Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

Malnutrition among women: Indicators to estimate prevalence

Kathleen M. Rasmussen and Jean-Pierre Habicht


The availability of data about the extent of maternal malnutrition and its consequences for women, their children, and the rest of the family varies. Data on the birth weight of the newborn are rather extensive; those on the well-being of the breast-feeding child are becoming more readily available. Data are almost completely lacking, however, on the consequences for the performance and survival of the mother herself throughout her life cycle, or for those who depend on her in the family or in society at large.

This paper contributes to both the short-term and long-term efforts to help women and the societies in which they live to improve women's nutrition by addressing technical considerations on four issues:

- What are the characteristics of the indicators of nutritional status that are used to ascertain the prevalence of malnutrition among women?
- How important for the roles women play are the kinds of malnutrition that have a high prevalence?
- What are the determinants of the types of malnutrition that impair performance?
- Given these determinants, what kinds of interventions are appropriate and possible?

The last two questions are considered in less depth than the first two. What we know about women's nutrition has been the subject of recent reviews [1; 2], and the implications of this knowledge for choosing appropriate interventions also have been studied [3; 4].


The prevalence of malnutrition among women

To determine the prevalence of malnutrition among women, it is necessary to estimate the number of women whose nutritional status is unsatisfactory. This is usually done by counting the number of women who fall below some agreed-upon cut-off point (a reference value) of an indicator of nutritional status.

Prevalence estimates from cut-off points

The use of cut-offs to estimate the prevalence of anaemia is a good illustration of how this counting is done. Panel A of figure 1 presents the total population distribution (solid line) of haemoglobin values. It is skewed (longer tail) toward lower haemoglobin levels as compared to a healthy population (dotted line). Most of these non-healthy persons are below the usual cut-off point (labelled X) for defining anaemia.

A number of healthy persons also are below the usual cut-off point, however, and a few who are not healthy have haemoglobin values higher than the cutoff point. The proportion below the cut-off point rises and falls as the prevalence of non-healthy persons rises and falls. The proportion is only equivalent to the prevalence of non-healthy women when the number of healthy women misclassified below the cut-off as non-healthy equals the number of non-healthy women misclassified above the cut-off as healthy. No cut-off point will solve this problem, because the relative number of these misclassifications varies with the prevalence itself. Nor is there any other practical solution to correct for these misclassifications [5]. The result is that, although one can rank populations according to their prevalence of anaemia with this method, one cannot estimate the prevalence of the non healthy population by using a cut-off point.

In cases like this, where the healthy normative distribution of the indicators is well known (which is not true of most indicators), another approach can be used to estimate the prevalences of the non-healthy [6]. This method cannot yet be used for many indicators of women's nutritional status, however, because the necessary data are not available. Thus, one priority should be to define the distribution of indicator values for healthy, well nourished (but not over nourished) women.

Choice of indicators

It would be wise to make this effort on the more promising indicators. To identify such indicators, one should first find the ones that are most useful and then select those that can be used ethically and are feasible. It is a mistake to prejudge ethics and feasibility until one has determined utility, because good but impractical indicators often lead to feasible adaptations once the usefulness of the indicator is demonstrated. In general, indicators of malnutrition are much less well understood and characterized for adults than for young children. Among women, more attention has been paid to indicators useful for assessing nutritional status and determining potential medical risk during pregnancy than for any other period of life. A major problem in this field is that there is little agreement about which indicators to use to assess women's nutritional status during most of the other periods.

An important initial consideration is the purpose for which a particular indicator is to be used, because different indicators are needed for different purposes. For example, as is discussed below, those used to estimate the prevalence of malnutrition may not be effective in screening women to target nutritional interventions. As another example, indicators to predict a need for service do not have the same characteristics as those that assess response to intervention. The list of potentially useful indicators that is developed below must be understood with this caveat in mind.

Potentially useful indicators

Concurrent reflective, and predictive indicators

Indicators of current nutritional status, which are in some way a measure of an existing pool of nutrients, and those that reflect past or predict future nutritional status, which in general are not a measure of a pool of nutrients, are available. A reflective (also called "lagged") indicator is usually an outcome that is not itself identical with the woman's nutritional status but rather is a product of her poor nutritional status. A predictive indicator may be able to be taken as a proxy of the present when it can be presumed to be stable (e.g., dietary intake, education, socio-economic status).

