Cultural adaptation, translation, and pretesting (completed in the first quarter of 2026):

The questionnaire draws from validated Norwegian instruments, including those used in the North Norwegian Youth Study (NNYS) [45], a longitudinal epidemiological study conducted in the three northernmost counties of Norway between 1994–1995 and 1997–1998; the Norwegian Arctic Adolescent Health Study (NAAHS) [30], which was carried out in the same region beginning in 2003; and the National Norwegian Youth Survey, to assess ethnicity, ethnic identity, and experiences of discrimination. The PIU questionnaire has been culturally adapted for Sámi and non-Sámi students in Finnmark, while items measuring social norms were developed based on relevant theories and prior research. Input from group interviews with 15 Sámi and non-Sámi high school students informed the adaptation of PIU and social norms items, ensuring cultural relevance and comprehensibility. Items not previously translated into the Northern Sámi language—particularly those related to PIU and social norms—were translated from Norwegian using a forward–backward translation method to ensure linguistic accuracy and conceptual equivalence.

The draft questionnaire will undergo a pretest with approximately 30 Sámi and non-Sámi students to evaluate the clarity, comprehension, and usability of the items. The pretest aims to identify potential issues with question interpretation, the estimated completion time, and any participant burden. Feedback from participants will guide the refinement of wording, layout, and structure to improve the overall quality of the questionnaire.

Pilot testing and finalization of the questionnaire:

Following the pretest, a pilot test will be conducted with a separate group of approximately 30 Sámi and non-Sámi students. The goal of the pilot test is to assess the preliminary reliability and validity of the questionnaire. Internal consistency will be measured using Cronbach’s alpha, while exploratory factor analysis will be used to explore the construct structure of the instruments. The results from the pilot test, combined with participant feedback, will be used to make final revisions to the questionnaire.

Students participating in the pretest and pilot test will not be included in the main study sample (WP2: Data Collection), as these activities are conducted during the 2025–2026 school year primarily with VG3 students (equivalent to 13th grade), who will have graduated before the 2027 data collection begins. This ensures that there is no overlap between pilot participants and the total population sample.

Upon completion of the study and publication of results, the Norwegian and Northern Sámi versions of the questionnaire will be made openly accessible for research purposes.

Classification of Sámi and non-Sámi adolescents:

Ethnicity is a multidimensional construct that encompasses ancestry, language, cultural affiliation, and self-identification. As emphasized by Bhopal (2004), ethnicity reflects “a mix of cultural and other factors including language, ancestry, and self-identity,” and researchers are encouraged to classify ethnic groups using “one or a mix” of these indicators [57]. This conceptualization aligns with long-established Sámi health research traditions, which operationalize Sámi ethnicity using both linguistic and self-identification dimensions [58].

The combined use of language indicators and self-identification is particularly crucial in the Sámi context due to the historical effects of Norwegian assimilation policies. These policies contributed to the loss of Sámi language transmission and to shifts in ethnic self-identification across generations, resulting in situations where individuals may have Sámi ancestry but do not identify publicly or subjectively as Sámi. Findings from the North Norwegian Youth Study [45] illustrate this divergence clearly: fewer than half of adolescents with a Sámi ancestral background reported Sámi self-identification. Consequently, relying solely on either ancestry/language background or self-identification would systematically misclassify a substantial proportion of Sámi youth.

In line with this evidence, the present study classifies adolescents as Sámi if they meet either of the two criteria, or both. The first criterion is Sámi language use at home. Respondents will indicate spoken language (Sámi, Norwegian, Kven, Finnish, Swedish, or other), with multiple responses permitted, and participants will be classified as Sámi when Sámi is reported as a language they have learned or use at home. The second criterion is self-identification, defined by selecting Sámi—either alone or in combination with other categories—in response to the item “I regard myself as…,” which offers the same set of ethnic options and allows multiple responses.

Those who report no Sámi language affiliation and do not identify themselves as Sámi will be classified as non-Sámi. This group comprises primarily majority Norwegian adolescents and ethnocultural Kven youth, reflecting the population composition of Finnmark. A small number of adolescents with other non-Sámi backgrounds (e.g., immigrant or minority backgrounds not Norwegian or Kven) will be included in the non-Sámi category; these groups will not be examined separately due to limited sample size and will not be the focus of analyses.

To enhance transparency and acknowledge potential heterogeneity within the Sámi group, descriptive analyses will distinguish adolescents who meet only the language-use criterion, only the self-identification criterion, or both.

