Development and validation of the Self-Efficacy in Addressing Menstrual Needs Scale (SAMNS-26) in Bangladeshi schools: A measure of girls' menstrual care confidence.

Schoolgirls in low- and middle-income country (LMIC) settings cope with a variety of challenges in meeting their needs regarding menstruation [ 1 , 2 ]. While typically provided little education about menstruation, adolescent girls contend with menstrual stigma, inadequate sanitation and disposal facilities, limited access to reliable menstrual materials, and poor social support for menstrual issues [ 1 – 3 ]. Over the past decade, there has been increasing advocacy to implement school-based programs to ensure girls can adequately and comfortably address their needs related to menstruation in the school environment—both because it is critical for girls’ human rights to dignity and reproductive health, and also because it is essential for achieving gender equality in education [ 4 – 8 ]. School-based menstrual health intervention studies have conventionally aimed to improve menstrual knowledge and management practices with a view to reducing school absenteeism, but they have often stopped short of rigorously assessing other important components of girls’ menstrual experiences [ 9 , 10 ]. The relationship between menstrual practices and school attendance is an indirect one—likely mediated by menstrual factors (e.g., dysmenorrhea, menorrhagia, etc.) and psychosocial factors such as agency and confidence, shame, distress, perceptions of one’s environment, and perceptions of one’s own menstrual practices [ 1 , 10 ]. Measurement of these important psychosocial factors in menstrual health research and program evaluations has been weak, in part due to a dearth of appropriate tools specific to this domain [ 11 ]. Validated measures for the assessment of psychosocial components of adolescent girls’ menstrual experiences are critical for robust evaluation studies to establish strong evidence for school-based interventions that effectively improve menstrual experiences. Prior qualitative studies in LMICs have described women and girls’ varying levels of confidence in managing their menstruation, with greater confidence hypothesized to positively impact health, education, and social participation outcomes [ 1 ]. Menstrual care confidence is an important psychosocial factor to assess in the menstrual health and hygiene domain because beliefs in one’s capabilities are known to influence how individuals feel, think, motivate themselves, and act [ 12 ]. Guided by Bandura’s self-efficacy theory [ 12 ], we conceptualize menstrual care confidence beliefs as “self-efficacy” and define the construct as a girl’s beliefs in her capabilities to carry out the tasks required to address her menstrual needs . An individual’s self-efficacy beliefs are constructed through processing information drawn from their previous enactive experiences, vicarious experience through observing others similar to themselves, verbal persuasion (e.g., encouragement), and their own physiological and affective state [ 12 ]. In menstrual health intervention studies, assessing self-efficacy beliefs may provide a more comprehensive picture of girls’ lived experiences than objective assessments of the presence of sanitation facilities in schools or availability of menstrual materials [ 1 ]. Self-efficacy theory suggests that even among those who share the same physical and economic environments, girls with lower self-efficacy in addressing their menstrual needs will perceive opportunities and constraints differently than those who have greater self-efficacy. Those with low self-efficacy in addressing their menstrual needs may be likely to experience greater stress and anxiety while attending to their menstruation in the face of challenges. Such individuals are theoretically more likely to avoid challenging situations altogether [ 12 ]—such as attending school during menstruation. Bandura’s self-efficacy theory conceptualizes the self-efficacy construct as comprising three dimensions: strength, level/magnitude, and generality—which has implications for its measurement [ 12 ]. The strength dimension is assessed by having participants indicate on a Likert-type response scale how sure they are that they can perform a particular task [ 13 ]. Along the second dimension, the level/magnitude of self-efficacy beliefs refers to the level of difficulty of task demands at which individuals feel they are capable of succeeding [ 12 ]. Self-efficacy beliefs also differ along a dimension of generality—or the “degree to which the expectation is generalized across situations” [ 12 , 14 ]. When measuring self-efficacy beliefs, it is thus important to measure the strength of an individual’s perceived capability across varying degrees of challenge or impediments to successful performance in a variety of situations [ 12 , 13 ]. Doing so necessitates multi-item measures. The purpose of the present study was to develop and refine a pool of potential items for a Self-Efficacy in Addressing Menstrual Needs Scale and formally test the items to reduce the pool and preliminarily assess the measure’s psychometric properties with schoolgirls in Bangladesh.