Reflective indicators are the irreversible results of malnutrition, such as death or stunted growth. Reflective indicators among survivors, such as stunted growth, give evidence about the proportion of women who were ever malnourished in childhood. but give no information about whether they are now malnourished. Differentiating between the prevalences of present malnutrition and of ever having had malnutrition is important because the prevalence of the latter is always greater than that of the former in populations where malnutrition is endemic.

Reflective indicators may themselves affect other indicators. For example, stunted mothers tend to have low-birth-weight babies. Birth weight is affected both by stunting during the mother's childhood and by malnutrition during pregnancy [7]. Therefore, alleviating malnutrition only during pregnancy will not completely reduce low birth weight. For a full reduction in low birth weight due to malnutrition, one must wait at least a generation for the effect to become apparent, because birth weight acts, in part, as a reflective indicator of maternal malnutrition.

It is important to identify the kind of indicator being used in a particular situation. Even concurrent indicators, which are biologically the most direct, can be strongly affected by other, non-nutritional factors; for example, weight gain during pregnancy can be affected by medical problems of pregnancy that are unrelated to nutrition. Nonetheless, concurrent indicators are those most able to generalized to a variety of situations. Proximal predictive indicators, such as food intake, are more able than distal predictive indicators to be generalized, especially with respect to cut-off points. Similarly, when the prevalences of non nutritional influences are similar, these reflective indicators can be more generalized than when the prevalences are different. For instance, malaria is a potent confounder of birth weight as an indicator of nutritional status in areas where the disease is present but, of course, is not important where malaria is rare.

Some examples of potentially relevant indicators are provided in table 1, classified as predictive, concurrent, or reflective. This list provides several possibilities from which choices can be made, based on the intended use of the indicator and the period of a woman's life to which it is applied.

Normative, risk, benefit, and response indicators

Another categorization of potentially useful indicators characterizes them as normative [5], predictive or reflective of risk [8], or predictive of benefit [9]. With the normative approach, the non-healthy population is defined by exclusion relative to a distribution of values for a particular indicator among healthy women, as in panel A of figure 1 (see FIG. 1. Characteristics of an indicator that predicts benefits from an intervention.).

A. normative approach. X is the cut-off point below which an individual is considered to have an undesirable condition (e.g. anaemia).
B. risk approach. The shaded area indicates those whose performance (measured here as oxygen in take. or "aerobic power") is below their potential.
C. benefit approach. The shaded area indicates those who could benefit from an intervention. The proportion of the total population represented by this group may be different than in either A or B.

Normative data, however, may be poor at identifying women who are at risk, because they do not predict the likelihood that these women may have a bad outcome. This latter concern is what motivated the development of indicators of risk [10]. With the risk approach, one relates a range of cut-off points for a particular indicator to the probability of having some bad outcome. The best indicator is the one with the highest probability of correct classification across the range [5]. This approach requires that the badness of the outcome be clearly understood and that it be associated with poor nutritional status. Unfortunately, relatively few such indicators are available for women; those that have been used include low birth weight and maternal or neonatal mortality. In contrast to the risk approach, which tries to prediet bad outcomes, the benefit approach examines the effectiveness of an indicator in predicting the benefit from an intervention to improve nutritional status [9].

TABLE 1. Indicators of nutritional status that are potentially applicable to women during various periods of their lives


Value measured

Protein-energy malnutrition

Particular nutrients

Concurrent weight tissue and cellular biochemical indexes (level of substance, enzyme activity, etc.)
height (if subject is still growing)
weight for height or other index of body mass
arm circumference
skin thickness(es) haematologic indexes (haemoglobin, ferritin, mean cell volume. etc.)
arm muscle area
weight change during specific period (e.g.pregnancy)a
milk volume (?composition)a
infant growth during exclusive breast feeding
Reflective height (if subject has stopped growing)  
birth weightsa
duration of lactation
length of postpartum menorrhoea
age at menarchea
age at menopausea
length of fecund perioda
Predictive dietary intake socio - economic status dietary intake

a. Useful only for certain periods - e.g., age at menarche is relevant only after the subject has archieved menarche, birth weight only after a pregnancy.

