The goal of study 1 was to first create a comprehensive list of candidate items to assess a range of social media behaviors. The authors first drafted items to assess both positive and negative aspects of social media use, including addictive characteristics (eg, How hard is it to resist the urge to check social media?), self-branding (eg, How much do your profiles represent your “real life”?), adaptive coping (eg, How often do you turn to social media to relax?), and positive uses (eg, How often do you use social media for activism purposes?). Additional candidate items were then created after reviewing previous research and measures examining social media behaviors [33-37] to ensure a representative list of survey questions. Psychology professors at a small liberal arts college from a variety of disciplines (eg, cognitive, school, and clinical) informally reviewed the items and provided feedback. Since many young people have multiple social media accounts [1,2], all items use flexible language that is not specific to one social platform’s culture or cues. To respond, participants indicate the extent to which they agree with an item’s statement using a Likert scale ranging from 1 (not often or not at all) to 10 (very often or very much), such that a higher score indicates greater perceived engagement in a specific social media behavior. Our item generation process created a total of 42 candidate items. We then aimed to conduct an exploratory factor analysis to identify latent variables that would guide the structure of the CASM.

The first exploratory factor analysis produced a solution that included 12 factors that accounted for 67.5% of the variance. Using the criteria outlined above, a total of 16 items were eliminated, resulting in a 6-factor, 26-item solution. We deviated from the established cutoff criteria on 1 occasion: factors 7 and 9 were retained as we felt they would combine into 1 factor since both contained items related to online social communication. Thus, 8 factors and 30 items were retained from the first exploratory factor analysis.

A second exploratory factor analysis was conducted to confirm the identified factors. Using the same criteria outlined by Costello and Osborne [38], the second exploratory factor analysis resulted in a 7-factor solution with 29 items that accounted for 61.1% of the variance. One minor adjustment was made to the solution: item 29 (How often do you post to make others think you are feeling better than you actually are?) was moved to factor 6, as it had previously loaded onto factor 6 and aligned more with its theme of inducing negative affect. The factors and their internal consistency reliability are as follows: self-branding (0.829), compulsive use (0.798), disruptive use (0.759), impulsive sharing (0.701), social engagement (0.755), induce negative emotions (0.768), and induce positive emotions (0.735). Table 2 lists all retained items and their factor loading, mean, and SD.

The goal of study 2 was to conduct a confirmatory factor analysis to evaluate the model fit of the 29-item CASM measure created in study 1 (see Multimedia Appendix 1). A secondary goal was to assess the discriminant, convergent, and criterion validity of the CASM.

Data collection occurred from December 2018 to December 2019 on the online research platform Qualtrics. After providing informed consent, participants completed a demographic survey and the 29-item CASM. Participants then completed the B-FNES, DASS, MEIS, BFAS, and MoVEE. After completing the study measures, participants received a debriefing form with a list of local and national mental health resources.

Descriptive statistics were calculated to characterize sample demographics. Model fit indices calculated included root mean square error of approximation (RMSEA) [47], comparative fit index (CFI) [48], Tucker-Lewis index (TLI) [49,50], and standardized root mean square [51]. RMSEA provides an estimate of how far a hypothesized model is from a perfect model, CFI and TLI compare the performance of a hypothesized model against a null model, and standardized root mean squared residual measures the average difference between observed and predicted correlations.

Discriminant validity was examined by calculating correlations between the CASM, the MoVEE, and self-reported frequency of phone video game use. The MoVEE scale and the frequency of phone video game use question were chosen to assess discriminant validity, as both measure smartphone uses not related to social media. Construct validity was determined by calculating correlations between CASM latent factors, the BFAS, and MEIS-CF. Criterion validity was evaluated by calculating correlations between CASM latent factors and 2 indicators of well-being: fear of negative evaluation (measured by the B-FNE) and depression, anxiety, and stress (measured by the DASS).

