Consideration of the Nutrition Components of the Sick Child Initiative (1995)

The SCI algorithm is divided into three parts: assessment, classification, and treatment. Information collected in the first two parts is used as a diagnostic screen that then determines the treatment. A major innovation of the SCI is the development of an approach to create diagnostic screens across diseases, including malnutrition.

Diagnostic screens need to be evaluated in terms of their sensitivity (coverage) and efficiency. In the literature available to the CIN, preliminary analyses have been performed to evaluate some of the screens for malnutrition. Examples include the sensitivity and specificity of pallor to diagnose anemia and of low weight-for-age to diagnose malnutrition, based on the assumption that low weight-for-height (wasting) is the true indicator of the prevalence of malnutrition. In this section, comments are offered on: the need for analyses that measure the quality of the nutrition screens; a more appropriate method for comparing diagnostic screening tools according to these qualities; and the need to consider different outcomes to test the validity of these screens.

Diagnostic screens are customarily described by their sensitivity and specificity. In the context of the SCI's nutrition algorithms, sensitivity is the proportion of truly malnourished individuals who are properly defined as malnourished according to the screen, and is the coverage of the screen to identify those who are truly malnourished. Specificity is the proportion of truly well-nourished individuals who are correctly defined as well-nourished according to the screen. The quality of the screen is determined by its sensitivity and specificity.

It is also useful to determine the positive predictive value (PPV) of the screens, which is the proportion of those diagnosed as malnourished who are truly malnourished. The PPV depends not only on the sensitivity and specificity of the screen, but also on the prevalence of the malnourished. It follows that 1 minus the positive predictive value is the proportion of individuals who will be wrongly diagnosed as malnourished and wrongly treated for malnutrition. Therefore it is the positive predictive value of a screen, and not its sensitivity and specificity, that determines its efficiency. Analyses of efficiency will need to be made after the highest quality screens have been selected, and should include cost-benefit analyses that consider the prevalence of malnutrition and the resource constraints across settings.

A common way to compare screens for their ability to diagnose illness is to compare the differences in prevalences of the truly ill (for example, those who will die if untreated) or by comparing the differences in the relative risks for true illness (for example, the ratio of deaths of those screening positive over those screening negative). These approaches have been used by the IMCI—to evaluate the validity of severe wasting and edema for predicting mortality, for example. Both of these avenues of comparison are directly related to the positive predictive value, however, and therefore will change across populations even when the sensitivity-specificity characteristics are constant. For this reason, the more appropriate way to compare diagnostic screening tools is the Receiver Operating Characteristics (ROC) method. ROC compares the relationship of sensitivity to specificity, which is much more constant across populations than the positive predictive value and indexes derived from it (such as, relative risks) (Green and Swets, 1966).

The ROC method was originally developed as a dichotomous screen (Green and Swets, 1966), such as the health worker's perception of wasting to pick up life-threatening marasmus. Its use requires data from many health workers in order to have a range of sensitivity and specificity. The sensitivity and

specificity for each worker's assessment of a group of individuals are plotted as Z scores. This methodology has been adapted to compare indicators of nutritional status (for example, anthropometry) and health that are continuous variables, where ROC is plotted for the sensitivity-specificity at each value of the individuals measured (Brownie et al., 1986). The resultant curves are then compared across screens to identify the best. This must be done by first examining the ROCs for parallelism between the screens to be compared in the Z score transformations, rather than plotting ROC curves in the original percentage values. Only ROC curves that do not cross can be ranked. Rankings are then compared across settings to assess their reliability. In summary, it is the committee's opinion that the development of the nutrition screens requires that data from more health workers be used to develop ROC curves to compare the screens in different settings, and that Z score transformations be used when plotting these curves.

