Contents - Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
Assessment of the impact of a hygiene on environmental sanitation, childhood diarrhoea, and the growth of children in rural Bangladesh
Nasar U. Ahmed, Marian F. Zeitlin, Alexa S. Beiser, Charles M. Super, Stanley N. Gershoff, and Manira A. Ahmed
A community-based hygiene intervention was developed and implemented in five villages of lowland Bangladesh with the active participation of members of the target group, with the objective of reducing childhood diarrhoea by altering ground sanitation and personal and food hygiene practices such as the washing of hands with ash before handling food and after defecation-related activities, cutting fingernails, removing faeces from the child's body and from the yard, using tube-well water for preparing baby food, and reducing supplementary feeding contamination by proper cleaning of bottles or avoiding bottle-feeding.
The project area, typical of Bangladesh, was selected because of its pour hygiene and sanitation conditions and its high rates of diarrhoea and malnutrition. Households with children 0-18 months old in five contiguous villages were targeted for the intervention. Households with children in the same age range in a comparison (control) site selected for observational study without intervention were exposed to about the same amount of contact with the researchers.
Baseline surveys of the subset of households with children 9-18 months old were conducted at the control site in July 1985 and at the intervention site in September. The intervention activities were carried out from January to July 1986. A final survey was conducted at both sites in August 1986, using the same questionnaire as for the baseline survey and the children who were then in the same age range, 9-18 months.
Both sites had higher cleanliness scores, lower diarrhoeal morbidity, and better growth status at the end of the study period, but the improvement was greater at the intervention site. The effect at the control site may be attributed to the intensive observation exposures, mothers' education, and socio-economic conditions of the households, whereas the intervention site effects were most likely due to the intervention activities.
For evaluation of the effect of interventions, the repeat cross-sectional survey may be adequate for measuring relatively stable outcomes such as knowledge and practices, as well as cumulative growth status, but inadequate for fluctuating morbidity.
One of the deadliest childhood diseases in the world is diarrhoea . Scrimshaw  and others established that it has a synergistic relationship with malnutrition. This vicious cycle results in an adverse effect on growth [3-6]. In Bangladesh, 90% of preschool children suffer from some degree of malnutrition , and, as in many other countries, diarrhoea is one of the most important causes of malnutrition  and child mortality  Poor hygiene and sanitation are major contributors to the diarrhoea .
Many hygiene interventions that attempted to reduce childhood diarrhoea failed to demonstrate any effect, mainly because they were culturally unsuitable and often developed without understanding the problem in the target community [11-13]. In this project, a positive-deviance research approach  was used to develop a community-based intervention. The researchers sought to identify local adaptive behaviours that could be modified by a trial process  and implemented as culturally acceptable and low-cost interventions in five villages in rural Bangladesh. Five similar villages were used as a control site for evaluating the intervention.
The purpose of the study was to assess the impact of this intervention on mothers' knowledge and behaviour, and on the diarrhoeal morbidity and nutrition status of children 918 months old as measured by rapid cross-sectional surveys carried out before and after the intervention. A six-month longitudinal study was conducted, and its analyses are presented in a separate report . The repeat cross-sectional evaluation permitted an investigation as to whether this relatively inexpensive technique measured the impact of interventions adequately, compared with more costly longitudinal methods.
The project was carried out at two rural sites, each consisting of five contiguous villages in Harirampur subdistrict, Manikgonj district, about 100 km northwest of Dhaka on the northern bank of the Padma River. This area, typical of rural Bangladesh [ 17]. was chosen because of its poor sanitary conditions and high diarrhoeal and malnutrition rates .
A census was conducted in five villages in October 1985 to recruit all available households with children under 19 months of age for an intervention, and 185 households (98%) were targeted. In the same month, a similar census targeted 200 households (97%) for structured observational studies at the control site. The intervention site was five kilometres away from the control site and was accessible by a two-hour boat ride most of the year, and by foot over a narrow path in about one and a half hours during the driest winter months.
More than half the households in these villages did not own cultivable land, since much farm land had been submerged by the river. Household heads (almost exclusively men) were primarily engaged in farming, small-scale trading, salaried jobs, fishing, and day labour. Most of the farmers had small landholdings, growing mainly deep-water rice, jute, and seasonal vegetables. The traders sold grocery items and agricultural produce such as rice, wheat, vegetables, fruits, and date-palm molasses. Mothers spent their days processing food, cooking, cleaning, and caring for children.
