Contents - Previous - Next

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

Application of incremental growth standards

Gail Grigsby Harrison, Department of Family and Community Medicine and Department of Pediatrics, University of Arizona, Tucson, Arizona, USA

EDITOR'S NOTE

Current estimates of nutrient requirements for growth are based on the variation about average daily growth rates calculated from annual growth increments However, the report Protein-Energy Requirements under Conditions Prevailing in Developing Countries: Current Knowledge and Research Needs (UN University, Tokyo, 1979, WllTR-1/ UNUP- 18) called attention to an issue that has received very little notice - namely, that even normal, well-nourished children do not grow continuously at the average rate. Instead, they may be growing at two to three times this rate at one period and not growing at all during another. The following article provides the most extensive data to date on this phenomenon. Clearly, recommended nutrient intakes must be sufficient to sustain growth rates that are sometimes far above the mean plus two standard deviations.

Growth, as reflected in height and weight, is the most commonly used indicator of nutritional status in children. Weight or height for age reflects the accumulated growth experience of an individual but, except for the very young child, tells little about current or recently past growth rate, which should be a more sensitive indicator of nutritional experience.

Growth velocity has long been used in research situations to describe the characteristics of growth patterns, but the mathematical intricacies of curve-fitting have precluded its use in practical evaluation of individual children until the recent publication by Roche and Himes (1) of incremental growth charts based on six-month growth increments from one of the few longitudinal studies of growth of healthy children. These reference charts have been made available to clinicians, and thus their application in evaluating the growth of both normal and malnourished children needs evaluation.

The incremental growth charts mentioned above are based on measurements of healthy Caucasian children enrolled in the Fels longitudinal growth study between 1929 and 1978. All lived in south-western Ohio, in the United States, at the time of birth and were measured using carefully standardized methods six times in the first year of life and at six-month intervals thereafter until age 18 years. Data for six-month increments of height and weight were plotted as percentiles to construct the incremental standards, and the authors give instructions for adjustment for intervals which are not exactly six months between measurements. The total sample size for each six-month increment on which the standards are based ranged from 150 to 298, and averaged about 200.

The architects of the incremental reference charts are careful to point out that there are fundamental differences between these and the more commonly used attained-growth charts (1). Specifically, they mention that incremental charts are more sensitive than attained-growth charts, and that therefore children may be expected to "change channels," or move from one percentile to another, more commonly. They state that it is not possible at this time to provide criteria for clinical significance of changes in percentiles, for two reasons: First, although the standards are derived from data on healthy children, the reference increments make no claim to reflect "optimum" growth rates. And, second, the clinical significance of changes in increment percentiles must be related to the actual percentiles. Roche and Himes speculate that a change in percentiles near the extremes of the distribution should be of more clinical concern than equivalent change near the median, both because more absolute change in attained growth status results and because the observed change is occurring in a relatively slowly or rapidly growing child.

In this paper, we will examine weight increment data from two groups of infants in relation to the incremental growth standards, with a view to beginning to develop the data base necessary for rational clinical interpretation of growth increments.

METHODS

The data presented here come from two longitudinal studies conducted in our laboratory. The first, which will be referred to as the study of "normal infants," included babies selected at birth to be healthy singleton products of uncomplicated pregnancies. All were term infants (37 weeks or more gestational age, assessed by Dubowitz examination); all were of Anglo-American or Mexican-American descent; and all were born at the University of Arizona Hospital between December 1977 and August 1980. These infants were followed with multiple anthropometric measurements, health and dietary information, and family environmental data every six months until age three years. Sixty-two infants had complete data; the sample size for each six-month interval ranges from 138 to 269. Birth weights were taken from hospital nursery records; they were recorded to the nearest gram by nursing personnel using a Detecto beam balance that was checked with standard weights at regular intervals. Follow-up measurements were made in the subjects' homes, using a portable beam balance calibrated at frequent intervals. All follow-up measurements were obtained within plus or minus two weeks of the day on which the infant attained six months or a multiple of six months of age.

The second study, which will be referred to as the study of "fat infants," consisted of 27 babies studied from six to twelve months of age in a study of early-onset obesity. These infants were selected to represent the extreme of fatness at age six months. They were referred to us by paediatricians and other health-care providers. All were above the 75th percentile of weight for length in relation to the NCHS reference standards (2) and had triceps skinfold measurements above 7.5 mm at age six months. All were otherwise healthy infants who had experienced no significant illnesses. Methods and scales were the same as in the previous study, except that weights were taken within plus or minus one week of the monthly recurrence of the date of the infant's birth.

For both groups of infants, exact intervals between measurements were converted to 182-day intervals according to the guidelines provided by Roche and Himes ( 1).