These three approaches to the development of I indicators - normative, risk, and benefit are diagrammed using haemoglobin as an example of an indicator in figure 1. Note, however, that in panel A the "not healthy" are defined only relative to a normative population of healthy persons, not because they necessarily suffer or risk any ill health or can benefit from intervention.

Panel B presents the hypothetical distributions of haemoglobin values from a population according to whether or not an individual's oxygen uptake values (VO2 max) are satisfactory. Note that the prevalence of demonstrated risk is always lower than the prevalence of estimates from the normative approach.

Panel C presents the hypothetical distribution of haemoglobin values from a population according to whether or not individuals respond to iron therapy [11]. The prevalence of those who benefit is, in this case, larger than the estimate from the risk approach; in Sweden it is almost identical with the estimate from the normative approach. This is not so among black women in the United States; here the benefit approach identifies fewer women as having iron deficiency anaemia than the normative approach because black women suffer from anaemia for reasons other than iron deficiency. Thus, the normative, risk, and benefit approaches to defining nutritional status may result in different estimates of prevalence of malnutrition, and these differences may be different across populations.

Panel B also shows that haemoglobin is much better as an indicator for separating those at risk of impaired oxygen uptake from those not at risk than it is as a normative indicator or as an indicator of benefit. This means that indicators must be compared relative to the same approach, using the same tests or normality.

The use of the three approaches has been so poorly formulated in the past that it is not surprising that they have not been examined comparatively. Furthermore, the existence of such very different approaches to the development and use of indicators suggests that it is indeed important to determine the purpose for which a given indicator is to be used before applying it. In general, we favour using the benefit approach to estimate the prevalence of malnutrition in women and in screening women for nutritional interventions. We favour the risk approach for estimating the health consequences and economic costs of malnutrition, and for identifying indicators to predict bad outcomes where something can be done to mitigate the outcomes even though nothing can be done to prevent them. An example is the identification of malnourished women who are seen for the first time at parturition so as to screen for potential small-for-date babies.

Where does that leave us today in assessing whether malnutrition among women is an important problem? Unfortunately, from a practical point of view, we have only imperfect tools with which to address this issue. Even when a standard is agreed upon, as in the case of evaluating the iron status of girls and women by using haemoglobin values, prevalence estimates are likely to be only approximate. An example of this challenge is seen in the difficulty of assembling and interpreting the data on dietary intake and nutritional status from values for weight, height, and skinfold thickness provided by McGuire and Popkin [12].


What we know about indicators relevant to the nutritional status of women

We actually know very little about the assessment of nutritional status among non-pregnant, non-lactating women. Although reference normative data are available for biochemical indexes of nutritional status in this group, the anthropometric reference data are particularly difficult to extrapolate to developing countries. This is because the two databases that are available and have been used for this purpose (Metropolitan Life Tables [13] and NCHS data [14] on adult women) include very few individuals who are as short in stature as women living under poor circumstances. The ones in these reference populations who are the shortest are also, as a group, generally older and heavier than women of reproductive age.

Assessment of the nutritional status of pregnant women has been approached many ways, and several of the indicators during this period, such as gestation al weight gain, contribute to the identification of those who may be at risk for adverse pregnancy outcome. Although gestation al weight gain is commonly used in this way, its exact form (i.e., as a simple cut-off point or as a graph to evaluate pattern of weight gain) remains controversial. For example, although it is implicit in the construction of many weight-gain charts (e.g., that of Rosso [15]), it is not actually known whether the same or different patterns and amounts of weight gain should be expected of women of different physical size who live under different environmental circumstances.

Furthermore, it is not clear who should serve as the reference population. To date, the Metropolitan Life Insurance tables have most often been used [13; 15]. They are not representative of the current or past United States population, however, and also, as discussed above, are not representative of women of reproductive age. More appropriate data for this purpose are urgently needed. How best to use those presently available is the subject of deliberations now being held at the US National Academy of Sciences, the Centers for Disease Control, and the World Health Organization.

Unfortunately, the situation is much worse for lactating women. No established standard exists by which to declare an individual lactating woman to be malnourished [16]. No chart of expected weight loss (if, indeed, weight loss really occurs) that parallels the chart of expected weight gain for pregnant women has been developed. After the initial postpartum period during which much of the excess fluid gained in pregnancy is lost, it has been assumed that the an thropometric characteristics of lactating women are similar to those of non-pregnant, non-lactating women of the same age. Even if this assumption is correct, it must be understood that this approach suffers from the same inadequacies described above for non-pregnant, non-lactating women. Although there is extensive literature on the volume and composition of milk produced by women living under a variety of circumstances, these data have never been used to develop an indicator of maternal nutritional status for lactating women. Given the variability in these measures and the sensitivity of milk volume to infant factors (e.g., suckling intensity), creation of an indicator of maternal nutritional status using this approach would be a challenging task.


