Community-based approaches to combating malnutrition and poor education among girls in resource-poor settings : report of a large scale intervention in Pakistan

Introduction: Malnutrition and low levels of education continue to be major problems in many developing countries, especially for female children. Methods: In Pakistan, a large-scale school lunch program was implemented in 29 of the poorest rural districts through a publicprivate partnership. The project provided freshly prepared meals in 4035 government primary girls’ schools over a 2 year period. The primary strategy was empowerment of women in the community who volunteered to plan the meals, purchase the food, and cook and serve the meals. The project collected data from growth monitoring, attendance records, preand post-intervention community based surveys, focus group discussions, and the use of other ethnographic methods. A study on changes in the levels of


Introduction
In resource poor settings, malnutrition and low levels of education are still major problems, especially for girls.A recent national survey in Pakistan revealed that in rural areas of Sindh-Pakistan, 18.3% girls and 14.6% boys were stunted 1 .In 2005, the literacy rate among females aged 15 years or over in Pakistan was 36% 2 .Gross primary enrolment for 2005 was 77% among girls, compared with 94% among boys 2 .In rural settings only 22% of girls aged over 10 years have completed primary level or higher schooling, compared with 47% boys 3 .Low literacy and educational attainment is a complex problem associated with the low social status of women, and little perceived benefit from educating girls.
The literature suggests that malnutrition among those 5-12 years is high in Pakistan 1,4 .The consequences of malnutrition and poor literacy rates for women and girls in resource-poor settings have been well documented.
Childhood malnutrition and infectious diseases not only account for a major proportion of mortality 5 , but also for two-thirds of the disease burden 6 .Malnutrition and the resultant lowered immunity are the major underlying factors in approximately half of the 740 000 child deaths annually, worldwide 7 .The protein-energy, malnutrition-infection complex is a strong predictor of functional impairment in a wide range of biological, behavioural and social dimensions among children and adults, especially those in developing countries 8,9 .Developmental impairments, the long-term sequel of malnutrition 10 , include diminished learning ability in school 11 , a high risk of mortality among women during childbirth 12 , and of poor pregnancy outcomes, such as intrauterine growth retardation and infant mortality 13 .There is also evidence that malnutrition in childhood impairs the capacity for physical work later in adulthood 14 .In Pakistan, malnutrition is not simply a personal health hazard, but also a serious impediment to national development.
In the face of these challenges the Government of Pakistan implemented a large scale intervention to combat malnutrition and increase school enrolment among primary school girls, based on a successful pilot 15 .The program conducted baseline training in 4383 schools that catered for 439 852 children in 29 of the poorest districts in the country.
However, feeding was conducted in only 4035 schools for 417 665 girls.The majority of schools where feeding was not initiated were included late in the project.While communities were mobilized to send girls to schools, feeding could not be initiated as funds were not released by the government.
The major implementation strategy was to empower women in villages near the project schools.Rural women volunteered to serve on school committees that managed the

Pakistan
Pakistan is situated in the north-western part of South Asia.
The state consists of four provinces (Fig1): Punjab, Sindh, North West Frontier Province (NWFP), and Balochistan; and some federal units including Northern Areas, and Azad, Jammu and Kashmir (AJK).
Pakistan is the world's seventh most populous country and has one of the highest population growth rates.The population density of the country is estimated to be 166 persons per square kilometre; and 45% of the population is under 15 years of age 16 .The country's economy is mainly agrarian, with agriculture contributing 22% to the gross domestic product 17 ..Community women who volunteered were trained to manage the feeding program, which included managing funds, purchasing goods, planning meals, preparing the meals and serving the food.A substantial amount of training was provided to these volunteers to ensure the quality of the meals, as described in the Final Report of the Tawana Pakistan Program 19 .The average cost of a meal was between 6.07 and 7.14 Pakistani Rupees, approximately US10 cents.

