The development of the resilience to health misinformation in the context of social media instrument followed standardized criteria (eg, selecting and formulating items, scoring issues, pilot testing, and field testing) [21], allowing the development of a theoretically grounded and context-specific measure of individual resilience to misinformation encountered in digital environments, particularly social media.

The formulation of items was based on the examination of the resilience instruments [18,19,22], which helped identify relevant domains and behavioral expressions of resilience. Drawing from these sources, we identified core competencies and attitudes relevant to critically evaluating and resisting misinformation, including cognitive reflection (eg, I think carefully before I create a new post), emotional self-regulation (eg, I easily give up on seeing or reading the content that is on social media), and personal agency (eg, I like to ensure that my posts on social media are not dubious). These dimensions were operationalized into an initial pool of 18 candidate items, written in plain language and tailored to reflect everyday social media experiences. Each item was scored on a 5-point Likert scale: (1) strongly disagree, (2) disagree, (3) neutral, (4) agree, and (5) strongly agree.

To ensure content validity and appropriateness for the target population, a panel of 5 experts was convened. The panelists were selected through a structured nomination process based on their expertise in communication, psychology, and health, along with input from parents of school-age children to incorporate lived experience into the review.

The panel was asked to evaluate each item for comprehensibility, relevance, and completeness. The review process included 2 iterative rounds: in the first round, panelists provided individual ratings and comments on each item via a web-based survey; in the second round, a synchronous discussion was held to reach consensus on the proposed revisions. Because of this process, 3 items were deemed redundant or unclear and were removed, while 5 additional items were reworded to improve precision and reduce ambiguity.

Following the expert review, the revised 15-item version was pilot-tested with 15 participants from the target population—parents of school-age children (aged 6 to 10 years). The participants were recruited through local parent associations using a purposive sampling approach to ensure variation in education level and digital media use.

The participants were asked to complete the instrument and provide feedback regarding item clarity, interpretability, and response burden. Feedback was collected through brief follow-up interviews. This user-centered pilot test allowed the identification of minor structural and linguistic adjustments, which were incorporated into the final version of the instrument. No items were removed during the pilot phase, although adjustments were made to the wording of 3 items to improve clarity and alignment with the reading level of a general adult population.

The final instrument comprised 15 items. Each item contributes equally to the total score, which ranges from 1 (indicating low resilience or high susceptibility to misinformation) to 5 (indicating high resilience or low susceptibility to misinformation). Items 6, 9, and 11 to 15 were negatively worded and thus reverse-coded before computing the composite score. Higher scores indicate greater resilience to misinformation on social media.

The instrument was originally developed in European Portuguese. For this publication, the items were translated into English using a forward and backward translation process, following the best practices recommended in the literature to ensure conceptual and linguistic equivalence (Multimedia Appendix 1) [23].

This study was embedded within the BeE-school project (clinical trials registration: NCT05395364), a pragmatic cluster randomized controlled trial conducted in the north of Portugal and designed to evaluate the effectiveness of a school-based intervention on children’s health [24].

For the present psychometric evaluation, a cross-sectional study design was used, based on the baseline data from parents participating in the trial. Eligible participants met the following criteria: (1) were Portuguese speaking, (2) were aged ≥18 years, (3) were a parent or legal guardian of a school-age child (aged 6-10 years), and (4) provided consent to participate in the baseline data collection.

All parents of children enrolled in the trial (N=735) were invited to complete a structured, paper-based questionnaire, which included the newly developed resilience to misinformation in the context of social media instrument. A total of 511 parents completed the questionnaire, yielding a response rate of 69.7% (511/735).

A nonprobabilistic convenience sampling approach was applied within the context of the larger trial. The schools participating in the BeE-school project facilitated the distribution of the questionnaires to families, and the completed surveys were returned via sealed envelopes through the schools. This strategy enabled access to a diverse sample of parents from the local community.

