Introduction

Over the last few years, Fear of Missing Out (FoMO) has been the focus of an increasing amount of research (Elhai et al., 2021; Tandon et al., 2021). Keeping up with one's social circle and their activities is a key component of FoMO (Tandon et al., 2021). This higher need and desire to stay connected might lead to excessive social media use (e.g., Elhai et al., 2021), a problem that has been of interest to researchers in the field (Baker et al., 2016; Tandon et al., 2021).

A review by Tandon et al. (2021) found that FoMO research has focused on four major themes: compulsive media use, well-being, unsafe behavior, and intention to use (media). Studies have shown an association between FOMO and higher social media use, anxiety, depression, negative affectivity, and lower subjective quality of life (e.g., Elhai et al., 2021). It is still necessary to investigate FoMO using objective measures of internet, social media, and smartphone use, through longitudinal studies and from a neuroscience perspective (Elhai et al., 2021).

Nevertheless, several self-report questionnaires have been successfully used to measure FoMO (Elhai et al., 2021). As FoMO research advances, it is important to continue adapting and validating scales that measure this construct and study its psychometric properties. The current study aims to validate the most widely used FoMO's measure (e.g., Li et al., 2021), Przybylski and colleagues’ Fear of Missing Out scale (FoMOs) (Przybylski et al., 2013) to European Portuguese, and analyze its reliability (i.e., internal consistency) and validity [i.e., factor structure (including internal structure invariance) and other aspects of convergent and discriminant validity]. This study is part of a larger study focusing on the relationships between sleep quality, nighttime social media use, FoMO and maladaptive cognitive emotion regulation (Almeida et al., 2023). We will refer to this literature throughout the introduction to contextualize the nature of the present study and the variables used to analyze FoMOS-P validity. Still, for a more extensive literature review on this topic and research project, the reader is advised to consult Almeida (2020) and Almeida et al. (2023).

Fear of Missing Out (FoMO): definition

FoMO can be considered a type of social anxiety (Can & Satici, 2019), characterized as “(…) a pervasive apprehension that others might be having rewarding experiences from which one is absent” (Przybylski et al., 2013, p. 1841). Accordingly, this apprehension is a psychological state of anxiety concerning the possibility that others in one’s social sphere might lead more interesting and socially desirable lives. FoMO is typically associated with irritability, anxiety, loneliness, and feelings of social inadequacy (Abel & Buff, 2016; Przybylski et al., 2013). FoMO can also be conceptualized as a self-regulation strategy used to satisfy certain psychological needs, such as competence, autonomy, and relatedness (Beyens et al., 2016; Przybylski et al., 2013). This fear of being left out of rewarding experiences, excluded, or rejected by not being “up-to-date” with what happens in the social circle the individual belongs to, is also associated with a higher need and desire to constantly keep in touch with these social connections and their activities (Przybylski et al., 2013; Roberts & David, 2020).

Despite being relevant to study FoMO beyond the context of social media, social media plays an essential role in the understanding of this construct, as FoMO is intrinsically connected to the use of social networking sites (SNS) (Abel & Buff, 2016). For this reason, throughout this article, we will be referring to FoMO within the context of social media use and start by exploring its relationship with social media in detail.

Fear of Missing Out (FoMO)‘s relationship with variables of interest

FoMO and social media use

Social media is present in most people’s daily lives, with its use exponentially growing (e.g., Sette et al., 2020). As matter of fact, social media appears to be the main online activity of young adults and college students, mediating their social interactions and allowing them to establish and maintain social relationships (Beyens et al., 2016; Dempsey et al., 2019; Oberst et al., 2017; Sette et al., 2020). This intense use led to concerns over the potential impact of social media on mental health and stimulated research on the psychological processes related to (problematic) social media use, such as FoMO (Abel & Buff, 2016; Fioravanti et al., 2021; Gupta & Sharma, 2021; Przybylski et al., 2013).

In this respect, FoMO has also be conceptualized as “a preoccupation of SNS users with being deprived of interaction while offline” (Alutaybi et al., 2019, p. 3758). In fact, individuals high in FoMO might feel compelled to check their social media more often in order to stay constantly connected with the plans and activities of their social circles and not miss out on experiences others are having (Oberst et al., 2017). Social media, by allowing this continued access to what an individual might be missing out on (e.g., social events, career opportunities), can trigger FoMO and open the possibility for upward social comparisons to happen, which, in turn, can have a negative impact on users’ well-being (Abel & Buff, 2016; Li et al., 2021). Being constantly connected to social media, and accessing this kind of information through updates, can also promote compulsive checking behaviors and excessive engagement, which is typically associated with negative health consequences (Baker et al., 2016; Oberst et al., 2017).

Following this line of thought, we can conclude from the literature on the topic that there is a well-established connection between FOMO, social media, and mental health outcomes (e.g., Fioravanti et al., 2021; Gupta & Sharma, 2021). FoMO has been consistently associated with high social media engagement (Baker et al., 2016; Beyens et al., 2016; Casale & Fioravanti, 2020; Oberst et al., 2017; Przybylski et al., 2013; Roberts & David, 2020), and negative health outcomes such as depression, anxiety (Baker et al., 2016; Oberst et al., 2017) and poor sleep (e.g., Scott & Woods, 2018). Likewise, FoMO has been considered a risk factor for social media addiction (e.g., Oberst et al., 2017), and an explanatory mechanism of the differential impact of social media use on young people’s well-being (Beyens et al., 2016; Roberts & David, 2020).

Given that FoMO has negative affect and anxiety as key conceptual components (Beyens et al., 2016; Przybylski et al., 2013), it becomes important to explore its relationship with other cognitive processes involving negative affect, such as rumination, to fully understand its relationship with social media and the motivations behind (excessive) social media use.