What we know about women's nutritional status from application of the indicators

From the data that are available about the differences between children and adults and between pregnant and lactating women and non-pregnant, non-lactating women, it is clear that these three groups of females (children, and pregnant and lactating women) require separate treatment. Less obvious is the rationale for subdividing non-pregnant, non-lactating women. This subgroup consists of women whose nutritional needs may differ and whose nutritional status has very different implications for future reproductive success. If a young woman's first conception occurs close to menarche, it is likely that she still is growing herself and will have greater nutritional needs than an older woman would. Inasmuch as the period before conception is an important one for interventions designed to improve maternal nutritional status and modify adverse life-style practices to prevent foetal malformations [17], the time of first conception deserves separate consideration. For many women, the choice of whether or not to breast-feed is made then, and this time becomes an important one for educational interventions focused on the choice of infant feeding method.

The period between one reproductive cycle and the next also should be considered separately. Results of animal studies [18] suggest that the interval between peak lactation and subsequent conception is important for renewing maternal resources. Some attention has been paid to this period in the demographic literature on birth spacing, in which the length of time between births can be thought of as a proxy for the potential for biological renewal. In these studies [19] birth spacing was an important determinant of infant mortality; however, it is not clear whether short intervals are associated with this outcome because maternal resources are not replenished or because short intervals are more likely to include premature infants who are small and have a greater risk of dying in the first year of life. Clearly, this period of a woman's life must be better understood biologically and represents a potentially important period for nutritional intervention [20].

The third subgroup of non-pregnant, non-lactating women, those who have experienced their last conception (called menopausal in the tables discussed below), probably have the lowest nutritional requirements of any group of women. It should be recognized, however, that, although the nutritional status of these women no longer has any implications for their future reproductive success, it is important for their future health. Interventions to prevent or ameliorate the debilitation associated with chronic diseases are important. In many areas of the world, older women have essential roles in their families and their communities that also are dependent on their health and nutritional status. At the same time, their ability to meet their nutritional needs depends on their health and social status in their community and household. For example, those who are valued by their families and who still have most of their teeth are more likely to be able to meet their nutritional needs than women who are not valued by their families and who have lost their teeth.

Application of these indicators to populations of women permits one to estimate the relative prevalences (the number of women positive for the indicator divided by the total number of women in the population) of malnutrition. A matrix for such data is shown as figure 2. As mentioned above, absolute prevalences cannot be estimated using cut-off points. The choice of nutritional problems to include is admittedly somewhat arbitrary but reflects those problems most likely to be prevalent and/or important. For any given geographical area, however, it is necessary to consider the possibility that other problems also may occur.

FIG. 2. Prevalence of various nutrient deficiencies (and excesses) during periods of a woman's life. Prevalence is rated on a scale from 0 (not prevalent) to ++++ (highly prevalent). An asterisk (*) indicates that insufficient information is available for assessment of prevalence.

(or excess)
Life period
Childhood Menarche
to first
Pregnancy Lactation Repletion Menopausal

The matrix in figure 2 is empty for two reasons: (1) few data are available on worldwide prevalences of nutritional deficiencies (or excesses), either for women as a group or (especially) for subgroups of women, and (2) the matrix must be completed for each community or geographical area about which programme or policy decisions are being made. This latter is important because a number of nutritional deficiencies, even when they are widespread (e.g., iodine-deficiency diseases), have distinct geographical patterns. It is also important because some nutritional deficiencies may have a low prevalence on a worldwide basis but be a significant problem when they occur; an example of this is beriberi from thiamine deficiency.

Even when data about young girls are not specifically available, some inferences about their nutritional status can be made from the more readily available information on children (except in countries in which there is discrimination against females).

