Food insecurity is a common problem among adolescents and households in Southwest Ethiopia [13, 27]. The present study shows that food insecure adolescents and adolescents who were members of severely food insecure households were more likely to be absent from school and have a lower educational attainment in terms of the highest grade completed after 1 year of follow-up in Southwest Ethiopia. Food insecure adolescents had on average a lower grade completed (5.5) compared to food secure adolescents (6.0). The proportion of food insecure adolescents who completed primary education (24.1%) was also significantly lower than that of food secure adolescents (31.5%), although both proportions were lower than the national completion rate of primary schools in 2005-2006 . Food insecure adolescents are likely to have a lower educational attainment due to several reasons including high absenteeism, illness, poor academic performance, academic delays, poor social functioning and behavioral problems [2, 3, 6]. According to Food and Agricultural Organization of the United Nations, hungry children start school later, drop out sooner and are also more likely to be absent and learn less while they do attend .
The effect of food insecurity on physical health and wellbeing is one of the ways through which food insecurity can erode educational attainment. As adolescence is a critical period of human growth and development, it can be affected by nutritional constraints resulting from food insecurity. There is an increasing recognition that ill health and malnutrition among school aged children have a major impact of on their cognitive development, learning and educational achievement. Improved health and nutrition are positively associated with enrolment at younger age, reduced absenteeism, less grade repetition and higher performance on test scores [30, 31] due better cognitive development . Cross-sectional studies from high income countries show that when food supplies are constrained at the household level, both adults and children use coping strategies such as reduction of meal size and meal frequency and eating lower quality foods , which in turn might induce malnutrition . Similarly, cross-sectional  and longitudinal studies from high income countries [3, 7] and a cross sectional study from a low income country  also showed that malnutrition is associated with poor cognitive development and decreased school performance in children. Malnourished children are less able to concentrate in school [3, 18, 29], which could lead to lower educational attainment.
In this study, adolescents who reported an illness during the last month were twice as likely to be absent from school. A report from the study area showed that food insecure adolescents had a higher frequency of illness than food secures ones [9, 13]. Food insecurity not only jeopardizes the right to health but also has serious implications on education and schooling of adolescents. Secure access to food may enhance school attendance and overall health and well-being through decreasing use of negative coping strategies [30, 31].
Food insecurity can also act as a psychological or emotional stress factor [15, 34, 35], affecting adolescent behavior and aspiration for further education. Our findings show that adolescents who aspire to complete higher grades were 18% less likely to be absent from school. The effect of food insecurity on children's well-being is related to its effect on the physical and socio-emotional aspects [3, 4] that are linked to developmental consequences through nutritional and non-nutritional mechanisms . While the nutritional effects might explain the negative consequences of inadequate nutrient intake on physical health and cognitive development, non-nutritional aspects of food insecurity relate to psychological effects including worry, anxiety or sadness about the family's food supply, feelings of having no choice in the foods eaten, shame/fear of being labeled as poor which are beyond the nutritional effects of food insecurity on adolescents . Cook and Frank reported that food insecurity is a common risk to the growth, health, cognitive, and behavioural potential of children in constrained situations . Children living in constrained environments were reported to have significantly higher cortisol levels due to stress . Prolonged elevated cortisol level in humans have been associated with depression, cognitive deficits, and atrophy of brain structures involved in learning and memory [16, 17] which may also lead to lower educational attainment. As a result of stress, anxiety and disrupted household dynamics , food-insufficient teenagers are more likely to have behavioral problems which makes them incompatible with school norms  negatively impacting on their educational attainment. Any of these mechanisms, alone or combined, could explain how food security affects the educational attainment of adolescents in the study area. Interventions focusing on improving the food insecurity situation such as conditional cash and/or food transfer programmes  need to consider these consequences on adolescents in identifying target groups.
After stratifying for gender, the bivariate analysis showed that in food insecure situations a higher proportion of girls are more likely to be absent from school compared to boys. This might be related to the social norms in the community that give low value female to education  that are potentially exacerbated when resources are constrained. A significantly larger proportion of girls was from food insecure households, food insecure themselves and reported an illness during the past one month compared to boys. This may be due to the selective buffering of boys from food insecurity by adult household members that is common in the study area . The higher rates of food insecurity in girls could result in their school absenteeism through mechanisms discussed above. When entering gender as a variable in the multivariable model for school absenteeism however, no significant association was observed (results not shown), which indicates that the effect of gender is mediated by the other variables in the model.
School absenteeism in this study was also determined by a high frequency of work at the household level that the adolescent was engaged in. Adolescents who had to work longer hours per week were 1.4 times more likely to be absent from school. Children become absentees or dropout of school to help household labor as one of the coping strategies of food insecurity .
Our results imply that secure access of adolescents to food supply needs to be given attention to achieve the targets of the Millennium Development Goal. The findings provide arguments to incorporate food security in interventions that aim to address school attendance. Interventions including food stamp programs  and school breakfast programs  were reported to have beneficial effects for children on academic learning through improving dietary intake and/or reduction of stress. School supplementation program were also shown to have beneficial effects in reducing scholastic difficulties of adolescents . In Ethiopia although such programs are being implemented by relief actors as part of an emergency response  there is no school feeding programme in non-emergency scenarios. There is a need to consider such interventions in potentially food constrained areas to prevent attrition of food insecure adolescents from school.
We acknowledge a number of limitations in our study. Although the adolescent and household food security scales were adapted from household food security scales after thorough discussion with the interview team who are residents of the study area, we might not rule out the possibility of some misclassification. The scales used assessed acute food insecurity as they covered only the last three months prior to the interview. The total survey period involved two seasons (rainy/peak hunger season) and spring season  which is better off in terms of food security and this matches a follow up period of one year. The proportion of food insecure households was 39.4% which was similar to estimations for urban areas of the region . When interpreting the results of the study, it is important to note that our findings regarding food insecurity relate mainly to the study season and might not reflect the total year, but this is not a problem for the absenteeism variables. In addition, household food insecurity might be associated with adolescent food insecurity, which could introduce co-linearity in the models. The correlation coefficient between household food insecurity and adolescent food security was 0.2 and variance inflation factors in both logistic and linear regression models were generally low (the highest observed was 1.08), which indicates no considerable co-linearity in our analysis.
As the study involved adolescents who are at the different stages of academic status, we used the highest grade completed as a measure of educational attainment that can serve across all age groups. However, there are other measures of educational attainment that were not captured. The fact that we did not have data from the school records regarding the academic performances is also a limitation of our analysis.