Intervention
The specific foods chosen and the actual menus were determined by the community and monitored by the supervising and training NGOs.The volunteers were trained to plan balanced meals which included one food from each of the three food groups: for energy carbohydrates (grains); for protection fruits/vegetables; and for growth protein (meat, dairy or pulses).The project carefully monitored the menu composition and found improvement over the project period (more menus with the three food groups represented).
However other issues related to the quality (specific quantities of micronutrients) or quantity of the food served to each girl were not monitored.Based on average prices of foods and the average caloric needs of primary-school aged girls, it was determined in the planning phase that the girls would receive one-third of their daily caloric intake from the school lunch program.Volunteers were trained to serve equal portions of food to the girls and to follow specific program hygiene practices.from the median weight-for-height of the reference population were defined as wasted 20 .The changes in nutritional status were calculated by the number of girls for whom there was a change in the standardized nutritional status scores (z scores).
There were a total of 417 665 girls inducted into the program over the project period.Of these, 203 116 girls received at least two sets of body measurements at least 6 months apart.
The major reasons for this drop included late entry to the program; less than 6 months between measurements; and absenteeism on the days of the second measurement.There were no statistically significant differences in the levels of stunting and wasting between the two groups of girls.
Neither were there important differences in nutritional status at province/area or district level between the girls in the analytical group and the other group.It was concluded that the analytical sample was not biased in relation to a major analytical variable of the study, malnutrition.
Enrolment and anthropometric data: Enrolment was noted before and after the project on 4383 schools where COs were trained.Anthropometry was conducted when each girl was inducted into the program and then quarterly.The anthropometric analysis was conducted on 203116 girls in 3356 schools for which both baseline and post-Tawana data was available.
Percent increase in enrolment is calculated by subtracting the enrolment numbers in a school at the end of the project from the enrolment at the beginning, and dividing the difference by the enrolment before the intervention.

Pre-and post-intervention community surveys:
Consecutive cross-sectional household surveys were conducted at baseline between April and October 2003, and at the end of the project, between August and October 2005.
For the surveys, 5% of the estimated 5000 target project schools were selected, yielding 250 sample schools.Twenty households were selected in the communities adjacent to these selected schools, providing 5000 households.The same communities were sampled in the pre-and post-intervention surveys.
A three-stage sampling strategy was utilized.First, half of the districts (15 of 29) serviced by each of the NGOs were selected at random.Each district is divided into tehsils, and a 5% sample of the schools was selected in each tehsil.
Households in each community adjacent to the selected schools were numbered at the time of the survey, and 20 households were selected through systematic random sampling for interview for both the pre-and postintervention surveys.

Statistical methods
Differences in anthropometry and enrolment data at the baseline and at the end of the project were tested using McNemar's test.Differences between the two sample-based pre-and post-survey data were tested by a weighted 2 test.
All statistical testing was performed using SAS v 9.1 (SAS Institute Inc., Cary, NC, USA).Survey estimates and statistical testing were calculated using weights and a design effect estimated to reflect the complex sample design.

Results
The number of districts, schools and girls by area reached by the program are presented (Table 1).At baseline, 23.2% girls were stunted and 14.3% wasted.This baseline nutritional status may be more fully understood with ethnographic reports of hunger from these communities.Reports of hunger from the schools were common.
The intervention decreased malnutrition, increased enrolment and improved dietary knowledge in the communities.Over the study period, improvements were observed in the three measures of nutritional status (Fig2).
Wasting, or acute malnutrition, decreased by 45% *Number of schools in before-and-after analysis divided by total schools where feeding was conducted.†Number of girls in before-and-after analysis divided by the total girls in school where feeding was conducted.

Discussion
The project demonstrated impressive improvements in nutrition and school enrolment among primary-school girls.
The major limitation of the study is the lack of a control group to determine if other factors could explain the observed improvements.The authors believe that having meal at school can generate happiness in children, even if it does not have an effect on malnutrition.Therefore, depriving one group while providing food to another is an ethical issue that makes justifying a control arm in school interventions difficult.However, it was possible to compare the girls inducted early into the feeding program with those inducted late.It was found that both groups had similar levels of malnutrition.This provided an 'internal control' and supported the conclusion that the program was responsible for their improvement.In addition, there were no major changes in Pakistan's development indicators during the 2 years of the study that could be associated with observed changes in nutritional status in this large group of school girls.The fall in school attendance after the feeding program was suspended supports the link between the program and improvements in enrolment.
Another limitation of this analysis was the lack of data on one critical outcome: the learning achievements of the girls.
The project was not designed to evaluate learning.An independent evaluation of the project in one province (Sindh), which included control schools, documented improved learning among schools where girl students received feeding 22 .While improved learning was demonstrated, this should not be a criterion for judgment of the school feeding program because good nutrition is a fundamental human right.
The utility of this large intervention is that it permits evaluation of the capacity of a model to be used at national level.The scale and success of the intervention demonstrates that Pakistan has the human capital necessary to mount an effective program on a national scale.The improvements in nutritional status in this large cohort were achieved despite students in some schools receiving very few feedings, due to local resistance and administrative problems.A fully implemented school feeding program (with a longer implementation periods) may have demonstrated an even stronger impact.The results of this study cannot easily be compared with small-scale pilot studies that have addressed nutrition of primary-school children 23,24 .There are many models for school feeding, each with a different set of outcome measures.For example, a program that is not based on empowerment and nutritional messages would not be expected to make changes in community involvement and knowledge about nutrition.Balanced meal was defined as that which includes the three food groups essential for energy, growth, and protection from illnesses.Locally available food items were categorized into three groups for easy understanding by community women.

