Menstrual characteristics, menstrual anxiety and school attendance among adolescents in Uganda: a longitudinal study

The parent study was prospective, with pre-post evaluation of the MENISCUS intervention. The intervention is shown in Fig. 1. Full details of the design and results of the evaluation have been reported [22, 23]. All participants self-completed a questionnaire at baseline (October 2017) and endline (August 2018) on tablet computers using OpenDataKit software. At both surveys, information was collected on socio-demographics, knowledge of puberty, menstruation and pain management methods, menstrual myths, perceptions of menstruation, including menstrual anxiety, menstrual management practices at last menstrual period (LMP) and experiences of pain and leakage at LMP and behavioural and conduct problems using the 25-item Strengths and Difficulties questionnaire (SDQ) [24]. An additional file shows the measures from the baseline survey which were used in this study to answer objective 1 (Additional file 1). In addition, a sample of 50 female students per school were randomly chosen from those self-reporting having started menstruating in the baseline questionnaire, to participate in a nested cohort and self-complete daily diaries throughout the 10 months of follow-up. Diaries were provided for every term and holiday and participants were asked to record on each day whether they were menstruating and to self-define the flow as light, moderate or heavy, whether they had ‘period pain’, as well as school and class attendance. Data on pain, school and class attendance were collected each day, irrespective of whether participants were menstruating that day. An additional file shows the diary developed for the study which was self-completed by female students (Additional file 2).

Data for the current analyses focus on menstrual experience in the absence of an intervention, so we analyse baseline survey data from the parent study (objective 1), the 10 months of longitudinal data from the diaries to characterise menstrual patterns and characteristics (objective 2); and the 3 months pre-intervention data from the diaries to assess associations of menstrual characteristics with school attendance (objective 3).

Survey data captured electronically were stored on the LSHTM ODK secure server and transferred to Stata 15 for analysis. Diary data were paper-based and were double-entered using Microsoft Access and transferred to Stata 15 for data management and analysis.

Data from the baseline survey were used to examine the association between reported anxiety about next menstrual period (outcome) and menstrual factors (exposure) using logistic regression (objective 1). Measures are described in Additional file 1. Anxiety about next menstrual period was defined according to participants stating that they agreed or strongly agreed with the following statement: “I feel anxious about having my next period”. Using the Hennegan model of menstrual experience [6], we conceptualised anxiety about the next menstrual period as representing an aspect of menstrual confidence, and therefore a core part of the menstrual experience (Fig. 2). We first examined associations of anxiety with each variable. Then, aligned with the Hennegan model, we examined the most distal antecedents (level 1) i.e. those related to resource limitations (socio-economic status) and socio-cultural context (school, ethnic group, religion, living with mother, living with father), and retained factors independently associated with anxiety (using a cut-off p < 0.1). We created a similar model for the next level of antecedents (level 2), which relate to behavioural expectations (disagreement with the statement ‘it is healthy for a girl to run, dance or cycle during her period’) and knowledge and beliefs around puberty and menstruation, retaining factors from this level independently associated with anxiety (p < 0.1). We then created a model assessing the association between anxiety and variables related to menstrual practices, individual menstrual factors, confidence around menstruation, and shame and distress around menstruation (level 3), retaining factors independently associated (p < 0.1). We present multivariable models containing (a) level 1 and 2 variables (model A); and (b) level 1, 2 and 3 variables (model B). In all analyses we allowed for clustering by school by including school as a fixed effect.