Correctly defining the outcome is crucial to the validity of the this approach to developing appropriate risk indicators. For this purpose, the best outcome is that which is to be prevented (for example, death). This was used in the IMCI validation to test the “severe wasting” indicator for screening. Sometimes the outcomes are too costly or time-consuming to measure, so that surrogate “gold standards ” are used. Thus weight-for-height, an anthropometric proxy for wasting, has been used in the SCI as the outcome for defining the sensitivity-specificity characteristics of the weight-for-age screen to diagnose mild to moderate malnutrition. The “gold standard” proxy is only as good as its relationship to the outcome to be prevented. A potential problem noted by the CIN is that in most populations with mild to moderate malnutrition, death as the outcome to be prevented is probably better proxied by active stunting (inadequate growth in stature) than by the wasting (which is a more useful indicator of severe malnutrition) used in the SCI. This conclusion is based on the recent insight that case fatality rates for common illnesses increase with decreasing weight-for-age (Pelletier et al., 1993) in populations that do not demonstrate significant wasting (Victora, 1992). This strongly implies that present wasting may be a poor proxy for present growth stunting. If so, the validation studies that WHO's IMCI program have done on the screens for moderate malnutrition may be valid for parts of the world where low weight-for-height is prevalent, but not where stunting occurs without low weight-for-height, as is the case in most of the world. In the latter case, active stunting, rather than wasting, may need to be the “gold standard” for validating the screen.

Rather than using risk screens for a deleterious outcome such as death, the most useful screen would properly identify children who would benefit (Habicht and Pelletier, 1990) from the nutrition and other interventions proposed. The prerequisites to investigate this question are: a defined intervention and a defined response, application of the intervention, and measurement of the response. The response is the difference in outcome (in mortality, for example) between those who receive the intervention and those who do not. With this “benefit approach,” it would be possible to relate the sensitivity-specificity characteristics of the diagnostic screens to the benefits (including nutritional status) of the integrated management of childhood illnesses. This should be the long-term goal of the SCI. In the meantime, WHO's IMCI program should continue with its current approach of developing screens for risk.

Research to develop indicators that predict benefit requires a research program that goes beyond the present time frame and resources of WHO's IMCI program. It is important, however, that the IMCI program and the larger SCI community mobilize the research community to address these issues in relation to the SCI diagnostic screens for malnutrition, so that they are ultimately validated relative to the SCI interventions. These interventions would also simultaneously serve to validate the SCI interventions themselves, which is an ultimate necessity.

The SCI's management of childhood diseases, including undernutrition, can be divided conceptually into three major areas:

• The diagnosis and treatment of the sick child with severe malnutrition or severe anemia.

• The diagnosis and treatment of the sick child with moderate malnutrition or moderate anemia.

• The nutritional management of the sick child who is not malnourished.

This section of the report is structured around the nutrition components of the algorithms that are to be used for the diagnosis and treatment of children in these three categories. Responses incorporate answers to the questions that USAID asked the committee to address.

Children are classified as having severe malnutrition or severe anemia if they have any of the following symptoms: visible severe wasting, severe pallor, clouding of the cornea, or edema of both feet.

The use of clinical signs of severe wasting is innovative and appropriate for detecting severely marasmic children. As discussed in the previous chapter, WHO has begun the necessary task of investigating the validity of these indicators. The results are very encouraging, although, as suggested, different methods could be used to test the specificity and sensitivity of the screens. The performance of these indicators is likely to be dependent on the expertise of the observer, so that studies at other sites will be important to confirm these results, which so far reflect the performance in the hands of only one user. In addition, training materials to systematize the recognition of visible severe wasting will need to be developed (C. Bern, CDC, Atlanta, personal communication, 1995).

There is a long history of the association between edema and risk of mortality, which has been confirmed in early tests of the validity of edema as a screen in the IMCI. Because many malnourished children present with kwashiorkor rather than, or in addition to, marasmus, inclusion of this screen is appropriate.

