MATERIALS AND METHODS

Settings and participants. Official results of national academic tests between 2017 and 2020 were analysed. The tests are applied by the Colombian Institute for the Evaluation of Education and are aimed to evaluate the skills and academic achievement of Colombian secondary school students in conjunction with their background, non-cognitive characteristics and school context ¹³. The exam is applied since 1968 every six months and is required for college and university admission; it includes questions regarding familiar socioeconomic traits such as household belongings and parents' education level; students' ethnicity, working status and food frequency intake. The academic achievement is assessed through the domains of Critical Reading, Mathematics, Social and Citizenship, Natural Sciences and English as foreign language; a global score which is the weighted average of the five scores is calculated as follows: Global Score = 5 * ((3^*Math + 3*Reading + 3^*Natural science + 3^*Social + 1^*English) / 13) ¹³.

Variables. Scores of each domain and the global score were considered as the dependent variables. Scores varied from 0 and 100 for the five domains, and from 0 to 500 for the global score ¹³. Independent individual-level variables were ethnicity, defined as belonging or not to an ethnic minority; adolescent labour, defined as the self-report of working outside the house and school the previous week of the exam; food frequency intake per week of dairy products, meat and eggs, and cereals, nuts or legumes through a Likert scale, going from never or almost never to daily or almost daily; and socioeconomic status (SES) by means of a composite indicator that includes the educational level of parents, material of floors and walls of the house, toilets and proper sanitation facilities, internet and television services, and other amenities including computer, DVD, refrigerator and washing machine ¹⁴. This indicator ranges between 0 and 100, being 0 the less favoured. At collective level, Colombian states were considered as natural grouping variables, and rurality was included in the model by means of the proportion of people living in rural places, i.e., the total of people living in rurality divided by the total population of the state by year ¹⁵.

Statistical analysis. Global and specific scores for each academic domain were reported using means and standard deviations according to sociodemographic characteristics. A crude evaluation of the effect of individual-level variables was obtained using Wald tests, and associations between rurality and academic domains were assessed using Pearson correlation coefficients. Two-level structural equation models (SEM) were used to evaluate the effect of individual-level variables and rurality on academic scores. Covariance among individual variables and rurality were considered in the model. P-values less than 0.05 were considered significant. All analyses were carried out via Stata V.17 (Stata, Inc., College Station, TX, US) ¹⁶.

During the study period, a total of 2,147,062 adolescents were included distributed as follows: 546,255 (25.4 %; 95 % CI 25.4 %-2 5.5%) in 2017, 549,933 (25.6 %; 25.6 %o-2 5.7 %) in 2018, 546,077 (25.4 %; 25.4 %o-2 5.5 %) in 2019, and 504,797 (23.5 %; 23.5 %o-23.6 %) in 2020 (Table 2). 54.4 % (95 % CI 54.3 %-54.5 %) of the people included in this study were females, 6.6 % (6.5 %-6.6 %) belonged to an ethnic minority, 38.6 % (38.5 %-38.6 %) reported not having access to internet and 34.6 % (34.5 90-34.7 %) worked the previous week to the exam. The SES score varied from 12.5 to 84.8 (mean=50.3, SD=9.5), and less than the 50 % of students reported a daily or almost daily intake of meat or eggs (39.9 %; 39.9 %>-40.0 %), dairy products (30.6 %; 30.5 %o-30.6 %), and cereal, nuts, or legumes (16.6 %; 16.5 %-16.6%). Table 1.

A total of 580,717 persons resided in the 4 biggest cities in the country: Bogotá, Medellin, Barranquilla and Cali; 338,459 (15.8 %>; 95 %o CI 15.7 %>-15.8 %o) were from Bogotá, the capital city. In contrast, a 15.6 % (95 % CI 15.6 %-15.7 %; N=334,997) of the scholars studied in institutes located in rural areas.

Average academic scores by domain varied from 48.22 (SD = 11.84; 95 % CI 48.20-48.24) in citizen social science to 52.52 (SD=10.17; 95 % CI 52.51-52.54) in critical reading. The highest average scores, both global and specific domains, were reported in Bogotá, also the area with the lowest proportion of rural residents (0.2 %). Correlation between rurality and average scores are shown in Table 2.

Crude comparisons showed significant effects of ethnic minority, child labour, and low food frequency intake on each of the five domains and the global score (Table 1). In the adjusted analyses, individual-level variables remained significant; in particular, belonging to an ethnic minority reduced the scores on average by 2.83 (95 % CI 2.77-2.89) and 4.61 (4.55- 4.67) in English and Mathematics tests, respectively. Likewise, global score was 20.07 (19.8120.33) points less among minority students. Adolescent labour was also associated with lower test results, with 1.72 (1.69-1.75) and 2.35 (2.32-2.38) points less in mathematics and social sciences, respectively; and an average reduction of 10.62 (10.49-10.76) points in the global score. Regarding the SES punctuation, an average increase of 2.21 (2.20-2.22) points in the global results is shown with an increase of one point in the SES indicator. Food intake effects varied from a reduction of 0.06 (0.04-0.07) points in social science scores with higher consumption of cereals and legumes to an increase of 1.29 (1.22-1.37) in global score with higher dairy consumption. Dairy intake did not had a significant effect on English test results (p=0.068). Rurality was positively associated with all academic domains, except for English language. As expected, individual level variables showed a significant correlation (Table 3).

