Participants
Female medical and economic students from Universiti Malaysia Sarawak (UNIMAS), which is a public university in the state of Sarawak in Malaysia, were recruited voluntarily for this study. Sample size was estimated using the population-based sampling method by Krejcie and Morgan (1970) [8]. Based on the total number of approximately 500 female medical students from Year 1 to Year 5 (with confidence interval of 95% and the margin of error 0.05), the estimated sample size for female medical students was 210. And based on the total number of approximately 400 female economic students from Year 1 to Year 3 (with 95% confidence interval and the margin of error 0.05), the estimated sample size for female economic students was approximately 190.
Materials
Exploratory factor analysis (EFA) was performed using Statistical Package for the Social Sciences (SPSS) statistical software with principal axis factoring as the extraction method. Partial least squares structural equation modelling (PLS-SEM) using the SMART-PLS software was performed to measure the impact of various sociocultural factors, religious factors as well as biological symptoms on the quality of life (QoL) during menstruation. The list of common biological symptoms during menstruation was adapted from past study by Wong and Khoo [6] on menstrual-related symptoms in a similar population of Malaysian female adolescents. The validated Quality of Life Enjoyment and Satisfaction Questionnaire – Short Form (also known as Q-LES-Q-SF) by Endicott et al. [9] was adapted to measure quality of life, which is the dependent variable in this study. Q-LES-Q-SF had been similarly used before to measure the impact of menstrual pain [10] and the impact of pre-menstrual related disorders on quality of life. This instrument had also been shown to have good internal consistency and reliability [5].
Procedures
Stage 1: Generation of preliminary list of items using Modified Delphi Technique
Adopting the classification by Young and Bacdayan [3] as described above, personal interviews with female friends and family members were conducted by authors KEP, AKH, NZ and NALY to identify common sociocultural and religious beliefs and practices during menstruation in our local setting. Besides that, literature search was also conducted using keywords such as “menstruation”, “menstruating”, “menstrual”, etc. to skim for information in academic journals, webpages, blogs, etc. From this initial search, a preliminary list of socio-cultural and religious beliefs during menstruation among the various ethnic groups in Malaysia were listed and categorized according to the categories described by Young and Bacdayan [3].
Following that, opinions and suggestions were sought from five female lecturers in UNIMAS to validate and to further improve and refine our preliminary list using the modified Delphi technique. Modified Delphi technique is a structured iterative process aimed to obtain consensus from individuals through a series of communication until the group consensus is reached [11, 12]. In this regard, a group of female lecturers of different ethnicities in Malaysia were invited to participate in a modified Delphi technique through email communications. We asked them how representative our preliminary list was on common sociocultural and religious beliefs during menstruation. After we obtained some initial responses, we further refined our list and added in items based on the inputs from these lecturers. Our edited list was then emailed again to these same female lecturers whether they would agree to this edited list. A cut-off point of 70% agreement was set as the minimum level for an item to be included in the edited questionnaire [11].
Stage 2: Exploratory Factor Analysis (efa)
From this edited list of items intended generated from Stage 1, a preliminary set of questionnaires was developed. In this stage, 100 female medical students and 100 female economic students were asked to volitionally and anonymously rank their agreement on these items in a Likert scale of five, ranging from “1 = strongly disagree” to “5 = strongly agree”. The construct validity was then determined using principal axis factoring as the extraction method via SPSS software. An initial run of factor analysis was performed in order to determine the number of factors to be extracted. Factors with eigenvalues greater than one would be retained.
Once the number of factors was determined, repeated runs of factor analysis were then performed to determine the factor loadings of the items as well as to identify problematic items that may need to be removed. Varimax rotation was used with a cut-off factor loading value of 0.4 as the criteria to determine whether an item was to be removed or not [13]. Pattern coefficient values of less than 0.5 were suppressed. The communality value, which indicates convergent validity of the items, was set at 0.25. Finally, the Cronbach’s alpha coefficients were then checked to evaluate the degree of internal consistency of the items in each construct or factor. A cut-off point of Cronbach’s alpha > 0.6 was set as the criteria of a satisfactory degree of internal consistency [14]. Based on the EFA in this stage, the questionnaire was edited, and some items were deleted as dictated by the indicators in EFA.
Stage 3 Confirmatory Factor Analysis (CFA) and Structural Equation Modelling
Confirmatory factor analysis was then performed on the edited questionnaire that we have obtained from the EFA in Stage 2. In this stage, another 300 females medical and economic students were asked to rank their agreement on these items in a Likert scale of five, ranging from “1 = strongly disagree” to “5 = strongly agree”. Reflective measurement model was performed using partial least square (PLS) method in SMART-PLS software. For internal consistency of the items, three parameters were analyzed, i.e., i) Cronbach alpha; ii) composite reliability (CR) index; and iii) the rho A (ρA) coefficient (also known as Dijkstra Henseler’s rho [15–16]). For convergent validity, the factor loadings of all items were obtained, as well as the Average Variance Extracted (AVE) of each factor or construct. Factor loading of > 0.7 was considered as acceptable, whereas factor loading of < 0.4 was deleted. For a loading with values between 0.4 to 0.7, the AVE would then be used to determine whether the item should be accepted. AVE refers to the grand mean value of the squared loadings of all items associated with a factor. AVE of > 0.5 is generally acceptable for an item to be included even if its loading is between 0.4 to 0.7 [13]. For discriminant validity, Fornell and Larcker criterion [17], cross loadings of items as well as the Heterotrait-Monotrait ratio of correlations (HTMT) proposed by Henseler et al. [18] were obtained. All these measurements were generated from the SMART-PLS software.
Structural modelling was then performed to evaluate the influences of these various aspects of sociocultural and religious beliefs and practices on the students’ quality of life. In this regard, the practices of these various beliefs were taken into consideration as the mediating effect on the quality of life (as measured using Q-LES-Q-SF). The reason to include practices of beliefs as the mediating effect is because beliefs without practices, may not affect their quality of life. Furthermore, not all beliefs would be translated as practices. From a biopsychosocial perspective, the impact of biological symptoms of menstruation on quality of life was also taken into consideration. We adopted the five common biological symptoms of menstruation (i.e., fatigue, abdominal pain/cramp, mood swing, headache and irritability) from a previous study on quality of life [6] as measured on a Likert scale of 5 where 1 = strongly disagree that this symptom is common for me, and 5 = strongly agree that this symptom is common for me.
Institutional ethics approval was obtained prior to starting this research (reference no UNIMAS/NC-21.02/03 − 02 Jld.3(51)). All participants were assured that their data would be kept confidential, no personal identification data such as name, personal identity number, etc. would be revealed and their data would only be used anonymously solely for the purpose of this research. Participants were recruited voluntarily and they were informed that they could withdraw their participation at any time.