Severe pallor is proposed as a simple, innovative indicator of severe anemia that avoids the need for laboratory assessment. To date, testing of the anemia algorithms has focused on the sensitivity and specificity of pallor for the detection of severe anemia (J. Tulloch, CDR, WHO, Geneva, personal communication, 1995). Field performance of indicators for the clinical diagnosis of severe anemia have been fairly extensively evaluated in malarious areas, including western Kenya, The Gambia, and Uganda, where the prevalence of severe anemia (hemoglobin < 8 g/dL) reaches 50 percent (C. Bern, CDC, Atlanta, personal communication, 1995). In Siaya, Kenya, health workers assessed pallor at various anatomic sites and demonstrated that palmar pallor was 90 percent sensitive and 63 percent specific for the detection of moderate

anemia (hemoglobin < 8 g/dL), while severe palmar pallor was 60 percent sensitive and 98 percent specific to detect severe anemia (hemoglobin < 5g/dL) (C. Bern, CDC, Atlanta, personal communication, 1995). Of the anatomic sites tested, palmar pallor had the best combination of specificity and sensitivity for detecting severe and moderate anemia.

While the use of palmar pallor to detect severe anemia is innovative and should be encouraged, in the IMCI studies to determine specificity and sensitivity, only one or two clinic personnel made the evaluation (J. Tulloch, CDR, WHO, Geneva, personal communication, 1995). In future evaluations, at least ten individuals at each site should provide specificity and sensitivity estimates, so that the ROC methodology can be used appropriately.

Clouding of the cornea indicates severe vitamin A deficiency and is an appropriate diagnostic tool.

The IMCI algorithms direct the clinic health worker to give vitamin A to children in this category in areas where vitamin A deficiency is prevalent, and to refer them urgently to hospital. The recommendation to provide vitamin A in the clinic seems appropriate. The recommendation to simply refer such children elsewhere, however, assumes that primary care facilities are available, and that these facilities will have the information and resources to provide the appropriate nutritional treatment, as well as to manage the illness. Feeding assessment and the provision of appropriate nutrition advice for these children also become the responsibility of the hospitals. Thus, the feasibility of these recommendations needs to be systematically evaluated during the adaptation and early implementation phases of the SCI.

Although severe malnutrition is a major concern because it can be immediately life-threatening, the majority of mortality associated with malnutrition occurs among children with moderate malnutrition because of its higher prevalence (Pelletier et al., 1993). Thus, it is most important to correctly identify

and treat moderately malnourished children in order to have a significant impact on mortality rates.

Children are placed in this category if they have low weight-for-age, foamy patches on the white of the eye, or pallor.

Initial visit. Diagnosis of the moderately malnourished child is more problematic than that of the severely malnourished child. Most moderately malnourished children are not wasted (Beaton et al., 1990; Victora, 1992), so that wasting is not useful in determining which of these children need nutritional management.

Anthropometric indicators for the detection of marginal malnutrition must be developed while considering the pattern and timing of growth failure in developing countries. Malnutrition during the first few (4–6) months of life generally manifests itself as an earlier and more pronounced deficit in length growth rather than in weight growth (Calloway et al., 1988; Ruel et al., 1995).This means that length-for-age may be a better outcome than weight-for-length for validating the screen to detect children who are undernourished in early infancy, and that failure in length growth is likely to reflect current undernutrition at this age.

Between about 4–6 and 18 months of age, children in poor regions of the world are likely to suffer both poor weight gain and poor linear growth, brought about by a combination of inadequate nutrient intake and infections (Tomkins and Watson, 1989). This will be reflected in low weight-for-age and low length-for-age, but not in low weight-for-height in most regions of the world, unless the child is severely malnourished. Because deficits in length and weight are cumulative over this period, as the child becomes older, low height-or weight-for age are progressively more likely to reflect past malnutrition than the current feeding situation. In nonwasted populations, after about 18–22 months of age, growth rates become more similar to those of well-nourished children, although the child usually remains lighter and shorter than the international reference values (Allen, 1993; Beaton et al., 1990).

One implication of this progression is that a different threshold for low weight- or height-for-age might be considered at different ages. Consideration should be given to developing a growth chart with a reference curve based on varying cutoffs depending on the age of the child. If the appropriateness of different cutoffs is not carefully considered and tested, this could lead to inappropriate and inefficient use of resources, given that the IMCI recommends feeding assessment, advice, and follow-up for every child whose weight falls below a given threshold. Intervention will most benefit the child who is currently growing poorly, and may not change the outcome for older children

who have low weight-for-age because of previous poor growth (stunting) (Beaton et al., 1990). The current plan is to determine actual thresholds during the in-country adaptation phase of the SCI. This is an area that needs more thought and research by the SCI community (WHO, in press).