GROWTH INCREMENTS OF NORMAL INFANTS

Figure 1 shows the distribution of growth increments among children in the "normal" study for each six-month interval between birth and 36 months of age. The distribution closely approximates the reference standard between birth and six months but gradually develops a relative excess of increments at the extremes (below the 3rd and over the 97th percentile) at older ages. The distribution is significantly different from the reference population (chi-square p < .0001) at all ages over six months, and the effect is most pronounced at 30-36 months of age.

FIG 1. Distribution of Growth Increments among "Normal" Children

When we look longitudinally at individual infants' growth patterns, we see that the usual experience in this group of normal babies was to have at least one interval between birth and 36 months which fell at the extreme (3rd percentile or lower or 97th percentile and greater; see fig. 2). Indeed, 35 of the 62 infants who had complete data for all intervals experienced two or three of the six intervals as "extreme" growth increments. A closer look at these 35 infants reveals that the most common experience among them was to have an extreme growth interval (either slow or rapid) followed by a compensatory interval of extreme growth in the other direction. Only three infants experienced two or more periods of extreme growth in the same direction (either slow or rapid) without a compensatory interval either immediately following or following within a year. The initial extreme interval was as likely to be rapid as slow; thus fall-back growth seems as common as catch-up growth. It should be remembered that none of these infants experienced a documented severe insult such as malnutrition or serious illness during this period; rather these growth patterns represent the dynamics of growth under relatively ideal conditions. We examined the data in relation to season of the year and found no consistent seasonal pattern; thus, whether these alternating fast and slow growth periods are primarily endogenous or environmental in origin, they seem to be quite normal.

FIG. 2. Six-Month Periods with "Extreme" Growth Increment (N = 62)

ATTAINED GROWTH OF CHILDREN WITH DIFFERENT VARIABILITY IN GROWTH INCREMENTS

Table 1 shows the percentage of children whose attained weight at age 36 months was greater than one standard deviation (,-score) from the median of the reference standard (2), classified by the number of extreme growth increments. Those children who experienced no extreme increments all fell within plus or minus one standard deviation from the reference median. About one-quarter of those experiencing one or two extreme increments fell outside one standard deviation of weight for age at 36 months, and half of those who had experienced three such periods had weight for age more than one standard deviation from the reference median (either above or below). Thus it appears that at least some children who may be relatively heavy or light for age may not represent steady rapid or slow growers but rather inconsistent growers. It should be remembered that these were all basically healthy children; only one fell outside two standard deviations of weight for age at 36 months.

GROWTH INCREMENTS OF FAT INFANTS

The study of relatively heavy and fat six- to twelve-month olds provides the opportunity to study the growth patterns of babies selected for a previous extreme growth increment. Figure 3 shows the weight gain increments for these 27 infants from birth to six months and from six to twelve months, plotted with the data from the study of normal babies. As we would expect, the birth-to-six-month interval for these infants is greatly skewed toward rapid growth; all were above the 75th percentile for weight gain, and almost all were above the 90th percentile. In contrast, the second six months of life for these infants was characterized by low weight gain velocity, with more than 80 per cent below the 50th percentile and one-third below the 3rd percentile. It is possible that the study itself produced these low velocities of weight gain from six to twelve months, since all these babies had been identified by their doctors as obese or potentially obese and consistent although conservative nutrition education was provided to the mothers throughout the study. However, much previous work has shown that most fat infants revert to normal weight status with advancing age; this phenomenon may simply be an example of normal alternating growth velocity.

TABLE 1. Attained Growth of Infants with Different Numbers of Extreme Increments

FIG. 3. Weight Gain Increment Patterns of "Fat" Infants Compared with Those for "Normal" Infants in the First Two Six-Month Intervals

CONCLUSIONS

The speculation that changes in incremental growth percentiles should be of greatest clinical concern at the extreme edges of the distribution appears to be an oversimplification. These data seem to show that most healthy infants growing under favourable conditions experience not only changes in growth velocity but changes from one extreme of the distribution to the other. This compensatory change in growth rate, catch-up or fall-back growth, appears to be at least as common as steady growth velocity. Whether these periods are primarily environmental or endogenous in origin is not evident; none of these infants experienced clinical malnutrition or serious illness which could account for slow growth periods. It may be that the child who should elicit our clinical concern is not the one with a single extreme growth interval, or shifting growth velocity in the absence of other signs of malnutrition, but rather the child who exhibits steady growth velocity but at the extreme edge of the distribution. This child for whatever reason is not experiencing the opportunity to catch up or to fall back.

ACKNOWLEDGEMENTS

The studies from which this paper is drawn were supported in part by grants from the Nutrition Foundation, the Arizona Agricultural Experiment Station, and Ross Laboratories. These data were presented at meetings of the American Anthropological Association, November 1983, and the American Institute of Nutrition, April 1984.