FoMO and rumination

As previously mentioned, FoMO can be thought about as a form of social anxiety that involves negative affect, such as apprehension, unease, and fear (Dempsey et al., 2019; Vally et al., 2021). Similarly, to FoMO, rumination, which can be defined as a maladaptive repetitive thought pattern, is associated with negative affect (e.g., Vally et al., 2021). Furthermore, like FoMO, rumination is a cognitive aspect of social anxiety (Dempsey et al., 2019) and an important mechanism for managing (online) social interactions (Elhai et al., 2021; Vally et al., 2021).

Still, there is a lack of literature focusing specifically on the relationship between social media and maladaptive cognitive processes other than FoMO, and on the relationship between FoMO and these processes (e.g., worry, rumination) (Almeida et al., 2023; Dempsey et al., 2019).

We will specifically examine the relationship between FoMO and rumination. Most studies investigating FoMO and rumination mainly explored their relationship with anxiety and depression. These studies focused particularly on the mediator effects of FoMO and rumination on the relationships between these psychopathological symptoms and (problematic) smartphone, internet, and/or social media usage (e.g., Dempsey et al., 2019; Elhai et al., 2021; Vally et al., 2021). Consequently, there is a limited amount of literature focusing specifically on the relationship between FoMO and rumination. Several studies involving these variables have found that they are associated with one another. However, these results were typically secondary to the main hypotheses being tested. Accordingly, Dempsey et al. (2019) found a significant association between these two variables, in a study that analyzed the mediator effects of these variables on social anxiety and problematic Facebook use. Vally et al (2021) concluded that FoMO and rumination were significantly associated and mediated the relationship between anxiety, depression, and problematic smartphone use. Bayın et al. (2021) confirmed this association and concluded that ruminative thought style mediated the relationship between FoMO and the fear of COVID-19. Lastly, Fitzgerald et al. (2022) conducted two studies regarding the relationship between FoMO, loneliness, anxiety sensitivity, and rumination. On the first study, rumination only predicted FoMO when anxiety sensitivity was not considered in the model. On the second study, both rumination and anxiety sensitivity predicted FoMO. Even though the results were mixed, the authors concluded that rumination appears to be an important variable for understanding FoMO that should be considered in future studies.

Despite the lack of research on the relationship between FoMO and rumination, this cognitive process is usually regarded as a "negative cognition(s) associated with FoMO" (Dempsey et al., 2019, p.2), possibly because of FoMO's initial theoretical conceptualization and relation with anxiety and negative affect (Przybylski et al., 2013).

Therefore, based on the current state of the literature, some researchers recommend further investigation into the relationship between this transdiagnostic process, FoMO, and social media use. (e.g., Casale & Fioravanti, 2020; Dempsey et al., 2019). It is unclear if focusing on what one is missing out increases rumination and worry or whether these cognitive processes lead to a higher focus on the social experiences and activities one is not taking part in (Almeida et al., 2023). Further research is necessary to explore the directionality of these associations and understand whether a causal link may be established.

It is clear that FoMO and rumination can both be considered “transdiagnostic psychopathology constructs” (Dempsey et al., 2019, p.6), associated with negative affect (e.g., Bayın et al., 2021), indicating similar constructs. Considering the above, we investigated the convergent validity of the Fear of Missing Out Scale (FoMOs-P) by analyzing FoMO’s relation with rumination and social media use.

FoMO and sleep quality

As mentioned previously, previous research linked FoMO to several negative health outcomes, including poor sleep. In this regard, FoMO has been associated with increased sleep arousal, shorter sleep duration, longer sleep onset and decreased sleep quality (Adams et al., 2020; Almeida et al., 2023; Gupta & Sharma, 2021; Scott & Woods, 2018). Recently, Adams et al., (2020, p.1) also concluded that "(…) insomnia partially mediated significant associations of interpersonal stress and FoMO with mental health".

Understanding how FoMO itself and its association with higher social media use can contribute to poor sleep quality in higher education students is relevant, given the prevalence of sleep deprivation in college students and the known importance of adequate sleep for mental/physical health and cognitive, emotional, and social functioning (e.g., Adams et al., 2020).

This topic is beyond the scope of the present study, but has been investigated in previous research (e.g., Almeida et al., 2023). It must be noted, however, that in social media research, sleep variables, including sleep quality, appear to be important, but are considered distinct constructs from FOMO. Therefore, FoMOs-P's divergent validity was tested by examining the association between FoMO and sleep quality.

Measuring FoMO: Fear of Missing Out scale (FoMOs) validations, factorial structure, and psychometric properties

So far, we have examined FoMO's relationship with social media and (mental) health outcomes in an effort to justify the importance of validating FoMO's main measure, namely in higher education students. Two additional reasons can be provided to support this issue's importance. The first reason is that FoMO is considered an important and prevalent phenomenon among adolescents and young adults, with studies indicating that between 70 and 75% of young adults experience FoMO at least to some degree (Przybylski et al., 2013; Rozgonjuk et al., 2021). In Portugal, higher education students are in their vast majority young adults. A second reason supporting the importance of validating FoMOs is that this measure might be of use in practical contexts as a mean of screening people with disruptive levels of social media engagement (Casale & Fioravanti, 2020).

For all that has been mentioned, it appears to be important to validate psychological instruments that allow FoMO to be measured and to develop research that improves our understanding of this construct and its prevalence in the Portuguese higher education student population.

