Table 1 shows the mean SAS-SV scores of the participants according to sociodemographic data. The highest scores were detected amongst male participants, single, from the Northern region of Brazil, and those university students. In the adult population, those with a higher level of education presented lower scores of PSU. In all analyses, we found a low effect that suggests significant observed differences may be due to high sample size.

Table 1

Variables	SAS - Short Version
	N	%	M	SD	p	η²
Sex					**	0.00
Male	262	36.50	31.6	9.1
Female	456	63.50	29.7	9.2
Marital status					***	0.03
Single	535	74.50	31.9	8.7
Married	183	25.50	27.9	9.9
Brazilian Region					0.8	0.00
South	40	5.60	29.9	10.1
Southeast	540	75.20	31.0	9.1
North	5	0.70	33.4	9.2
Northeast	54	7.50	30.2	8.6
Midwest	79	11.00	30.7	9.8
Characteristic of population					***	0.01
University Students	387	53.90	32.0	8.6
Adults	331	46.10	29.7	9.6
Adult education level					**	0.02
Elementary	4.0	1.20	32.0	5.9
High School	88	26.60	32.2	10.2
University, MSc, PhD	239	72.20	28.7	9.3

Note: ** p < 0.01; *** p < 0.001. η²: Effect size calculated by using Eta Squared Test.

Prevalence of Problematic Smartphone Use

The total amount of participants with Prevalence of Problematic Smartphone Use was 39.4% (n = 283). When the prevalence was analysed by gender, there was no significant difference (χ²= 0.45,p= 0.49, Cramer's V test = 0.02) found in the proportion of women (n= 184, 40.4%) and men (n = 99; 37.7%) with problematic smartphone use. On the other hand, a statistically significant difference (χ²= 4.25,p= 0.04, Cramer's V test = 0.08) was observed in the prevalence of PSU when comparing the adult population (35.3%) to university students (42.9%).

Table 2 shows the frequency of symptoms when comparing the adult population to university students in the total sample and only among those who presented PSU. In the overall sample, university students had a higher frequency of symptoms than adults, but only in three of them there was a significant effect. When comparing only those with PSU, the frequency of symptoms between the two samples was quite similar, and only the withdrawal’s symptom had a significantly higher frequency in adults compared to the students.

Table 2

Symptoms	Total Sample
	Adults		Students		Total		Test	p	Effect size
	N	%	N	%	N	%
Total of sample	331	100	387	100	718	100
Loss of control	102	30.8	167	43.2	269	37.4	χ2 = 11.6	***	0.12
Disruption	76	23.0	122	31.5	198	27.5	χ2 = 6.55	**	0.09
Disregard	52	15.7	70	18.1	122	16.9	χ2 = 0.71	0,39	0.03
Withdrawal	105	31.7	105	27.1	210	29.2	χ2 = 1.82	0,17	0.05
Preoccupation	94	28.4	132	34.1	226	31.4	χ2 = 2.70	0,10	0.06
Tolerance	246	74.3	335	86.6	581	80.9	χ2 = 17.3	***	0.15
Symptoms	Sample with PSU
	Adults		Students		Total		Test	p	Effect size
	N	%	N	%	N	%
Total of sample	117	100	166	100	283	100
Loss of control	83	70.9	129	77.7	212	74.9	χ2 = 1.67	0.19	0.07
Disruption	62	53.0	101	60.8	163	57.6	χ2 = 1.73	0.18	0.08
Disregard	43	36.8	57	34.3	100	35.3	χ2 = 0.17	0.67	0.02
Withdrawal	74	63.2	86	51.8	160	56.5	χ2 = 3.66	*	0.11
Preoccupation	69	59.0	99	59.6	168	59.3	χ2 = 0.01	0.91	0.00
Tolerance	116	99.1	163	98.2	279	98.6	χ2 = 0.44	0.50	0.04

Note: * p < 0.05, ** p < 0.01, *** p < 0.001. The symptom was considered present if the mean of the scores from each dimension was greater than 4.00, as follows: Loss of control (Q1, Q8); Disruption (Q2, Q10); Disregard (Q3, Q7); Withdrawal (Q4, Q5); Preoccupation (the raw score of Q6); Tolerance (the raw score of Q9). Effect size was calculated by using Cramer’s V Test according to a df = 1.

Item analysis and reliability

Considering the 10 items of the SAS-SV, the fit parameters of the unifactorial model from the confirmatory factor analysis, were inadequate (CFI = 0.806; TLI = 0.751; RMSEA = 0.124). For this reason, an exploratory analysis was performed using the PCA method (Table 3). With this method, the average item score varied from 2.55 (question 7) to 4.56 (question 9), and all items were maintained in SAS-SV because they presented a factorial load higher than 0.30. Also, all items on the scale presented good internal consistency, and overall scale reliability was high (α = 0.819).

