Reliability, validity, and factorial structure of the Turkish version of the Bradford Somatic Inventory (Turkish BSI-44) in a university student sample

ABSTRACT Objective: Mumford and colleagues developed the Bradford Somatic Inventory (BSI) that examines the somatic symptoms of anxiety and depression, which has transcultural applications. The objective of the current study was to establish the psychometric properties and factorial validity of the Turkish version of the BSI-44 in a healthy Turkish population and obtain normative data. Methods: The study was conducted at the Marmara University School of Medicine with a sample of 201 healthy students (18–30 years old). In order to estimate the test–retest reliability of the Turkish BSI, 53 participants from the original sample were asked to fill in the questionnaire one month after the initial testing. Socio-demographic data of the participants were collected and the Turkish BSI, Somatosensory Amplification Scale (SSAS), Whiteley Index (WI-7), and somatization subscale of the Symptom Check List (SCL-90-R) scales were administered. All statistical analysis were performed by using SPSS version 23 for Windows. Results: The mean age of the study participants was 22.9 ± 1.95 years; 57.7% (n = 116) of participants were female; 42.3% (n = 85) were male. BSI scores were normally distributed. The scores of the BSI were categorized as high (>40), middle (26–40), and low (0–25); no statistically significant differences were found between males and females. The Cronbach’s alpha coefficient for the scale was 0.90 and the test–retest correlation coefficient was found to be 0.75. A positive and statistically significant correlation was found between the Turkish BSI and the WI (r = 0.38, p < .01), the SSAS (r = 0.48, p < .01) and the SCL-90-R (r = 0.79, p < .01) scales. A principal components analysis was performed on the BSI responses of the participants, which yielded 14 factors with an eigenvalue greater than one, representing 65.2% of the total variance. Conclusions: Our results suggested that the Turkish BSI was a valid and reliable tool with a robust factorial structure to use in clinical populations in Turkey.


Introduction
Somatization, in general, is defined as the tendency to express emotional dysphoria with somatic symptoms [1]. Most people with psychiatric disorders express their mental distress by somatic symptoms rather than psychological symptoms [2]. It is clear that somatic symptoms are a very common way of presentation of psychiatric illness throughout the world [3].
Although somatization is seen with many psychiatric disorders, it is most commonly associated with anxiety and depression in psychiatric disorders [4][5]. There are many questionnaires and inventories that measure psychological symptoms in psychiatric situations, but none of them measures somatic symptoms of anxiety and depression [3]. Furthermore, many wellconstructed scales consist of limited items about somatic symptoms of anxiety and depression. Other inventories including somatic symptoms, which have been developed in recent years, have been designed to be only proper to particular ethnic groups. Therefore, Mumford et al. determined the need for a scale that would measure somatic symptoms associated with psychiatric illness in detail and can be applied to multiple ethnic backgrounds. They emphasized the need for a systematic research to obtain a comprehensive list of somatic symptoms often seen in each group and build the scale with an appropriate construct to each ethnic group [3].
Mumford and colleagues developed the Bradford Somatic Inventory (BSI) that examines somatic symptoms of anxiety and depression and comprehends multiple ethnicities [3]. Symptoms of this scale were obtained from psychiatric case notes of Pakistani and British patients with a clinical diagnosis of anxiety, depression, hysteria, and hypochondriasis [3]. The inventory was constructed in two languages: Urdu and English. The pilot version of the BSI was checked against psychiatric case notes of patients in the other parts of the India-Pakistan subcontinent. The revised BSI covered 90% of all somatic symptoms registered in each centre. The linguistic equivalence of the Urdu and the English versions was established in a bilingual student population in Pakistan [3]. The conceptual equivalence of the BSI was explored using factor analysis of responses by functional patients presenting to medical clinics in Britain and Pakistan. Four principal factors including head, chest, abdomen, and fatigue were obtained from the results of the factor analysis [3]. In a British primary care population, Cronbach's alpha coefficient of the BSI was found to be 0.86 with good test-retest reliability [2].
The BSI consists of 46 items, 2 items applying to men only. It inquires about a wide range of somatic symptoms during the previous month and if the subject has experienced a particular symptom, and whether the symptom has occurred on more or less than 15 days during the month [3]. The BSI was constructed simultaneously in Urdu and English and has been translated into several languages, namely, Arabic, Bengali, Chinese, French, Italian, Polish, Romanian, Russian, Spanish [6][7][8], German, and Turkish [9]. Normative data and psychometric properties of the Russian version have been reported [10].
The objective of the current study was to establish the psychometric properties and factorial validity of the Turkish version of the BSI-44 (two items for men only in the original form were excluded) in a healthy Turkish population and obtain normative data for future clinical and epidemiological studies in psychiatric patients in Turkey.

