Stress can be regarded as one of the most influential construct in understanding health and wellbeing. A major shift in conceptualizing stress occurred when Lazarus [1] pointed out the importance of individual’s appraisal and its interaction with environmental events in understanding stress. Inclination towards measurement of appraisal of stress rather than stressful life event itself has since been targeted. Cohen et al. [2] developed Perceived Stress Scale (PSS) to measure the degree to which an individual appraises his/her life situations as stressful. This appraisal based scale has been demonstrated to outperform life event based measures in assessment of stress [3]. Three different versions of PSS with 14, 10 and 4 items have been developed, and validated.

With evidence of sufficient validity and reliability indicated by studies conducted over different cultures and populations all over the world, PSS has been established as a robust tool for measuring perceived stress. PSS 10 demonstrated a stable two-factor structure where items 1, 2, 3, 6, 9 & 10 constitute the first factor and the remaining four items constitute the second factor [4–9]. The first factor is often termed as ‘general stressors’ [10], ‘perceived helplessness’ [11] or ‘negative perception’ [12]. The second factor is termed as ‘ability to cope’ [10], ‘perceived self-efficacy’ [11] or as ‘positive perception’ [12]. PSS has been translated and validated in more than 30 languages and on different populations including, general people [13], pregnant women [7], elderly workers [14], cancer patients [15], and psychiatric patients [16].

Compared to general population, lesbian, gay, bisexual and transgender (LGBT) people are more prone to experience stressful life events including stigma, discrimination, financial hardship, abuse (physical, sexual & psychological), and legal difficulties, [17–19]. Section 377 of Bangladesh penal codes clearly prohibits carnal intercourse in the ground of being “against the order of nature”. This same principle may be used against all variations of sexuality. However, the stress faced by LGBT population stems more from socio-cultural and religious rather than legal aspects. Mental health impact of minority stress among LGBT has been established for a long time [20, 21]. Studying stress among these sexual minority populations is essential. To ensure accurate interpretation and usability of research findings, contextually validated tools are required. Perceived Stress Scale can be a suitable option to measure stress among LGBT population. Although developed primarily on general population sample, its robustness in measuring stress among different population subgroups has been widely demonstrated [7, 14–16, 34]. The generic items of the PSS seem to fit well with appraisal of any stressors including the circumstances of LGBT populations. Perceived stress scale (mostly the 4-item version, i.e., the one with poorest psychometric property) has been used in studies conducted to assess stress among LGBT population [22–27]. However, no study seemed to be concerned about performing factor analysis or validation of PSS on LGBT population. Risk of drawing faulty conclusion from poorly- or non-validated tool is too great. A tool developed in one culture may measure completely different construct for other cultures especially for a minority group (see [28]). The present study was aimed at validation along with exploration and comparison of factor structure of PSS 10 among LGBT population in Bangladesh.

A recent estimate suggests the size of MSM (men who have sex with men) population to be between 21,833 to 110,581 and the transgender population to be between 4504 to 8882 in Bangladesh [29]. The size of lesbian and bisexual population in Bangladesh is still unknown. Although commonly coined together as LGBT, the four sexual minority population have differences in terms of their psychosocial circumstances in Bangladesh. LGBT population, except for the transgender group, are largely hidden in Bangladesh due to lack of socio-cultural and legal acceptance. Transgender people commonly known as Hijras in Bangladesh usually live and roam around in well connected groups segregated from mainstream society with a distinct culture of their own (see [30]). They are easily identifiable with their distinctive appearance (feminine dress up with masculine facial features, voice tone, and body built) and activities (claiming and collecting money from the shops; singing, dancing and claiming money from parents when babies are born). Among the other members of LBGT, a portion of the gays have feminine behavior pattern which often leads to their identification and subsequent victimization. The remaining two groups i.e., lesbians and bisexuals, kept themselves unidentified and face lesser daily hassles associated with disclosed identity. While the often married bisexuals face difficulties in maintaining dual relation, the lesbians face stress when their parents apply force to marry them off to men. These differences necessitate the need to check whether or not PSS10 have comparable psychometric properties among the lesbian, gay, bisexual and transgender population.

Published data on stress and mental health of LGBT population in Bangladesh are limited. However, different studies equivocally reported presence of extreme stressors in their life [30–32]. Utility of a validated measure of stress for LGBT population in Bangladesh would be enormous.