The 404 girls who participated in the feasibility testing of the tool’s intended format, instructions, and response options ranged in age from 10 to 18. Our observation of the feasibility testing and debriefing with the survey enumerator team indicated that enumerators found it easy to explain the instructions for the self-efficacy items and the item format made sense to the schoolgirls. Enumerators found that the “pen test” enabled them to quickly identify and clarify girls’ misunderstandings about the response options. A total of 51 girls ages 11–16 years participated in the nine FGDs (socio-demographic information available in S1 Table ). Five FGDs (3 urban, 2 rural) involved girls who were two years or less post-menarche while four FGDs (2 urban, 2 rural) involved girls who had been menstruating for more than two years. A summary of example menstrual care tasks discussed by our FGD participants, along with the conditions under which those tasks might be made more difficult or less difficult to enact is available in S2 Table . This chart guided the development of our initial pool of 50 draft scale items. We had not included “being prepared for menstruation” as an a priori category of menstrual care tasks in the participatory activities, yet its salience to schoolgirls was identified in FGDs. We therefore also drafted items within this category (e.g., items regarding a girl’s confidence in her capability to track her menstrual cycle or estimate approximately when her next period will begin—which involve body literacy). Four out of the five invited experts participated in the content validation exercise. The overall average relevance rating for items was 3.52 out of 4. Experts did not suggest the inclusion of additional categories of tasks, therefore indicating our initial item pool appeared to have good content validity. No items were dropped solely based on expert feedback. Cognitive interview participants were ages 11–16 years (socio-demographic information available in S1 Table ). Thirteen girls (9 urban, 4 rural) were 6–24 months post-menarche, and 8 girls (6 urban, 2 rural) had reached menarche more than two years prior. Most item revisions performed after each round of cognitive interviews concerned word choice and sentence structure. However, most items in the category of “keeping bodies clean during menstruation” were dropped due to lack of variation in responses during cognitive interviews. All girls interviewed felt fully confident they could wash their genital area and body as needed during their menstrual periods; therefore, such items would not contribute meaningfully to a scale that is intended to differentiate among girls with varying self-efficacy beliefs in this context. We also dropped items referring to the washing and drying of menstrual materials, because girls who only used disposable materials were unsure how to respond. We retained items referring to disposal of menstrual materials since even those who use cloth eventually dispose of it. Revisions based on cognitive interviewing reduced the pool of draft items from 50 to 35. The items were intended to be relevant for girls regardless of the type of menstrual materials they use, but cognitive interviews revealed the difficulty in selecting a generic term to encompass any type of material used for absorbing or collecting menstrual blood. Therefore, each time the term “menstrual material” appeared in an item, we added a list of the most commonly used materials in this context so girls could more easily understand what was being referenced (e.g., pad, cloth, cotton, tissue, etc.). Table 1 displays two selected examples of item revisions based on cognitive interview findings. During cognitive interviews we noticed that girls were better able to focus on, process, and respond to the items when they could read them themselves. We therefore shifted from developing an interviewer-administered self-efficacy scale to a self-administered one, which would reduce administration time. The average time to complete the 35 items during final field piloting was 15.5 minutes with a range of 10 to 22 minutes. Girls requested little clarification about item meanings during questionnaire administration, and retrospective probing indicated that most girls had understood the items as intended. After piloting, we further clarified wording in items about pain management and removed an item that asked about confidence to reduce abdominal pain “by a medium amount” because of its similarity to other items and the need to reduce administration time. Of 382 girls who indicated on the main study’s endline survey they had reached menarche, 381 completed the self-efficacy sub-study survey. No items were missing values among the 381 completed surveys. S3 Table displays the