Contents - Previous - Next

This is the old United Nations University website. Visit the new site at http://unu.edu

Controlling for socio-economic status

No two households will maintain exactly the same per capita income. Moreover, income and wealth are notoriously difficult to measure accurately. Assumptions underlying the concept of positive deviance are that (1) the level of income is more or less the same in very poor neighbourhoods; (2) fairly simple indicators, such as land ownership, housing, and visible possessions can be used to identify families living at approximately the same level of poverty; and (3) additional statistical adjustments made for income can more or less remove its confounding effects.

The studies classified as true positive-deviance research in table 2 did find some variability in family wealth, although they tried to compare homogeneous groups. Morley and associates (1968) concluded, for example, that the fathers of the well nourished were "better farmers" than those of the malnourished. Such variability may be critical to positive deviance when additional small amounts of income are earned by an enterprising mother and used to feed her child.

The point to be made here is that positive-deviance studies must make the attempt to control for socio-economic status. The Kanawati and McLaren (1973) study of failure-to-thrive in Lebanon, for example, was not included in table 2 because large socioeconomic differences were found between the malnourished and well-nourished groups. Given the difficulties of measuring income, total household expenditure is often used as a proxy variable for income and can be analysed by subcategory.

Limiting the number of covariables: restriction by age and topic

As noted many times in this document, an extremely broad range of covariables contribute to child growth amidst poverty. A large number of these interrelate in ways that make it impossible to omit them from consideration without biasing the results of research. Therefore, it is necessary to seek ways to confine research studies in positive deviance to manageable proportions. Studying a narrow age-group is one method of restricting variables that also makes sense from a child-development perspective. Young children change so rapidly that the environmental and behavioural factors that contribute to positive deviance at one developmental stage may differ vastly from those that are important earlier or later. Seven-month-olds have very different needs and characteristics from two-year olds.

Suggested age ranges are: pre-pregnancy; pregnancy; neonates (0 to 1 month); first three months; 4 months until the child starts to crawl; the crawling stage, with its high exposure to dirt from the ground; first walking to start of nutritional recovery (15 to 21 months depending on population): and the early recovery period (21 to 30 months). The exact range should be specified to fit the research question and the location.

Restricting the main topics of investigation to one (or two) item(s), such as breast feeding, is a necessity, although it doesn't remove the need to measure or control for other related factors, such as supplementary food intake, etc.

As with all research on living subjects, it is necessary to ignore factors having effects that are relatively independent of the main relationships being tested by the research study. It is never possible to study or control for all variables.

Rationale for existing behaviours and social structures

Behaviours that appear to have negative effects on the growth and health of young children may serve other purposes related to the survival of the household, or may be caused by constraints that cannot be changed, or by historical developmental disorders within the family that do not respond to short-term programme approaches. Behavioural and social factors that contribute to positive deviance must be tested for transferability before they can be incorporated into programme design. Behavioural trial methods developed by social-marketing and communications professionals are suitable for tests of behaviour-change messages recommended as part of stage 2 in our research model.

Timeline for change

Some factors contributing to positive deviance go back to the childhood of the parents or to earlier circumstances in the history of the parents or to earlier circumstances in the history of the family or the subculture. Cross-sectional research approaches should not ignore these historic factors. Different types of change also require different lengths of time to achieve. Programmatic trials should not assume that attitudes and behaviours that prove resistant in the short run will not yield to change over longer periods.

Nutrition and infection

Because nutrition and infectious illness are so closely related, the resistance of the mother-child dyed to infection. the degree to which the environment exposes the dyed to infection, and the mother and other caretaker's preventive and curative health related behaviours may greatly influence the child's growth status and vice versa.

Aspects of Health Research that are Necessary for Research in Positive Deviance in Nutrition

The mother's overall health-seeking behaviours- how soon she goes for treatment, where, in what order, how persistently she follows treatment instructions, whether administering treatment herself, for the prescribed period, etc. - are topics for research on positive deviance in nutrition. Child-feeding during illness and convalescence belongs to this research domain.

Scrimshaw and Scrimshaw (1980) found, for example, that the mothers in their positivedeviant families persisted in seeking health care for their infants until the child had recovered. They also were more likely to take care of the child personally when he was sick, staying home from income-earning activities in order to do so.

Morbidity cannot be ignored in nutritional studies. Similarly nutritional status cannot be disregarded in morbidity studies. At the very least, the effects of morbidity must be controlled or adjusted for the studies of growth outcome.

The micro-behaviours and environmental factors associated with the transmission and cure of specific diseases fall outside the domain of this paper. A positive-deviance approach should definitely be taken to disease-specific studies of infant and young child morbidity. However' the growth of the child will be a secondary outcome variable in such studies. The primary outcome should be the incidence and duration of the particular infectious illness that is under scrutiny. It is possible and desirable to conduct combined studies of both nutrition and morbidity outcomes. Information for the design of the morbidity component of such studies must be obtained from appropriate medical researchers and epidemiologists.