The present study assessed factor structure, reliability and validity of PSS 10 among lesbian, gay, bisexual and transgender population in Bangladesh. The findings revealed a two-factor structure of the scale suitable for all four populations, which is consistent with findings from the original study [13] as well as other studies conducted in Asia, Africa, Europe, and Americas on PSS 10 factor structure [5, 6, 11, 15, 51]. In three instances, (item 8 for gay, item 9 for transgender, and item 10 for bisexual population) items demonstrated loading at 0.30s range on the alternative factor, however, this is not uncommon in the PSS 10 literature (see [7]). The dominant loading was taken into consideration in deciding factor structure in such cases. The amount of total variance explained by the two factors ranged from 43.55% - 51.44%, which is consistent with published studies on PSS 10 [7, 13]. In line with Örücü and Demir [11] the two factors in the Bengali PSS 10 can be termed as ‘perceived helplessness’ and ‘perceived self-efficacy’.

Multiple fit indices consulted during CFA indicated good fit of the two-factor model among the lesbian, bisexual and transgender population. Model fit for gay population was only supported by χ²/df criterion. Additionally, SRMR value (0.094) can be considered as indicative of acceptable fit of the model for gay population if a less stringent criterion (SRMR ≤0.10) is used (see [49]). Modification of the model could lead to better fit, however, no modification was performed to maintain comparability of the model among the four populations. One challenge in interpreting these model fits is the small sized samples used in analyses especially for the lesbian group. As RMSEA and SRMR values are inflated with smaller sized samples, it can be assumed that larger samples would indicate better fit. However, CFI and TLI indices used here are claimed to be less affected by sample size and hence can be considered reliable indicators of fit in the present study.

PSS 10 was internally consistent for lesbian, gay and bisexual population. Poor Cronbach’s alpha was found for transgender population. Additional analysis revealed negative correlation (− 0.233) between the two factors in transgender population. Lack of internal consistency in transgender population can therefore be interpreted as result of distinctiveness of the two factors. It should be noted here that all the fit indices suggested good fit of the two-factor structure of PSS for the transgender population (Table 3). Internal consistency for Factor 1 was sufficient for all groups while poor Cronbach’s alpha (0.536–0.671) was revealed for Factor 2. It should be noted that Cronbach’s alpha is heavily influenced by the number of items in the instrument. Lower alpha on Factor 2 can be caused by having only four items in the factor. Inter-correlation between the two factors for two groups (Lesbian and Gay) was very low and non-significant which is rarely found in PSS 10 literature [14] while for the two other (Bisexual and Transgender) it was at moderate level and significant (one correlated positively and the other negatively; see Table 4). These findings can be indicative of differences in response pattern and therefore in perception of stress for the four sub-groups of LGBT.

Due to sensitive nature of the topic, no names, addresses or other identification data were collected from the participants. Anonymity was also crucial for averting putative legal implications associated with identification of the participants [42, 43]. This anonymity prevented planning for assessing test-retest reliability of the PSS 10 for LGBT population. Testing for criterion validity could not be carried out due to lack of alternative gold standard for assessing perceived stress. Few studies claimed criterion validity of PSS 10, using mental component of SF-36 as the criterion measure [33]. However, as the two tools measure different constructs, it is likely that correlation of PSS 10 and SF 36 mental component is actually indicative of construct validity for PSS 10 instead of criterion validity.

The findings indicated construct validity of Bengali PSS 10 through convergent method. This study used SRQ 20 for validation instead of common practice of using GHQ, depression inventory, anxiety inventory or life event scales [34]. SRQ 20 seemed more suitable as it is known to directly measure overall psychological morbidity and has also been validated in Bangladeshi community sample [52]. High correlation was found between PSS 10 and SRQ 20 scores for full scale and Factor 1, while poor correlation was found for Factor 2.

Poor psychometric properties for Factor 2 have been reported in previous studies [4, 5, 14] however, these were not so drastic as in the case of present study. Psychometric properties for Factor 2 i.e., ‘perceived self-efficacy’ part of PSS 10 suggest against possibility of using the two factors of scale as two subscales for LGBT population in Bangladesh.

These findings reinforced the robustness of PSS 10 by providing evidence of comparable factor structure, validity and reliability among lesbian, gay, bisexual and transgender population. Thus it suggests usability of PSS 10 among sexual minority population that is LGBT. Evidence of validity and reliability of PSS in these populations are likely to increase confident use of the 10-item perceived stress scale as a measure of stress appraisal among LGBT population worldwide and in Bangladesh. However, factor level analysis indicated poorer psychometric properties for Factor 2, restricting usability of Factor 2 as a sub-scale in assessing perceived stress among LGBT. Moreover, small sample size for the LGBT populations (especially for the Lesbian) raised concern regarding robustness of the estimates, requiring readers to uses caution in interpreting the findings.