characteristics of participants disaggregated by urban and rural. The mean age was 14 years (SD = 1.5), and mean age at menarche was approximately 12 years (SD = 1.2). Regarding type of menstrual materials used, 61.7% of girls reported using disposable pads while in their homes, and 58.3% reported using pieces of cloth and/or reusable pads to absorb menses (participants could select multiple options). While away from home, 70.3% reported using disposable pads and 41.2% reported using pieces of cloth and/or reusable pads. The majority (66.4%) of girls reported experiencing menstrual pain during their last period, with 41.1% of those respondents describing their pain as “severe.” For those girls experiencing menstrual pain, most (55.3%) reported a one-day duration of pain. We present item response mean, standard deviation, skew, and kurtosis for all 34 self-efficacy items in S4 Table . We display the items we dropped during the psychometric testing and our rationales in S5 Table . We dropped four self-efficacy items because they showed very little variation in responses, and thus would not contribute meaningfully to a scale [ 35 ]. We dropped two additional items because they did not correlate with any other items at a level of at least 0.30 [ 37 ]. The Bartlett’s test of sphericity [ 31 ] for the retained 28 items was significant ( p < 0.001) and the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was 0.88, indicating our data were appropriate for factor analysis [ 32 , 37 , 38 ]. We employed iterated principal factors as our method of factor extraction due to the data not being multivariate normal. Performing a PCA on the retained 28 items showed seven components with eigenvalues greater than one. The scree plot showed an elbow at three factors, and the parallel analysis test suggested a 3-factor solution. In the subsequent 3-factor EFA, the item that assessed a girl’s confidence in her capability to remove a bloodstain from her clothes while at school failed to load at least 0.30 on any factor, so we dropped this item and re-ran the EFA. Upon assessing internal consistency of the three factors, we determined that dropping an item that asked about a girl’s confidence in her capabilities to use hot fomentation to reduce menstrual pain would increase the second factor’s internal consistency. The 26-item, 3-factor final model accounted for 43.5% of the total variance. Table 2 displays the rotated factor loadings (pattern matrix) and uniqueness for this final model, while Table 3 displays the structure matrix consisting of correlations between items and the rotated factors. Table 4 displays the correlations among the three factors. Notes: Factor loadings over 0.30 in bold; MHPM Menstrual hygiene preparation and maintenance, MPM Menstrual pain management, EST Executing stigmatized tasks * aya refers to the women who worked in the study schools as janitors and caretakers. Schoolgirls typically had friendly relationships with these women, and if the school provided any menstrual pads for emergencies, it was typically the aya who could give one to a student. * aya refers to the women who worked in the study schools as janitors and caretakers. Schoolgirls typically had friendly relationships with these women, and if the school provided any menstrual pads for emergencies, it was typically the aya who could give one to a student. Note: MHPM Menstrual hygiene preparation and maintenance, MPM Menstrual pain management, EST Executing stigmatized tasks All 17 items loading on factor 1, which we labeled Menstrual Hygiene Preparation and Maintenance (MHPM) , measured girls’ confidence in their capabilities to accomplish tasks related to obtaining, using, and changing menstrual materials in a variety of contexts; seeking assistance for menstrual hygiene when needed; and anticipating days of bleeding (factor loadings: 0.31–0.60). All five items loading on factor 2, labeled Menstrual Pain Management (MPM) , measured confidence in mitigating menstrual pain (factor loadings: 0.59–0.87). The four items loading on factor 3, labeled Executing Stigmatized Tasks (EST) measured girls’ confidence in their capabilities to accomplish menstrual care tasks that are heavily affected by menstrual stigma—particularly involving the risk of disclosing menstrual status in the presence of males (factor loadings: 0.41–0.95). We refer to the finalized tool ( S3 and S4 Files) as the Self-Efficacy in Addressing Menstrual Needs Scale (SAMNS-26). The mean MHPM sub-scale score for all 381 participants was 74.4 (SD = 16.6) with a range of 23.5 to 100. The mean MPM sub-scale score was 64.7 (SD = 26.5) with a range of 0 to 100. The mean EST sub-scale score was 45.5 (SD = 28.6) with a range of 0 to 100. Table 5 displays mean scores disaggregated by geography and use