Regarding FoMO’s assessment, different instruments for its evaluation have been used and developed (e.g., Abel & Buff, 2016; Sette et al., 2020). The instrument most widely used to measure an individual’s general FoMO is the Fear of Missing Out scale (FoMOs), developed by Przybylski et al. (2013) (Casale & Fioravanti, 2020; Li et al., 2021). FoMOs is a brief measure, containing only 10 items with a 5-point likert response scale (Can & Satici, 2019). In the original study of this scale, an initial 32-item version of FoMOs was tested in a sample of 1013 fluent English speakers using Principal Component Analysis (PCA), Confirmatory Factor Analysis (CFA) and Item Response Theory (IRT) analysis. Based on these preliminary findings, the final 10-item one-factor version was obtained. This version exhibited good internal consistency (α = 0.87).

Following the original English version of FOMOs developed by Przybylski et al. (2013), the scale has then been translated and validated for the Arabic (Al-Menayes, 2016), Spanish (Gil et al., 2015), Turkish (Can & Satici, 2019; Gökler et al., 2016), Chinese (Li et al., 2021), Korean (Joo et al., 2018), Italian (Casale & Fioravanti, 2020), German (Wegmann et al., 2017), Peruvian (Correa-Rojas et al., 2020), Indonesian (Kaloeti et al., 2021)and Colombian (Núñez et al., 2022) populations.

The single-factor structure found in FoMOs’ development study was supported in the Spanish version of FoMOs (Gil et al., 2015). Additionally, the Turkish version of FoMOs, validated in two different studies, as well as the Peruvian and Colombian versions, also supported the original 10-item one-factor structure (Can & Satici, 2019; Correa-Rojas et al., 2020; Núñez et al., 2022; Gökler et al., 2016). Nevertheless, the factor structure of the scale has yielded conflicting results. The Italian FoMOs version (Casale & Fioravanti, 2020) found a 10-item two-factor structure. Factor analysis of the Arabic version of the scale (Al-Menayes, 2016) also indicated a two-factor structure, with eight items instead of the original ten. An 8-item two-factor structure was also identified in the validation study for the Chinese version (Li et al., 2021). The German validation study conceptualized FoMO as a two-dimensional construct (i.e., attribute-FoMO and state-FoMO) and found that their modified 12-item FoMOs had a two-factor structure (Wegmann et al., 2017). Using both EFA and CFA procedures, the Indonesian and Korean FoMO versions revealed a three-factor structure with 12 and 8 items, respectively (Joo et al., 2018; Kaloeti et al., 2021). Lastly, the different versions of scale across countries have revealed good psychometric proprieties, namely reliability and validity (including measurement invariance) (e.g.,Casale & Fioravanti, 2020).More specifically, in terms of internal consistency, in studies that obtained a unifactorial solution, including the original version of the scale, Cronbach alfa ranged from α = 0.78 to α = 0.90 (Can & Satici, 2019; Gil et al., 2015; Przybylski et al., 2013). In the validation of the Arabic, Italian and Chinese versions of the scale, which obtained a two-factor solution, Cronbach alpha values ranged between α = 0.72 and α = 0.82 for both factors (Al-Menayes, 2016; Casale & Fioravanti, 2020; Li et al., 2021). Table 1 summarizes the main results of FoMOs validation studies.

Table 1 Main results of FoMOs validation studies (organized by factor solution results)
Full size table

The present study

Considering the absence of validated FoMO measures available to Portuguese researchers and that FoMOs is the principal measure of FoMO in Psychology (e.g., Casale & Fioravanti, 2020), the main aim of the present research was to validate the European Portuguese version of the Fear of Missing Out scale (FoMOs-P), by ensuring the scale´s cultural adaptation to Portugal and assessing its internal consistency, internal structure, validity and invariance across sexes. This initial validation study focused on higher education students, since previous research showed that they use social media more, when compared to other social groups (e.g., Zhou, 2019) and that the prevalence of FoMO in this age group is higher (e.g., Rozgonjuk et al., 2021). At the same time, and despite the phenomenon being typically associated with younger generations, the literature on FoMO and age differences is not consensual, with several studies founding no differences in FoMO levels across age groups (Akbari et al., 2021). Social media use might relate with FoMO in individuals of all ages as well (Przybylski et al., 2013; Roberts & David, 2020). Therefore, a broad array of higher education students, in terms of variables such as age and occupational status, was included in this study. Invariance with respect to age was not tested because most students from out sample were 29 years old or younger.

Lastly, other aim of this study was to explore an “optimal” cutoff score that differentiates students with high social media engagement, since FoMOs has previously been deemed a potentially useful screening tool (Casale & Fioravanti, 2020).

Material and methods

This research, which is part of a larger study regarding the relationships between sleep quality, nighttime social media use, FoMO and maladaptive cognitve emotion regulation (Almeida et al., 2023) was approved by the ethics committee of the Faculty of Psychology and Educational Sciences at the University of Coimbra (ref: CEDI 30/01/2020).

Data collection and sampling procedures

Data collection took place from February to April of 2020. The participants were recruited through convenience sampling in psychology classes from a Portuguese public university or through social media channels (i.e., Facebook student groups). Participants (n = 500) were eligible if they were Portuguese, had at least 18 years old and attended a Portuguese higher education institution. Students were always shown an informed consent, which provided information about the nature of the study and ethical considerations, such as confidentiality and anonymity, before voluntarily agreeing to participate.

FoMOs-P was administered along with measures of demographic and health data, social media use, rumination, and sleep quality. Participants replied to this cross-sectional survey, online (n = 390) or, in pencil-paper format (n = 110), before the COVID-19 pandemic. No compensation or any other type of benefit or incentive was offered for their participation.