Participants
The study was conducted at the Marmara University School of Medicine with a sample of 201 healthy students (85 males, 116 females, mean age 22.9 ± 1.95 years, range [18][19][20][21][22][23][24][25][26][27][28][29][30]. To estimate the test-retest reliability of the Turkish BSI, 53 participants from the original sample were asked to fill in the questionnaire one month after the initial testing (see Appendix). The study participants were aged 18-30 years old; able to read and write Turkish; and free of psychiatric disorders such as psychosis, autism, mental retardation, and substance abuse. Participants who had neurological disorders such as cerebrovascular disorders, convulsions, meningitis, or encephalitis, with any history of abnormal computed tomography or magnetic resonance imaging scans, or who were on psychotropic medications were excluded.
The current study was approved by the Ethics Committee of Marmara University Hospital and all the subjects gave written informed consent before participation.

Psychometric measurements
Socio-demographic Data Form. This form (prepared by the researchers) includes demographic variables, including gender, marital status, alcohol use, substance usage, psychiatric diseases, and medical diseases.
Bradford Somatic Inventory (BSI-44). BSI is a 44-item inventory for psychosomatically expressed psychological distress. It has cross-cultural validity as shown by studies carried out in Great Britain, Pakistan, India, Nepal, and Russia. The BSI asks the subject about a wide range of somatic symptoms during the previous month, and whether or not the subject has experienced a particular symptom, on more or fewer than 15 days during the month (scoring 1 or 2, respectively). For the present purpose, the scoring was based on Mumford, where a score >40 was considered to be high range; 26-40, middle range; and 0-25, low range.
Somatosensory Amplification Scale (SSAS). The SSAS is a 10-item scale developed by Barsky et al. [11] and its validity and reliability have been demonstrated. Respondents score each item from 1 (not at all true) to 5 (extremely true). Most items describe a physical discomfort, which does not indicate a disease. In the original version, by adding the scores, a total amplification score is obtained (ranging from 10 to 50). Its adaptation into the Turkish form was shown by Gulec et al. and the Turkish version of the SSAS had good internal reliability with a Cronbach's alpha of 0.80 [12].
Whiteley Index (WI-7). The WI is a widely used instrument developed by Pilowsky which finds hypochondriac worries and beliefs [13]. Factor analysis of the WI yielded three separate factors: disease fear, disease conviction, and bodily preoccupation. The WI has been widely used in studies of hypochondriasis and provides a useful screening measure [14]. The Turkish version of the WI-7 which has been prepared by Gulec et al. showed good reliability in the Turkish population, with a Cronbach's alpha of 0.76 [15].
SCL-90-R somatization subscale. The somatization subscale of the SCL-90-R is a multidimensional selfreport measure of psychopathology widely employed in psychiatric and medical populations with well-established reliability and validity [16]. The SCL-90-R somatization subscale is a 12-item list of common somatic symptoms and has been demonstrated to be reliable in the Turkish population, with Cronbach's alpha = 0.75 [17].

Statistical analysis
Data analysis was performed using SPSS for Windows Version 23.0 (SPSS Inc., Chicago, Illinois, USA). Cronbach's alpha coefficients were calculated for each item to identify the internal consistency of the Turkish BSI. Correlation analysis between test and retest data were performed using Pearson's correlation coefficients. To compare the average mean of the categorized BSI in terms of gender, a crosstab analysis was used. Convergent and discriminant validity were examined by correlation coefficients between the BSI scale scores and total WI, total SSAS, and total SCL somatization subscale scores. Based on the theoretical structure, exploratory factorial analyzes were performed. Principal factor analyzes with Promax rotations were used. A p-value less than .05 was considered statistically significant.

