Validating the Turkish version of the Weinstein noise sensitivity scale:effects of age, sex, and education level

Background/aim Subjective noise sensitivity (SNS) is not only one of the most important predictors of perceived noise annoyance for the subjects exposed to the noise, but also strongly related with some psychological conditions and poor health perception of the subjects. Weinstein’s Noise Sensitivity Scale (WNSS) is widely used for measuring of SNS. Materials and methods In this study we adapted WNSS to Turkish (Tr-WNSS) and evaluated its psychometric properties and effect of age, sex and education level on SNS. Results The study included 105 males and 105 female subjects between 18 and 55 years of age. Conclusion Our data demonstrated that Tr-WNSS was a reliable, valid, and invariant scale for the Turkish population, and not affected by sex, age and education level.


Introduction
Noise has been described as unwanted sound. The American Hearing and Speech Association (ASHA) points out that one person's music is another person's noise [1]. However, a part of population appears to be more prone to perceive sound as a negative condition and present strong emotional reaction to noisein the same acoustical conditions [2][3][4][5][6][7]. It has been reported that the subjects with negative tendency to sound are resistant to adaptation to noise for longer periods of time [3][4][5]7], and that it is related with a genetic or familial tendency [8]. This concept has been named as subjective noise sensitivity (SNS) [5,6] or noise annoyance susceptibility [2].
The researchers clearly presented that SNS was a predictor of noise annoyance [5,7,9]. It is known that some clinical conditions such as head trauma, migraine headaches, facial paralysis, otologic infections and surgeries such as stapedectomy, and some medicines are also related with increasing annoyance from aloud sound (hyperacusis) [10,11]. Although the association of SNS with hyperacusis or phonophobia is still an unanswered question, the SNS concept summarized above appears to be a consistent and stabile personality traitin the subjects without such a clinical problem. No relationship of SNS with auditory acuity was reported [2,5,12,13]. It was also documented that noise sensitive people did not have more experience of exposition to noise [14].
However, it has been reported that SNS is related with behavioral differences, tendencies towards some psychological conditions, poor health perception, and social differences. Weinstein [3]reported that theywere lower in scholastic ability, felt less secure in social interactions, and had a greater desire for privacy. Weinstein [4] and Stansfeld [5,12] pointed out that people with SNS were more sensitive to sensory stimuli around them and hence more critical of their environmental quality. Stansfeld [5,15], Öhrström et al. [16], Belojevic and Jakovlevic [6], and Stansfeld and Shipley [17] pointed out a tendency toward neuroticism, depressive symptoms, and psychological distress. Furthermore, Babisch et al. [18], Fyhri and Klaeboe [19], and van Kamp and Davis [11] reported higher cardiovascular problems, while Stansfeld and Shipley [17] documented only a meaningful association between SNS and angina in people from lower employment grades. Besides, Shepherd et al. [20] found that SNS was negatively associated with self-reported healthrelated quality of life scores. Stansfeld and Stanley [17] recently documented that SNS has been identified as a predictor of mental ill-health, while Schreckenberg et al. [21] reported an association of SNS with physical health. These data point out that SNS is an important predictor not only for noise annoyance but also for health perception of the subjects and their predisposition to the psychological problems.
Although the studies presented that environmental high noise appears to be important problem in Turkey [22][23][24] and there are some noise-annoyance questionnaires adapted to Turkish [23,25] and developed in Turkish [24], noise sensitivity has never been subject to any research in the Turkish population as much as we found in the English and Turkish literature. For the measurement of SNS, long and short questionnaires have been proposed [7]; the Weinstein noise sensitivity scale (WNSS) has been one of the most widely used standard questionnaires with high validity, reliability, and internal consistency [3,4,26,27]. In this study we aimed to adapt WNSS to Turkish and looked for its reliability, internal consistency, factor structure, and furthermore its stability regarding sex, age, and educational level in the Turkish population without any otologic and neuro-otologic problem.

Material and method
This study was approved by the clinical research ethics board in Gazi University (Approval date and no: 23.03.2015, 176). This research was conducted at the audiology department of the university hospital.