Problematic internet use (PIU):

The Problematic Internet Use Questionnaire, Short Form (PIUQ-9), will be used to measure problematic internet use (PIU) [59,60]. The PIUQ is a widely used and psychometrically robust instrument, making it a preferred choice in research assessing PIU [61,62]. This validated instrument assesses three dimensions of problematic internet use: obsession, neglect, and control disorder, with three items per dimension [59]. The obsession dimension captures obsessive thinking about the internet, including daydreams, fantasies, and mental withdrawal symptoms when access is restricted. The neglect dimension reflects the disregard of essential needs and everyday activities, such as sleep, schoolwork, and in-person interactions. The control disorder dimension assesses difficulties in regulating internet use, including unsuccessful attempts to reduce or control online behavior. Each item is rated on a 5-point Likert scale (1 = “never” to 5 = “always”), resulting in a total score ranging from 9 to 45, with higher scores indicating greater levels of problematic internet use.

For descriptive prevalence reporting, a cut-off score of ≥ 22 will be pre-registered to classify participants as being “at risk” of problematic internet use. This threshold is based on proportional scaling from the established cut-off for the PIUQ-18 (≥ 41) [59]. Although this cut-off will be used to estimate prevalence, PIU will be treated as a continuous latent variable in all structural models to capture the full range of variability in participants’ responses. Both the PIUQ-9 and PIUQ-18 have demonstrated satisfactory reliability and validity across diverse cultural contexts [59,61]. However, no universally established cut-off score exists for the PIUQ-9, and prior studies have highlighted the need to develop such thresholds to enhance its applicability [61,63]. Given the variability in cut-off recommendations across study countries, the cut-off of ≥ 22 will be used as a descriptive benchmark rather than a diagnostic criterion.

To ensure contextual relevance to contemporary internet use—particularly considering changes in youth internet behavior such as widespread smartphone use—the PIUQ was adapted based on input from group interviews with Sámi and non-Sámi high school students. These revisions updated the original questionnaire, which was developed over a decade ago. Permission for adaptation to the Finnmark context was obtained from the original author.

Positive internet use:

In addition to assessing problematic internet use, the questionnaire includes items on positive internet use. Positive internet use is conceptualized as a conceptually and empirically informed construct derived from previous research on Indigenous youth. These items are informed by previous research showing that digital technologies can support social connection, emotional support, cultural engagement, and identity affirmation among Indigenous youth in Australia and Canada [21–23]. Community belonging and cultural connectedness have also been identified as protective factors associated with well-being and resilience among Indigenous adolescents [25]. Therefore, positive internet use in the present study is conceptualized as adolescents’ use of digital technologies to maintain social connections, seek social support and emotional connection, and engage with cultural identity. These domains were further refined through group interviews with Sámi and non-Sámi adolescents to ensure cultural relevance in the Finnmark context.

The pilot phase will be used primarily to assess item clarity, relevance, and feasibility rather than to establish the psychometric properties of the construct. The dimensionality, reliability, and construct validity of the positive internet use items will subsequently be evaluated in the full study sample prior to inclusion in structural analyses. If the items demonstrate a unidimensional structure, they will be modeled as indicators of a single latent construct. If a multidimensional structure emerges, positive internet use will be modeled as a higher-order latent construct.

Social norms:

All social‑norm items used in this study are operationalized as perceived norms—that is, adolescents’ subjective perceptions of others’ behaviors (descriptive norms) and approvals or disapprovals (injunctive norms) within family, peer, and school contexts, rather than objective norms obtained directly from those referent groups.

Social norms related to internet use behaviors will be measured using items adapted from theories of social norms [31,32] and previous research on adolescent risk-taking behaviors [33,34,36–38]. Descriptive norms measure how frequently participants believe their reference group engages in internet‑related behaviors; reference groups are defined as the cultural and social networks that adolescents consider when making behavioral decisions [31,32]. Injunctive norms will assess perceived approval or expectations from their reference group regarding internet use. Social norms items were developed through group interviews with Sámi and non-Sámi high school students to ensure cultural relevance and alignment with current internet use. The study focuses on three key reference groups influencing adolescents use of internet and mobile phone: (1) Family—parents and older household members; (2) Peers—both in-person and online friends; and (3) School — teachers and rules regarding internet and mobile phone use. Assessing norms across these domains provides a comprehensive view of the social influences associated with adolescent internet behaviors.