Checkclist of Nutrition-Infection Relationships

What follows is a checklist of nutrition-infection relationships that are important for the design of research in positive deviance. For the sake of brevity, this checklist is not referenced.

• Nutrition-infection interactions start in utero. Sexually transmitted infections of the placenta and chorionic membranes reduce nutrient transfer and lead to premature birth. Other maternal infections must also be considered.
• Breast-milk of well-nourished mothers has been found to contain more immune factors than that of the malnourished.
• Breast-feeding contributes to resistance to infection in a large number of physiologically complicated ways.
• Protein-energy malnutrition impairs numerous aspects of the immune system. Cellmediated immunity is most severely affected.
• Iron deficiency and vitamin A deficiency reduce immunity function.
• Resistance to infection increases greatly with age. Malnourished children aged two years plus have greater immunity than younger better-nourished children. Well nourished exclusively breast-fed infants below four to six months can be expected to be relatively healthy.
• Infant and young-child morbidity is highly seasonal. Respiratory infections tend to peak in cool dry seasons, and diarrhoeal infections in hot and/or wet ones. Different regions have distinctive seasonal patterns. Infants selected for comparison in positive-deviance studies should be matched both for age and season of the year in which the data are collected.
• Most studies have found that the incidence of infection does not differ significantly between the well-nourished and the malnourished, but the severity and duration are greater for the malnourished.
• Diarrhoeal infection is much more strongly associated with poor growth than respiratory or other types of infection.
• Measles, pertussis (whooping cough), and tuberculosis may precipitate malnutrition.
• The effect of parasites on growth status depends on the degree of infestation, with light worm loads having almost no effect, but heavy worm loads being quite serious, for example.
• Age of exposure determines the virulence of some infections. Measles and pertussis are both believed to cause higher rates of malnutrition and mortality in Africa than in the Asian subcontinent, for example, because age of incidence is significantly lower in Africa.
• Children lose their appetite and almost invariably lose weight when they are sick. Exceptions to this rule are very minor colds, and skin and eye infections.
• During recovery the child's appetite increases and its catch-up growth rate may be up to 19 times the reference rate, depending on its degree of depletion and the nutrient density and quantity of the food it receives. Protein requirements increase during recovery because the stress of illness catabolizes protein stores.
• The rate and duration of catch-up growth during recovery from illness is critical to the child's long-term growth rate. For this reason, positive-deviance studies may wish to weigh infants frequently and monitor their diets during illness and convalescence.
• Morbidity recall for a longer period than the two weeks prior to interview is not considered reliable. One week recall is better. Daily checks are most accurate, although frequent visits to the home produce a Hawthorne effect. Health history questions regarding infants should go back to pregnancy and birth.
• If morbidity data are collected from the mother's recall, more educated mothers will give answers that differ not only in the type of illness entity reported but in the amount of ill-health they report. Educated urban samples may appear to be more sick than traditional rural ones because the educated mothers report more minor problems.
• Perceived illness entities differ greatly with degree of modernization and ethnic tradition. Therefore, the list of symptoms on a morbidity questionnaire must be thoroughly pretested to assure that it elicits consistent answers from different subgroups in the sample. Mothers should not be asked open-ended questions of the sort: "What health problems did your child have last week?" Rather, a list of commonly understood symptoms should be read to them and they should be asked if the child had any of the symptoms on the list.
• Physicians do not usually make the best data collectors or morbidity or anthropometry data. The doctor's training is focused on evaluating the state of the whole individual with a view to prescribing treatment. Data collection, by contrast, requires a level of precision and attention to minute detail that is irrelevant in clinical diagnosis and curative care. This difference is a source of misunderstanding that tends to lead to high levels of recording error. Data collection also is the tedious and repetitive type of work best assigned to lower-level but welltrained and standardized technicians. Physicians are indispensable, however, for diagnosis of unusual conditions and for emergency treatment of severely ill infants.
• Mortality rates are difficult to relate to nutritional status. In most cases official figures cannot be trusted to apply to a given study population. The most common source of error stems from failure to report either the birth or death of children who die shortly after they are born at home.
• Very large samples are needed to estimate accurately birth- and death-rates, although reproductive-history questions can be used to approximate these rates.

Management of multidisciplinary teams

The different disciplines relating to positive deviance have different vocabularies, theoretical assumptions, and standards for quality control. Therefore, when persons representing these disciplines attempt to work "in committee" the negative aspects of "committee" efforts (slowness, lack of agreement, etc.) are greatly exaggerated. For these reasons it is desirable to have strong leadership for such teams. A strong principal investigator (a single person representing one discipline) should personally integrate the contributions from each discipline into a single study design. Alternately, a strong project director or contract manager with a technical background may provide sufficient leadership to ensure that principal investigators from different disciplines integrate their work effectively.

Contents - Previous - Next