The danger of interpreting anthropometric data out of context

Mark Myatt is a consultant epidemiologist and senior research fellow at the Institute of Ophthalmology. His areas of expertise are infectious disease, nutrition and survey design. He is currently working on a rapid assessment procedure for trachoma prevalence.

Gwenola Desplats is a nutritionist with an initial background in food science. After working in India and Bangladesh, she developed an interest in nutrition in complex emergencies and worked in DRC, Afghanistan and Ethiopia, with ACF, Concern, and Save the Children-US respectively. She is a PhD candidate at Tufts University and is undertaking her dissertation research in the area of Community Therapeutic Care.

Dr. Steve Collins is a Nutrition Consultant working for Valid International and has worked in many complex emergencies focusing on assessments and establishing emergency nutritional interventions and evaluations. His research interests include severe adult malnutrition and community therapeutic feeding. (Photos of Steve are welcome!)

The Concern programme in North Eastern Afghanistan (in non-Taliban held areas) has been operational since 1998 in two provinces Badakshan with a population of 842,702 and Takhar with a population of 883,910. In this area there was an estimated internally displaced population of around 100,000 people. The Concern programme focused on food security (food distribution and food for work projects), water/sanitation and educational infrastructure. The rationale for the survey reported on in this article developed in response to the influx of IDP's into the area due to the ongoing conflict between the Northern Alliance and the Taliban, and also the ongoing drought conditions, which resulted in crop failure, displacement and general food scarcity. It was felt that the nutritional status was poor and an in depth nutrition and food security survey was required to help direct future programming needs.

In August 2001 Concern Worldwide commissioned Valid International to conduct a nutrition, health, and food security assessment in Badakshan Province, NE Afghanistan. The survey used both quantitative and qualitative methods to estimate the prevalence of acute undernutrition in children and their mothers. The survey also provided information about agriculture, household food security/economy, coping mechanisms and information about patterns of morbidity and mortality¹.

Based on the survey findings, this article argues that, in the absence of context, the results of nutritional surveys are almost meaningless. Traditionally, children under five years of age are viewed as a sentinel group reflecting the nutritional status of the population. In the Khosh Valley of Badhakshan Province of NE Afghanistan we found a prevalence of acute wasting undernutrition of 11.5% (95% CI = 8.7%, 14.9%) in children under five using standard cut-off points for weight-for height Z-scores. Taken alone, these figures do not suggest impending or ongoing famine. For example, Table 1 compares our data with data from two nutritional surveys in Africa both of which were classified by the RNIS as grade III or IV (i.e. no pending or impending disaster).

In this article we describe the circumstances surrounding the headline prevalence figure and demonstrate that, when other factors are considered, the people of the Khosh Valley are on the verge of famine.

Choice of indicator

In addition to measuring children under five years of age, we assessed maternal undernutrition using a MUAC (< 21.5 cm) cut-off point corresponding to a BMI of below 16.0 kg/m2. which defines severe chronic adult undernutrition for women. We found 21.2% (95% CI = 16.0%, 27.1%) of mothers to be severely undernourished. Qualitative investigation revealed preferential feeding of children, with mothers reducing their own intake before reducing the intake of their children or husbands. This information was verified by interviewing a variety of groups using different methods (in-depth interviews with mothers and community leaders, focus group discussions with mothers and community leaders, 'natural' focus group discussions with men at mosques and guest house sites). The qualitative and quantitative results are consistent with each other indicating that, in this population, prevalence of acute undernutrition in the traditional sentinel group alone is not a good choice of indicator for community level nutritional stress.

Public health environment

Communicable diseases and undernutrition constitute the principal health problems in the area. Marked seasonal variations were described by community members and health staff. These are summarised in Figure 1.

The public health situation in the Khosh Valley was poor. Diarrhoea and ARI were the most common causes of childhood mortality. Both preventative and curative services were limited in scope and inadequate. All villages complained of a lack of potable water. The water table had dropped considerably with recently constructed wells in the middle of the valley now drying up. Spring discharges were approximately half their usual levels. This had an effect on the quantity of water available for irrigation as well as for drinking and hygiene purposes. The lack of potable water was often cited as the cause of diarrhoea but faecal disposal methods were far from ideal and this, combined with a lack of water for washing, is probably as important a cause of diarrhoea as contaminated water.

