1. Introduction
Eating disorders (ED) are a group of increasingly prevalent psychiatric pathologies that, having a relative bad prognosis, account for significant morbidity and mortality in Western countries [
1,
2,
3]. In this vein, the fact that around a third of women and half of men who meet the clinical criteria for ED do not seek professional help [
4] means that the social and economic burden of ED extends beyond healthcare costs [
5]. As a result, identifying modifiable risk factors for ED on which to focus treatment and prevention efforts has been a major research priority [
6,
7,
8].
A cognitive process that has been consistently identified as a modifiable risk factor for ED is the tendency towards body/appearance-related social comparison [
9,
10,
11,
12]. However, an emerging line of research has begun to demonstrate that further consideration of comparisons focused on actions aimed at improving body appearance, such as eating or exercise behavior, may provide deeper insight into the cognitive processes underlying ED [
13,
14,
15,
16]. This body of research has been built around the only self-reported quantitative measure proposed to date for the purpose of assessing the three facets of ED-related social comparisons outlined above (i.e., the Body, Eating, and Exercise Comparison Orientation Measure, BEECOM) [
17].
Findings from several research studies have been consistent in supporting the psychometric properties of the BEECOM scores in terms of both reliability and convergent/incremental validity [
14,
16,
17,
18]. In contrast, the available evidence on the factorial validity of the instrument is far less conclusive. For example, evidence has been provided that supports the original long 18-item structure consisting of 6 items each for the three lower-order factors (body-, eating-, and exercise-related comparison orientation) and a higher-order factor (ED-related comparison orientation) on a sample of non-clinical U.S. college females [
17]. However, this factor structure is not supported by the results of two subsequent studies. Firstly, the one conducted in a non-clinical sample of Iranian university male and female students with the aim of providing a Farsi translation of the original long version of the BEECOM (hereinafter BEECOM-L) revealed gender differences in the distribution of several items across the lower-order factors [
18]. Secondly, the one conducted on a sample of U.S. females in their young adulthood revealed that several items did not load clearly on their theoretically relevant lower-order factors in the clinical subsample in terms of ED [
16]. This finding led the authors to propose a short version of the instrument (BEECOM-S) consisting of 9 items (3 for each lower-order factor) instead of 18 items, whose adequate fit was subsequently corroborated in a second non-clinical subsample of otherwise similar socio-demographic characteristics to the first one [
16].
Taken together, the available evidence suggests that (i) the content of BEECOM might be subject to gender-specific interpretations, which would make this instrument unsuitable for making comparisons across groups based on this variable and (ii) that the revised short version of the instrument is likely to outperform the original long version in psychometric terms. However, the limited number of studies from which such evidence is drawn, coupled with the particular socio-demographic profiles of the participants included in these studies, precludes generalizing the above conclusions to other populations of interest in the context of ED. A clear example of such a population would be adolescent girls and boys under 17 (the minimum age of individuals on which the psychometric properties of the BEECOM have been tested so far [
17]), who represent an important target group for ED prevention efforts [
19]. The latter is clear in the light of the evidence suggesting that (i) 25–50% of ED onset occurs before the age of 17 years and (ii) that the differences in prevalence levels traditionally observed in favor of females are decreasing [
4].
Simultaneous examination of the psychometric properties of both versions of the BEECOM in a wider range of populations in terms of age and gender than hitherto considered may provide valuable information on the applicability of the instrument and, in particular, with regard to the comparative psychometric performance of the two versions. Thus, should evidence confirming the superior performance of the short version be obtained, then its use could be recommended. This would mean halving the application time of the BEECOM, which may decrease the response burden and increase the response rate of surveys including this instrument [
20]. In addition, evidence derived from examining whether the BEECOM scores are invariant across populations consisting of individuals of different gender and age would allow recommendations to be made concerning the appropriateness of using these scores in two main situations: firstly to make reasonably unbiased comparisons between these population groups [
21] and secondly in the context of conducting regression analysis involving moderation testing, which—as a result of considering group membership as the moderating variable instead of conducting subgroup analyses—would benefit from an increased level of statistical power [
22].