Ethnic identity and ethnic-related victimization:

Ethnic identity will be assessed using the revised version of the Multigroup Ethnic Identity Measure (MEIM) [64], which captures two core dimensions: ethnic identity search (cognitive exploration) and affirmation, belonging, and commitment (affective attachment). This instrument has been widely applied in diverse cultural settings [65,66], with the Norwegian versions applied in NNYS [50] and NAAHS [30]. Items assessing ethnic-related victimization, such as experiences of discrimination, will be adapted from NAAHS [30,67]. Permission to use questionnaire items from NNYS and NAAHS has been obtained from the original authors.

Control variables:

To reduce potential confounding, covariates were selected a priori based on prior research on adolescent internet use [19,42,45]. Individual‑level control variables include gender, grade level (VG1–VG3; equivalent to 11th–13th), living arrangement (with parents vs. dormitory), depressive symptoms, alcohol use, smoking/nicotine use, and ethnic community context.

Depressive symptoms are measured using five items assessing emotional distress during the past week, rated on a four‑point scale and treated as a continuous measure. Alcohol use is assessed by frequency of consumption (never to weekly) and intoxication during the past 12 months (never to >10 times). Smoking/nicotine use is measured by current smoking frequency (never smoked to weekly or more). All items are drawn from the National Norwegian Youth Survey (Ungdata) and included for statistical adjustment in the structural models.

Ethnic community context is included as a school‑level control variable, operationalized by whether the school is located in a municipality included in the Sámi Language Act administrative area [68] (used as a proxy for higher Sámi density). Schools located in Kautokeino, Karasjok, Lakselv (Porsanger), and Tana are included in this administrative area and are therefore classified as higher Sámi density, whereas schools in all other municipalities are classified as lower Sámi density. This variable is entered as an exogenous covariate with direct paths to the outcome variables and is not used for subgroup or multigroup analyses due to sample size considerations.

Data preparation:

Prior to analysis, the collected data will undergo a thorough cleaning and preparation process to ensure accuracy and reliability. This process includes examining the distribution of values, identifying potential collinearity, and addressing missing values, outliers, and inconsistencies.

The extent and pattern of missingness will be examined across all variables included in the PLS-SEM measurement and structural models. Following published PLS-SEM recommendations, respondents with more than 15% missing data across all model-relevant indicators will be excluded prior to imputation. This threshold aligns with guidance that variables or cases with excessive missingness should be removed to maintain data quality. The percentage of missingness per variable and for the full dataset is transparently reported, as recommended [69].

For remaining missing values, multiple imputation by chained equations (MICE) will be used under the assumption of missing at random (MAR), generating m = 30 imputed datasets. This approach is consistent with best practices, which recommend multiple imputation (or maximum likelihood-based methods) over deletion when missingness exceeds minimal levels (~5%) and strongly advocate for imputation when missingness surpasses ~10% to preserve statistical power and reduce bias. The imputation model included all variables used in the analysis as well as auxiliary predictors of missingness to improve the accuracy of imputations [69–71].

Because the study involves multigroup PLS-SEM (PLS-MGA) to compare Sámi and non-Sámi adolescents, imputations will be conducted separately within each group. This ensures that group-specific distributions, means, variances, and covariances are preserved, which is critical for valid multigroup comparisons and measurement invariance testing (MICOM). This procedure follows PLS-SEM guidance on carefully handling grouping variables during preprocessing to avoid introducing biased or artifactual differences [69,70].

Outliers will be identified and treated following PLS-SEM data quality recommendations. Suspicious response patterns (e.g., straight-lining, alternating extremes) will be screened using descriptive statistics and visual inspection, and any removals will be documented. Univariate outliers will be flagged using boxplots (values beyond three times the interquartile range [IQR]), while multivariate outliers will be assessed using Mahalanobis distance and leverage diagnostics appropriate for SEM preprocessing. Clearly erroneous outliers, such as data entry artifacts, will be removed. However, extreme values that could be meaningfully justified will be retained, consistent with recommended practices. The final analytic sample size (N) is reported [69,70].

To ensure consistency across numerical variables, all data will be standardized to a zero mean and a standard deviation of 1 prior to analysis.

Descriptive statistics:

Descriptive statistics (e.g., means, standard deviations, frequencies) will be calculated for all key variables, including PIU (total PIUQ-9 score), positive internet use, and demographic characteristics and control variables. Prevalence estimates for PIU will be reported using the pre-registered cut-off score of ≥ 22 on the PIUQ-9 to classify participants as being “at risk” of problematic internet use. Group comparisons (e.g., Sámi vs. non-Sámi adolescents) will be conducted using non-parametric tests, such as the Kruskal-Wallis H test, to account for potential differences in sample sizes and distributions.