of disposable pads while away from home during the most recent menstrual period. Girls in urban schools tended to have slightly higher scores than girls in rural schools across all three sub-scales, although the differences were not significant. Girls who ever used disposable pads while away from their home during their most recent menstrual period tended to have slightly higher scores on all three sub-scales than girls who only used other types of menstrual materials, although the differences were small and not statistically significant. Note: Numbers are mean (SD) (No differences between groups were significant at a level of p <0.05 based on Wilcoxon rank-sum tests) Each of the three SAMNS-26 sub-scales showed good internal consistency in our sample. The MHPM sub-scale had a Cronbach’s α of 0.86 and an average interitem correlation of 0.26 (range from 0.26 to 0.27), the MPM sub-scale had an α of 0.87 and an average interitem correlation of 0.58 (range from 0.55 to 0.63), and the EST sub-scale had an α of 0.77 and an average interitem correlation of 0.46 (range from 0.37 to 0.52). The SAMNS-26 sub-scales showed good temporal stability after a mean interval of 15 days ( Table 6 ). The concordance correlation coefficient for the MHPH sub-scale at time 1 and time 2 for the 42 girls who retested was 0.80 (95% CI = 0.69–0.91), for the MPM sub-scale 0.70 (95% CI = 0.56–0.85), and the EST sub-scale 0.71 (95% CI = 0.56–0.86). Bland-Altman plots indicated that differences between individuals’ scores at time 1 and 2 tended to be smaller as average scale score increased, thus the measure had poorer temporal stability for those at the very lowest scale scores. SAMNS-26 sub-scores performed generally as expected in relation to validation measures ( Table 7 ). The Beck Anxiety Inventory for Youth (BAI-Y) was negatively correlated with the MHPM sub-scale at -0.33, MPM at -0.17, and EST at -0.19. The Beck Self-Concept Inventory for Youth (BSCI-Y) was positively correlated with the MHPM sub-scale at 0.32, MPM at 0.19, and EST at 0.18. Our Bengali version of the Social Self-Efficacy Scale showed questionable internal consistency (α = 0.66) in our sample and was positively correlated with the SAMNS-26 sub-scales. The MHM sub-scale—which included the most items that required enlisting assistance from others—correlated the most strongly at 0.40. The EST sub-scale to a lesser degree contained items that required interaction with others, and it correlated with the Social Self-Efficacy Scale at 0.26. The MPM sub-scale correlated at 0.21. All correlations with validations measures were statistically significant with p < 0.01. * p < 0.01 Table 8 displays SAMNS-26 sub-scale scores according to girls’ responses to three validation items from the main study’s endline survey. Wilcoxon rank-sum tests indicated that those who endorsed the item “During my last period, I felt anxious at school because of my menstruation,” had lower scores across SAMNS-26 sub-scales than those who disagreed with the statement. Conversely, those who endorsed the item “During my last period, I felt comfortable at school,” had higher scores across SAMNS-26 sub-scales than those who disagreed with the statement. Those who endorsed the item “I prefer to stay at home during my period,” had lower scores across SAMNS-26 sub-scales than those who disagreed. All differences were significant at level p < 0.01 except for the differences in MPM sub-scale scores for each validation item, which were not statistically significant. SAMNS-26 sub-scale scores were not significantly correlated with months since menarche (MHPM: ρ = 0.01, p = 0.80; MPM: ρ = -0.02, p = 0.71; EST: ρ = -0.07, p = 0.18). *Difference significant at level p < 0.01

We developed a 26-item scale comprising three sub-scales that measure schoolgirls’ self-efficacy in addressing their menstrual needs. In our testing sample of 381 girls attending rural and urban schools in Bangladesh, the scale demonstrated favorable reliability and construct validity. The SAMNS-26 can be used for further intervention research and menstrual health program evaluations with schoolgirls in Bangladesh from the time of menarche up to Class 10 (age 17), providing a way to assess changes in self-efficacy beliefs over time. The tool enables incorporation of self-efficacy into multivariate models for exploring the relationships among antecedents to menstrual experiences and hypothesized impacts on health, wellbeing, and education attainment. Additional testing of the tool in new samples in Bangladesh and other contexts globally is recommended to strengthen evidence of its validity. We recommend researchers and program evaluators take a similar iterative approach to pretesting scale items when adapting and revalidating the tool for new contexts.