In view of the above, the main objective of the present study was to compare the psychometric properties of the scores derived from the BEECOM-L and the BEECOM-S in terms of factorial validity, reliability, measurement invariance according to age and gender, and convergent/incremental validity. Given the paucity of research on the topic and the inconclusive nature of the available evidence, no hypothesis was advanced concerning which version of the instrument would show better overall psychometric performance. According to findings from previous research, it was hypothesized that the BEECOM scores would show high levels of reliability in terms of internal consistency (
α) or construct reliability (ρ) (i.e., >0.80) [
14,
17,
18]. It was also expected that, providing evidence of convergent validity, the BEECOM scores would be positively correlated with two main groups of variables: firstly with comparisons strictly focused on physical appearance, with effect sizes expected to fall within a range of
r = 0.53 to
r = 0.76) [
17] and secondly with several ED outcomes, these including (i) overall ED symptoms, with effect sizes expected to fall within a range of
r = 0.37 to
r = 0.71) [
16,
18]; (ii) weight and shape concerns, with effect sizes expected to fall within a range of
r = 0.38 to
r = 0.77) [
13,
17,
23]; and (iii) dietary restraint, with effect sizes expected to fall within a range of
r = 0.20 to
r = 0.46) [
13,
18]. Finally, it was hypothesized that, consistent with the results of previous research providing evidence of incremental validity for the BEECOM scores [
16,
17], these would account for additional variance in all three ED outcomes under consideration over and above that accounted for by both body mass index (BMI) and the scores derived from a measure strictly focused on appearance-related comparisons. In the event that evidence was obtained to support measurement invariance of the BEECOM scores across age and gender groups, a secondary goal of the present study was to examine gender and age differences in ED-related comparison orientation. It was hypothesized that, in line with previous research examining appearance-focused comparisons [
10,
12,
24], small-to-intermediate-sized differences would be found in favor of groups consisting of females and younger adults.
4. Discussion
The present study elaborates on previous research on the psychometric of the BEECOM [
16,
17,
18] by examining this is issue in both the long and short versions of the instrument in four different samples in terms of gender (i.e., male and female) and age (i.e., adolescents and young adults). The results supported the use of the BEECOM-S over BEECOM-L in the four population subgroups under consideration. This recommendation is based on the fact that although we found evidence that supports both versions in terms of reliability, convergent/incremental validity, and measurement invariance of their scores across gender and age groups, this same evidence clearly emerged in terms of factorial validity only in the case of the BEECOM-S. These findings are largely consistent with those reported for clinical and non-clinical samples in terms of ED consisting of females in their young adulthood [
16] in suggesting the improved performance in terms of factorial validity of the BEECOM-S over the BEECOM-L. These findings make it possible to extend previous recommendations for the use of the BEECOM-S in clinical and non-clinical females in their young adulthood to other relevant populations in terms of ED prevention efforts, such as adolescents of both genders and young adult males not clinically diagnosed with ED [
4,
19]. The main findings of the present study are further elaborated below.
A first notable finding of the present study concerns the apparent mismatch of some of the items included in the instrument with their theoretical ascription factors. This was evident from the fact that the model mis-specification of the BEECOM-L was largely due to the ambiguous factor loadings shown by most of the items not included in the BEECOM-S. A possible explanation for these findings could be drawn by examining the content of this group of items. For example, the three items showing unexpected cross-loadings on the exercise comparison orientation factor include content that, despite alluding to comparison features from the original factors (i.e., the body or eating habits), could reflect a presumably healthy lifestyle focused on the attainment of a certain body ideal throughout dietary and exercise behavior [
44,
45]. This was the case with Item 17 (which alludes to having a toned body) and Items 8 and 16 (which allude to healthy eating and consuming junk food, respectively). The latter suggests that part of the content included in the items present in the BEECOM-L but excluded from the BEECOM-S may cover comparisons referred to more than one of the theoretical objects of comparison (e.g., food and exercise or body and exercise). In this vein, the similar correlation patterns observed between the scores derived from (i) both versions of the BEECOM and (ii) the variables included in the convergent validity analyses would suggest that the omission of items present only in the BEECOM-L does not undermine the comprehensiveness of the BEECOM-S scores. This means that despite some items being removed (these including the items potentially referred to more than one comparison feature), the key aspects of the constructs being assessed would not be missing, which would support the use of the reduced version of the instrument. Further studies should nevertheless examine whether complex comparisons incorporating elements of more than one of the objects of comparison under examination would be relevant within the context of studying the onset and maintenance of ED.