According to the nutrition algorithms, the follow-up visit for a child who meets the anthropometric criteria for poor growth on the initial visit is to be scheduled at 14 days. This is appropriate for enabling the health worker to check whether the caretaker was able to implement the recommendations, to provide support to continue correct practices, and to reorient the caretaker in cases of difficulty or failure. Fourteen days, however, is too short for growth to be determined accurately, and may lead to an incorrect assessment of improvement. In addition, since no new intervention is recommended at 14 days, there is no concrete incentive for a mother to return.

Alternative indicators are to be addressed in the “Adaptation Guide,” including length-for-age, weight-for-height, mid-upper-arm circumference, and growth faltering. Each of these indicators has advantages and disadvantages compared with weight-for-age.

Mid-upper-arm circumference (MUAC) is more likely to identify younger children than older children as malnourished, when used with a fixed cutoff. It identifies different children as malnourished from those identified either by weight-for-age or weight-for-height. Also, measurement errors in MUAC may be more likely to shift a child from the “not malnourished” category to the “malnourished” category (or vice versa) than for other measurements, because the margin for error is smaller.

While this indicator is appropriate for diagnosing vitamin A deficiency, the committee suggests adding “won't open eyes” and a question about nightblindness history to the classification of nutritional status. The committee also recommends that children who “won't open their eyes” or “whose eyes are pussy” be included under measles, requiring referral, in addition to clouding of the cornea. It should be made clear that the child's eyes must be forced open if necessary, because health personnel are often reluctant to do this.

Further attention needs to be paid to the problem of diagnosing less severe anemia (hemoglobin 8–11 g/dL) using pallor. The IMCI has tested the sensitivity and specificity of palmar pallor for diagnosing moderate anemia in only one nonmalarious site in Bangladesh (J. Tulloch, CDR, WHO, Geneva, personal communication, 1995). Hemoglobin concentrations below 8 g/dL are relatively rare in countries where malaria is absent, whereas concentrations in the range of 8–11 g/dL are very common in most areas of the world (ACC/SCN, 1991). Concentrations in this range, however, are well-established as having detrimental consequences for the cognitive performance and other functions of the child (Scrimshaw, 1991). Whether or not clinical signs are useful to detect moderate anemia remains to be seen.

Assessment of feeding problems. The caretakers of all infants ages 1 week to 2 months are asked questions about whether there is any difficulty feeding; whether the child is breastfed, and if so, how many times a day; and whether the infant usually receives any other foods or drinks, and if so, how often. An assessment of breastfeeding is then recommended in the algorithm if the infant has any difficulty feeding, is feeding less than six times a day, has a low weight-for-age, or is taking any other food or drinks, but has no indications for urgent referral to hospital.

The part of the nutrition algorithm concerned with assessment of the feeding of older children contains three sections: usual breastfeeding practices, usual consumption of other foods and fluids, and any modifications of these feeding practices during the current illness. The questions on breastfeeding refer to the frequency of daytime breastfeeding and any occurrence of nighttime breastfeeding. The questions on other foods and fluids solicit information on the kinds of items, their frequency of administration, and the use of feeding bottles; for low weight-for-age children, information is also sought about the portion size, who feeds the child, and the methods of feeding.

Nutrition counseling. The feeding recommendations encompassed within the algorithm are divided into five age categories: birth to 4 months of age, 4–6 months, 6–12 months, 12–24 months, and older than 24 months. Exclusive breastfeeding at least six times daily is suggested for the youngest age group, and introduction of complementary foods is recommended for those 4 to 6 months, depending on the presence of specific indicators of apparent need.

These indicators included poor weight gain (not defined), hungry appearance after breastfeeding, and demonstration of interest in semisolid foods. During the next age period, continued breastfeeding on demand is recommended, along with at least three to five adequate servings daily, depending on breastfeeding status, of locally available foods rich in nutrients and energy. The recommendations specify the appropriate order of feeding: first, breast milk; then semisolids; and finally other fluids.