Participants

In total, a sample of 500 Portuguese higher education students (84.6% female and 15.4% male), residing in different regions of Portugal, enrolled in the study. The average age of the participants was 22.37 years (SD = 5.68, range: 18–64). The majority of the participants were full time students (82.8%), single (94.2%) and did not have any children (97%). In terms of academic background, participants were primarily first year (29.3%) and third year (34.9%) students, from different courses and institutions. Social media was used daily by 96.8% of students. Among these students, 7.6% used social media for less than an hour, 19.4% for 3 to 4 h, and 9.0% for more than 6 h. Also, 14.2% of the participants reported having a physical and/or mental health issue, with anxiety (3%), depression (4.2%) and asthma (2.2%) being the ones most commonly reported. Over half of the data (56%) was collected before the first Covid-19 home confinement.

Measures

The research protocol used in this study comprised a demographic and health information section, followed by FoMOS-P and a set of self-report measures on social media use, rumination, and sleep quality, that were used to examine FOMOs-P validity.

Demographics and general health issues

Demographic information was collected on age, gender, marital status, nationality, location, employment status and profession (if applicable). Self-reported physical and/or mental illnesses were identified by asking if issues were present and, if so, which ones.

Fear of Missing Out (FoMO)

FoMO was assessed using the translated European Portuguese version of FoMOs (Przybylski et al., 2013, Appendix Table 5), which is a brief 10-item self-report instrument that measures the extent of people’s fear missing out on social events, information and gratifying experiences that might be happening in their social circles. Each item is rated using a 5-point Likert scale that ranges from 1 (not at all true of me) to 5 (extremely true of me). The total score is obtained from the sum of the score obtained in all 10 items and ranges between 10 and 50 points. Higher scores correspond to higher levels of FoMO (Przybylski et al., 2013).

Social media use

Social media use was assessed using a 7-item scale developed by Woods and Scott (2016) (Portuguese version: Almeida, 2020), which assesses how often respondents use social media, how much time they spend on it, when they engage with social media, which devices are used to assess social media and how many different social media platforms they use. An example of an item of this scale is “How often do you use social media?”. Six of the seven questions are answered using a 6-point Likert scale (each scored from 0 to 5 points), with only one “yes–no” question (scored 1 or 0). Total scale score is determined by summing the scores of the seven items and ranges from 0 to 31. Higher scores correspond to higher levels of social media engagement. Mcdonald’s omega for this scale was ω = 0.58. This measure was employed in order to analyze FoMOs-P’ convergent validity (i.e., association of FoMOs-P scores with other variables) and to explore the scale cut-off through ROC analysis).

Rumination

Rumination was assessed using the Ruminative Responses Scale – short form (RRS-S) (Portuguese version: Dinis et al., 2011), a 10-item instrument that quantifies the tendency to engage in ruminative thinking, when experiencing depressed mood. The response scale is a 4-point Likert scale that varies between 1 (almost never) and 4 (almost always). This measure is composed of two subscales: “brooding” and “reflection”. It is possible to obtain both the total scale score and subscale scores. This study only considered the total scale score, which is derived from adding up the individual item scores and ranges between 10 and 40. Higher scores correspond to higher levels of ruminative thinking tendency. McDonald’s omega for RRS-S total score in this study was ω = 0.88. This measure was employed to analyze FoMOs-P’ convergent validity (i.e., association of FoMOs-P scores with other variables).

Sleep quality

Sleep quality was measured using the Basic Scale on Insomnia Symptoms and Quality of sleep—BaSIQS (Gomes et al., 2015). This self-report 7-item scale encompasses items on sleep latency and continuity, deepness, and quality, of sleep, and the total score is largely correlated with the Pittsburgh Sleep Quality Index overall score (Gomes et al., 2015). BaSIQS uses a 5-point Likert scale, with item scoring ranging from 0 to 4 points (except for two reversed items), and the total score varying between 0 and 28 points. Higher scores translate more insomnia complaints/poorer sleep quality. McDonald’s omega in the current study was ω = 0.65. This measure was used to analyze FoMOs-P’ discriminant validity.

Translation and cultural adaptation procedures

At the beginning of this study, the original authors of FoMOs (Przybylski et al., 2013) gave consent for the scale to be validated for European Portuguese speakers on November 28, 2019. FoMOs was initially translated by the first and last authors of the current study. All items were considered relevant to the Portuguese context. This initially translated and semantically adapted version of FoMOs, from English to Portuguese, was then reviewed by a group of four experts, who were Portuguese native speakers fluent in English with broad research experience and whose academic background was psychology. These experts were not affiliated with this study. After expert revision, some minor alterations (e.g., phrasing, sentence structure) were considered. Reviewers also mentioned the comprehension of the meaning of question 5. could be problematic, since there is no direct translation of the expression “in jokes” to Portuguese.

We then conducted a pilot study using two versions, with variations in the wording and phrasing of items 1, 2, 3, 4, 5 and 6. These two versions were filled individually and voluntarilyFootnote 1 by 12 individuals of different ages (from school children to older adults) and educational backgrounds (from elementary school to university education) to ensure the compressibility of the scale’s items. These participants were individually interviewed by using a combination of the cognitive interviewing techniques of thinking aloud and verbal probing and asked which version of the scale was easier to understand. This feedback was used to refine each item wording and define the final version of FoMOs, which was used in the current study.

Statistical analyses

Statistical analyses were made with IBM SPSS (version 27.0). When the percentage of missing items in each scale was under 20%, the missing response for that examinee was replaced by the average of their available responses at the scale (i.e., person mean imputation. In this study, it was preferred to use this method over others, namely multiple imputation, since it is an adequate method for item-level missingness. To ensure statistical power, this method was also preferred over listwise deletion. Significance level was set at p < 0.05 for all analyses. Descriptive statistics (i.e., frequencies, mean, standard deviation) were determined to characterize participants’ demographics, social media usage and FoMO. Univariate normality was analyzed for all variables using the criteria of skewness < │3│and kurtsis < │7│ (Kline, 2016). The distribution of age was substantially different from normal, so a logarithmic transformation was used (log [10]). Correlation analysis was used to explore the association between FoMO and age (log10). Pearson correlation coefficients were interpreted as: small if 0.10 ≤ r < 0.30, moderate if 0.30 ≤ r < 0.50 = and large if r ≥ 0.50 (Cohen, 1988).