FIG. 3. Biological and other productive roles of women that may be influenced by their nutritional status


Determinants of malnutrition that impair performance

Inasmuch as the prevalence and importance of nutritional deficiencies (or excesses) are likely to vary greatly among groups of women at different stages of their lives who live under vastly different circumstances, it is difficult to make general statements about the determinants of problems that might be judged important in a particular area. The only relevant area in which information is sufficient to make judgements of this sort is for the determinants of low birth weight [21; 22]. Assuming that the factors that operate during pregnancy to produce a low-birth-weight infant may be common to other time periods in a woman's life, we can propose, as a conceptual framework, the following grouping of factors that are determinants of women's nutritional status:

- less proximal family and community factors: availability of public services, education, ethnic/cultural factors, economic factors, ecological factors;
- proximal behavioural and immediate environmental factors: reproductive practices, utilization of health services, maternal work/working conditions, adverse life-style practices, environmental pathology and oxygen;
- proximal biological-factors: past nutritional status, parity, infection, medical problems of reproduction, activity, dietary intake, toxicity.

It must be emphasized that the data necessary to test this assumption are not available; this framework thus can serve as a basis for developing hypotheses about factors that may be relevant and important. As is appropriate to the lack of supporting data, we have not attempted to specify connections or directions of influence between the groups or among the factors within each group.

This conceptual framework also can serve as a basis for making inferences about what kinds of interventions are likely to be successful and how they might work. For example, grouping the factors into those more proximal to and those more distal from the outcome of interest draws attention to the fact that an intervention aimed at improving education is likely to have more wide-ranging effects (because many of the more proximal factors may be affected) than one aimed at changing dietary intake (because only nutritional status is likely to be affected). It also must be emphasized that improving maternal nutritional status requires knowledge of the underlying biological mechanisms. Only when these mechanisms are understood will it be possible to predict which interventions are likely to be important. It is noteworthy, however, that predicting the success of a particular proposed intervention requires more than knowledge of the relevant biological factors because behavioural change also is required for the intervention to succeed.


Interventions to improve the nutritional status of women

As shown by the list of determinant factors above, many interventions are possible that could have a beneficial effect on the nutritional status of women. At some level, however, they all operate through the factors shown in figure 4 (see FIG. 4. Points of intervention between food that is available in the community and the nutritional status of the individual). Therefore, they are reasonable targets for interventions. How effective a particular intervention is likely to be, however, depends on many things. Important among these are some of the biological factors shown in figure 4 (e.g., nutrient losses, competing demands for nutrients) as well as some not shown (e.g., participation in and compliance with the intervention). These are particularly poorly understood and deserve additional research.



Women have different roles and different nutritional needs during various period of their lives, a fact that must be considered in designing inerventions. The information presented here makes it clear that interventions before the first or later conceptions have the potential for an important impact on subsequent reproductive success.

It would be desirable to select indicators of benefit rather than of risk or those based on normative values. They should (1) have high enough positive predictive values for benefit (not necessarily for risk) that an intervention is likely to be justified, (2) provide an estimate of the prevalence of need, and (3) provide an estimate of the potential and actual benefit of an intervention. Thus, one would avoid using an indicator (e.g., height) that shows that a population is needy when it may not be suitable for targeting (because the majority of women were stunted as children) or for assessisng the benefit of an intervention aimed at adult women (because height will not change in response to an intervention directed toward this age group). It also needs to be noted that normative standards cannot be used to predict the probability of benefit for individual women.

It would be desirable to use indicators of risk (1) to screen women when, even though it may be too late to prevent harm, it is still not too late to predict and treat bad outcomes (e.g., low birth weight) and (2) to provide an estimate of harm and economic costs.

The same indicators may be used for both purposes, even though the way in which they are used may be different. The development of such all purpose indicators may be possible inasmuch as many types of indicators are available, but most are not fully developed for use in the subgroups of women identified as important here. In that case, the choice of which one to use should be highly dependent on the circumstances under which it is to be employed.

To go beyond viewing reproduction as the sole or primary role of women, it is necessary to examine other outcomes that are important to women themselves and to society, and to ascertain whether or not they are influenced by nutritional status. A high priority area is the social role of women; this is important now and is likely to be more important with future socio-economic development. This view makes obvious the importance of interventions, such as education, that are aimed at girls and young women.



Highest priority should be given to developing standards using concurrent indicators and combining the normative, risk, and benefit approaches. This requires both theoretical and basic scientific work. Much of this work could and should be done in the context of action programmes to improve women's nutritional status. This scientific component will not be inexpensive because it must document reliably both the intervention and the response to it.