During the period from 12–24 months, continued breastfeeding is recommended, as well as five adequate servings of family foods or other recommended foods. The suggested order of serving semisolids and breastmilk is reversed from that stated for the preceding age group. For the oldest age group (>24 months), three servings daily of family foods and two nutritious snacks are recommended. A footnote providing examples of the kinds of foods that might be offered for the periods from 6 to 24 months suggests an energy-rich food (possibly a thick cereal with added oil), meat, fish, eggs or pulses, and fruits and vegetables.

It will be necessary to develop appropriate age-specific feeding recommendations (considering issues of food availability, cultural beliefs, and prevailing nutritional problems) for each country or subregion within the country before successful implementation of the nutritional treatment algorithm will be possible at new locales. The IMCI proposes to do this within the local adaptation activities. Once these recommendations are developed, the assessment of child feeding within the context of the SCI visit can be simplified to examine the extent to which the locale-specific recommendations are already being used for an individual child, and to reinforce recommended practices.

Infants found to have low weight-for-age between 1 week and 2 months of age will be targeted for a feeding assessment. Low weight-for-age during the first few weeks after birth, however, may be caused primarily by low birth weight. A substantial proportion of infants may therefore be assessed unnecessarily for breastfeeding problems, a possibility that will need to be investigated. A 1-week-old infant who weighs less than at birth may have feeding problems, whereas a small infant who has gained weight probably does not. Thus, weight in relation to birthweight may be a better indicator of feeding problems during this period; information on birth weight, however, is often unavailable.

The committee felt that more thought needs to be given to the section on “Assess Breastfeeding” of the young infant. As stated, breastfeeding will not be assessed for infants who have fed in the previous hour, regardless of identified

low weight or feeding problems. If the infant has not fed for at least an hour, there is to be 4-minute period when breastfeeding is observed. Assessment of attachment probably needs to be done within a few days after birth—if it is not adequate at this time, the infant will be at serious risk of malnutrition and dehydration. Even making the assessment at 1 week postpartum may be too late. The assessment would require careful training of personnel because it is not easy to do well. The committee also questions what will be done if poor attachment is identified and how successful any intervention is likely to be.

The questions about breastfeeding will, appropriately, provide information on whether the child is currently breastfed during the day and at night. It is not certain, however, whether all mothers will be able to report their frequency of nursing. Because the feeding recommendations specify that the child should be fed as often as he or she wants, it would be desirable to obtain information on the signals used to initiate a feed (mother's preference or child's demand). Finally, mothers should be asked whether they are aware of any problems with breastfeeding.

The questions on other foods may need to be reformulated in each locale. If, for example, there are common local preparations that are suitable for young children, mothers might be questioned specifically about their knowledge and use of these foods. It is not clear how information on serving size will be obtained, especially when the children do not receive their own servings and when there are multiple caregivers. Each of these questions must be developed during the locale-specific adaptation phase. Because the committee was unconvinced of the importance of the order of feeding breast milk and semisolids, the lack of questions pertaining to serving order was not viewed as problematic.

The committee felt that several aspects of the feeding recommendations are unnecessarily complicated. In particular, it should be possible to reduce the number of age groups presented, and it may be possible to eliminate the recommendations regarding the order of individual foods served. At the same time, much more effort should be devoted to defining what is meant by a “good daily diet” in a given setting. This will require adaptation of the algorithm in each country, and a process must be developed to prepare country-specific recommendations based on the general guidelines.

The emphasis on local adaptation of the Food Box is strongly endorsed by the CIN. In some countries, existing governmental and nongovernmental nutrition programs have already acquired valuable information on child feeding practices, feeding problems, and feasible solutions to some of these problems that could be incorporated into the modified feeding recommendations. Nevertheless, the CIN believes that the available information needed to serve as the basis for feeding recommendations will be poor in many countries and regions. It is unrealistic to expect that resources would necessarily be available to fill these information gaps. Therefore, the CIN strongly suggests that the mechanisms for generating this information, and the resources to do so, be clearly identified and developed prior to full in-country implementation of the SCI.