Previous studies have reported conflicting results regarding the factor structure of FoMOs. Given the amount of evidence underlying these models and to ensure comparability between different studies, Confirmatory factor analysis (CFA) was conducted to assess FoMOs-P' factor structure and test/compare the goodness-of-fit of the two main competing models: Model 1 was the one-factor model initially proposed by Przybylski et al. (2013) and validated by Can and Satici (2019), Gil et al. (2015), Núñez et al. (2022) and Correa-Rojas et al. (2020). Model 2 was the 10-item two-factor model proposed by Casale and Fioravanti (2020). This model is only partially similar to those proposed by Al-Menayes (2016) and Li et al. (2021), in which items were removed (cf., Introduction, Sect. 1.2, Table 1). A number of reasons supported the decision to only select Model 1 (one-factor model) and Model 2 (correlated two-factor model with 10 items) for CFA testing. The first reason is that researchers selected models that contained only the original FoMOs items (i.e., no additional items), since validation was conducted on this version of the scale (cf., Table 1). Second, since all 10 items were deemed relevant to the Portuguese cultural context, CFA was not used to test two-factor models with 8 items. For the same reason (i.e., excluding items), the Korean model, which is the only three-factor model from prior studies that solely contains the original items of the scale, was not tested. There are also other reasons for not performing CFA to test this model, including the limited empirical support for a three-factor model and cultural differences between collectivistic (i.e., Korean) and individualistic cultures (i.e., Portuguese). CFA was performed using JASP (Version 0.18.1.0) [Computer software] (JASP team, 2022). This analysis relied on a specific JASP module, the Structural Equation Modeling (SEM) module, derived from R's package lavaan. Parameters were estimated using the Robust Diagonally Weighted Least Squares (RDWLS) estimation method, which accounts for polychoric correlation matrices of Likert-type scales, and it is robust to deviations from normality. CFA model adjustment was assessed through the following global model fit indices and reference values: Chi-Square test (χ2); Chi-Square Critical Ratio less than or equal to 3 (χ2/df ≤ 3); Comparative Fit Index (CFI) and Tucker Lewis Index (TLI) greater than or equal to 0.95 (CFI ≥ 0.95; TLI ≥ 0.95); Root Mean Square Error of Approximation (RMSEA) less than or equal to 0.07 (RMSEA ≤ 0.07); RMSEA's 90% confidence interval lower limit less than 0.05, and its upper limit less than 0.08; and Standardized Root Mean Square Residual (SRMR) less than 0.05 (SRMR ≤ 0.05) (Brown., 2015). The two competitive models (Model 1 and 2) were compared using the chi-square difference test (Δχ2) (i.e., if significant, Model 2 is a better fit than Model 1), the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), with preference being given to the model with the lowest AIC and BIC value (Kline, 2016). Also, multi-group CFA was conducted to determine whether the best fitting model was invariant across participants’ sex [males (n = 77) versus females (n = 423)], considering three levels: configural, metric and scalar. The model was considered invariant compared to the less-restrictive model based on the chi-square difference test (Δχ2), the CFI and TLI difference (ΔCFI ≥ 0.010; ΔTLI ≥ 0.010) and RMSEA difference (ΔRMSEA ≤ 0.015) between the models (Chen, 2007).

For the total scale and factors, internal consistency reliability was examined using McDonald's omega (ω), composite reliability (CR), and, for comparability purposes, Cronbach alpha (α). Alpha, Omega, and CR values greater than or equal to 0.70 were considered adequate (Brown, 2015). Corrected item-total correlations and McDonald’s omega (ω) when item deleted were used for item analysis. The generally accepted threshold for corrected item total correlations is 0.3.

Convergent/discriminant validity were tested by examining the pattern of Pearson’s correlations between social media use, rumination, sleep quality and FoMOs scores. We expected non-significant or low associations between FOMOs scores and sleep quality scale scores, (since FOMO may be related to poorer sleep quality), suggesting poorly correlated or dissimilar constructs (i.e., discriminant validity). Similarly, if Model 2 provides a better fit to the data, a factor intercorrelation under 0.85 would also imply good discriminant validity, indicating that the latent factors represent different constructs (Brown, 2015). It was also expected that there would be a moderate correlation between FoMO and the selected measures regarding convergent validity, since they are related (but not overlapping) constructs—hence, convergent validity was assessed using the RRS-S and the social media scale scores. Convergent validity was also examined using the Average Variance Extracted (AVE). The accepted threshold for this measure was AVE ≥ 0.50 (Brown, 2015).

Receiver Operating Characteristic (ROC) analysis was used to analyze the scale’s accuracy and explore the ““optimal”” cutoff point for screening students with high levels of social media engagement. Since the ROC curve evaluates the performance of a binary classification method (e.g., Unal, 2017), the targeted state variable had to be dichotomized. We used a similar dichotomization approach to that described in MacCallum et al. (2002). The test variable was the FoMOs-P total scale score, and the state variable was a binary variable that defined two groups of social media use intensity (n = 182). More specifically, based on the sample's mean social media usage (M = 17.94; SD = 3.67), all cases scoring below the M-1SD threshold were considered low social media users, while those scoring above the M + 1SD threshold were considered high social media users. These points were chosen to define two groups at the extremes since, based on the normal distribution curve, most observations are within one standard deviation of mean. To achieve the best balance between sensitivity and specificity, we selected the ““optimal”” cutoff point based on the following criteria: “(…) the value whose sensitivity and specificity are the closest to the value of the area under the ROC curve and the absolute value of the difference between the sensitivity and specificity values is minimum” (Unal, 2017, p. 1). An Area Under the Curve (AUC) of 0.7 was the minimum acceptable threshold. An AUC value greater than 0.9 was classified as high accuracy and between 0.7 and 0.8 as moderate accuracy (Unal, 2017).

