The study was carried out in the states Chhattisgarh, Maharashtra, and Tamil Nadu, representing the diverse cultural and socio-economic spectrum in India (Table S1 in Online Supplementary Document ). Chhattisgarh is a state from central India with a predominantly Tribal population with 2.7 million adolescent girls. Maharashtra is a more developed state in the western part of India with 9.9 million adolescent girls. Tamil Nadu is a southern state having one of the highest levels of development with 6.1 million adolescent girls. Tamil Nadu has implemented a free sanitary pad scheme since 2011, making pads free of cost for girls living in rural areas, those in government schools, and new mothers. Girls can receive three packs of pads once every two months, in addition to iron tablets, and may receive education about menstruation from an “aganwadi” (female community health) worker [20]. Similar programs in Maharashtra and Chhattisgarh are less developed.

A total sample size of 1800 adolescent girls (600 girls per state, about 75 girls per school), would be sufficient to measure a state-based prevalence of 50% with 95% confidence interval and 5% of margin error, taking clustering into account and using a design effect of 1.5. Multi-level stratified sampling was used for each state, by first randomly selecting one district in each of the three states. In each of the selected districts, one block was then randomly selected, and then in each of the selected blocks, a list of all schools was prepared in collaboration with the state government education department in the respective districts. In each of these, schools were then randomly selected from all government middle and high schools (regular) after excluding boys’ only schools, solely residential, and private schools (Figure S1 in Online Supplementary Document ). Adolescent girls above 12 years of age in class (called grade in India) 8-10 (comparable to school year 8-10 in the United States) were selected to maximise the likelihood they had reached menarche and could provide information on menstrual management and water sanitation and hygiene. One class was randomly selected if there was more than one class in grades 8-10. In Tamil Nadu, more schools were included than in the other states because girls were younger and fewer girls had reached menarche compared to the other states. In addition to regular schools, schools that received support on menstrual hygiene from external sources (“better practice” or model schools) to represent the best case scenario in MHM were purposively chosen with the help of the UNICEF team in the respective state to assess if this resulted in significantly better menstrual management practices. In the model schools, external experts (from UNICEF or other NGOs) regularly provided information sessions on puberty, menstrual hygiene and on how to use menstrual pads.

Schools were visited, and meetings were held with the head teachers. Parental consent forms were then distributed by study staff with the help of school staff. Meetings with target girls who had parental consent were conducted to discuss the study and respond to questions before girls assented. Pre-tested structured self-administered questionnaires in the local languages of each state were used to elicit information on the sanitation status of the school, knowledge about menstruation, pre-menarche, menstrual practices and beliefs, and the effect of menstruation on school life. The data collection was carried out from June to December 2015. Three senior research officers supervised the field data collection team who received intensive one-week training before the start of the study.

For this analysis, only girls who reported they had started menstruating were included. We tabulated results for model and regular schools by state for the following themes: awareness about menarche and source of information, menstruation-related restrictions, menstrual absorbents, effect of menstruation on the school experience, and barriers and facilitators of menstrual management at the school level. Missing data was included as a separate category of the variables of interest. Significant differences were explored at the state level and model vs regular schools overall and within states (χ² test). To assess factors associated with school absence during menstruation, we used generalized linear regression with a log link and binomial distribution for multivariate analyses. Poisson regression with a robust variance estimator was used for models which did not converge (Stata v14.2, StataCorp LLC, College Station, USA). The following factors were explored in univariate analysis: age, state, model vs regular school, menstrual item used, education or program on menstrual hygiene in school and factors related to sanitary situation in school. Factors with a P-value <0.1 in the univariate model and model vs regular schools as a focus of interest were included in the multivariate model, whereby factors with a P-value >0.05 were removed from the multivariate model using backward elimination. Univariate and multivariate models were adjusted for clustering at the school level, and interactions between significant variables were examined.

The study was approved by the Tata Institute of Social Sciences, Mumbai, and the Liverpool School of Tropical Medicine, UK, after fulfilling all the ethical requirements. Participant information sheets, describing the study and the activities involved for study participants were prepared. Written informed consent from the parents and assent from the girls was obtained before the study, in compliance with national and international ethical committee requirements. The survey questionnaire had ID numbers and had no names on it. Consent forms and questionnaires were translated into local languages of the states involved.

Of the 3617 girls who participated, 2564 (70.9%) reported they had begun menstruating and were included in this analysis. These menstruating girls attended 43 randomly selected regular schools (N = 2072) and 10 model schools (N = 492) in the 3 states ( Table 1 ). Over half of schools were co-educational (58%), a third (31%) was girls’ only, and the remaining had some girls and some mixed classes. Girls’ average age was 14.1 years (standard deviation (SD) = 1.1), with girls from Tamil Nadu slightly younger than other states, and girls in model schools slightly older (14.2 vs. 14.0 years, P < 0.01). Most girls were in grade 10 (51%), while 35% were in grade 9, and 13% in grade 8. Participating girls were mostly Hindu (93%), with 3% Muslims, and 2% other religions.