REFERENCES

1. A.F. Roche and J.H, Himes, "Incremental Growth Charts," Amer. J. Clin. Nutr.. 33: 2041 (1980).

2. P.V.V. Hamill, T.A. Drizd, C.L. Johnson, R.B. Reed, and A.F. Roche, NCHS Growth Curves for Children Birth-18 Years, United States, DHEW Publication No. PHS 78-1650, Vital and Health Statistics Series 11, no. 165 (Government Printing Office, Washington, D.C., 1977).

Nutrition education: Lack of success in teaching Papua New Guinea mothers to distinguish "good" from "not good" weight development charts

Stewart J. Forsyth, Psychological Services Branch, Public Services Commission, Papua New Guinea

EDITOR'S NOTE

The suggestion that illiterate mothers could readily learn to interpret growth charts of their infants and should be allowed to keep such charts ;in the home was initially greeted with scepticism and opposition by health professionals. However, actual experience in many different countries indicates that most mothers can readily distinguish the significance of a gain or loss in weight of their children from one weighing to the next. The following article from Papua New Guinea suggests that this generalization may be too optimistic. Since it is unlikely that Papua New Guinea mothers are less competent than those of other developing countries, the teaching method employed may have been the problem. Alternatively, this study may be more thorough. It is hoped that publication of this article will stimulate letters or other articles that will provide additional information on both the successes and limitations of using growth charts kept in the household as a guide and stimulus to the adequate feeding of the young child.

Poverty is the principal cause of malnutrition in the Third World (1). Populations such as slum dwellers with limited economic self-sufficiency and poor housing are those most likely to be malnourished, and within these disadvantaged populations children of less than three years are most likely to suffer sickness, developmental retardation, or death (1, 21. Nevertheless, organizations such as UNICEF consider that some cheap and simple technologies together with educational efforts can reduce child malnourishment and associated problems. The educational strategy favoured by UNICEF is to teach mothers to understand weight or height growth charts with the goal that such understanding will promote the mothers' involvement in the nourishment and health of their children. Surprisingly little effort has been made to study how mothers in Third World countries perceive such charts. Similarly, there is little evidence of evaluation of efforts to teach understanding and use of the charts to mothers (David Morley, personal communication, 21 April 1983).

A survey of mothers attending maternal and child health clinics in a Papua New Guinea city found that, while 81.5 per cent of mothers could correctly identify their own child's weight graph as either "good" or "not good," only 34.5 per cent were able to label correctly four out of four varied "test" weight graphs (3). This suggests that mothers' accuracy in identifying trends in their child's development was not the result of knowledge of how to identify trends shown in the child's weight graph. The graph then was presumably not functioning to motivate mothers to participate in the nourishment of their children.

The mothers' understanding of the test weight graphs was not related to the nutritional status of their children. Malnourished children were more likely to have unemployed fathers, with the family living in self-built homes. Nutritional status of children also tended to decline after 12 months of age. There was a relationship between the understanding of weight graphs and their children's nutritional welfare for mothers with children over the age of 12 months; mothers with children over this age were more likely to have well-nourished children if they correctly identified all of the four test weight graphs. This led to the conclusion that mothers of children at risk of poor nutrition were able to put their knowledge of nutritional deveIopment measured as accuracy in identifying weight charts - to work in improving their children's nutrition.

In the present study, the mothers and their children were followed up four months later, after some had received instruction in how to identify different examples of weight charts. The effects of this nutrition education we're assessed in terms of changes in mothers' accuracy in identifying test charts, and in terms of their children's nutritional development.

METHOD

Subjects

One hundred forty-two women attending maternal and child health clinics with 143 children were interviewed again after a four-month survey. This sample represents 45 per cent of the mothers originally interviewed. At this second interview, 7.1 per cent of the children were 6 months of age or less; 33 per cent were in the age range 7 to 12 months; 20.4 per cent were 13 to 18 months old; and the remaining infants (39.4 per cent) were older than 20 months but less than five years old.

The majority (55.8 per cent) of the women interviewed did not know their own age. The mothers who did know their age tended to be younger; 64 per cent were less than 25 years. For 26.5 per cent of mothers this was the first child; overall the mothers had a mean 3.02 living children. Infant deaths were reported by 20 per cent of mothers. For 13.7 per cent, one child had died; 4.8 per cent reported the death of two children, and 1.3 per cent the death of three children. About a third (37.5 per cent) of mothers had no education; 42.3 per cent had some schooling but had not completed primary school. The remaining 20.1 per cent had graduated from primary school or had further education.