Conclusion
This study demonstrates the potential success and scalability of school feeding programs in Pakistan.The success of the program can be explained by the considerable synergies that are found when communities are encouraged to work across sectors (health, education and empowerment).The major challenges encountered in this large-scale project reflect this same intersectoral nature of the projects.
The intervention, the Tawana Pakistan Project (TPP), was funded by the Government of Pakistan.The project was implemented and managed by a consortium including: the Aga Khan University, Department of Community Health Sciences (AKU/CHS); the National Implementation Unit; and the Pakistan Bait-ul-Maal, the latter two being units of the Ministry of Social Welfare and Special Education.In a public-private partnership, AKU/CHS provided technical support for the design of the project, and trained and managed 11 non-governmental organizations (NGO) that supported local governments in project implementation.Comprehensive training was provided to 663 fieldworkers, 4383 community organizers (COs), 4336 school teachers, and 94 134 community women, many of whom volunteered in the School Tawana Committees (STC).The project was implemented in all four Pakistan provinces, Northern Areas and AJK between September 2002 and June 2005.Extensive administrative records were maintained relating to costs, volunteers, training and school committee meetings.The project provided freshly prepared, balanced school lunches made from locally available food items.Community development strategies were used to mobilize the local communities.The model used was similar to the 'Focusing Resources on Effective School Health' (FRESH) model suggested by WHO, UNESCO, UNICEF and the World Bank

Figure 1 :
Figure 1: Map of Pakistan showing the project's districts.
Survey teams were mounted by the NGOs and trained by AKU/CHS.Field interviewers and supervisors received 6 hours theoretical training and one day of field-based practical training.Semi-structured interviews were conducted with women, preferably the mothers of school children, in the selected households.The questionnaire included modules concerning nutrition, education and community participation.The participants were asked about the food consumed in their last lunch.The food items reported for each meal were coded into one of the three food groups, using standard nutritional criteria to determine whether or not meals were balanced.Focus group discussions: Focus group discussions (FGDs) were conducted in villages selected for the community survey.Each group consisted of 8-10 participants, including school committee members, mothers, local community women, 'lady health workers' and elected female members selected in a convenience sample.These FGDs collected perceptions about health, nutrition and the schools (teacher attendance, schools timings, discipline in the school and school hygiene).One trained field worker used a semistructured guide to lead each focus group.A second trained fieldworker took notes during the discussion and made verbatim transcription of key responses; the notes were coded by AKU staff.
Malnutrition is measured by three indicators: stunting, underweight, and wasting.Children below -2 standard deviations from median height/length-for-age World Health Organisation/National Center for Health Statistics (WHO/NCHS) reference population are stunted.Similarly, children below -2 standard deviations from median weight-for-age of the WHO/NCHS reference population are underweight.And children below -2 standard deviations from the median weight for height of the WHO/NCHS reference population are wasted.Pre-intervention anthropometric measurements were taken in the schools at the time of registering a girl in the program.Post-intervention is defined as a measurement closest to the end of the project and at least six months after the pre-intervention.Data source: project, years 2002/2003 and 2005.

Figure 2 :
Figure 2: Percentage girls malnourished at the baseline and end of the project (n = 203 116).
Percent increase was calculated by subtracting the enrolment at the end from the enrolment at the beginning, and dividing the difference by the enrolment at the beginning.Data source: project, years 2002 and 2005.
Data source: project household community surveys, years 2003 and 2005.

Table 1 : Districts, schools and beneficiaries included in the before-and-after analysis by province Provinces Districts Total schools Schools in before-and- after analysis Total schools* (%) Total girls Girls in before-and- after analysis
21lunteers.In contrast, Balochistan, where the improvements were less, has the lowest literacy level and the highest gender gap in the country21.AJK, Azad, Jammu and Kashmir; NWFP, North West Frontier Province.