Partial Least Squares Structural Equation Modeling (PLS-SEM):

PLS-SEM analysis will be performed to investigate the associations between exogenous and endogenous variables. PIU will be modeled as a continuous latent construct with three reflective dimensions (obsession, neglect, and control disorder), while positive internet use will be modeled as a separate latent construct indicated by items reflecting social connections, emotional support, and cultural engagement.

As the first step of the PLS-SEM analysis, model evaluation will be conducted for both the measurement and structural models. Assessing the measurement models included evaluating factor loadings, reliability and convergent validity. Factor loadings indicate the variance explained by a measured variable on a specific factor, with factor loadings of 0.5 or higher considered acceptable [54]. Internal consistency reliability will be evaluated using composite reliability (CR) for each latent variable. Unlike Cronbach’s alpha, which assumes equal indicator loadings, CR offers more flexibility and is thus more suitable for PLS-SEM. [54]. Convergent validity will be evaluated using the average variance extracted (AVE), with values of 0.4 or higher indicating adequate convergence [54]. Once the measurement model is validated, the structural model will be analyzed to test the study’s hypotheses. Bootstrapping with 5,000 resamples will be used to generate confidence intervals and determine statistical significance (p < 0.05) [54]. Multicollinearity among predictor constructs will be assessed using the variance inflation factor (VIF), with values above 5.0 indicating potential collinearity [54].

PLS multi‑group analysis (PLS‑MGA):

After evaluating the structural model, we will examine whether the relationships among the study variables differ between Sámi and non-Sámi adolescents through Partial Least Squares Multi-Group Analysis (PLS-MGA). Because meaningful multigroup comparisons require that constructs are measured equivalently across groups, the analysis begins with an assessment of measurement invariance using the Measurement Invariance of Composite Models (MICOM) procedure [72,73]. In this step, we first ensure that both groups share the same basic measurement configuration, including identical indicators, model specifications, and data treatment procedures. We then evaluate whether the composites are formed in a comparable way across groups by testing the equivalence of their indicator weights through a permutation-based approach. Finally, we assess whether composite means and variances differ statistically across groups, which allows us to determine the level of invariance established before proceeding to the substantive group comparisons.

The primary preregistered multigroup analysis uses a combined definition of Sámi identity, classifying adolescents as Sámi when they meet either the language-use or self-identification criterion, or both. To enhance transparency and assess robustness, descriptive analyses will distinguish between adolescents who meet only the language-use criterion, only the self-identification criterion, or both. In addition, planned sensitivity analyses will restrict the Sámi group to adolescents who meet both criteria.

Following establishment of measurement invariance, multigroup structural comparisons will evaluate whether associations among social norms, ethnic identity, ethnic victimization, and problematic or positive internet use differ between groups. These comparisons draw on established PLS‑MGA procedures [73,74], including permutation‑based assessments of whether observed between‑group differences exceed what would be expected by random reassignment, bootstrap‑based evaluations of group‑specific sampling distributions, and complementary parametric tests that examine differences under assumptions of equal and unequal variances. Together, these procedures allow for a comprehensive assessment of potential group-specific differences in both measurement and structural relations.

By integrating MICOM with PLS‑MGA, the analysis provides a rigorous basis for determining whether associations among social norms, ethnic identity, ethnic victimization, and problematic or positive internet use vary between Sámi and non‑Sámi adolescents.

To reduce potential confounding, the primary PLS‑SEM specification will include the following exogenous covariates with direct paths to PIU: gender, grade level (VG1–VG3), living arrangement (with parents vs. dormitory), depressive thoughts, alcohol use, smoking, and Sámi community context.

Risk-benefit assessment.

The study involves no physical intervention and is fully pseudonymized. However, topics such as ethnic identity, discrimination, and problematic internet use may evoke emotional discomfort, particularly among Sámi adolescents. To mitigate this, participation is voluntary, questions may be skipped, and support staff will be available during school-based data collection. All data are securely stored and accessible only to authorized researchers, as described above. The study follows ethical principles for Indigenous research, emphasizing relevance, respect, and voluntary involvement.

The potential societal benefits—particularly improved understanding of internet use, social norms, and ethnic identity—outweigh the minimal risks. Findings may inform culturally adapted support programs in schools and health services, promoting inclusion and well-being.