Little information is available on the relationships between dietary intake and functional outcomes in children beyond two years of age. A number of studies suggest, however, that the growth velocity of children of this age in developing countries approximates that of children in more affluent nations (Allen, 1993). Therefore, specific feeding recommendations may no longer be necessary for children of this age range in developing countries when local data suggest adequate growth and nutritional status. The committee agrees that the suggested provision of two nutritious snacks in addition to three meals of family foods a day seems acceptable and is unlikely to produce any harm. Only recommendations that are important for child nutrition and health should be articulated, however, to avoid any unnecessary expenditure of health worker time for counseling on nonessential matters. In addition, it may be better to give intensive age-specific advice to mothers of children in the age group when growth faltering tends to occur, rather than attempt to target those with anthropometric deficits. It will be important to be able to tap into resources for feeding advice and support already operating in a community, or to develop them. This may be especially possible in, for example, Latin America (as compared to many parts of Africa). The counseling and support provided by the IMCI algorithms should be carefully linked to that provided by other programs to ensure complementary services.

It appears that the component of the algorithm that will be most difficult to implement will be the country-specific recommendations for special foods for infants during the period from 6 to 12 months. The committee recommends that general guidelines be prepared first regarding ideal characteristics of the special foods and a process to allow translation of these guidelines into specific recommendations in a given setting. In particular, quantitative guidelines are needed on the recommended energy density, feeding frequency, and nutrient content of special foods for this age group. The committee was not provided with information on these issues and felt that a great deal of additional thought

will need to be devoted to them. The current recommendations on the kinds of foods that comprise a good daily diet are too vague to be very useful in planning adequate diets. In many instances there may be several ways to increase the quantity or quality of complementary foods that are equally adequate from a nutritional viewpoint, but not equally feasible or acceptable from the viewpoint of the family. In order to increase the likelihood that families will be able to follow the feeding recommendations given during the counseling session, it is critical that the recommendations be sound from both a biomedical and a social science perspective.

The optimal age for introduction of complementary foods is controversial. While some recent evidence suggests that these foods may not be necessary before 6 months of age (Cohen et al., 1994), more information is needed on this question, especially for populations with large numbers of low-birth-weight infants or undernourished mothers. If these recent observations are confirmed, separate feeding recommendations for infants from 4 to 6 months of age may become superfluous.

Although it is recognized that individual children may require complementary foods at different ages, the committee recommends, from an operational perspective, that the age at which additional foods should be introduced be stated. Moreover, the currently proposed indicators of apparent need for complementary foods were of great concern to the committee for two reasons. First, the definition of inadequate weight gain needs to be specified. This is a major challenge, especially given the current debate about the appropriate reference data to use for breastfed infants. Second, the appearance of hunger after breastfeeding may easily be misinterpreted. It is conceivable that infants who are uncomfortable after a feed for any of a number of possible reasons will be assumed—erroneously —to be receiving an inadequate amount of breastmilk. Thus, if advice on introduction of complementary foods is to be individualized, better indicators must be developed.

The committee was not aware of sufficiently compelling scientific evidence to warrant specification of the order of breastmilk, foods, and other liquids in the course of a single feeding episode. These recommendations, as currently stated, may be especially difficult to communicate to caregivers because of the different suggested orders of breastfeeds and semisolids before and after the age of 12 months. Unless additional evidence is forthcoming, the committee recommends that the advice be simplified to the following: other foods and liquids should be offered at least three times a day, after or between breastfeeds. The committee was uncertain about the specific age at which children could consume customary family foods in place of specially prepared infant foods. This may vary in individual countries, depending on the types of family foods available and prevailing cultural beliefs. Until further information is available, the recommended transition at 12 months seems reasonable.

Milk should be mentioned in the footnote that describes a good daily diet. The recommendation to add oil to thick cereals should be made with caution, because this practice can easily dilute the concentration of protein and micronutrients in the food, and consequently the amounts of these nutrients consumed by the infant. This is an area of current research that needs to be encouraged.