It is self-defeating to try to standardize cut-off points for indicators now because the necessary theoretical and practical work has not yet been done. A better investment would be to determine what kinds of indicators are needed. It would be wise to do this on the basis of a specific, reflective indicator such as weight for height.

It is essential to identify nutritional interventions that benefit women so that indicators of benefit can be developed and used to screen them for the intervention. Using these indicators will also have the advantage, not presently possible, of directly translating prevalence estimates of malnutrition into predictions of benefit if the intervention is implemented.



  1. Adair L. Nutrition in the reproductive years. In: Johnston FE, ed. Nutritional anthropology. New York: Alan R. Liss. 1987:119-54.
  2. Soysa P. Women and nutrition. World Rev Nutr Diet 1987 ;52: 1 -70.
  3. Winikoff B. Women's health: an alternative perspective for choosing interventions. Stud Fam Plan 1988;19:197-214.
  4. Ghassemi H. Women, food and nutrition: issues in need of a global focus. A report prepared for the Division of Family Health. Geneva: World Health Organization, 1988.
  5. Habicht J-P, Meyers LD, Brownie C. Indicators for identifying and counting the improperly nourished. Am J Clin Nutr 1982;35:1241-54.
  6. Meyers LD, Habicht J-P, Johnson CL, Brownie C. Prevalences of anemia and iron deficiency anemia in black and white women in the United States estimated by two methods. Am J Public Health 1983;73:1024-49.
  7. Habicht J-P, Yarbrough C, Lechtig A, Klein RE. Relationship of birthweight, maternal nutrition and infant mortality. Nutr Rept Int 1973;7:533-46.
  8. Habicht J-P. Some characteristics of indicators of nutritional status for use in screening and surveillance. Am J Clin Nutr 1980;33:531-35.
  9. Rothe GE. Determinants of response to supplementation in malnourished Guatemalan children. M.S. thesis, Cornell University, Ithaca, NY, USA, 1988.
  10. Backett EM, Davies AM, Petros-Barvasian A. The risk approach in health care (with special reference to maternal and child health, including family planning). Public health papers, 76. Geneva: World Health Organization, 1984.
  11. Garby L, Irnell L, Werner I. Iron deficiency in women of fertile age in a Swedish community. Acta Med Scand 1969;185:72-107.
  12. McGuire J. Popkin BM. Beating the zero-sum game: women and nutrition in the third world, part 1. Food Nutr Bull 1989;11(4):38-63.
  13. Jelliffe DB. The assessment of the nutritional status of the community. Geneva: World Health Organization, 1966.
  14. Abraham S. Johnson CL, Najjar MF. Weight by height and age for adults 18-74 years: United States, 1971-74. Vital and health statistics, series 11, no. 208. DHEW publication no. (PHS) 79-1656. Washington, DC, 1979.
  15. Rosso P. A new chart to monitor weight gain in pregnancy. Am J Clin Nutr 1985;41:644-52.
  16. Rasmussen KM. Maternal nutritional status and lactational performance. Clin Nutr 1988;7:147-55.
  17. Report from international conference on prepregnancy nutrition, 9-10 March 1987 in College Park, MD. White Plains, NY, USA: March of Dimes Birth Defects Foundation, 1988.
  18. Rasmussen KM, Fischbeck KL. Effect of repeated reproductive cycles on pregnancy outcome in ad libitum-fed and chronically food-restricted rats. J Nutr 1988; 117: 1959-66.
  19. Social and biological effects on perinatal mortality. Report of an international comparative study. Sponsored by the World Health Organization. Budapest: Statistical Publishing House, 1978.
  20. Rasmussen KM. Maternal depletion syndrome: what is it biologically and what are the implications for modeling its consequences? Presented at the International Center for Research on Women, Washington, DC, Jan 1989.
  21. Rasmussen KM, Mock NB, Habicht J-P. The biological meaning of low birth weight and the use of data on low birth weight for nutritional surveillance. Working paper no. 43. Ithaca, NY, USA: Cornell Nutritional Surveillance Program, 1985.
  22. Kramer MS. Intrauterine growth and gestational duration determinants. Pediatrics 1987; 80:502-11.

Contents - Previous - Next