In the first step of the study, WNSS [3] was translated into Turkish by 2 of the authors and translated back to English. Then, the preliminary Turkish version was evaluated in a pilot study including 20 interviewers, and it was noticed that only small changes were needed, because all items were found to be clear in Turkish and understanding of the interviewers was in accordance with the original target of each question. After these small corrections, the Turkish version of WNSS (Tr-WNSS), which is comprised 21 items with a 6-point scale ranging from 1 (agree strongly) to 6 (disagree strongly) was completed. Hence, summing up the scores of all items created a noise sensitivity score (NSS) of each subject. For test reliability, Tr-WNSS questionnaire was refilled out by 64 subjects after a 15-day interval. The test/retest reliability was assessed with Pearson's test. Then, validation of the scale was calculated with confirmatory factor analysis (CFA). The internal consistency for Tr-WNSS was calculated with Cronbach's alpha (α) coefficient. In order to test homogeneity of variances, Levene's test was used. According to result of Levene's tests, homogeny or nonhomogeny group of variances decided and independent 2 sample t-test was conducted. By using analysis of variance test, the differences between groups were assessed.
In the second step of the study, Tr-WNSS was applied to 210 subjects (M: 105, F: 105) aged between 18 and 55 years. Inclusion criteria were as follows: no chronic health problem causing functional impairment in daily life (only exception was use of eyeglasses), no craniofacial abnormality or congenital disorders, and normal hearing as personal perception without history of any otologic and/or neuro-otologic disorders.
NSS was calculated for each subject and compared between i) males and females, ii) age groups (age subgroup-1: the subjects younger than 35 years of age; age subgroup-2: the subjects between 35 and 55 years of age), and iii) education groups (education subgroup-1: elementary, secondary and high school; education subgroup-2: 2-years of higher education after high school or Bachelor's degree). Furthermore, correlation of NSS with age was tested.
Then according to NSS, the lowest (lowest 30%) and highest (highest 30%) SNS subgroups were found, and the lower and upper SNS groups were compared regarding sex, age, and education level in addition to NSS.
At the first step of the study, by using the data of 64 subjects (37 males with mean age of 33.95 ± 3.37 months and 27 females with mean age of 34.23 ± 9.88), it was found that test reliability was high by Tr-WNSS (Pearson's test, r: 0.92, P < 0.001). Levine test presented that the data were distributed in accordance with normal distribution.
The factor structure of the 21-item scale was tested by using CFA. The CFA was conducted using IBM SPSS AMOS 26 software. We follow the same strategy for the factor model case with Senese et al. [28]. Items 2,4,5,6,7,10,11,13,16,17,18,19, and 21 were specified to the first factor, while items 1, 3,8,12,14,15, and 20 were specified to the second factor. From this setting of the items, we summarized the following results in Table 2.
According to Table 3 the default model fit has the following parameters. χ² = 267.621 with df = 169 which means that the default model is statistically significant. We can also support the result from CMIN/DF ratio is below 3, which is a good metric for an acceptable model (CMIN/DF = 1.584). We do note that the RMSEA is below the accepted threshold (<0.10) for the default model (RMSEA = 0.054). The CFI value is 0.888 which is between traditional bounds and it is acceptable for the default model. The GFI value is a little below than the acceptable value but it is in satisfactory limits (GFI = 0.892). The AGFI value is above the accepted threshold (>0.80), which is 0.866. The SRMR value is way below than .09 for our model (SRMR = 0.00). The TLI parameter value is close to 1 (TLI = 0.874), which indicates a very good fit for the default model. Also, according to the AIC values, the default model has the minimum value among the others (AIC = 349.621). This means that the default model has the best fit. All other values for the parameters of the CFA results can be seen in the Table 3. Figure is the diagram of the CFA analysis which shows the relations and the correlations coefficients between the factors. Our results support the conclusion that the 2 latent factors are strong reflections of the associated observed variables.
When participants were asked 21 questions, the CFA was used for validity of Turkish scales. CFA results showed that 2-factor model fits the model according to the significance value P < 0.05 (Table 2). These 2 factors are highly correlated to each other (r = 0.656; P < 0.05). As supported by different researches, the scale can be considered substantially unidimensional because the second factor can be accounted for as an effect of the direction of item wording more than an expression of a different dimension. This was also confirmed by the strong correlation between the 2 factors.