Among the respondents, the majority (418/511, 82.6%) were mothers, with a mean age of 41.2 (SD 6.1) years. Most participants were married or cohabiting (392/511, 78.1%), had completed university-level education (219/511, 44.9%), and were employed at the time of the study (390/511, 81.4%). The sociodemographic characteristics of the sample are presented in Table 1.

This study was approved by the ethics committee for Life and Health Sciences of the University of Minho (CEICVS 009/2022) and was conducted in accordance with the Declaration of Helsinki for human studies. The study implemented rigorous measures to protect the collected data (eg, General Data Protection Regulation) and minimize risks associated with data processing, including pseudonymization and restricted-access control to ensure confidentiality. Before data collection, parents provided written informed consent. No compensation was provided to participants.

Data analysis was performed using SPSS software (version 29.0; IBM Corp), Factor 9.2 (Lorenzo-Seva & Ferrando, Rovira i Virgili University) [25], and JASP (version 0.18.1.0; JASP Team, University of Amesterdam). A level of significance of .05 was considered. The analysis followed the COSMIN (Consensus-Based Standards for the Selection of Health Measurement Instruments) and COMET (Core Outcome Measures in Effectiveness Trials) guidelines on the taxonomy of measurement properties [26]. Initial data screening revealed no missing values. However, a few moderate univariate outliers were identified using Mahalanobis distance [27]. These cases were retained in the dataset because of their limited influence on the overall results. To ensure robust model testing, the sample was divided into 2 distinct subsamples based on 2 pre-established groupings of schools participating in the study. The parents from group 1 schools completed the version of the questionnaire used for the exploratory factor analysis (EFA), while parents from group 2 schools completed the version used for the confirmatory factor analysis (CFA; Figure 1).

Psychometric sensitivity was assessed through the descriptive statistical analysis of item response distributions, focusing on central tendency, skewness, and kurtosis. Items were flagged as potentially problematic if they showed skewness >3 and kurtosis >7 [28].

The EFA was conducted on the first subset to explore the factor structure of the resilience to misinformation on social media questionnaire. Analyses were based on a polychoric matrix, using robust diagonally weighted least squares as the extraction method [29]. The number of factors was determined via the parallel analysis technique with random permutation of the observed data [30], and the rotation method used was robust promin [31]. The model fit was assessed using the root mean square error of approximation (RMSEA), comparative fit index (CFI), and Tucker-Lewis Index (TLI), with the accepted thresholds being RMSEA<0.08 (95% CI not exceeding 0.10), and CFI and TLI ≥0.90 [32].

Factor stability was assessed using the H-index. Values >0.80 indicated a well-defined latent structure, suggesting that the factor would likely remain stable across different samples [33]. The CFA was then conducted on the second subset to confirm the factor structure derived from the EFA. Three models with different configurations were tested: (1) a unidimensional model, (2) a 2-factor model from the EFA, and (3) a refined 2-factor model with residual covariances based on modification indexes.

The model fit was assessed using the chi-square test, the χ² to df ratio (χ²/df), the CFI, the TLI, and the RMSEA. Acceptable model fit was defined for χ²/df <5, CFI and TLI ≥0.90, and RMSEA <0.08 [32]. The best-fitting model was visually represented using Structural Equation Modeling diagnostic tool [34,35].

To establish the factorial validity, the standardized factorial weights were examined. Items with loadings ≥0.50 (λij ≥0.50 and λij² ≥0.25) were considered valid indicators of their respective constructs [27]. The convergent validity was evaluated using the average variance extracted (AVE) [36-38], with values >0.50 indicating adequacy [36]. The discriminant validity was assessed by computing the heterotrait-monotrait ratio of correlations (HTMT), with values <0.90 suggesting acceptable discriminant validity [39].