Diarrhoea and fever were significantly associated with acute undernutrition. This finding agrees with other reports that cite diarrhoea as an important cause of undernutrition. Figure 2 illustrates this relationship using data from previous nutritional surveys conducted amongst non-displaced populations throughout Afghanistan between January 2000 and August 2001.

Crops

The rainfed winter and spring wheat crops have failed for the second consecutive year. The small quantity of grain harvested was of poor condition, contaminated with smut, and not of seed quality. Yields per unit of seed (i.e. the number of kilograms of grain harvested for each kilogram of seed planted) ranged between zero and 3.8 averaging 1.4. Yields per hectare ranged between zero and 416 kg / hectare averaging 133 kg / hectare. The irrigated winter and spring winter wheat harvests were also poor due to lack of irrigation water. Yields per unit of seed ranged between 1.7 and 17.5 averaging 5.5. Yields per hectare ranged between 292 kg / hectare and 2450 kg / hectare averaging 940 kg / hectare. The harvest of barley, the second commonest staple, was equally poor. The poor grain harvests over the previous two years are reflected in steadily rising cereal prices (Figure 3).

Potato and onion harvests appeared to have been reasonably good although farmers reported lower than average yields. Harvests of other crops were strongly dependent upon the area of irrigated land cultivated and the diversity of crops planted. Crop diversity was low with only a small minority of farmers planting vegetables such as carrots, tomatoes, aubergines, turnips, and pumpkins other than in small irrigated areas (kitchen gardens). Production from these areas will be consumed before the start of winter.

All farmers interviewed reported planting opium poppy. Yields were low, averaging 0.45 Kg per farming household. Combined with the collapse of labour markets, this will probably lead to an inability to compensate for production shortfalls using purchased food.

Figure 4 presents a calendar (May 2000 - July 2002) that provides an estimate of the proportion of food needs that will be met by stocks of staple foods. The percentage shown on the y-axis of each chart represents the percentage of farmers who report that their food stocks for a particular staple will be sufficient to meet their requirements during a particular month.

Livestock

Mark and Gwenola on the road with colleagues in Afghanistan

Most poorer farmers have already sold their livestock in order to compensate for production shortfalls. Markets operate as a cartel. Livestock prices were low and wheat prices high. Sale of livestock was associated with a reduction of dietary variety due to the loss of sources of meat, animal fat, eggs, and milk. Livestock are an important source of winter fuel and for maintaining soil fertility. Livestock sales have left people with small herds sufficient, on average, to provide only 230kg of cereal if sold at the local market. This quantity of grain would feed a family of eight for approximately 2 months. Fodder production (winter-feed) has been affected by lack of water and it is likely that a fodder shortage will occur during the winter months. This may seriously affect livestock with decreased milk yields and may further depress the livestock market reducing the viability of selling livestock as a coping strategy.

Fruit and wild foods

Fruit was available at the time of the survey and was an important source of both food and income. Some families gather wild cumin for sale to petty traders who sell it on at a small profit to traders in the market towns. Collection and consumption of wild grasses and vetches was also reported. Both fruit and wild food will not be available during the winter.

Labour

Wage labour (agricultural labour, mining, donkey driving, service) was a common source of income. Demand for labour is highest during the harvest of rainfed wheat and barley as well as during the earlier poppy harvest. Little work has been available. The supply of casual labour exceeds demand. Wages levels were insufficient to meet household food requirements. Child labour was a source of family support with children 'leased' to wealthy persons as servants in return for a ration. Migrant labour was common with young men migrating to Pakistan and Iran to work and remitting money to their families. Regular seasonal labour migration was also reported with male household members travelling to Pakistan to work in the autumn and returning in the spring. Returning workers often return with goods such as clothes and shoes and engage in petty trading.