Data collection took place in eight schools in Bangladesh—four in urban Dhaka, and four in rural Manikganj District—that were participating in the larger study Piloting menstrual hygiene management interventions among urban and rural schools in Bangladesh (henceforth referred to as the “main study”). The main study comprised a formative research phase in four of the schools (two urban, two rural) to inform the development of an intervention package, followed by a six-month piloting period in four other schools (two urban, two rural) to evaluate the intervention. The intervention aimed to promote supportive school environments for menstruating girls. Intervention components included provision of schoolteacher-led puberty and menstruation education, improved waste disposal facilities, and distribution of menstrual materials and menstrual cycle tracking calendars (among other activities). Details about the study context, school selection and characteristics, and the main study’s intervention activities and evaluation methods have been presented separately [ 15 , 16 ]. The work described in this paper was a self-efficacy sub-study commenced halfway through the main study ( S1 Fig ), after the formative research phase but immediately prior to implementation of a baseline survey in intervention schools. We first leveraged the main study’s baseline survey conducted in the four intervention schools to assess feasibility of our tool’s instructions and response options. We then conducted qualitative research to inform the content of the scale items in the four schools that had earlier participated in the main study’s formative research phase. Lastly, we leveraged the main study’s endline survey in the intervention schools to collect quantitative data for the self-efficacy scale development. The Ethical Review Committee of icddr,b approved the study protocol (PR15115). The Dhaka Zonal Office, Directorate of Secondary and Higher Education; the Dhaka Divisional Office, Directorate of Primary Education; and School Management Committees provided permission to conduct research in the schools. School leadership further provided approval to conduct study activities on school property, primarily during school snack breaks to reduce disruption. All participants assented to participate and had written consent from a parent or schoolteacher as their guardian ( in loco parentis ). Fig 1 summarizes our four-stage process model for developing the Self-Efficacy in Addressing Menstrual Needs Scale between April 2017 and April 2018. In Phase 1, we designed the questionnaire format and created an initial pool of draft items. In Phase 2, we assessed the content validity of the item pool through an expert validation exercise. In Phase 3, we iteratively pretested and refined the item pool and conducted final field piloting of the tool. In Phase 4, we assessed the psychometric properties of the tool through testing items on a survey of randomly selected schoolgirls in urban and rural Bangladesh. Guided by existing measures of self-efficacy in other domains [ 13 , 17 ], we developed a generic format for items that directs a respondent to indicate the strength of her confidence in her capability to perform a particular task involved in addressing her menstrual needs. We chose an 11-point Likert-type response option to allow enough variation in responses as recommended by Bandura for self-efficacy scales [ 13 ]. We trained 14 female professional health survey enumerators to pilot 10 test items on the main study’s baseline survey in the four intervention schools to qualitatively assess acceptability and feasibility of our proposed self-efficacy questionnaire format, instructions, and response options. During training, survey enumerators recommended using response options 0 to 100 (in intervals of 10) because of schoolgirls’ familiarity with the 0–100 scale used in academic grading. Among several options for visual cues to aid in understanding response options, survey enumerators recommended an “X” over the “0” response option, and a check/tick that increased in size over subsequent response options. Word labels anchored either end of the scale where 0 = “No, I absolutely cannot do it” and 100 = “Yes, I am absolutely sure I can definitely do it” ( Fig 2 ). We devised a “pen test” to check participants’ comprehension of the instructions before collecting data [ 13 ]. A survey enumerator would place a pen on the table near to a participant and ask, “ How confident are you that you can reach the pen (while remaining in your seat) ?” with the expectation that the response should be at or near 100. The enumerator would proceed to move the pen progressively farther away and ask the same question. If responses did not move in a reasonable manner along the response options, then the enumerator would provide further explanation of the instructions to clarify misunderstandings. In July 2017, we observed the main study’s baseline survey implementation with 527 randomly selected girls in Classes 5–10 (404 of whom had reached menarche and so participated in the feasibility testing of the self-efficacy questionnaire’s instructions and response options) and debriefed with the survey enumerator team about their experiences conducting the feasibility testing. Having confirmed the acceptability and feasibility of the intended format for our scale, we compiled a list of actions involved in menstrual self-care to write a comprehensive pool of potential self-efficacy items. To this end, we first reviewed global literature on girls’ menstrual experiences to identify categories of tasks related to addressing menstrual needs: obtaining menstrual materials; using, changing, disposing, and/or washing menstrual materials; keeping bodies clean during menstruation; reducing menstrual pain or discomfort; seeking support, help or advice related to menstruation; and coping with stress or anxiety related to menstruation. Then, between August 2017 and February 2018, we conducted nine focus group discussions (FGDs) with schoolgirls from Classes 4–10 to explore how they understood the meaning of menstrual management