Results

Descriptive statistics, item analyses and overall reliability

The main descriptive statistics related to FoMOs-P are presented in Table 2. Item means ranged from 1.46 (item 4) to 2.85 (item 9). Skewness ranged from 0.75 (item 4) to 1.30 (item 5), while kurtosis ranged from—0.07 (item 1) to 2.89 (item 8). Neither skewness nor kurtosis exceeded the acceptable limits for univariate normality. FoMOs-P mean score was 22.06 (SD = 7.52; min = 10; max = 47). FoMO was significantly associated with age (log10) (r =  − 0.29; p < 0.001, a small to moderate association). All corrected item-total correlations were above 0.3. In terms of overall reliability, Cronbach’s alpha was α = 0.87 and McDonald’s Omega was ω = 0.87 for the total scale. The value of Mcdonald's Omega would only be marginally increased by eliminating item 3, so the item was retained.

Table 2 FoMOs-P items analyses (n = 500)
Full size table

Factor structure

Considering previous conflicting results regarding FoMOs' factor structure, it was decided to test the two most empirically supported factor structures using CFA: the one-factor (Model 1) and the two-correlated factor model (Model 2).

Regarding the overall fit of Model 1, results show a poor fit across all indices considered (χ2 (35) = 218.36, p < 0.001; χ2/df = 6.23; CFI = 0.94; TLI = 0.92; RMSEA [90 CI] = 0.102 [0.090;0.116]; SRMR = 0.107). For Model 2, the following results were found: χ2 (34) = 100.59, p < 0.001; χ2/df = 2.96; CFI = 0.98; TLI = 0.97; RMSEA [90 CI] = 0.063 [0.049;0.077]; SRMR = 0.07). Models 1 and 2 were then compared. Model 2 demonstrated a better fit in comparison to Model 1, according to the chi-square difference test (Δχ2 (1) = 495.01, p < 0.001) and AIC and BIC’s values (AICModel-2 = 13004.58 < AICModel-1 = 13497.59) and (BICModel-2 = 13135.23 < BICModel-1 = 13624.03). These findings indicate that Model 2 demonstrated an adequate fit to the data. The two factors were termed “Internalizing FoMO” (Int-FoMO) and “Externalizing FoMO” (Ext-FoMO). In Fig. 1, the factor loadings and correlation between factors of Model 2 are illustrated in a Path Diagram, which is in accordance with the Reticular Action Model (RAM) symbolism (Kline, 2016). In terms of reliability, internal consistency measured by Cronbach’s alpha was α = 0.85 for the Int-FoMO factor and α = 0.82 for the Ext-FoMO factor. McDonald’s Omega was ω = 0.87 for Int-FoMO and ω = 0.83 for the Ext-FoMO. CR was 0.89 for the Int-FoMO factor and 0.90 for the Ext-FoMO factor. Lastly, the two factors are strongly associated (r = 0.65, p < 0.001).

Fig. 1
figure 1

Path diagram for the CFA of FoMOs-P (Model 2- standardized solution)

Full size image

Model invariance across sex

Measurement invariance is necessary to future examination of FoMOs-P scores across groups. We conducted multi-group CFA to determine whether the best fitting model, the two-related factors model, was invariant across sex (female versus male). Previous studies have found invariance of the internal structure of FoMOs regarding sex (e.g., Casale & Fioravanti, 2020). Goodness-of-fit indexes supported configural invariance (CFI = 0.99, TLI = 0.98, RMSEA = 0.05 (90% CI [0.03, 0.07]) of the best fit model across sex (cf., Table 3). Then, metric invariance or weak invariance was also tested and showed a good fit to the data (CFI = 0.98, TLI = 0.98, RMSEA = 0.05 (90% CI [0.03, 0.07]) When comparing the configural and metric models, the chi-square difference test (Δχ2) was non-significant, thus supporting metric invariance. Metric invariance was also supported by a minimal change on fit indices (ΔCFI = -0.002; ΔTLI = -0.001; ΔRMSEA = 0.001) in comparison with the configural invariance model. Lastly, scalar invariance or strong invariance was not supported by the chi-square difference test, despite model fit being adequate (CFI = 0.98, TLI = 0.98, RMSEA = 0.05 (90% CI [0.03, 0.06]) and the change on fit indices being minimal (ΔCFI = 0; ΔTFI = 0.002; ΔRMSEA = -0.003) in comparison to the metric invariance model. Table 3 summarizes the results of this analysis.

Table 3 Measurement Invariance of FoMOs across sex: male and female
Full size table

Discriminant and convergent validity

In terms of convergent validity, significant and positive correlations were found between social media use, FoMOs-P total score (r = 0.314; p < 0.001), Int-FoMO (r = 0.290; p < 0.001) and Ext-FoMO (r = 0.274; p < 0.001) factors scores. Rumination tendency and FoMOs-P overall score (r = 0.485; p < 0.001), Int-FoMO (r = 0.407; p < 0.001) and Ext-FoMO (r = 0.454; p < 0.001) scores were significantly and positively correlated. AVE was 0.68 for Int-FoMO and 0.60 for Ext-FoMO. With respect to discriminant validity, FoMOs-P total score and scores on the Int-FoMO and Ext-FoMO factors did not significantly correlate with sleep quality scores.