The traditional homestead consisted of one or more thatch, jute straw, or tin houses surrounding an earthen courtyard. Huts made of thatch or straw generally served as living quarters but sometimes also as kitchens, animal sheds, or grain-storage areas. The more desirable and expensive tin-roofed and tin-walled houses were mainly owned by the richest households.
Children were cared for in and around the courtyard by the mother, grandmother, and siblings. The earth surfaces of the house and yard were used for domestic work, for raising chickens, ducks, goats, and cattle, and for child care, including child toileting. Ground sanitation' personal cleanliness, and food hygiene were difficult year-round, with the additional problems of insufficient water in the hot dry season (April-May) and too much water in the flood season (July-September).
A community-based trial model, described in detail elsewhere  was used to develop interventions. Briefly, specific hygiene practices and sanitary conditions associated with diarrhoea were identified through baseline surveys, field observations, in-home problem diagnosis, and focus-group discussions and other methods recommended by Scrimshaw and Hurtado . These assessments showed that villagers believed supplementary feeding of infants, teething, evil eye, bad air, and spirits caused diarrhoea . Only 4% of the mothers at the intervention site had heard about germs. The connection between faecal contamination and diarrhoea was not recognized by most of the villagers. To develop an understanding of the occurrence of diarrhoea, the germ of theory of disease was taught to the participants.
One Tufts University doctoral candidate and two master's-level field supervisors worked with local project workers and the community to develop and test informational messages and teaching aids. A core of ten project workers were chosen from the community on the basis of at least ten years of education, ability, willingness to work in the field, and trustworthiness in the community.
Hygiene practices were proposed during working-group sessions and tested and revised through trial and practice at three levels, the first of which was in the homes of the ten project workers.
For the next level, the community was divided into five geographical blocks, each supervised by a project worker. From each block, five volunteers, who were themselves mothers targeted for the intervention, were selected using the following criteria: ability to articulate messages, willingness and family support for volunteering, and a friendly relationship with neighbours; it was prestigious to be a volunteer. Five workers taught the practices and supervised the trials at the second level in the homes of these 25 community volunteers.
After the volunteers' trials, the messages were modified again, and the volunteers taught the rest of the community. The teaching was interactive, by demonstration, emphasizing question-and-answer and discussion sessions. Volunteer training and community-level teaching were done at least twice a week in groups of up to five. For the sessions, the mothers chose their less busy hours, late morning and late afternoon.
For the third level, the working-group members visited five of the poorest households in each block to assess their ability to use the messages. The final messages were established on the basis of this assessment.
The three trial levels thus addressed the needs and understanding of the different groups in the community, including the poorest. Successful behavioural advice was composed into simple, feasible, direct, and motivating verbal messages created to resemble locally popular proverbs, poems, and folk songs. The intervention themes and messages, after final testing and revisions during the education campaign in the community as a whole, were as follows, presented in the order in which they were developed and integrated into community practice.
» Ground sanitation themekeeping babies from touching and eating disease-causing matter on the dirt surface of the compound:
» Personal hygiene themereducing the transmission of germs from defecation and other related activities:
» Food hygiene themereducing the transmission of germs during bottle-feeding and supplementary feeding:
Evaluation of the intervention
While the intervention targeted all available households with children 0-18 months old, the age range 9-18 months was chosen for the repeat cross-sectional evaluation. The crawling and toddler stages of development were considered crucial in terms of hygiene and sanitation practices, diarrhoeal morbidity, and malnutrition; and, at the crawling stage, infants' exploratory behaviours lead them to touch and taste faecal matter on contaminated surfaces 
The same 9-18-month age range was used in bath the baseline survey in July and September 1985 and the final survey in August 1986. The 11 to 13 months between the two therefore meant that the children in the first sample were excluded from the final sample; four completely independent groups were thus obtained. The mothers of almost all the children in this age range participated in the surveys. There were 111 households at the intervention site in the 1985 survey and 90 in the 1986 survey; at the control site the sample sizes were 96 in 1985 and 78 in 1986.
The field workers who took part in the longitudinal research or intervention at the two sites switched locations to administer the final survey. In addition, personnel who had implemented the intervention were not involved in monitoring the outcomes of the intervention.
The same questionnaire was used for both surveys. Information was recorded on the socioeconomic and demographic characteristics of the households as well as on basic practices and beliefs with regard to hygiene, sanitation, and infant feeding. The interviewers asked each mother to indicate the two locations where she most commonly placed the infant on the ground to play while she worked. The interviewers inspected the sanitary conditions of these areas and recorded the presence of faecal matter, spoiled food, garbage, other dirty things, dust, and wetness.
