closer to the “induced” insightpole, whereas spontaneous request for treatment along liecloser to the “spontaneous” pole of this dimension. Thisdistinction between “spontaneous” and “induced” insight bearssome similarities to the distinction between “integrativeinsight” and “passive insight” of illness described by Roeet al. [26].

For the test–retest reliability, it is considered that the “r”Pearson's coefficient values ranging from 0.36 to 0.84represent reliability from “regular” and “good” while valuesover 0.75 are considered to be excellent. The test–retestreliability in our study for 2 weeks is satisfactory (r = 0.8)and indicates the stability of the patient's response during

this time period [27]. In the Greek validation study, ICC wasbetween 0.79 and 0.91, and the total ICC was 0.88 [25].

Regarding inter-rater reliability, the maximum value forICC is 1.0 and occurs if all the raters perfectly agree on therating for each patient [21]. In our study, the ICC valuesfound for the individual items can be considered satisfactory,and for the total score it was excellent (ICC = 0.84). In theoriginal reliability study of SAI, from which SAI-E derives,the ICC found was 0.72 [22]. In the Brazilian and Greekvalidation studies, the overall ICC value was excellent(ICC = 0.90) and comparable to the one we found [24,25].

4.3. Validity

The factor analysis performed for the original version of theSAI-E indicated that three factors accounted for 70.2% ofvariance. Our result was similar to the factor analysis of theoriginal version of the SAI-E, as well as the Brazilian andGreek versions [14,24,25] and had revealed the presence ofthree factors (awareness of having a mental illness, ability torecognize as pathological events such as unusual delusions orhallucinations and adherence). In the original English andBrazilian studies, the three factors were responsible for 66.5%and 71.7% of the variance [14]. The Tunisian version of theSAI-E appears in close correspondence with the insightconcept established by David et al. [14].

However, in the factor structure found by David et al. [14],the item “awareness of the need for treatment” (item 6) wasrelated to the treatment compliance factor whereas in the studyby Dantas and Banzto [24] and in the Greek validation studythis item was found related to the awareness of illnesscomponent. This may be due to the differences in the studysamples, since the former used a sample of first-episodepsychosis patients while the two latter used samples consistingof chronic patients [14,24,25].

In our study, factor 2 (16.6% of variance) consisted of thetwo items related to awareness and attribution of symptoms, theitem of hypothetical contradiction and item 4, “explanation ofmental condition,” and it corresponds to the relabeling ofsymptoms dimension. In the original version of SAI-E and theGreek validation, the authors did not find the implication of item4 in this dimension [14,25].We can conclude that the emergenceof item 4 in this dimension is related to its low alpha coefficient(0.42), while items 7, 8 and 9 were between 0.8 and 0.89.

For the third factor, which consisted of the two itemsregarding treatment compliance, our results were similar tothose found in the Brazilian and Greek validation studies[24,25].

In relationship to the validity of convergent construct, weobserved a negative correlation between the score of the SAI-Eand G12 item of the PANSS (lack of judgment and awarenessof the disease). This shows the correspondence between the twoscales. In the literature, Sanz et al. [28] had found correlationbetween SAI-E and the Insight Scale (IS). In the Greekvalidation study, the insight item of the PANSS was stronglycorrelated with SAI-E total score (r = 0.72, p = 0.001) [25].

1054 J. Nakhli et al. / Comprehensive Psychiatry 55 (2014) 1050–1054

5. Conclusion

The Tunisian version of the SAI-E presented good internalconsistency and excellent inter-rater reliability for the totalscore, and featured a three-factor structure similar to theEnglish original version, in close correspondence with thedimensions of insight which the SAI-E is intended to assess.

References

[1] Mintz AR, Dobson KS, Romney DM. Insight in schizophrenia: a meta-analysis. Schizophr Res 2003;61:75-88.

[2] Erickson M, Jaafari N, Lysaker. Insight and negative symptoms aspredictors of functioning in a work setting in patients withschizophrenia. Psychiatry Res 2011;189:161-5.

[3] Raffard S,BayardS,CapdevielleD,Garcia F, Boulanger JP,Gely-NargeotMC. Lack of insight in schizophrenia: a review; part i: theoretical concept,clinical aspects and Amador's model. Encéphale 2008;34(6):597-605.

[4] Amador XF, David AS. Insight and psychosis. Oxford UniversityPress; 1998.