A second main implication from the present study derives from the evidence that supported the measurement invariance of the BEECOM-S scores across males/females and adolescents/young adults. This finding confirms with empirical data the applicability of the instrument to male individuals hypothesized by the authors who developed the original version of the BEECOM [
17]. This also implies that the instrument can be recommended for the purpose of conducting reasonably unbiased comparisons between the population groups under consideration in terms of self-reported levels of body, eating, and exercise comparison orientation. In this vein, the pattern of differences observed in body-related comparisons between the age and gender groups considered in the present study is consistent with prior research [
10,
12,
24]. We refer to the fact that body-related comparison scores favored females and individuals in their young adulthood over males and adolescents, respectively. In this sense, the results of the present study suggest that, while showing appreciably smaller differences, such a pattern could be equally present in comparisons concerning eating and exercise. Taken together, these findings suggest that females in their young adulthood may be a priority target for prevention efforts aimed at reducing body-focused social comparisons and, as a result, their potentially harmful outcomes [
15,
23,
46].
A third important implication from the present study is that the three forms of social comparison assessed by the BEECOM may differ in their contribution to explaining specific ED outcomes across gender- and age-based population subgroups. In this vein, the results broadly suggest that comparisons focused on body, eating, and exercise may be particularly detrimental in terms of their potential influence on different ED outcomes in female individuals and, among these, in those in their adolescence. Consequently, this latter population emerges as a clear target for prevention in terms of the three forms of comparison under consideration. In this case, this is not due to the higher comparison frequency of comparison observed in these groups (as was found when analyzing the differences in the BEECOM scores between groups) but because of the positive and particularly strong relationship between this variable and the ED outcomes found in these groups. These results suggest call for further research on the role of gender and age in the relationship between the different forms of social comparisons proposed as particularly relevant in the context of ED and the symptoms inherent to this group of pathologies. The results derived from such investigations could inform both the content (e.g., by identifying priority comparison features) and the target population (e.g., in terms of age groups and gender) of ED prevention efforts and intervention development.
Several limitations of the present study should be acknowledged. Firstly, the data came only from the Spanish version of the BEECOM, which prevents the results obtained here from being generalized to all language versions of the instrument (i.e., the original English version [
17] or the Farsi translation [
18]). This is a relevant limitation given the lack of cross-country and cross-cultural equivalence shown by a number of instruments assessing body image-related constructs [
47,
48]. In the same vein, the fact that the study population was limited to adolescents and young adults does not allow the results obtained to be generalized to other populations of interest in terms of ED prevalence (e.g., middle-aged and older women [
49]). Secondly, evidence suggesting the limited contribution of general comparison tendencies relative to physical appearance social comparisons in explaining variability in ED outcomes over and above the scores from the BEECOM subscales [
17] led us to omit the former from the incremental validity analyses. However, it cannot be ruled out that the inclusion of a score reflecting general comparison trends would have led to slightly different results. A third and final limitation is the lack of testing of a relevant psychometric property such as reliability in terms of temporal stability. As the evidence on this issue is so far limited to that reported for college women in the case of the original English language version of the BEECOM-L [
17], future research aimed at examining the reliability of the BEECOM-S in terms of temporal stability is warranted.