Accuracy

The obtained AUC and its 95% confidence interval were AUC [95 CI] = 0.742 [0.67;0.82]; (p < 0.001). Considering the previously presented criteria, FoMOs-P ““optimal”” cutoff point was set at 20 (i.e., sensitivity was 70.1%; specificity was 68.4%). Figure 2 depicts the ROC curve associated with this analysis. Table 4 summarizes the specificity and sensitivity values for all possible FoMOs-P cutoff scores.

Fig. 2
figure 2

ROC curve for FoMOs-P (total score) against “high/low social media usage problems”

Full size image
Table 4 FoMOs-P sensitivity and specificity as determined by ROC analysis (N = 182)
Full size table

Discussion

Considering previously reported widespread use of social media and FOMO by higher education students, this preliminary study developed a Portuguese version of FOMOs (FoMOs-P) and examined its factorial structure, reliability, and validity on this population. The current study has several strengths such as providing additional cross-cultural psychometric data regarding FoMOs, using CFA to further establish the scale’s factorial structure (including measurement invariance across sex) and determining a cutoff point for screening students with potentially high social media usage, thus providing FoMOs-P screening utility.

In general, this study supports the psychometric properties of the European Portuguese version of FoMOs. Particularly, this research offered initial evidence regarding FoMOs-P reliability and validity. In terms of reliability, FoMOs-P showed adequate internal consistency – overall and by factor. Other empirically supported two-factor versions reported similar Cronbach's alpha values (Al-Menayes, 2016; Casale & Fioravanti, 2020; Li et al., 2021). Omega's reliability and CR were also satisfactory.

The Portuguese version of FoMOs also demonstrated acceptable discriminant and convergent validity. Discriminant validity was supported by non-significant correlations between FoMO and sleep quality scores and by a correlation bellow 0.85 between the two factors (Brown, 2015). FoMOs-P also showed acceptable convergent validity as small to moderate correlations were found between social media use, rumination, and the factors and total scores. Convergent validity was also demonstrated by the values obtained for the Average Variance Extracted (AVE), in line with prior validations.

Regarding FoMOs-P factorial structure, CFA revealed that two factors were underlying students’ responses to FoMOs-P, supporting previous research suggesting a scoring model based on two factors (Casale & Fioravanti, 2020). Besides the chi-square statistic, which is known to be highly susceptible to a large sample size, several other model fit indices were also utilized in assessing model fit. Further evidence of the goodness-of-fit of the two-factor model (Model 2) is that its results were consistent across multiple fit indices (Kline, 2016). The comparison of competing models also added to model selection, by contributing to overcome the disadvantages associated with the use of cutoff points in CFA (e.g., Tomarken & Waller, 2005).

It is worth noting that the final model showed only configural and metric invariance across sex. In this regard, the model might not be fully invariant across sexes. Assessing correlation differences between sexes appears feasible, however greater caution should be used when comparing mean scores between men and women (Casale & Fioravanti, 2020). The high percentage of female participants in our sample could explain our results, as the distribution of males and females is uneven. Therefore, model invariance across sexes needs to be further tested in future studies using a more balanced sample. Also, more research to determine model invariance across age is needed since previous research on FoMO’s age and sex differences is conflicting (e.g., Rozgonjuk et al., 2021).

The results of this study differed from those of the original scale development study and those of validation studies conducted in other countries, such as Spain and Turkey, where only one factor was retained based on the same number of items, and Korea and Indonesian studies that indicated a three-factor structure (Can & Satici, 2019; Gil et al., 2015). However, the validation studies of the Arabic, Italian, and Chinese versions of FoMOs also concluded for a two-factor model (Al-Menayes, 2016; Casale & Fioravanti, 2020; Li et al., 2021). For the most part, the factorial structures found in these three studies were not markedly different from the one proposed in the Portuguese validation of the scale, despite only 8 items being retained in both the Arabic and Chinese versions. Specifically, in the Chinese and Arabic versions, the fifth item and forth items were eliminated due to low and high factor loadings on both factors, respectively (Al-Menayes, 2016; Li et al., 2021). Comparatively, the Italian version has the same factorial structure as the Portuguese version, and the two-factor solution obtained in this study closely corresponds to the Italian version. Both factors include the same items and, as such, seem to measure the same dimensions of the FoMO’s construct. The intercultural similarities between Portugal and Italy could probably explain the similarity obtained in the factorial structure of the scale.

Additionally, the differences that do exist between the other two-factor versions of the scale and FoMOs-P might exist due to the use of different statistical methods [(e.g., some of these validation studies resorted to Exploratory Factor Analysis (EFA)], item modification/addition] and to cultural differences (e.g., the experience of FoMO might vary across countries and cultures). The presence of discrepancies in terms of cultural background (e.g., Asian countries are collectivistic cultures and European countries are individualistic cultures) might also help explain the observed differences in the factorial structure in various versions of the scale (Casale & Fioravanti, 2020; Li et al., 2021).

In line with Wegmann et al., (2017, p. 35), who considers that “(…) FoMO is not a unitary phenomenon but rather a more complex construct (…)”, our findings suggest that FoMO is a theoretical concept composed by two dimensions: factor 1 items seems to represent a cognitive-emotional/internalizing dimension (Int-FoMO) and factor 2 a behavioral/externalizing dimension (Ext-FoMO), based on the Portuguese version’s item content. Int- FoMO seems to be related to the internal experience of negative emotional states (by definition, apprehension, fear, stress, and anxiety) regarding the possibility of missing out on important information and rewarding events of their social networks (Dempsey et al., 2019). The first factor of the Italian version also represents the fear that others are having more rewarding experiences. Similarly, in the Portuguese version of the scale, all items included in this factor also start with “I fear…” and “I get worried…”, which is very much in line with the original concept of FoMO (Casale & Fioravanti, 2020).