[5] Amador XF, Flaum M, Andreasen NC, Strauss DH, Yale SA, Clark SC,et al. Awareness of illness in schizophrenia and schizoaffective andmood disorders. Arch Gen Psychiatry 1994;51(10):826-36.

[6] Weiler MA, Fleisher MH, Mac Arthur-Campbell D. Insight andsymptom change in schizophrenia and other disorders. Schizophr Res2000;45(1):29-36.

[7] Wilson WH, Ban TA, Guy W. Flexible system criteria in chronicschizophrenia. Compr Psychiatry 1986;27(3):259-65.

[8] McEvoy JP, Apperson LJ, Appelbaum PS, Ortlip P, Brecosky J. Insightin schizophrenia. Its relationship to acute psychopathology. J Nerv MentDis 1989;177(1):43-7.

[9] David AS. Insight and psychosis. Br J Psychiatry 1990;156:798-808.[10] Kemp R, David A. Insight and compliance. In: & Blackwell B, editor.

Treatment compliance and the therapeutic alliance; 1997. p. 61-84.[11] Amador XF, Strauss DH, Yale SA, FlaumMM, Endicott J, Gorman JM.

Assessment of insight in psychosis. Am J Psychiatry 1993;150(6):873-9.[12] Birchwood M, Smith J, Drury V, Healy J, Macmillan F, Slade M. A

self- report Insight Scale for psychosis: reliability, validity andsensitivity to change. Acta Psychiatr Scand 1994;89(1):62-7.

[13] Greenfeld D, Strauss JS, Bowers MB, Mandelkern M. Insight andinterpretation of illness in recovery from psychosis. Schizophr Bull1989;15(2):245-52.

[14] David AS, Morgan KD, Mallet R, Leff J, Murray RM. Insight: unitaryor multidimensional phenomenon? Schizophr Res 2003;60(1):14.

[15] Gigante AD, Castel S. Insight into schizophrenia: a comparisonstudy between patients and family members. Sao Paulo Med J2004;122(6):246-51.

[16] Fabrega H. On the significance of an anthropological approach toschizophrenia. Psychiatry 1989;52(1):45-65.

[17] Mapi Research Institute. Linguistic validation of a patient reportedoutcomes measure. Available online at http://www.mapigroup.com/Services/Linguistic-Validation2005.

[18] Kay SR, Fiszbein A, Opler LA. The Positive and Negative SyndromeScale (PANSS) for schizophrenia. Schizophr Bull 1987;13(2):261-76.

[19] American Psychiatric Association. DSM-IV-TR. Diagnostic andstatistical manual of mental disorders. Washington: APA; 2000.

[20] Wallace C,Mullen P, Burgess P. Criminal offending in schizophrenia overa 25-year periodmarkedbydeinstitutionalization and increasing prevalenceof comorbid substance use disorders.AmJPsychiatry 2004;161(4):716-27.

[21] Shrout PE, Fleiss JL. Intra class correlations: uses in assessing raterreliability. Psychol Bull 1979;86(2):420-8.

[22] Triandis HC, Brislin RW. Cross-cultural psychology. Am Psychol1984;39(3):1006-16.

[23] Schuman H. The random probe: a technique for evaluation of quality ofclosed questions. Am Sociol Rev 1966;31:218-22.

[24] Dantas CR, Banzto CEM. Inter-rater reliability and factor analysis of theBrazilian version of the Schedule for the Assessment of Insight–ExpandedVersion (SAI-E). Rev Bras Psiquiatr 2007;29(4):359-62.

[25] KonstantakopoulosG, PloumpidisD,Oulis P, SoumaniA,Nikitopoulou S,Pappa K. Is insight in schizophrenia multidimensional? Internal structureand associations of theGreek version of the Schedule for theAssessment ofInsight–Expanded. Psychiatry Res 2013;209(3):346-52.

[26] Roe D, Hasson-Ohayon I, Kravetz S, Yanos PT, Lysaker PH. Call it amonster for lack of anything else: narrative insight in psychosis. J NervMent Dis 2008;196:859-65.

[27] ShuttleworthM. The test–retest reliabilitymethod of the simplestways oftesting the stability and reliability of an instrument over time. Availableon URL: http://explorable.com/test-retest-reliability 2009 [April].

[28] Sanz M, Constable G, Lopez-Ibor I, Kemp R, David AS. A comparativestudy of insight scales and their relationship to psychopathological andclinical variables. Psychol Med 1998;28(2):437-46.