Diary data were used to describe menstrual cycles during the 10 months of follow-up (objective 2). Initial analyses were at menstrual period-level, to estimate menstrual patterns (mean duration of cycle: number of days between the start of one menstrual period and the start of the next menstrual period; mean duration of menstruation (with standard deviation (SD)); perceived menstrual flow; and pain. Repeated-measures linear regression with robust standard errors was used to compare duration of menstrual cycle and menstrual period, by age group (14–16; 17–19 years) and years since starting menstruation (< 3; ≥ 3 years). To capture differences in participants’ experience of menstruation, the following participant-level summary statistics (summarised over all menstrual cycles per participant) were calculated: (i) percentage of days of menstruation with self-reported heavy flow and (ii) percentage of days of menstruation with pain. Further analyses were at day-level, adjusting for clustering by participant using random-effects logistic regression to estimate odds ratios (OR) and 95% confidence intervals (CI) for the association between pain (outcome variable) with flow (light/medium/heavy) and day of menstrual period (1st day vs. not 1st day).

The relationship between menstrual cycle characteristics and school/class attendance (objective 3) was analysed using the diary data collected during the first term of data collection (October-December 2017) i.e. prior to implementation of the intervention, which began in February 2018. Non-school days (i.e. weekends and school holidays) were excluded. Outcomes were (i) school attendance (attended a full day of school—including if they missed some classes) and (ii) class attendance (i.e. attended a full day at school and all classes). Random-effects logistic regression (adjusting for within-participant clustering) was used to estimate ORs and 95%CI for associations of each outcome with: day of menstrual period (1st day, not 1st day); menstrual flow (light, moderate, heavy) and pain (yes/no).

Overall, 222 girls (95.7%) reported that they had started menstruating at baseline and were included in analyses of menstrual anxiety. At baseline, over half (58.6%; 130/222) of participants reported being anxious about their next period. Table 1 shows associations with menstrual anxiety. There was no evidence of an association between anxiety about the next period and socio-economic status or living with their father. However, girls who did not live with their mother were more likely to report being anxious about their next period (68.1% vs. 48.6% of girls living with their mothers; OR = 2.25; 95% CI 1.30–3.90).

Anxiety about the next period was more common among those who disagreed that it is healthy for a girl to run, dance or cycle during her period (65.2% vs. 51.4%) and among those who agreed that when a girl has her period she is unclean (63.9% vs. 49.1%). Anxiety was less common among participants with poorer knowledge about menstrual biology (51.9% among those with ≤ 7/9 questions answered correctly vs. 68.5% among those with 8/9 questions answered correctly). These variables remained associated with anxiety after adjustment for each other and whether girls lived with their mother (Model A). There was no evidence of an association between anxiety and knowledge of the menstrual cycle, effective pain management methods or agreement that painkillers cause problems having children or sanitary pads can cause sickness or infection (Table 1).

With respect to the level 3 variables, there was no evidence that anxiety about the next period was associated with reporting using manufactured materials (i.e. reusable or disposable pads, tampons or menstrual cups) during the last menstrual period (LMP), blood leaking through clothes at LMP or pain during LMP. However, anxiety was associated with variables that may represent lower confidence around menstruation. Those agreeing with the following statements were more likely to be anxious: during my period I feel less confident than on other days (63.7% vs. 41.2%); during my period I avoid doing physical activity (71.4% vs. 45.5%); I prefer staying at home during my period rather than going to school (74.5% vs. 53.3%); boys/girls tease me about my period (78.4% vs. 52.6%). Girls who didn’t report feeling comfortable talking to other girls at school about their period (63.9% vs. 52.4%) or reporting that they worry about being teased during their period were also more likely to report feeling anxious (68.5% vs. 45.3%). In a model containing all level 3 variables, girls who agreed with the following statements were more likely to be anxious: during my period I avoid doing physical activity and boys/girls tease me about my period. Girls reporting that they do not feel comfortable talking to another girl about their period or who reported worrying about being teased during their period were also more likely to report feeling anxious.