Other coping strategies

Most people were reducing intake. Usual meal-times were kept but solid food was replaced by tea or milk. Reduction in dietary variety was a common consequence of the sale of livestock. Preferential feeding of children was consistently reported. Although unpopular with the men, both men and women reported that mothers preferred to reduce their own intake before reducing the intake of their children or husbands.

Borrowing has become increasingly common. The borrower buys food at between two or three times the market price that he agrees to repay in cash or grain at a later date, usually after the harvest. The loan is secured against property with irrigated land being preferred as security. Food is, therefore, purchased at a multiple of a peak market price that can only be paid for by sale or transfer of food immediately after harvest when food prices are likely to be low. The effective interest rate may be as high as 350% over a three month period. Defaulting on a loan usually leads to seizure of property with the borrower either becoming a sharecropper for the lender or being displaced. Loss of land was frequently cited as a reason for displacement with Pakistan and Iran being mentioned as the preferred destinations.

Sale of land was reported to be a last resort and was usually mentioned as a prelude to displacement. Land prices are low with irrigated land prices reported to be between fifty and sixty-five percent of 1998 land prices.

Most of the poorer households had already sold their household goods. The within-village market for household goods is exhausted and the prices offered for household assets in market towns were too low to allow this strategy to be anything but a stop-gap measure. It is unlikely, therefore, that this will be a viable or important coping strategy to the newly impoverished. Capital accumulation is both slow and an hereditary process and this strategy will lead to long term and continued vulnerability.

Reports of displacement were common. Displacement in Afghanistan is, however, substantially different from displacement in, for instance, Africa. It takes place before food and income sources are exhausted and is facilitated by a strong culture of hospitality to travellers. It is possible that displacement in the face of adverse economic circumstances is a common and longstanding coping strategy rather than a crisis strategy. At present it is unclear whether reports of displacement in areas away from the front-line represent crisis displacement or coping migration. It is also unclear what proportion of reports of displacement emanating from UNOs and NGOs refer to normal seasonal movements of Hazara and Kuchi peoples rather than to crisis displacement. More work is required to characterise displacement in the Afghan context. No UN or NGO interventions are targeted at nomadic groups. These groups may make up approximately 10% of the population and are known to be amongst the most economically disadvantaged people in the country.

The identified coping strategies may be viewed as lying along a continuum ranging from normal coping behaviours (e.g. cash crops, casual labour) through stress coping behaviours (e.g. borrowing, kinship support) to crisis behaviours (e.g. sale of highly portable valuables, displacement). Figure 7 and Table 2 show identified coping strategies, their status, and likely outcomes.

Food economy

We performed a basic food economy analysis converting sources of income and expenditure to cereal equivalents using a purposive sample of 39 farmers in the Khosh valley based around the WFP VAM methodology. Figure 8 presents four different analyses.

Figure 8 (A) presents the sources of food used by the sample population in 2000. It identifies a gap of 24% that the people filled using one or more of the coping strategies identified in Table 2.

Figure 8 (B) characterises the same sources of food for 2001. In this analysis, the un-specified deficit is 22% larger than was accommodated by the coping mechanisms used in 2000. The information from the focus group interviews presented in Table 2 indicates that in 2001 the population has less ability to cope and it is likely that many of these mechanisms are now exhausted. Even assuming that these coping mechanisms remain intact, the analysis indicates the farmers have an average food deficit that is at least 22% of their annual requirements (see Figure 8 (C)). Given the exhausted state of the coping mechanisms the actual deficit is likely to be higher.

Figure 8 (D) presents a similar analysis but with the addition of two crisis strategies. These are strategies that remove people's ability to live in subsequent years:

1. Any farmers who have remaining cows, sheep or goats sell them to buy grain.
2. Farmers eat their seed stocks.

If farmers adopt these crisis strategies, the unspecified food deficit drops to 12%. This is lower than the un-quantified coping mechanisms employed in 2000 (Figure 8 (A)), indicating that with the sale of most of their essential resources the majority of people will be able to survive the winter. After this, however, they will be incapable of surviving without interventions.