and to identify various tasks they viewed as part of meeting menstrual needs across the identified categories. We used participatory activities based on vignettes ( S2 Fig ) to facilitate discussion due to the research topic being socially proscribed and the potential for participants to feel shy discussing menstruation in a group. To recruit and select participants, we worked with female schoolteachers to explain the purpose of the research in classrooms of female students and compile a list of post-menarcheal girls by class level and approximate time since menarche. Schoolgirls voluntarily self-identified their eligibility in the absence of males. We then instructed eligible participants on how to request written consent from parents to participate in the study and requested students return signed consent forms to us the following day. We later purposively selected participants from this master list of eligible girls for each data collection episode according to class level and length of time since menarche. Focus group discussions were convened with privacy in empty classrooms. They were conducted in Bengali ( Bangla ) and lasted 84 minutes on average. The study team took detailed field notes and debriefed immediately after each FGD to expand field notes, chart findings, write analytic memos, and to make sampling decisions for subsequent FGDs. We charted findings from each FGD by completing a table with columns for 1) menstrual care tasks identified across various categories, 2) situations that would make those tasks more difficult, and 3) situations that would make the tasks less difficult. FGDs were audio recorded so we could check any gaps in field notes, but full transcriptions were not produced for the analyses presented in this paper. Data collection continued until we stopped identifying new tasks and felt confident we could write a sufficiently large pool of potential scale items [ 18 , 19 ]. Using these data and our review of the literature, we wrote an initial pool of 50 items, including items at varying levels of difficulty to avoid a ceiling effect in the final scale. We invited five global experts on menstrual health and hygiene to provide feedback in March 2018 via online survey to assess content validity of the initial pool of 50 draft items. Experts provided qualitative feedback regarding our conceptualization of the construct and then rated the relevance of each item from 1–4 where 1 = not relevant , 2 = unable to assess relevance without item revision , 3 = relevant but needs minor alteration , and 4 = very relevant and succinct . They also provided comments on the clarity of each item and made suggestions for revision. We revised items with average ratings of less than 4 according to experts’ suggestions. In February and March 2018, we conducted cognitive interviews to pretest and refine the pool of draft items. Cognitive interviewing enables survey developers to examine whether participants can easily understand questions and whether they interpret items the way they were intended—which aids in identifying concrete suggestions for improving items and response options [ 20 ]. We conducted 21 cognitive interviews in iterative rounds (2–6 interviews per round) with girls in Classes 5–10 in the same urban and rural schools where we had conducted FGDs. This iterative approach enabled us to identify and address issues with items and test revised versions until we achieved a good match between each item’s intent and participants’ interpretations [ 21 ]. We purposively selected girls from the master list of eligible students who had not previously participated in FGDs and according to time since menarche to ensure a range of familiarity with menstrual care. Cognitive interviews were conducted privately in empty classrooms. Study team members used a field guide that contained all draft items being tested, each with suggested probing questions and blank space for field notes. We encouraged participants to express their thoughts aloud as they interpreted the items, searched their memory to formulate an answer, and chose how to select a response from the response scale [ 22 , 23 ]. Study team members also probed concurrently to encourage participants to verbalize their thought processes and to explore any confusion that arose or uncertainty over word meanings or instructions [ 24 ]. We applied both proactive (anticipated and spontaneous) and reactive (conditional and emergent) probes [ 24 , 25 ]. Interviews were conducted in Bengali, typically lasted 1–1.5 hours, and were audio recorded for verification purposes but not transcribed. Led by the first author, the study team convened immediately after interviews to debrief, combine field notes from all interviews, and revise items to improve clarity. The revised list of items served as the field guide for the following round. Items were dropped before final field piloting if they were not comprehensible during cognitive interviewing despite multiple efforts at revision or if response frequency distribution charts showed very little variation in quantitative responses. We conducted a final field pilot in two rounds in April 2018 with a total of 13 girls selected by convenience from Classes 8–10 in an urban school to determine how long it took them to complete the revised pool of 35 items and if they could do so without assistance. The iterative approach provided an opportunity to make adjustments between rounds if necessary. We asked retrospective probing questions after girls completed the tool to check whether their interpretation of items matched our intention [ 20 , 24 ]. We dropped one item following field piloting, leaving a total of 34 items to be formally tested in Phase 4. In April 2018, we leveraged the main study’s endline survey in the four intervention schools (two urban, two rural) to collect data for assessing the dimensionality, reliability, and validity of the 34-item menstrual care self-efficacy tool and to determine which, if any, items should be dropped. The main study’s endline survey was administered to 528 girls randomly selected from class rosters. After