Information was also collected on diarrhoea among the children on the day of the interview and over the preceding two weeks. The children were weighed by trained anthropometry workers, using a beam-balance scale, and the weights recorded to the nearest 0.1 kg. The measurements were standardized by the Zerfas method .
The socio-economic scale used in the analyses was adapted from a previous scale developed by researchers at the Institute of Nutrition and Food Science, University of Dhaka . The scores for the components of the scale were as follows: housing materials and structuresall tin = 8, four-sided tin roof and bamboo walls = 6, two-sided tin roof and bamboo walls = 4, one-sided tin roof and bamboo walls = 2, all thatch-straw=1; radio ownership=3; tube-well ownership = 8; and mother's meals per daythree meals = 5, less than three meals = 0.
The sum of the mother's and father's years of schooling was used to measure education. The amount of land owned by the family was used as a measure of economic conditions.
Each mother was asked whether or not the following might cause disease if eaten by a baby: animal faeces, garbage, flies, baby faeces, adult faeces, spoiled food, and fresh food fallen onto the ground. The mother's knowledge of hygiene and sanitation was measured by the sum of her responses (incorrect = 0, correct = 1).
Hygiene practices and sanitary conditions
A sanitation scale was formed from the sum of the responses and observations of the following variables, with response categories ranked from low to high on the basis of their likelihood of preventing faecal contamination:
Two other variables to address hygiene and sanitary conditions were the baby's contact with faeces within the previous two weeks, based on the mother's recall (no contact = 0, physical contact = 1, physical and oral contact = 2), and the dryness of the play area as observed on the day of survey (all mud = 0, moist earth and some mud = 1, some moist earth and some mud=2, all moist earth = 3, some dry and some moist earth = 4, all dry earth = 5).
The dryness of the baby's play areas was one of the indicators of the ground-sanitation condition of a household because muddy areas trap contaminants and are not easily swept clean. Micro-organisms thrive in a moist environment, and moist or muddy soil is likely to adhere to crawling infants. Accumulated faeces, urine, and moist kitchen garbage also can make an area wet and muddy. The baseline survey results showed a negative association between the dryness of play areas and the prevalence of diarrhoea in the children. In the intervention campaign, the cleanliness and dryness of the play area were emphasized.
The point prevalence of diarrhoea and its prevalence over the preceding two weeks were measured by the mother's recall, as recommended in the WHO rapid assessment manual. A composite variable representing the diarrhoeal history of the child was also constructed by summing the responses to the following questions, to which mothers responded according to their own interpretation and recall:
A respiratory-infection variable consisted of the sum of the responses to the following queries:
Cold/cough and fever conditions were sought separately in the questionnaire. Analysis showed fever had a strong association with cold/cough but no significant association with diarrhoea. We created a composite variable including cold/cough and fever to reflect acute respiratory infection.
Weight-for-age Z scores (WAZ), used as a dependent variable in the regression analyses, were calculated using the WHO  and US National Center for Health Statistics 2 standard for each age and sex.
Internal consistency of the scales
The internal consistencies of the composite variables/ scales were confirmed using the SPSS reliability program  with Cronbach's alpha, yielding the following alpha values: socio-economic scale .60, mother's knowledge .74, sanitation scale .66, diarrhoea scale .61, and respiratory scale .52.
The objective of the analyses was to investigate the comparability of the intervention and control sites at the baseline (1985) and the difference between them at the end of the project (1986), and to assess the changes within each site between the two surveys. The analyses were performed using the SPSS-X package . Possible confounding variables, such as the baby's age and sex, the parents' education, the amount of land owned by the family, and the family's socio-economic status, were taken into account in the analysis.
All categorical variables were cross-tabulated to contrast the four samples. Pearson chi-square and Fisher's exact test were again performed to contrast the 1985 intervention and control samples (the baseline difference), the 1986 intervention and control samples (the final difference), and the difference between 1985 and 1986 at each site. As in multiple pairwise comparisons, it is necessary to control for type I error to reduce the chance of falsely finding significant difference in any pair. We used the Bonferroni adjustment for significance level  to address this problem in comparisons for several proportions.
All the socio-economic characteristics measured as interval or ratio variables were compared by ANOVA  using a multiple comparison test, the Student-Newman-Keuls procedure , to identify significant differences in mean values among the four independent samples.
A series of multiple regression analyses explored and identified the contributions of different independent variables to the outcome variables. Both 1985 samples were combined to represent the condition of the population before the intervention, and the 1986 samples were introduced as dummy variables to distinguish placebo effects from intervention effects in the following models:
Contents - Previous - Next