Finally, the internal consistency was measured using Cronbach α and McDonald ω. Values ≥0.70 were considered acceptable, indicating a good level of internal consistency [40]. The interitem reliability was measured by computing the mean interitem correlations, which were considered acceptable within the range of 0.15 to 0.50 [41]. The corrected item-total correlations were also analyzed, with acceptable thresholds set at ≥0.20 [27]. The Pearson correlation coefficients among the dimensions of the instrument were interpreted as follows: >0.80 (very strong), 0.60 to 0.80 (strong), 0.40 to 0.60 (moderate), 0.20 to 0.40 (weak), and <0.20 (negligible) [42].

This initial phase involved developing the item pool, evaluating item relevance, and refining item wording based on expert feedback and pilot testing. The result was a 15-item instrument reflecting two conceptually grounded dimensions: (1) stress resistance and resilience to misinformation and (2) self-control regarding misinformation.

This study provides satisfactory evidence regarding the overall reliability and dimensional reliability of the resilience to misinformation on social media instrument. The findings support the internal consistency of the instrument both as a whole and within its 2 distinct dimensions. Furthermore, item analysis demonstrated consistently good psychometric properties across all 14 items, with satisfactory item-total correlations and acceptable ranges of skewness and kurtosis, reflecting the precision and conceptual clarity achieved during the item development process. The interconnectedness among the items revealed a coherent structure, culminating in the identification of 2 underlying dimensions: stress resistance and resilience and self-control regarding misinformation. This dimensional structure not only enhances the comprehensiveness of the instrument, but it also aligns well with established theoretical frameworks of resilience [13-15,18].

The study supports a conceptualization of resilience to misinformation on social media as a multidimensional construct, encompassing both stable traits and dynamic states. The self-control dimension captures more trait-like characteristics [19], such as a consistent tendency to question sources, critically reflect before sharing content, and act responsibly in digital environments. These traits are generally stable over time. In contrast, the stress resistance and resilience to misinformation dimension reflects resilience as a process that focuses on emotional and psychological responses to stressors [14] triggered by misinformation exposure, including confusion, distress, or cognitive overload. These responses are context-dependent and may fluctuate based on situational factors.

Together, these dimensions provide a nuanced understanding of how parents of school-age children experience and manage misinformation. This perspective enhances the public health relevance of the instrument by enabling the design of tailored interventions that address both enduring vulnerabilities and context-sensitive reactions to misleading content.

Substantial evidence supported the construct validity and reliability of the instrument. Convergent and discriminant validity indicators were confirmed using AVE and HTMT, establishing that the 2 factors were conceptually distinct and adequately represented the underlying construct. Internal consistency was assessed using both Cronbach α and McDonald ω, acknowledging their distinct assumptions. While slight deviations were noted in α values, ω coefficients, interitem correlations, and corrected item-total correlations supported acceptable reliability.

This dual reporting allows comparability with future studies and provides a more complete picture of the performance of the scale.

The resilience to misinformation on social media instrument can serve as a valuable tool for identifying knowledge gaps in resilience to misinformation, evaluating individual susceptibility to misinformation, adversity, and uncertainty [45,46]. In addition, it enables the monitoring of the effectiveness of interventions in strengthening resilience to misinformation. By systematically gathering this information, evidence-based strategies can be developed to enhance individuals’ ability to navigate complex health information environments, resist misinformation, and adapt to uncertainty, ultimately fostering a more resilient and well-informed society.

Moreover, while the instrument was designed to assess resilience to misinformation broadly, it is important to address health misinformation, a growing public health concern [10]. A clearer understanding of resilience in this domain is crucial for designing interventions that foster critical thinking, emotional self-regulation, and responsible information-sharing behaviors, especially among parents tasked with evaluating health-related information for their children and families.