Socio-demographic characteristics of sample
The mean age of the study participants was 22.9 ± 1.95 years (X ± SD); 57.7% (n = 116) of the participants were female; 42.3% (n = 85) were male. The majority of the participants in the study were single (99.5%) and no one was married, and one participant was divorced. In all, 54.2% of the sample had never used alcohol before, and 38.3% were using alcohol. In the sample, 190 (94.5%) participants have no psychiatric diseases, but 5.5% were suffering from at least one. The socio-demographic characteristics of the participants are presented in Table 1.

Prevalence of somatization
The scores of the BSI were categorized as 40 and above as high, 26-40 as middle, and 0-25 as low range, and crosstab analysis used. We did not find any statistically significant differences between males and females in terms of the categorized scores (p = .452). Table 2 shows the frequencies and percentage scores on the BSI of the whole sample as well as for males and females separately. Only 1% of the sample scored in the high range, who are all females. That means 2 women and no men had a score of 40 or higher. In all, 14.9% of the whole sample scored in the middle range. Again, females (15.5%) outnumbered the males (14.1%). The majority of the subjects (84.1%) were in the low range. Here males (85.9%) outnumbered females (82.8%). These results revealed that although women in the sample showed a trend towards somatization, no statistically significant differences in the mean BSI scores were found between men and women (χ 2 = 1.587, df = 2, p = .452).

Internal consistency
The Cronbach's alpha coefficient for the Turkish BSI was found to be 0.90.

Test-retest reliability of the Turkish BSI
There was a period of one month between test and retest administrations and 53 students participated in the retest procedure. Total BSI scores were found to be highly correlated with total retest BSI scores (r = 0.75, p < .001). The highest correlation coefficient was found for Hands or feet pins and needles (r = 0.77, p < .001) and the lowest correlation coefficient was found for Trembling or shaking (r = −0.03, p > .05). Results of correlation coefficients between test and retest scores of all items are presented in Table 3 in detail.

Convergent validity
Convergent validity was examined by correlation coefficients between the BSI scale scores and total WI, total SSAS, and total SCL somatization subscale scores. A positive and statistically significant correlation was found between total BSI and total WI (r = 0.384, p < .001), total SSAS (r = 0.482, p < .001), and SCL somatization subscale (r = 0.793, p < .001). Correlations between the Turkish BSI, age, and other scales are presented in Table 4.

Factor structure of the Turkish BSI
To examine the factor structure of the BSI scale, an exploratory factor analysis (EFA) was performed using various methods. Kaiser-Meyer-Olkin (KMO)   43) loaded onto more than one factor, and 1 item (Item 7) failed to load at least 0.4 on any factors. However, in the original study of the scale, it was reported that an eight-factor structure provided a strong fit. An eight-factor solution was rotated by using Promax rotation and minimized the number of variables that have high loadings on any one factor. When all the rotated solution was examined, the eight factors accounted for 49.75% of the total variance. The eight-factor solution presented in Tables 5 and 6 compares the eight-factor solutions of the Turkish, English, and Urdu versions.