Finally, a bivariate correlation was calculated to test whether there is a significant relationship between hours per day spent on social media and symptoms of depression, anxiety, and stress. For all analyses, missing data were handled using listwise deletion. All statistical analyses were conducted with SPSS.

This study received institutional review board approval from Iona College (reference number: 201718‐22) in 2018. Before participating, individuals agreed to participation by electronically signing a consent form which outlined study aims, procedures, time requirement, potential risks, and rights for human research subjects. Individuals from special or protected groups (ie, individuals who are <18 years old, imprisoned, or cognitively impaired) with impaired ability to provide free consent were ineligible to participate. We did not screen for pregnancy, and pregnant women were considered eligible to participate.

Table 3 presents the standardized parameter estimates for the confirmatory factor analysis model as well as the factor loading and error variance for each item. A variety of indices of model fit were examined: the chi-squared test of model fit was significant (χ²₃₅₆=941, P<.001; RMSEA=0.064; CFI=0.855; TLI=0.848; standardized root mean squared residual=0.060). The factors and their internal consistency reliability are as follows: self-branding (0.817), compulsive use (0.771), disruptive use (0.699), impulsive sharing (0.750), social engagement (0.712), induce negative emotions (0.751), and induce positive emotions (0.746). As expected, all CASM factors were positively correlated with one another but not so strongly as to suggest a single factor (Table 4). Skewness and kurtosis of all items and subscales on the CASM are reported in Multimedia Appendix 2.

The results found all CASM latent factors to be weakly correlated with the MoVEE scores and frequency of phone video game use, providing initial evidence of discriminant validity (Table 5). Scores on the BFAS and MEIS-CF, which both assess social media behavior, were correlated with the CASM, providing initial evidence of construct validity (Table 5). All CASM latent factors were correlated with B-FNE scores. Moreover, all CASM latent factors were found to be correlated with scores on the depression, anxiety, and stress subscales from the DASS, providing initial evidence of criterion validity (Table 5). The internal consistency reliability for measures used in validity testing is as follows: MoVEE (0.385), BFAS (0.802), MEIS-CF (0.924), B-FNE (0.791), and DASS (0.937).

Finally, a bivariate correlation revealed no statistically significant association between hours per day spent on social media and symptoms of depression, anxiety, and stress (r₃₉₇=0.06; P=.21; R²=0.003), where less than 1% of the variability in symptoms of depression, anxiety, and stress could be accounted for by hours per day spent on social media.

This study developed and validated the CASM. Exploratory and confirmatory factor analyses resulted in a 29-item self-report survey measure that assesses 7 distinct types of social media behavior: self-branding, compulsive use, disruptive use, impulsive sharing, social behavior, induce negative emotions, and induce positive emotions. Each factor of the CASM demonstrates good internal consistency reliability. Further, following cutoff criteria outlined by Hu and Bentler [52], the CASM demonstrates adequate model fit with moderate discriminant, convergent, and criterion validity.

To meaningfully advance research on social media use and mental health, researchers require better measurement tools to capture a range of online behaviors and engagement patterns. Encouragingly, the CASM measures a range of social media behaviors identified in prior research [53-55] that are clinically meaningful. To summarize, the self-branding subscale evaluates how often an individual engages in self-presentation strategies to manage how others perceive them and is similar to digital status–seeking behavior identified in prior research [54]. Individuals with higher scores on the branding behaviors subscale are more likely to have their own social media “brand,” which they adhere to. The compulsive use subscale measures behavior that suggests an overreliance on social media, which prior research has framed as addictive social media behavior [53]. Individuals with higher scores in this category may find it hard to disconnect from social media. The disruptive use subscale measures how frequently an individual uses social media maladaptively. This factor is distinct from the addictive behavior subscale, as its items assess social media use in inappropriate contexts (eg, during class or while driving), rather than use that is excessive. The impulsive sharing subscale examines how often an individual impulsively posts on social media, which aligns with oversharing behavior characterized in prior research [55]. Individuals with higher scores on the impulsive posting subscale are more likely to “over-share,” or post in an unedited, spontaneous manner that results in feelings of regret. The social engagement subscale examines how often an individual uses social media to communicate with others. Individuals with higher scores in this category are more likely to use social media as a tool to communicate and plan with others. The induce negative emotions subscale explores how often an individual uses social media to worsen their affect. One item in this factor evaluates upward social comparisons, a behavior examined extensively by prior research [56-58]. Individuals with higher scores in this category are more likely to use social media to lower their mood. Finally, the induce positive emotions subscale examines how often an individual uses social media to improve their affect. Individuals with higher scores in this category are more likely to use social media as a coping strategy to feel more relaxed or connected with others.