The CIN feels that it is important in both the training protocols and the adaptation manuals to explicitly recognize the role of family members other than the mother in transmitting and implementing feeding advice. It is not always the mother who brings the child into the health facility. The caretaker accompanying the child will depend on such factors as the work patterns of the mother, cultural norms concerning travel outside the village and interaction with authorities; and perceived severity of the child's condition (both at the initial visit and at follow-up visits). While it may be convenient (as is currently the case) to frame the algorithm in terms of the mother/child dyad, the CIN recommends considering substituting “mother/caretaker ” for “mother” in the IMCI charts.

In addition to asking the mother/caretaker about current feeding practices and whether or not the child's feeding has changed during the illness, the committee also recommends that she be asked what she perceives as the main problem (if any) in feeding this child. Not only would such a question draw the attention of the health worker to constrained resources (such as time, food, and money) where they are as much or more of a barrier to optimal feeding as lack of information about correct feeding practices, but it would also convey to the mother/caretaker that her analysis of the feeding problem is considered valuable. In some cases, when the health worker understands what the mother/caretaker views as the most important feeding problem, together they can come up with a practical solution. In addition, however, a considerable body of research and programmatic experience suggests that when mothers are made to feel that their viewpoint is respected and considered valuable by health workers, they are more likely to follow the advice being given and to return for future visits to the health facility.

One component of the nutrition algorithm is to advise the mother/caretaker of infants or children diagnosed as having moderate malnutrition or feeding problems to return with the child in 14 days. A further follow-up visit after another 14 days is recommended if the child is not judged by the health worker to be doing well nutritionally at the first follow-up visit. The CIN feels that the practicality or feasibility of follow-up visits needs to be carefully assessed in terms of the distance from where families live to the primary level health facility; competing demands on the family member's time (particularly the time of the mother); and the benefits or additional services that the mother might

expect to receive on behalf of herself or the sick child at the time of the follow-up visit. The committee recognizes that the issue of the practicality of follow-up visits extends beyond the nutrition components of the algorithm, but because follow-up visits after 14 days are an important component of the nutrition advice given to the mother, the CIN wishes to raise the issue for consideration. In many settings over half of all children under 5 years of age are likely to be classified as malnourished or anemic (ACC/SCN, 1992), so it is probable that many follow-up visit recommendations will be triggered, thus making their practicality an appropriate concern.

The CIN would like to offer two suggestions concerning the issue of practicality of follow-up visits. The first is that further information should be obtained concerning the proportion of recommended follow-up visits that actually occur, as well as any factors that are strongly associated with an increased or decreased likelihood of the sick child being brought in for a follow-up visit. There are several potential sources of such information. One is systematic evaluation of the past experience of first-level health facilities with follow-up visits for sick children. Another source of information would be stand-alone research on follow-up visits. A third source would be careful evaluation of the follow-up visit element of the IMCI as it is implemented in the field to test if the health worker can go to the family rather than expecting the mother or another family member to bring the child back to the facility. In addition, where something of value to the family will be obtained at the follow-up visit (for example, the growth monitoring programs are linked to take-home food rations), the visit is more likely to occur. Anthropological research concerning child survival interventions leads to the conclusion that if the mother/caretaker expects only to receive further advice, and perhaps to be evaluated negatively on the progress that her child is making, the likelihood of a follow-up visit is significantly diminished (Coreil et al., 1994).

While iron deficiency is the primary cause of anemia, it is generally accepted that another important cause of anemia, especially during pregnancy, is folate deficiency. For this reason, combined iron-folate supplements are available from WHO. The IMCI recommends not using supplements containing folate in malarious areas because of competition with antimalarials. Also, the effectiveness of folate supplements for reducing anemia in adult women has been shown in several countries to be negligible. Consideration should be given to the potential benefits for children of combining other nutrients in iron supplements rather than, or in addition to, folate. For example, retinol (Mejia and Chew, 1988; Muhilal et al., 1988; Suharno et al., 1993) and riboflavin

(Powers et al., 1985) supplements have produced significant improvements in anemia compared with iron alone.

It has long been recognized by WHO that the appearance of xerophthalmia in a population indicates a high prevalence of vitamin A deficiency, which may provoke the need to provide vitamin A to all children seen in the clinic and/or country-level programmatic interventions.