Table 3

Smartphone Addiction Scale-Short Version
Questions	M	SD	Factor loading	Corrected item total correlation	Cronbach’s a if item deleted
Q1	3.34	1.52	0.630	0.512	0.804
Q2	3.30	1.53	0.664	0.544	0.801
Q3	2.73	1.60	0.340	0.260	0.830
Q4	2.92	1.50	0.598	0.485	0.807
Q5	2.96	1.45	0.684	0.568	0.799
Q6	2.69	1.44	0.735	0.626	0.792
Q7	2.55	1.38	0.555	0.440	0.811
Q8	3.20	1.56	0.686	0.570	0.798
Q9	4.56	1.40	0.672	0.554	0.800
Q10	2.66	1.51	0.610	0.492	0.807
Eigenvalue			3.920
Variance (%)			39.20

Note: The extract method was performed by using Principal Component Analysis, with Kaiser normalization. A cutoff of 0.30 was used for inclusion. Bartlett's Test of Sphericity: p < 0.001; KMO overall: 0.858. The overall Cronbach’s α of the SAS-Short version was 0.819.

Network analysis indicated the correlation between the all items of SAS-SV items between university students (Figure 1A) and adults (Figure 1B). All items (nodes) showed positive correlations (green edges), which indicates that the increase in the score of one item led to an increase in the scores of other items. The strength of correlations amongst students (Figure 1A) was higher between items 1 and 2 (r= 0.43); 4 and 5 (r = 0.42) as well as 5 and 6 (r= 0.38). Regarding the adults (Figure 1B), it was detected strongest correlations between items 1 and 2 (r= 0.52); 4 and 5 (r= 0.43), as well as 9 and 10 (r= 0.29). Concerning centrality measurements (Figure 1C), item 6 was one of the most important in both groups, since it had higher levels of connectivity and proximity to other items and showed a more central location in the system. Amongst adults, item 9 stood out in the two measures of centralities. On the other hand, item 3 had the lowest influence on the system with a lower correlation strength with the other items and was the most isolated one.

Figure 1 Note: Each line represents the zero-order correlation between two variables, and the thickness signifies magnitude. The green color represents a positive correlation. Those nodes (variables) with stronger inter associations appear in the center of the network. The colors of the nodes represent the symptoms: Red node: Disruption; Yellow node: Loss of control; Green node: Disregard; Light blue: Withdrawal; Blue (indigo): Preoccupation; Violet: Tolerance.

The color of the items (nodes) was based on the criteria proposed by Lopez-Fernandez (2017). The withdrawal symptoms (items 4 and 5) were the ones that presented the strongest correlations with each other, but items 1-8 (r _students= 0.04_; r _adults= 0.05; loss of control); 3-7 (r _students= 0.00;r _adults= 0.05; disregard); and 2-10 (r _students= 0.03;r _adults= 0.12; disruption) showed poor correlations. Regarding the proximity of items, only the connection between items 4-5 showed strong proximity to each other. Items 2-10 presented some proximity, and the other symptoms were very scattered in the network. As described above, item 6 (preoccupation) was the one that most connected with the other symptoms in both groups: university students (betweenness = 2.25) and adults (betweenness = 1.53).

The final version of the SAS-SV with all the items translated into Brazilian Portuguese is presented in Table 4. The instrument is under public domain and can be used freely, complying with its original version (Kwon et al., 2013a).

Table 4

INSTRUÇÕES: Por favor, assinale qual das afirmações abaixo de uma escala de 1 (discordo totalmente) a 6 (concordo totalmente) se aplica ao seu uso de celular.
Itens	Discordo totalmente	Discordo	Discordo um pouco	Concordo um pouco	Concordo	Concordo totalmente
1. Deixo de fazer tarefas ou trabalhos planejados devido ao uso do celular.	1	2	3	4	5	6
2. Tenho dificuldade para me concentrar na aula, nas lições de casa ou no trabalho devido ao uso do celular.	1	2	3	4	5	6
3. Sinto dor nos punhos ou pescoço enquanto uso o celular.	1	2	3	4	5	6
4. Não há nada mais difícil do que ficar sem meu celular.	1	2	3	4	5	6
5. Eu fico impaciente e irritado quando estou sem meu celular.	1	2	3	4	5	6
6. Fico pensando no meu celular mesmo quando não o estou usando.	1	2	3	4	5	6
7. Eu nunca vou deixar de usar meu celular, mesmo se este uso cause problemas ou efeitos negativos na minha vida.	1	2	3	4	5	6
8. Tenho que checar constantemente meu celular para não perder as publicações nas redes sociais (WhatsApp, Twitter, Facebook, Instagram, por exemplo).	1	2	3	4	5	6
9. Uso meu celular por mais tempo que pretendia.	1	2	3	4	5	6
10. As pessoas à minha volta me dizem que uso excessivamente o celular.	1	2	3	4	5	6

Source: Copyright ©Kwon et al. 2013a.