In the multivariable model including variables from all levels (Model B), the following factors were associated with being more likely to report anxiety about their next period: not living with their mother (aOR = 1.91 [1.01–3.60]), not agreeing that it is healthy for a girl to run, dance or cycle during her period (aOR = 1.83 [0.95–3.50]), agreeing that when a girl has her period she is unclean (aOR = 1.97 [1.04–3.73]), avoiding physical activity during her period (aOR = 2.49 [1.33–4.65]), boys/girls teasing her during her period (aOR = 2.89 [1.28–6.54]) and worrying about being teased during her period (aOR = 2.27 [1.21–4.27]). Those with lower general knowledge about menstruation biology were less likely to report anxiety about their next period (aOR = 0.44 [0.23–0.84]). The association between avoiding physical activity and anxiety was weaker in this model adjusted for level 1 and 2 variables suggesting some confounding by these variables. Associations between disagreeing that it is healthy for a girl to run, dance or cycle during her period and not disagreeing that when a girl has her period she is unclean were also attenuated in the final model (Model B) suggesting that they were mediated through the level 3 variables.

The Hennegan model suggests that menstrual experience impacts on psychological health. In our study, participants who reported being anxious about their next period had a higher mean SDQ score than those who did not report being anxious (11.0 (SD5.7) vs. 9.1 (SD5.5)) but there was no association after adjustment for level 1, 2 and 3 variables in the final model (aOR = 1.02, 95%CI 0.96–1.08) implying that they have similar causal factors.

Using data from October-December 2017 (prior to the intervention), menstruation was associated with reduced school attendance. Participants reported not attending a full day of school on 15.8% of period-days (70/443) vs. 8.6% of non-period days (235/2736) (OR = 2.36 [1.72–3.23]). Table 2 shows the menstrual-related characteristics associated with missing school. On period-days, pain was strongly associated with not attending a full day of school with 21.7% of girls missing school on days when they reported pain vs. 8.3% on days when no pain was reported (OR = 4.22 [1.88–9.51]). There was weak evidence for an association between day of the period and not attending the full day of school with 20.7% girls reporting missing school on the first period-day versus 14.1% on subsequent period-days (OR = 1.83 [0.97–3.45]). There was no evidence of an association between heaviness of flow and school attendance (Table 2). After adjustment for all menstrual variables, pain remained strongly associated with school attendance (adjusted OR = 3.82 [1.66–8.77]) but there was no evidence of an association between first day and not attending a full day. Similar associations were seen with not attending all classes.

The majority of participants in our study (58.6%) reported feeling anxious about their next period. These findings align with Hennegan’s meta-synthesis which reported that worry and fear featured prominently in women and girls’ menstrual experience [6]. In our study, menstrual anxiety was strongly associated with various aspects of the socio-cultural context, particularly not living with their mother, negative behavioural expectations, lack of menstrual confidence and shame (anxiety about being teased), possibly reflecting an internalised stigma. There was no evidence of an association with menstrual practices or individual menstrual factors (e.g. pain or blood leaking through clothes at last period), which may be due to a high proportion of this population reporting using manufactured menstrual materials during their last period.

We observed lower anxiety in girls living with their mothers, and those who reported feeling comfortable to talk to other girls at school about their period but no differences in anxiety according to whether girls were living with their fathers. Other studies have highlighted the challenges of sharing information about menstruation with fathers [8]. The association between menstrual anxiety and greater knowledge of menstrual biology is puzzling, but there was no evidence of an association between menstrual anxiety and knowledge about the menstrual cycle specifically and, reassuringly, no association between menstrual anxiety and greater knowledge of menstrual biology using data from the endline survey. Given the role boys play in teasing girls during their menstruation, also found in other studies [25, 26], interventions to improve girls’ menstrual experience should include both boys and girls to improve broad social support from the school community. We found an association between anxiety and a specific concern about being teased during menses but no evidence of an association of anxiety with either blood leaking through clothes during the last period, or with pain, despite the fact that in other studies teasing has been reported to be related to others knowing whether a girl was menstruating e.g. due to stains [4].