The CASM is novel in that it is one of the first validated measures to assess both positive and negative aspects of social media use simultaneously. Social media use measures have traditionally focused on overtly negative patterns of online engagement [27,28,32,41], such as addictive characteristics. The CASM extends this research by also measuring overtly positive aspects of social media use, such as social connection. This inclusion is especially relevant to advancing our understanding of social media and mental health; emerging research shows that social media may promote mental health through increased social support and peer connection [59]. However, existing evidence is still limited, emphasizing the need for improved measurement to assess potentially protective aspects of social media use.

Another advantage of the CASM is its emphasis on reporter agreement with items, rather than duration or frequency reports of social media engagement. When examining correlations between the CASM and indicators of mental well-being, self-perceived levels of depression, anxiety, and stress were significantly associated with the induce negative emotion and impulsive sharing subscales (Table 5). Notably, self-perceived levels of depression, anxiety, and stress as measured by the DASS were not associated with the self-reported hours per day spent on social media. This finding aligns with prior research showing that characterizing social media use as specific behaviors, rather than time spent online, is a better predictor of mental health [5,18,19].

Advancing our understanding of social media use is especially important for clinicians and parents tasked with promoting the well-being of young people. It is our hope that researchers will leverage the CASM’s ability to measure distinct aspects of social media behavior simultaneously to identify unique “engagement profiles” that more accurately reflect how people interact with social media. With a more holistic understanding of how an individual uses social media, it may be possible to gain deeper insight into what specific patterns of use may be adaptive, neutral, or detrimental to mental health. With this information, interventions and guidance aimed at promoting online hygiene could become more targeted and personalized. To accomplish this, future research should continue to advance the measurement of social media use, incorporating additional markers, such as passive sensing, to more accurately capture youth engagement with these platforms.

The findings from this initial validation study are important and yield ideas for further research. This study used a convenience sample of college students, limiting the generalizability of the findings beyond this population. This is especially important given demonstrable age differences in social media use [60]. Future research should examine the CASM’s application and validity across different age groups. Second, the CASM is a self-report measure and is therefore susceptible to response bias. This may have resulted in skewed responses, especially on items that conveyed negative social media habits. Future research should seek novel ways to assess social media use without the use of self-report measures. Third, while the 7 categories of social media use identified in this study cover a broad range of behaviors, it is likely that there are other clinically relevant dimensions of social media use. Future research should continue to examine and identify novel ways individuals interact and use social media, especially as these platforms continue to extend into news, gaming, and entertainment ecosystems and incorporate advances in artificial intelligence. Finally, this study predates the COVID-19 pandemic. Given documented increases in social media use during the pandemic [61], the behaviors identified in this study may not reflect social media engagement in more recent populations.

This study led to the development of the CASM, a 29-item scale that assesses 7 distinct social media behaviors: self-branding, compulsive use, disruptive use, impulsive sharing, social engagement, induce negative emotions, and induce positive emotions. The CASM demonstrates adequate model fit with moderate discriminant, convergent, and criterion validity. The CASM is one of the first validated measures to multiple, distinct aspects of social media behavior simultaneously, enabling researchers to more effectively examine associations between social media use and mental health outcomes. With this improved measurement, researchers can better assess how individuals use social media, and how this use may impact mental well-being.