Discussion
In this study, we aimed to examine the validity, reliability, and factor structure of the BSI in a Turkish sample. The main findings of the present study confirmed that the Turkish BSI was observed to have stable and reliable psychometric properties.
When the socio-demographic data are taken into consideration, no significant differences were observed between the male and female participants in terms of total BSI scores and there were no statistically significant correlations between age and BSI scores. In our sample, although the average mean BSI scores of women were higher than the men's, this difference was not statistically significant. Studies in the literature show that somatic disorders are affected by social position and are mostly seen in communities with lower urbanization and literacy level, and they are more common among women than men [18]. A study conducted by Aragona et al. [7] using the BSI-21 reported that female gender was a significant predictor of the frequency of 12 out of 21 symptoms and female participants showed significantly higher scores on the BSI-21 than men. However, studies that examined the relationship between age and somatic disorders were consistent with our findings [7,19]. Somatization has been reported to be more frequent among married females, aged 20-30 years, who are housewives [20]. But no women in this category participated in our study. Therefore, it can be speculated that our findings may have been affected by the marital status, education level, and/or the nature of our sample.
Although women in the sample showed a trend towards somatization, the difference between men and women was not statistically significant in terms of categorized BSI scores. However, in the sample, two women scored in the high range, while all men scored in the middle or low ranges. Therefore, an important implication may be that as found in most other studies in the literature, women are more  susceptible to somatization than men [21]. Another study conducted by Al-Lawati et al. also found similar results to our crosstab analysis for gender and categorized BSI scores [19]. However, in our study, this gender difference was not very clear and this may be mostly because we conducted our study with healthy participants and our small sample size was small.
Cronbach's alpha coefficient of the Turkish BSI was as high enough as in Mumford's original study [2]. In Mumford et al.'s original study, the internal consistency reliability coefficient of the BSI was 0.86 and in the present study, it is found to be 0.90. Due to the fact that the Cronbach's alpha coefficient was high enough (>0.60) in scale, the internal consistency of the Turkish BSI was considered to be sufficient. The present study also confirmed that the Turkish BSI has good test-retest reliability due to the fact that similar correlations were observed across a one-month interval, indicating the stability of the measure over time.
In our sample, the Turkish BSI was found to be positively correlated with the SSAS, which is a self-evaluating scale for measuring amplification during somatization, WI-7 which measures hypochondriac worries and beliefs, and the Somatization subscale of the SCL-90-R. The participants who received higher scores in the Turkish BSI also received higher scores in these scales that are specifically developed to examine somatic symptoms. These results confirmed that the Turkish BSI has a good convergent validity.
In the present study, in order to find out the number of dimensions and which items construct each factor, the EFA method was used. EFA can be quite useful for assessing the extent to which a set of items assesses a particular content domain and it is commonly used to reduce the set of observed variables to a smaller, more parsimonious set of variables [22].
The initial principal components analysis yielded 14 factors with an eigenvalue greater than one, representing 65.2% of the total variance. In the study of the original scale, 13 factors with an eigenvalue greater than one, representing 65.8% of the total variance, were yielded [2]. These results confirmed that the factorial for successive factors were displayed in a scree plot, which was used graphically determine the optimal number of factors to retain and suggested the extraction of six, seven, or eight factors [2]. The eight-factor solutions of the Turkish, British, and Pakistani samples yielded four similar factors: head, abdomen, chest, and fatigue. However, similarly, in all three versions the remaining factors were not stable across the solutions. An important difference between our study and the original scale study in terms of factorial analysis is that, unlike Mumford et al. [2], we used a Promax rotation instead of Varimax to obtain an eight-factor solution because the results of this oblique rotation were a set of loadings that typically reflect simple structure better than the Varimax rotation, especially when the latent traits are highly correlated [23]. In general, Varimax rotation in EFA is used when it is assumed that the factors extracted are not correlated with each other. However, Promax or other oblique rotations are generally used when it is assumed that they are orthogonal and correlated well.
The results reported in this study should be considered in light of certain limitations. First, the sample in this study was recruited from volunteer college students with a limited age range. This may to some extent affect the results and limit the generalization of the results to other samples. Another limitation is the fact that the cross-sectional nature of the study would not allow us to link the causality. Further prospective, longitudinal studies would help to establish a probabilistic causal relationship.
In conclusion, the Turkish version of the BSI had sound psychometric properties in our sample of Turkish healthy volunteers, including its internal consistency, test-retest reliability, concurrent validity, and factorial structure. The Turkish BSI will be useful for future studies in different countries to help better understand normalcy and psychopathology including somatization to examine the biological, social, and psychological differences in people from different cultures.

Disclosure statement
No potential conflict of interest was reported by the authors.