DISCUSSION

In this study, based on official records of the academic state exams from 2017 to 2020, we verified that recognized social determinants have also a significant negative effect in the academic achievement in Colombian youth, and in consequence, a poor performance in near future regarding personal fulfilment is expected. However, no differences in academic achievement were identified between pre-pandemic pandemic times. Similar to previous health studies in the region, variables like food insecurity, poor access to health services and poverty ¹⁷, are simultaneously related to increased morbidity, mortality, poor quality of life and decreased academic attainment. Indeed, our results indicated that a higher socioeconomic level confers an increase of approximately 2 points in the global test score with specific increases in each domain once compared with people living in less favoured conditions. Comparable findings have also been showed about socioeconomic status and mortality and morbidity elsewhere ¹⁸.

Previous adolescent labour was also identified as a negative predictor of academic accomplishment. In this study, the participants who reported having worked "during the last week", had worse scores in both the global and specific scores. A reduction over 10 points in the final rating was evidenced. As reported before, adolescents who have to work usually observe many other risk factors for their development, including poverty, mistreatment, violence and malnutrition ¹⁹. Social and family poverty, loss or incapacitation/illness of parents, lack of social security and protection, and ignorance about the value of, or limited access to, education are among the myriad reasons for the involvement of adolescents in the workforce, but all of them are simultaneously involved in a network of factors that at the same time affect the health and development opportunities of youths ²⁰.

Malnutrition during childhood is also a wide discussed health factor and its short- and long-term implications seem to be clear ²¹. Consumption of animal source foods in children has been associated with a reduced likelihood of stunting ²². In our context, as expected, malnutrition appears to be associated with lower competitive academic achievement as well as having lower self-esteem during adulthood ²³. Our results showed an important reduction in every score analysed among youth who eat meet or dairy products less frequently.

The biggest reduction, however, was associated to those who belong to a minority ethnic group. About 20 points less are expected in those who self-reported belonging to an ethnicity. Long-term barriers to health and well-being have been previously identified everywhere regarding minorities ²⁴, and this finding endorses the fact that this population faces not only difficulties related to their health and social exclusion, but also long-term disadvantages due to reduced academic options and fewer opportunities to reach college education, which, as is known, shall redound in poor health and life quality. Deep health inequities are expectable in minority ethnic groups all over the world, most severe in developing countries, and with several negative consequences, including, but not limited to, unequal access to screening and health services, poor working conditions, higher levels of forced migration and violence ²⁵, all of which are profoundly aggravated by educational limitations.

About rurality, our results showed only slight differences (0.01 to 0.06 points) in the scores, but a negative correlation exists among rurality proportion of population and the global score and specific for every domain. Likewise, rural areas scores are probably affected due that the figures only take into account the students that did not dropout school. Dropout rates have reported to be higher in rural contexts ²⁶.

Some limitations need to be declared to put in context the findings. First, the developed analyses supposes that state exams measure all Colombian youth, but the records are not relevant figures on people who did not finish secondary education. In the same way, a reliable level of population coverage of the exams is not offered in the official records. In the country, for 2019, the school dropout rate of the official sector was of 3.2 % ²⁷. Low academic performance has been described as one of the main predictors of school dropout ²⁸. Second, measures including student's age and about the institutions facilities, infrastructure, violence conditions, number or teachers per student, and many others at institution level are not available for this analysis and could affect our conclusions, nonetheless, we considered that these variables are indirectly related to the results. According to previous authors, disadvantaged and minority children are much more concentrated in highly poor schools ²⁹.

However, our results demonstrate the importance of widely recognized social determinants in explaining educational inequalities, and of the local educational context in determining the possibility of educational attainment of students from lower socioeconomic status or with minority backgrounds. Actions to manage those inequities, both within schools and in wider society, are urgently required to help address these educational disparities, which, in turn, may affect the adolescents in their own present health and their subsequent health in adult life. Despite the fact that pandemic had no effects on academic achievement among the youth in this study, some containment strategies, such as lockdowns and school closures, exacerbated health, social, economic, and academic inequalities ³⁰, therefore, it is possible that in younger children the effects will be seen in the future. It is also plausible that the effects among adolescents would be observed afterwards, in technical, technological or university studies, as well as in work as adults. Further studies with these student cohorts will be essential to understand the non-health problems of the pandemic, and define the appropriate actions to mitigate the effect ♠