This study is not without strengths. We devised the instrument by carefully crafting items in both positive and negative formulations, thereby mitigating the susceptibility of the scale to the influence of acquiescence response bias [47,48]. This approach aligns with the recommendation of Wagnild [48] to rephrase statements and incorporate negative keying in certain items. Similarly, adjustments to the Likert scale items to improve alignment with the questions were explored to enhance their use. Still, the scale would gain value from further investigation into its test-retest reliability, as well as its predictive validity, such as through its use in longitudinal studies. It has been established that community and cultural factors play a significant role in shaping the definition and typical expression of resilience among diverse populations [49]. Therefore, it would be beneficial for future research to investigate the concept of resilience to misinformation in various cultural and population contexts, aiming to determine the stability of the construct. We used a novel approach to determining discriminant validity, demonstrating superior performance in comparison to other criteria [36] and the assessment of (partial) cross-loadings [50].

Despite its strengths, this study presents several limitations that warrant consideration. First, there is currently no established gold standard for assessing resilience to misinformation, which limits direct comparisons with the instrument. Future research should consider exploring convergent and discriminant validity through comparisons with related constructs, such as resilience, digital health literacy, cognitive reflection, emotional self-regulation, or critical thinking. Second, resilience to misinformation on social media instrument relies on self-reporting. While this method facilitates large-scale data collection and enhances feasibility, it introduces potential response biases, particularly those related to social desirability and acquiescence tendencies [47,51]. Such biases may affect the accuracy of reported attitudes and behaviors, especially in sensitive or socially charged domains such as misinformation. Third, the psychometric testing was conducted using baseline data from the BeE-school project, in which the vast majority of respondents were mothers (82.6%). While this group is highly relevant, given its central role in mediating and guiding children’s media use and health decisions, the predominance of female participants may limit the generalizability of findings to other caregiver profiles, particularly fathers or nonparental caregivers. Previous research indicates that women may experience higher levels of perceived stress and lower levels of self-efficacy and resilience when compared with men [52-54], which could have influenced response patterns in this study. These gender-related differences may also affect individuals’ emotional regulation, capacity to critically engage with and process information, and ability to respond to adversity [54], including exposure to misinformation on social media platforms.

Future studies should focus on construct validity, ensuring the effective functioning of the test as a construct and accurately measuring its intended aspects through well-defined items. This form of validity essentially encompasses various types of validity [55]. The examination of construct validity involves analyzing correlations between the construct under investigation, in this case, resilience to misinformation, and variables known to be interconnected [56]. Although the translation process adhered to scientifically established procedures to ensure the accuracy and equivalence of the English version, it is important to note that the instrument was developed and validated in European Portuguese. Therefore, any application in different linguistic or cultural contexts should be preceded by a thorough content analysis and followed by an independent validation process to confirm the reliability and validity of the instrument in the new setting.

While the resilience to misinformation on social media instrument demonstrated satisfactory reliability when applied to Portuguese parents of school-age children, the findings should be interpreted with caution. Further research is needed to examine the performance of the instrument across different populations, age groups, and cultural contexts. Longitudinal studies are also recommended to explore the predictive validity and stability of the scale over time. The instrument has the potential to be applied in intervention research, particularly in educational and public health contexts, to help design and evaluate strategies that foster resilience. We posit that this study contributes to the initiation of a more profound exploration of resilience to misinformation associated with social media.

Capturing resilience to misinformation in the context of social media is critical for informing public health efforts to combat the infodemic. By identifying individuals and groups, such as parents of school-age children, who may be particularly vulnerable to misleading content, this instrument provides a valuable foundation for the development of tailored interventions.

The resilience to misinformation on social media instrument demonstrated satisfactory psychometric properties and offers a promising foundation for advancing both research and practice in misinformation resilience. Continued validation and application across diverse settings will strengthen its utility and impact.

The instrument demonstrated satisfactory psychometric properties, with strong evidence of construct validity, dimensional structure, and internal consistency. The 2 identified dimensions—stress resistance and resilience and self-control regarding misinformation—reflect the dynamic interplay between emotional regulation and cognitive control, and align with contemporary models of resilience. This work supports broader efforts to mitigate the impact of misinformation and promote resilience in an increasingly complex information ecosystem.