Nearly all (93%) menstruating girls had received some information about menstruation ( Table 2 ). Parents or guardians were the major source (68%), with friends the next most reported source. There were striking differences by state and type of school; approximately 1 in 10 girls said teachers were a common source, the highest proportion were in model schools in Tamil Nadu (51%), and lowest in regular schools in Maharashtra (3%). Half (48%) of girls did not hear about menstruation until their first period began; the proportion of girls who were informed before menarche was significantly higher in model schools compared to regular schools (56% vs 36%, P < 0.001).

Religious restrictions (not going to temple, etc.) were common, affecting 88% of girls overall, and nearly all (91%) girls in regular schools ( Figure 1 , and Table S2 in Online Supplementary Document ); girls reported less restrictions in the model schools in Maharashtra (64%) and Tamil Nadu (76%) compared to the regular schools (94% and 93%, respectively, P < 0.001). Restrictions during exercise were prevalent (83% overall); again this was significantly lower in model schools in Maharashtra (89% vs 50%, P < 0.001) and Tamil Nadu (92% vs 69%, P < 0.001) but not in Chhattisgarh (78% vs 84%, P = 0.205). Other reported restrictions due to cultural traditions were less common ( Figure 1 ); thus one in five reported different sleeping arrangements, 16% reduced social interactions within the home, 12% reduced social interactions outside the home, and 7% had restricted food choices.

Overall, 45% of girls used disposable sanitary pads, 28% used cloths, and 21% reusable pads. Menstrual cups and tampons were reported by 1% each, 2% of girls said they did not use anything, and 3% did not respond. There were considerable differences by state and school ( Figure 2 and Table S3 in Online Supplementary Document ). The majority of girls in Chhattisgarh used cloths (61%), whereas the majority of girls in Maharashtra and Tamil Nadu used disposable pads (47% and 66%, respectively). Reusable pads were mainly used in Maharashtra (37%) and Tamil Nadu (21%). Only in Chhattisgarh significant differences were present between model and regular schools; cloths were used less frequently in model schools where disposable pads were more common.

The majority of girls reported going to school during their menstruation (87%, Table 3 ), and this was higher among model schools (92% vs 86% in regular schools, P = 0.003). One out of five girls in regular schools in Chhattisgarh reported missing school during their period. The majority (65%) of girls reporting absence stated it was for 1 day, 22% said 2-3 days and 13% responded that it was throughout menstruation. Concentration problems at school during menstruation were common (45%), with differences noted by state ( Table 3 ). Other frequently mentioned problems in school included pain (stomach, head, hips and limbs, 36%), fear of staining or smell or losing the cloth or pad in school (15%), feeling unwell, tired, dizzy, and weak (11%). A few (5%) reported reduced mobility and comfort resulting in problems with sitting, walking, bicycling, and reaching school. Girls who used disposable pads were significantly more likely to report attending school during menstruation (95%), and less frequently reported concentration or other problems during menstruation (39%, and 47%, respectively) than girls who used cloths (81%, 53%, and 68%, respectively, P < 0.001 for all comparisons, Figure S2 and S2 in Online Supplementary Document ).

Overall, 34% of girls reported to have received education about menstrual hygiene in school; the proportion differed significantly by state, type of school and within states (Table S5 in Online Supplementary Document ). The majority of girls heard about it in a hygiene lesson (58%), during lessons separate from boys (82%). Written materials about menstruation were infrequently available (19%) and mainly present in model schools. Of the 1742 girls who heard about menstrual hygiene from their parents or guardians, 586 (34%) had lessons at school as well.

Numerous factors were associated with missing school during menstruation in univariate analysis ( Table 5 ); however, six remained in the multivariate model. State and type of school affected absence rates, and were significantly lower in Tamil Nadu (adjusted prevalence ratio (APR) = 0.24, 95% confidence interval (CI) = 0.14-0.40) and Maharashtra (APR = 0.56,95% CI = 0.40-0.77) compared to Chhattisgarh, and halved in model compared to regular schools (APR = 0.50, 95% CI = 0.34-0.73) The use of disposable pads, the availability of pain medication, and a space to wash in school were all associated with less absenteeism during menstruation. Dysfunctional toilets or long queues for toilets were associated with increased absenteeism. In a separate multivariate analysis including only variables related to sanitary facilities at school, “clean toilets”, “toilet breaks”, and “can wash in school” remained significant; however, “gender-separate toilets” was not significant in the multivariate analysis (Table S6 in Online Supplementary Document ).