The algorithms suggest that vitamin A be given to all children with measles, with “severe malnutrition or severe anemia,” or “malnutrition or anemia.” The committee felt that these recommendations were appropriate.

Under the algorithms for Identifying Treatment of the child with malnutrition or anemia, the recommendation is to “Make sure child is receiving foods containing Vitamin A.” This recommendation needs to be reexamined, and perhaps altered to refer more specifically to animal products high in vitamin A (retinol). Plant foods do not contain preformed vitamin A, and absorption of carotenoids (pro-vitamin A) may be poor in the many regions of the world where fat intakes are low. Increasing intake of plant sources of the pro-vitamin may therefore be less effective unless some oil is added to the local diet. The addition of oil, however, needs to be done with caution, as described under our comments on appropriate complementary foods for infants.

In most cases, few, if any, modifications of currently recommended feeding practices for infants and young preschool children are required during illness. Thus, the major concerns of the clinic-based health worker when examining a sick child should be to assure that the child 's usual feeding practices are appropriate for age, these practices are continued to the extent possible during illness, and that any necessary illness-related modifications are discussed with the child 's caregiver. When these feeding practices (either before or during illness) are inappropriate, the reasons should be explored to identify and resolve any constraints to implementation of the suggested feeding recommendations.

No specific recommendations are provided on feeding children during acute diarrhea. As indicated above, in most cases no modifications of the usual diet will be necessary. In settings where infants less than 6 months of age are fed

exclusively with nonhuman milks, however, caution is indicated when the diarrhea is severe and dehydration is present. Mothers should be advised to observe the child closely for worsening diarrhea, which may require that a local cereal or tuber be added to the milk drink to reduce the amount of milk (lactose) that needs to be consumed. The need for micronutrient supplementation is currently under study, and additional recommendations may be forthcoming in the near future. Mothers should also be warned of the possibility of reduced child appetite during diarrhea and the possible ways to manage this problem.

The recommendations for persistent diarrhea indicate that milk feeds should be modified to encourage more frequent breastfeeding of longer duration and the partial replacement of nonhuman milk consumption with fermented milk products or semisolid food. It is unlikely that women can actively increase the duration of breastfeeding unless they have been terminating feeds before the child was satisfied. It would be preferable to state that the infant should be allowed to terminate each nursing episode. The recommendation regarding nonhuman milks seems appropriate for children older than 6 months, but it may be difficult to implement for younger infants who have not had experience with nonliquid foods. Again, specific advice on micronutrient supplements may be appropriate once more information is available from ongoing clinical trials. Currently the IMCI does not recommend vitamin A supplementation in persistent diarrhea. Nevertheless, because vitamin A supplements do reduce the severity of diarrhea and diarrhea-associated mortality (Beaton et al., 1992), the rationale for this is unclear to the CIN.

No specific recommendations are provided on the dietary management of patients with dysentery. Some studies suggest that excessive losses of protein and other nutrients in fecal blood may necessitate supplementation with these nutrients. More research will be needed in this area before new recommendations can be implemented.

The nutritional recommendation for severe, complicated measles; complicated measles; and uncomplicated measles is to treat with vitamin A. This recommendation is based on a recent meta-analysis of community vitamin A supplementation trials, which shows that vitamin A reduced mortality and severe complications after the onset of measles (Beaton et al., 1992). Thus, the

recommendation to supplement with vitamin A is justifiable if there is a high prevalence of vitamin A deficiency or of measles fatality in the community, which must be decided prior to implementation of the SCI. Mothers of children with severe measles and uncomplicated measles are asked to return to the clinic with the child in 5 days if the child still has a fever; if there is no fever at 5 days, they are asked to return to the clinic in 14 days for a full assessment —including possible malnutrition—and for advice on how to feed her child based on the Food Box. We assume that the provision of feeding advice after a measles episode is thought to be useful here because malnutrition is a common complication of (that is, caused by) measles. The committee, however, questions whether it is necessary to provide the mother of a well-nourished child with information on how to feed her child after measles, especially if there are no complications.