The process of cross-cultural adaptation followed the guidelines17 for cross-cultural adaptation of self-report measures. A five-stage process is recommended for cross-cultural adaptation: translation, synthesis, back translation, expert committee review, and pre-testing. This process should adapt the concepts to the target population's culture and verify whether the connotative meaning is equivalent to the original scale.

The initial translation into Brazilian Portuguese was performed independently by two bilingual native speakers from Brazil, both language teachers. They were assisted by the researchers if they had any questions about the scale's content. The researchers were physical therapists familiar with the scale's concepts and smartphone addiction theory. They also had considerable experience with assessments using questionnaires. The two translators combined the translations into a consensus version based on the original scale. All decisions were made by consensus among the two translators and researchers. The common translated/adapted version was then back-translated into English by two bilingual native speakers from the United States, both language teachers, who worked independently.

A multi-professional expert committee consisting of five experts – one expert in methodology, one Brazilian language professional, one English language professional, and the two physical therapists who participated in the first stage – reviewed and compared the common translation, the two back translations, and the original scale to produce a pre-final Brazilian Portuguese version of the SAS-SV. The expert committee's decisions were made to achieve semantic, idiomatic, experimental, and conceptual cross-cultural equivalence.

Pre-testing was performed to verify if this version was equivalent to the original scale and if the target group could understand it properly. For the pre-testing, we used a sample of 33 university students who received a link by e-mail or text message to access the scale and all the instructions. All students were enrolled in random courses at a university in the state of São Paulo, Brazil. The students were asked to read the scale, fully explain their answers, and to report any issues. None of the students reported difficulty in understanding or interpreting the questions.

For the test–retest, 106 students enrolled in different courses at a university in the state of São Paulo, Brazil were invited to participate in the study. The students received a link to access the scale and all the instructions by e-mail or text message. Of the 106 students invited to participate in the study, 59 (18 men and 41 women; mean age 24.5; SD=4.2 years) responded to the scale once. Therefore, the questionnaire was sent back to this population after one week. The scale was completed by 53 of the 59 students, and six students did not respond to the scale the second time.

To test the measurement properties of the final Brazilian Portuguese version of the SAS-SV, another convenience sample including 160 students, also enrolled in random courses at a university in São Paulo state, were invited to participate in the study in this phase. Of the 160 students invited to take part, 130 students answered the scale and 30 did not respond. The students received a link to access the scale and all the instructions by e-mail or text message. They responded to the scale and received the score via email and they did not receive any incentives.

The study sample size was estimated, as proposed by Terwee et al.18 and Beaton et al.,17 to be at least 50 subjects for reliability and construct validity and at least 100 subjects for internal consistency.

The consistency of a measurement, i.e., reliability, can be defined when the related conditions are constant. Reliability can be estimated through stability and internal consistency and refers to the extent to which the same score is provided on two different occasions for the same conditions, denoting stability.18–20

The subgroup of 53 university students, who answered the scale twice, participated in this test. A one or two week interval between the tests was allowed as recommended by Terwee et al.,18 because it is considered long enough to prevent recall. A minimum of 0.7 is the recommended criterion for a reliable measure.18 Good internal consistency ranges from 0.7 to 0.95.18 Moreover, for validation purposes, correlations between the total SAS-SV score and time indicators (frequency per day in hours, time working on the smartphone, leisure time on the smartphone, and age) were used to examine the construct validity. No other instrument was used for construct validity analysis due to the absence of Brazilian Portuguese questionnaires assessing smartphone addiction during the period of data collection.

The online survey comprised three sections: (a) socio-demographic, (b) mobile phone usage patterns, and (c) the SAS-SV adapted into Brazilian Portuguese. The variables examined in the socio-demographic section included: sex, age, weight, height, civil status (i.e., single, in a partnership/cohabitating/married, separated/divorced, widow(er), other), physical exercise, and pain in wrist and/or neck. Participants were asked about the presence of pain in the wrist and neck, and they could answer “yes” or “no”. In the case of physical activity, those who answered that they exercised three times or more per week were considered active. Otherwise, they were considered sedentary.