Figure 9 show cumulative frequency plots projecting the percentage of farmers who will no longer be able to support household food needs over time given four different scenarios.

The 'coping' line shows the percentage of farmers who will not be able to feed their families using the same coping strategies as they did in 2000. By January 2002 nearly 80% of farmers will have exhausted their capacity to cope. The 'crisis' line show the percentage of farmers who will not be able to feed their families even if they adopt the crisis strategy presented in Figure 8 (D). In this scenario approximately 30% will have exhausted their capacity to survive even with selling their remaining herds and eating their seed stocks before the end of January.

At the time of the assessment, it was planned to distribute only 341.5MT of wheat through a FFW scheme. The 'existing FFW' line shows the effect of this intervention demonstrating that even with this intervention, over 70% of those surveyed will not be able to find sufficient food for their families without selling vital assets. The 'intervention' line shows the effect of a relief distribution meeting 22% of the average annual food requirement. With this intervention, 50% of the families will be able to meet their food requirements up until the end of January.

Access and humanitarian space

The Khosh valley is a high altitude valley inaccessible to motorised transport from mid- November until February because of snow. At present, grain is transported via Tajikistan, taking at least 2 months from time of purchase to time of distribution. Even before the start of the war, this delay meant that it was unlikely that sufficient quantities of grain could be transported and distributed before the winter began.

Timing of survey

This survey took place two months after grain harvests at a time when household food stocks should have been at their maximum. Low levels of household food stocks at this time of year represent a far more serious situation than equivalent stores immediately prior to harvest. In addition, within three months the winter will isolate the valley and low temperatures will increase the need for food, fuel and clothing. Indications are that all of these will be in short supply.

Conclusions

When interpreting nutritional surveys, it is vital to consider context. In isolation, 11.5% global undernutrition does not indicate a serious nutritional situation. Using this figure alone, it would be difficult to convince a donor to fund immediate emergency nutritional interventions. In the Khosh Valley, by contrast, the contextual data we collected demonstrate that the coping capacity of the population has been exhausted and, in the absence of immediate intervention, they face famine.

The overheads for the collection of contextual data are not great but its collection should be planned in advance. In retrospect, we made mistakes in the sampling system that we used and this caused some difficulty in data analysis². In future we will collect food security data using a larger random or systematic sample. Qualitative data is often seen as being somehow 'second rate'. This need not be the case if multiple data sources are used and triangulated with each other and if sampling to redundancy is used³. In contrast to other disciplines such as mental health, emergency nutrition guidelines covering the collection and analysis of qualitative data are over-standardised and proscribe a relatively limited number of data sources and methods of data collection. Often workers will resort to selecting various developmental techniques such as participatory rural appraisal (PRA) techniques which may not always be applicable or may require adaptation for the emergency context. We feel that more emphasis should be placed upon the collection and use of qualitative data during emergency nutritional assessments.

Note: This survey took place in August/September 2001 and was abruptly interrupted by the terrorist attacks in the US on September 11th, 2001. This report reflects activities and needs at the time of assessment only. For further information contact Mary Corbett at mary.corbett@concern.ie or contact the authors through steve@validinternational.org

¹Nutritional anthropometry, health, food security and agriculture assessment, Concern programme areas northeast Afghanistan, October 2001. Report compiled by Collins S, Myatt M and Desplats G. Available online at http://www.concern.net/

²For example, we used a simple sampling scheme whereby one 'rich', 'middle' and 'poor' farmer were selected from each village. This meant that the sample was small (39) leading to wide confidence limits and also that the rich and middle farmers were over-represented.We tried to get around this problem by weighting each observation during analysis (i.e. so that the data from poor farmers had more weight in the analysis).Weights were derived from secondary sources, i.e. VAM surveys, but we had no way of validating the weighting.

³Sampling to redundancy means that once no new information is being revealed and all potential sources of variation have been explored then sampling may cease. In other words all further sampling tells you what you already know and is therefore redundant.