In turn, Ext-FoMO seems to be related with a set of actions and attitudes whose purpose is to reduce the aversive experience of this kind of social anxiety, which might be negatively reinforcing in itself and promote certain behaviors such as the use of social media, as shown in previous research (Sette et al., 2020). The results also highlight that these gestures can be seen as self-regulation strategies that allow the satisfaction of psychological needs of autonomy, competence, connection with others and/or relatedness, reveling the presence of a higher need and desire to constantly be in touch with the members of one’s social group, conclusion supported by FoMO’s initial theoretical conceptualization (Przybylski et al., 2013). The second factor of the Italian version also represents strategies used to obtain knowledge about what is happening in an individual’s social circle (Casale & Fioravanti, 2020). This theoretical conceptualization seems to be in line with previous literature, which identified behavioral (externalizing) and/or cognitive-emotional (internalizing) components to FoMO (e.g., Casale & Fioravanti, 2020). For instance, Elhai et al., (2021, p.203) state that: “The first component [of FoMO] maps onto the cognitive aspect of anxiety (e.g., worry, rumination, etc.). The latter component [of FoMO] involves a behavioral strategy aimed at relieving such anxiety analogous to how compulsions aim (though maladaptively) to relieve anxiety in obsessive compulsive disorder”. Also, Scott and Woods (2018) concluded that FoMO predicted shorter sleep duration on adolescents through a behavioral and cognitive mechanisms. Our findings also seem to be aligned with the authors of the Italian version argument that this last dimension might correspond to a more behavioral component of FoMO (Casale & Fioravanti, 2020).

In the Chinese version, the two factors obtained were named Fear of Missing Out on social opportunities (FoMO-SO) and Fear of Missing Out on novel information (FoMO-NI). This theoretical interpretation does not fit the Portuguese version since, for instance, item 6 (“Sometimes, I wonder if I spend too much time keeping up with what is going on”) and 7 (“It bothers me when I miss opportunities to meet up with friends”) were grouped together. Also, item 5 showed good psychometric properties and this finding does not seem to confirm the hypothesis that understanding "in-jokes" is more aligned with adults than with young people. However, it does appear to be in line with the idea that this item might be more applicable in an individualistic occidental society. Item 4 was also suitable for this scale, contrary these authors’ conclusion (Li et al., 2021).

A further contribution of this study is that it explored a cutoff point for students with high engagement in social media. While previous literature has indicated a relationship between FoMO and social media (e.g., Fioravanti et al., 2021; Tandon et al., 2021) and FoMOs has been recognized as a possible screening tool for social media engagement (e.g., Casale & Fioravanti, 2020), only one study has previously determined a cutoff point for FoMOs. As indicated by the AUC value of 0.742 obtained in our study, FoMOs-P has a 74,2% chance of correctly identifying students with high social media engagement. The selected cutoff point minimizes the difference between sensitivity and specificity, making FoMOs-P a moderately sensitive and specific instrument with potential clinical/practical utility for higher education students in various settings.

Altogether, this study advances our understanding of the theoretical underpinnings of FoMO and concluded that the Portuguese version of FoMOs had good psychometric properties. This study, however, has several limitations. First, the use of person mean imputation as a method for handling missing data might have biased the present results.

Second, scale validation is a continuous process. Therefore, additional empirical validation of the Portuguese version of FoMOs is needed to examine the FoMOs-P psychometric proprieties and the stability of its factor structure. Considering that the sample used was a convenience sample composed mainly of young adults—all higher education students—with a high percentage of females, the findings are non-generalizable to the general Portuguese population, despite the inclusion of students of different ages, occupational status, and academic backgrounds. To support the current findings, more representative samples should be used in future studies. Additionally, the collection of data during the Covid-19 first home confinement might have impacted the results [(for a more detailed description of the first issue, cf. Almeida et al., 2023].

Another limitation of this study is that good fit does not necessarily imply that the model is correct or true, but just one that is plausible, since other models may also fit the data at a similar level of goodness-of-fit (e.g., Tomarken & Waller, 2005). For instance, the covariance between residuals was not specified and applying this procedure may have altered the results. Further, for the reasons mentioned above, the researchers decided not to test the 3-factor model obtained in the Korean validation study. As previously mentioned, one of these reasons was that important cultural differences exist between collectivistic (i.e., Korea) and individualistic (i.e., Portugal) cultural contexts. However, it would still be relevant to do so. Additionally, in future studies, CFA should also be used to further test the 2-factor model across other samples (Joo et al., 2018; Kaloeti et al., 2021).

Lastly, ROC analysis should be viewed as exploratory. The authors used another self-response scale as a criterion instead of a "gold standard" method to assess disruptive use of social media. Second, measures from the same group were used instead of measures obtained from two different groups (i.e., groups of similar individuals with and without the characteristic for which the cutoff point is being determined).

Despite the above limitations, the present findings contributed to the field by validating the main psychological instrument for measuring FoMO, expanding research in another non-English speaking country, demonstrating its potential practical application, and encouraging further research into FoMO's psychometric properties, especially regarding its factor structure. The Portuguese version could also be of use to researchers that pretend to develop cross-cultural comparations studies between different versions of this scale.

Conclusions

The present research supports the validity and reliability of the European Portuguese version of FoMOs in higher education. FoMOs-P appears to be a useful psychological instrument for assessing FoMO among Portuguese higher education students, encouraging its use in this population and future psychometric testing (e.g., test–retest reliability). Subsequent studies are needed to support application of FoMOs-P in the general population. The present findings shed light on the factor structure of FoMOs-P, and its relationship with other psychological constructs as well as practical application as a potential indicator of high social media engagement or even problematic social media use (Casale & Fioravanti, 2020).