The patterns of smartphone use measured were: average minutes per day spent on the smartphone, average minutes per day of smartphone use for activities related to work/study per typical weekday; average minutes per day of smartphone use for online leisure activities per typical week; and time that the participant has owned a smartphone. All patterns of smartphone data were obtained using a questionnaire prepared by the authors of this study. The same participants answered all of the online survey questions about socio-demographic data, mobile phone usage patterns data, and the SAS-SV adapted into Brazilian Portuguese.

All statistical analyses were performed using IBM SPSS Statistics 21, and a significance level of 0.05 was adopted throughout. The normality of data distribution was assessed with the Kolmogorov–Smirnov test (p<0.001). The test has a p-value lower than 0.05, which indicates non-normal data distribution.

Intraclass Correlation Coefficient (ICC) (ICC3,1; model: two-way mixed; type: absolute agreement – single measurement) and confidence intervals (CI) of the mean coefficients (95%) were used to verify the reliability of the SAS-SV scale in the test and retest. Internal consistency was assessed using Cronbach's alpha for the 10 items. For this test, the data from a subgroup of 53 university students, who answered the scale twice were used. A minimum of 0.7 is the recommended criterion for a reliable measure.18 Good internal consistency ranges from 0.7 to 0.95.18 The analysis was also conducted excluding one item each time to check the contribution of that item to the homogeneity of the scale.

All data related to the prevalence of smartphone use were analyzed by median, maximum, and minimum. The variables sex, civil status, physical exercise and symptoms in the cervical region and wrist were associated with smartphone addiction using the Chi-Square test. Moreover, for validation purposes, Spearman correlations between the total SAS-SV score and time indicators (frequency per day in hours, time working on the smartphone, leisure time on the smartphone, time that the volunteer has owned a smartphone and age) were used to examine the construct validity. The Spearman correlation values (r) considered were: r=0.10 (small effect); r=0.30 (medium effect); and r=0.50 (large effect).21

According to Nunnally22 the sample sizes (n=130) were sufficient to do the measurement properties, because of the SAS-SV, includes ten items, so at least 100 participants were tested.

During the translation process of the scale, few disagreements were found between translators and researchers. The most common problem was adapting the verbal tense of the questions. All differences were resolved by consensus between translators and researchers, and the most appropriate terms were chosen. During the application of the scale in the sample of 33 university students, no issues were reported regarding the understanding and interpretation of the questions.

The SAS-SV showed good reliability for the test–retest scores (ICC3,1=0.82; 95% CI 0.70–0.89) among 53 university students. The SAS-SV also showed good internal consistency considering all 10 items (Cronbach's alpha=0.95). Whenever one item was excluded, alpha values did not reveal changes in scale homogeneity (Table 1).

Regarding construct validity, the variables frequency per day, time working on the smartphone, leisure time on the smartphone and time that the participant has owned a smartphone were significantly positively correlated with higher SAS-SV scores (Table 2). The correlation between SAS-SV score and frequency per day and leisure time on the smartphone were classified as medium effect (Table 2). The correlation between SAS-SV score and time working on the smartphone and time that the participant has owned a smartphone were classified as small effect (Table 2). Age was not significantly correlated with the SAS-SV score (Table 2).

The variables pain in the neck and pain in the wrist were positively associated with the SAS-SV score (Table 2). Accordingly, the participants classified as not addicted to the smartphone showed a small prevalence of pain. In contrast, the variables sex, civil status, and physical were not significantly associated with the SAS-SV score (Table 2).

The socio-demographic and descriptive characteristics of smartphone use patterns are shown in Table 3.

Table 3 shows that the population evaluated consisted mostly of young individuals with median age of 21 years. The median time spent daily on the smartphone was 5h, and the largest amount of time reported was 19h per day. In addition, the present study found high rates of wrist pain (31.4%) and neck pain (44.6%).

Based on Kwon et al.,8 we adopted a cut-off point of 31 for boys and 33 for girls to classify smartphone addicts. The present study classifies participants as “not addicted” or “addicted” according to Kwon et al.,8 and the results show that 87 (66.9%) of the population were classified as not addicted and 43 (33.1%) of the population were classified as addicted. The median, maximum, and minimum SAS-SV scores obtained in the group classified as not addicted were 25 (32-10) and for the group classified as addicted were 39 (50-32).