The main goal of this study was to translate and adapt the SAS-SV and to evaluate its validity in a population of Brazilian university students and adults. To our knowledge, this is the first study in the literature evaluating the psychometric properties of the SAS-SV through network analysis. The scale maintained the semantic, idiomatic and conceptual equivalences of the original version. Initially, confirmatory factor analysis was performed, similarly to the SAS-SV validation for Italian (De Pasquale et al., 2017) population. However, the CFA did not show goof fit indexes (CFI = 0.80; Standardized Root Mean Square Residual (SRMR) = 0.06). In the study by De Pasquale et al. (2017), the authors found better fit indicators (CFI = 0.92, SRMR = 0.06).

A PCA exploratory analysis method with a one-dimensional solution was therefore used, in a similar way to other SAS-SV adaptation studies (Lopez-Fernandez, 2017; Noyan et al., 2015; Sfendla et al., 2018). This procedure showed good validity and adequacy of the instrument in the Brazilian version and explained 39.2% of the variance. This finding was lower than other versions, such Arabic (42.4%, Sfendla et al., 2018), Spanish and Belgian (respectively 49.0% and 54.0%;Lopez-Fernandez, 2017). The reliability level of scale in the present study wasα = 0.819, indicating good internal consistency (Streiner, 2003). Good levels of reliability were also obtained in the original version (α = 0.91,Kwon et al., 2013a), Spanish and Belgian (α = 0.88 and α = 0.90, respectively; Lopez-Fernandez, 2017), Arabic (α = 0.87; Sfendla et al., 2018), Turkish (α = 0.88; Noyan et al., 2015), German (α = 0.85; Haug et al., 2015) and Italian (α = 0.79; De Pasquale et al., 2017).

The NA analysis showed a strong positive correlation between all the items in the scale, particularly items 1-2 and 4-5, with high levels of correlation. Regarding centrality items, 6 and 9 were the most relevant on the network; however, the symptoms proposed for the 10 items, as done by Lopez-Fernandez (2017) were only partially confirmed in this analysis. Items 4 - "There is nothing more difficult than being without my cell phone" and 5 - "I get impatient and angry when I do not have my phone” (withdrawal) showed a strong correlation with each other, but items 1-8 (loss of control) and 3-7 (disregard) presented poor correlations amongst themselves, as well as being quite separated in the network. It should be noted that the original version of the SAS-SV does not evaluate the six symptoms since this was proposed later by Lopez-Fernandez (2017), and subsequently evaluated by other authors (Sfendla et al., 2018).

Regarding the SAS-SV scores, those participants male, single, and university participants had higher scores (p< 0.001), but in all analyses, we detected a small effect size. The total prevalence of PSU was 39.4%, and in university students, this rate was 42.9%. This finding is higher than that which has been found in another study conducted in Brazil (33.1%;Mescollotto et al., 2019) and in Swiss and South-Korea, with teenagers (21.6%; Haug et al., 2015; and 16.8%: Kwon et al., 2013a, respectively). Regarding the adult population (35.3%), the findings were also higher than those observed in Spanish and Belgian adults (21.5% and 12.3%, respectively;Lopez-Fernandez et al., 2017). When the frequency of symptoms among individuals with PSU was evaluated, most of the sample showed symptoms of tolerance (98.6%). In this case, similar data was found in Moroccan (93.6%; Sfendla et al., 2018) and Spanish samples (93.9%; Lopez-Fernandez, 2017).

The present study has some limitations that must be considered. A concurrent validity analysis of the SAS-SV was not performed due to the lack of another instrument validated in Brazil for the adult population until de period of the data collection. Test-retest analyses were also not performed. The main strength of this study was the adaptation and validation of the instrument for two populations, as well as the large sample size. The network analysis was also a major point in this work, although a few psychometric studies still use this statistical procedure. This type of analysis ensures better technical information because it estimates associations after multivariate control of all items in the system. Network analysis ensures that the remaining links in the system are the most stable.

The SAS-SV presented reliable semantic equivalence and factor structure for both university students and adults. This instrument is a suitable tool to detect the pattern of smartphone use for the populations studied. The prevalence of PSU was higher than the ones reported in other studies found in the literature, which corroborates the high frequency of use of digital media in Brazil found in international reports.