At the second step of the study, it was observed that the mean NSS value was 88.72 ± 17.97 for 210 volunteers, and 89.22 ± 16.21 (minimum: 48 and maximum: 124) for males and 88.23 ± 19.63 (minimum: 30 and maximum: 119) for females (Table 4). No statistical difference between males and females and between age subgroups and between education subgroups was detected (Student-t test, P > 0.05) ( Table 4). Pearson's correlation analysis declared that NSS was not correlated with age in either total group or male and female subgroups (Pearson's test, P > 0.05).

Discussion
Garrioch [29] reported that the prosperous Greek colony Sybaris in Italy prohibited industrial noises in residential areas in 720 BCE. The historical papers present that not only noise exposure but also noise annoyance increases in modern life since the time of Sybaris [29][30][31][32]. Belojevic et al. [7] say that noise affects the person who is "doing nothing" by annoying him, while it affects the person who works by impairing his performance or making it more costly in terms of increased effort. Therefore, during the last century, not only annoyance but also its impact on human performance and interpersonal behavior has been subjected to studies by social psychologists [33]. SNS has been documented as a predictor or modulator for noise annoyance level when exposed [2][3][4][5][6][7]9]. In this study, we adapted WNSS to Turkish (Tr-WNSS) to make it available for multidisciplinary researches related to noise. Our data documented that Tr-WNSS was a convenient tool with strong internal and external consistency in the Turkish population, as reported in other adaptation studies of WNSS to non-English languages including Swedish [26], German [34], Japanese [35], Persian [27], and Italian [28]. Hence, WNSS could be used in Turkish for predicting personal behavior in the case of noise exposure without language barrier. Besides, although it has been developed with college student samples by Weinstein [3], the studies documented that it is valid and convenient in other populations, such as the depressed subjects and hospital staff between 18 and 65 years of age [5], the adult residents living close to heavy traffic roads in Japan between 20 and 70 years of age [35], nonindustrial employees between 17 and 76 years of age in Iran [27], and adults living in the neighborhoods of southern Italy [28]. In our study, Tr-WNSS presented a convenient assessment of SNS in adults between 18 and 55 years of age so that the highest 30% could be significantly different from the lowest 30%.
Nevertheless, it could be said that the average score in our study revealed some differences from previous studies. While it was 54.6 ± 12.1 (with 1-6 rating) in the study of Weinstein [3], it was found to be 89.41 ± 17.38 (with 1-6 rating) in our study. The other studies conducted on university students in Switzerland and Germany reported 57.5 ± 12.61 (with 0-5 rating; it must have been 69 by 1-6 rating) and 63.08 ± 14.07, respectively [26,34]. Furthermore, Stansfeld [5] used WNSS (with 1-6 rating) in depressed subjects and control group composed of the hospital staff between 18 and 65 years of age, and the average score was found to be 56 and 57.21 for male and female controls, respectively. No average data was available in the Italian and Japanese studies. On the other hand, the study by Alimohammadi et al. [27] also reported a higher average score (74.93 ± 13.42 with 0-5 rating; it must have been 89.92 by 1-6 rating) for Persian nonindustrial employees. We may speculate that higher average scores found in Persian and Turkish studies could be related with the sociocultural differences in comparison to the Western countries; this aspect has never been questioned before and needs further research, which will be conducted in different countries.
Beyond this difference, our data was in accordance with the previous reports revealing that NSS was not different between males and females, and within the age subgroups [2,3,6]. However, Senese et al. [28] reported that females and people older than 45 years were more noise sensitive than males and younger people, respectively.
van Kamp and Davies [11] pointed out importance of focusing of noise research on vulnerable groups which could be defined by socioeconomic and/or education status.  Although noise exposure was found to be associated with income level [36], no direct data presenting a relationship between SNS and any measure of socioeconomic status was published. In our study, we used high school education as a cut-off and found that education after high school alone did not address any difference regarding increased or decreased risk of SNS, as reported by Moriera and Brian [2] and Belojevic and Jakovljevic [6]. Michaud  reported that education level had statistically significant associations with traffic noise annoyance. As a conclusion, it is said that WNSS is convenient tool for evaluation of SNS in Turkey even if the average score was higher than reported in the Western countries. No difference was detected between males and females, age subgroups (< 55 years of age), and education level (with cut-off of high school).