completing the survey, only girls who had experienced menstruation were directed to two study team members who oversaw the self-efficacy sub-study survey. Study team members provided brief verbal instructions to participants (one-to-one or in small groups) and conducted the “pen test” for comprehension of the instructions before asking each participant to complete the pen-and-paper questionnaire on her own. A random subsample of 42 girls completed the questionnaire again approximately two weeks later to test temporal stability. Data from the hardcopy questionnaires were entered into a spreadsheet by one study team member and checked by one or two others to ensure fidelity. The self-efficacy sub-study survey had two sections: the first comprised 34 self-efficacy items for formal testing ( S1 File ). The second section comprised three additional measures for use in the assessment of the self-efficacy tool’s construct validity: the Bengali ( Bangla ) versions [ 26 ] of the Beck Self-Concept Inventory for Youth (BSCI-Y) and Beck Anxiety Inventory for Youth (BAI-Y) [ 27 ], and our Bengali translation of the Social Self-Efficacy Scale [ 28 ]. Although often conflated, self-efficacy is a construct related to yet distinct from self-concept or self-worth [ 12 ]. Therefore, we expected not to see a strong correlation between the self-efficacy tool and the BSCI-Y. Informed by self-efficacy theory, we anticipated that self-efficacy scores would be negatively correlated with measures of anxiety. However, since the BAI-Y is not a specific measure of menstrual-related state anxiety, we hypothesized that correlations would be low to moderate. We included social self-efficacy as a validation measure because some of our tool’s items involve interacting with and obtaining assistance from others to address menstrual needs. Although we hypothesized there would be a positive correlation between scores on our tool and Social Self-Efficacy Scale scores, we wanted to ensure that our finalized measure was not so highly correlated with the construct of social self-efficacy that it would not provide any additional utility. Additional items from the main study’s endline survey were used for validity testing. The items “During my last period, I felt anxious at school because of my menstruation” and “During my last period, I felt comfortable at school” served as indicators of anxious arousal more specifically related to menstrual experiences than the BAI-Y. The item “I prefer to stay at home during my period” was used to validate whether girls with lower self-efficacy scores tended to desire avoidance of contexts that make addressing menstrual needs challenging. Participants’ responses to these items on a 6-point Likert-type response scale of 1 = strongly agree to 6 = strongly disagree were dichotomized as “agree” or “disagree” for this analysis. We hypothesized that girls’ self-efficacy scores would positively correlate with time (months) since menarche, calculated from two items on the main study’s survey: current age and age when menstruation first began. (See S2 File for additional details of the survey measures.) We applied factor analysis methods to empirically explore how many unobservable constructs underlie the set of 34 formally tested items. This was to provide an understanding of whether self-efficacy in addressing menstrual needs should best be measured in our sample as one broad construct or rather through multiple more specific constructs (or “sub-scales”) [ 29 ]. In preparation for the exploratory factor analysis (EFA), we assessed all items for missing values, examined charts of item response frequency distributions, and performed tests of multivariate normality using the STATA command “omninorm” [ 30 ]. We then examined the item correlation matrix and used Bartlett’s test of sphericity [ 31 ] to ensure the data showed mild collinearity. Lastly, sampling adequacy was assessed using the Kaiser-Meyer-Olkin (KMO) measure (with a priori minimum acceptable threshold set at 0.50) [ 32 ]. We applied three methods to determine the number of factors to extract in EFA. We first conducted a principal components analysis (PCA) to identify how many components had eigenvalues greater than one. We then examined a scree plot and performed a parallel analysis test. We tested whether the extracted factors were correlated above Tabachnick and Fidell’s threshold of 0.32 [ 33 ], and subsequently used oblimin oblique normalized rotation. We made an a priori decision to consider dropping items that failed to load at least 0.30 on any factor during EFA, or if any items loaded highly on more than one factor [ 34 ]. Internal consistency was assessed for each sub-scale by calculating Cronbach’s α and by examining the average and individual interitem correlations [ 35 , 36 ]. Items that contributed poorly to internal consistency were considered for elimination. We calculated sub-scale scores for each observation by calculating the sum of responses for each retained item divided by the total number of items in the sub-scale. Temporal stability was assessed by retesting a subsample of girls two weeks after the first administration of the survey and calculating the concordance correlation coefficient for scale scores at the two time points. Bland-Altman plots were also examined to assess test-retest agreement in individuals’ scores. The time interval between test and retest was chosen to minimize the chance of substantial real changes in self-efficacy beliefs. We examined initial evidence for construct validity in our sample by assessing relationships between self-efficacy scale scores and other validation measures. For validation measures treated as continuous, we created scatter plots to visualize the relationships and calculated Spearman correlations. For dichotomous variables, we performed Wilcoxon rank-sum tests to test for differences in distributions of self-efficacy scores between groups. All analyses were conducted using Stata, version 14.2 (StataCorp LP, College Station, TX).

(PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (DOCX) Click here for additional data file.