Our observed low levels of menstrual confidence among girls highlights the importance of addressing the socio-cultural context of menstruation. The importance of markers of social support and increased self-efficacy in managing anxiety around menstruation suggests that social cognitive theory (SCT) may be an appropriate framing for menstrual health interventions in schools [27]. SCT asserts that individuals learn by observing others, and that an observed behaviour is learnt and enacted if: (i) the individual has self-efficacy regarding the behaviour; (ii) individuals receive reinforcement for performing the behaviour; and (iii) the environment supports enactment. The theory suggests opportunities for providing social support, improving self-efficacy and using peer/observational learning to achieve behaviour change. After the pilot MENSICUS intervention, which is grounded in SCT, we found that the proportion of study girls reporting menstrual anxiety reduced from 58.6 to 34.4% [22]. However, there was evidence of persistent menstrual anxiety with about half of the participants who reported being anxious at baseline, also reporting anxiety at endline. Further research is needed to understand how best to address this.

Access to quality education, especially for girls, is a key sustainable development goal set by the United Nations in 2016 [28]. In previous qualitative and cross-sectional studies, girls reported menstrual pain to be a major contributor to school absence [29‐31]. By collecting prospective daily diary data, we were able to directly link school absence days with menstrual factors and examine the factors most strongly associated with school absence. Over the 10 months of follow-up, pain was reported on almost half of period-days, and with almost all participants (97.0%) reporting some menstrual pain during follow-up. Pain was more likely to be reported on days with moderate or heavy flow, and was the key factor associated with school absenteeism, as has been seen in previous studies [11, 13]. In our study, 1 in 5 girls reported missing school on period days when they reported pain, compared with 1 in 12 girls missing school on period days when no pain was reported. Further, we observed that the level of non-attendance on period days where there was no pain was similar to the level on non-period days, and after adjustment for pain there was no association between first day of the period and non-attendance suggesting that this association was mediated through pain. This finding highlights the need for improved education on pain management for adolescents, including addressing myths about use of painkillers and teaching alternative pain mitigation techniques such as stretching, exercise and use of hot-water bottles.

Previous studies show inconsistent evidence for a relationship between menstruation and school absenteeism [1, 26, 32‐37]. This may be due to differences in the association by setting, and also due to challenges in documenting school attendance itself, and documenting associations with menstrual characteristics. Prospectively-collected diaries work well to link data on menstruation directly with school attendance. As reported previously [22], we validated the prospective daily diary entry of attendance with unannounced observation of attendance (“spot-checks”). Girls were seen on 328/330 (99.4%) of days when their diary stated they were present and were not seen on all 37 days when the diary stated that they were absent. School registers were incomplete and rarely completed by teachers in this setting. However, spot-checks are not a ‘gold standard’ for school attendance as girls may not be visible (e.g. in the bathroom) or may attend for only part of a day (this would not be captured). The parent MENISCUS-2 study found that girls were more likely to attend school during menstruation at endline than at baseline, with reasons including the training on pain management, tracking their menstrual cycle, and having reusable pads [22].

A limitation of our study is the relatively small number of participants with detailed menstrual cycle data (n = 99). However, the multiple datapoints per participant resulted in good power to detect associations of menstrual factors with school attendance. Further, this was a random sample of all post-menarcheal girls in the study, with a high response and retention rate, and was unlikely to have suffered substantially from selection bias due to high levels of participation. Girls were provided with information about how to manage menstrual pain during the study so this might have affected the reporting of pain in the diaries. In our study we asked girls about period pain generally and we do not know the site of the pain. For example, previous studies have reported headache as an important menstrual symptom, particularly as part of the pre-menstrual syndrome [12, 13]. We do not know whether girls in our study would have considered headache as period pain and reported it. We were also unable to describe the association between detailed menstrual characteristics (e.g. percentage of menstrual days with pain or percentage of menstrual days with heavy flow) and menstrual anxiety since the detailed menstrual data were only available for the 99 girls who participated in the diary sub-study. We therefore used experiences during the last menstrual period as an imperfect proxy for this. Finally, our sample of two schools limits the external validity of our results, and meant that we were unable to look at associations between structural factors like school-based WASH facilities because we recruited from only two schools.