A large number of women were migrants into this major urban centre; 23.1 per cent were from other provinces. The mothers were all married; of their husbands, 31.2 per cent had work requiring education or training; another 57.2 per cent had unskilled work; and 11.1 per cent had no, or irregular, employment. The houses lived in by 73.8 per cent of mothers were self-built; the remainder lived in houses provided by employers.

Procedure

Four months before the survey women were asked to identify four test weight charts and their own child's chart as "good" or "not good" in terms of healthy development. Personal information and the weight and age of the child were also recorded. Mothers at two of the four clinics received individual training immediately after the survey to teach a discrimination between "good" and "not good" weight graphs. Training was provided by a white male or a Papua New Guinea female. The language of communication, as in the interview, was Tok Pisin (Melanesian Pidgin).

These mothers were shown paired black and white copies of photographs of two infants - one poorly nourished and one well nourished. The pictures and subsequent charts were arranged and presented such that order or placement could not be used as a clue to correct identification. The mothers were asked to identify the well nourished and the poorly nourished child by pointing. After correctly identifying two of the pairs of the pictures (all of the mothers were immediately successful on this task), they were shown pairs of weight graphs, showing good and poor nutritional development. Superimposed on the appropriate graphs were the previously present pictures of the well nourished and poorly nourished children. The mothers were asked to point to the marks on the weight charts to identify the well nourished and poorly nourished children. After succeeding in identifying two of these graph and picture pairs (as all did), the mothers were shown another set of pairs of weight charts of well nourished and poorly nourished children.

Smaller pictures of appropriately nourished infants were included on the chart, and close to the last weight mark the mothers were asked to trace the recorded weights for the well nourished and poorly nourished child. All mothers succeeded in correctly describing two pairs of charts. Next, mothers were shown weight charts in succession, each accompanied by small pictures separated from the weight chart. Each mother was asked to trace the record and to then describe the chart as showing a well developed baby or a poorly developing baby. If there were any mistakes the correct tracing and labelling responses were modelled. No mother mistakenly labelled more than one graph.

The next stage of the training was the showing of five successive charts without accompanying pictures. Again, the mothers were asked to trace the record and to describe the chart as showing a well nourished or poorly nourished child. Errors were corrected by the same modelling procedure. If a mother made two errors, the trainer went back to the previous stage of the training sequence that included pictures with the graph and proceeded through this stage to the charts presented without pictures. This training procedure resulted in all of the mothers being able to correctly identify five weight graphs without using clues from pictures. The procedure took about seven minutes for each mother.

The procedure used has been tried with young female school drop-outs and mothers attending a hospital maternal and child health clinic. The materials were developed with the assistance of a nutrition educator and a paediatrician, both with more than two years of experience with mothers and children in the city.

At one of the two clinics where mothers received this individual training, and also at a third clinic, the clinic staff received instruction on how to ask mothers to describe the meaning of the child's weight graph. This instruction took place immediately after the first survey had been completed for that clinic. The staff discussed why they should ask mothers to interpret weight charts, then decided how they would ask mothers. All staff then took turns at playing both the role of mother and clinic staff member talking to the mother and asking her about her child's weight graph. This training took about 30 minutes. A fourth clinic did not receive any training for mothers or for staff. The allocation of training is illustrated in table 1. This design can be described as quasi-experimental (4), since the opportunity to receive different training was not randomly allocated to different individuals but depended on what clinic was attended.

Follow-up

The follow-up survey was made at each health clinic four months after the first survey and training. Children's weights were recorded and the mother was presented with the four test graphs used in the first survey. Again, she was asked to describe the graphs as "good" or "not good."

RESULTS AND DISCUSSION

Weight for Age

Children's weight for age (W/A) was calculated from the standard weight chart used in Papua New Guinea. Over the four-month period between measuring weights there were changes in children's W/A according to their age. Figure 1 shows mean changes in children's W/A for different age cohorts (each point represents a mean from 11 to 20 children). Children of 2 to 14 months are likely to deteriorate nutritionally over a four-month period. After 18 months children are likely to remain at a constant nutritional level.

TABLE 1. Training for Mothers and/or Clinic Staff

A break-point of 18 months was selected to distinguish between those children most likely to lose weight and those more likely to remain at a stable weight. Children 18 months or younger showed a mean W/A loss of about 31/2 per cent, from 91.03 per cent of the standard (with a standard deviation of 13.71 per cent) to 87.41 per cent (s = 11.71 per cent). Children 19 months or older increased in mean W/A by about one-half per cent, from 82.77 (s = 12.06 per cent) to 83.39 per cent (s = 10.36 per cent). W/A was quite consistent for children over the four months of development. The correlation between first and second measurements of W/A for children of 18 months or less was 0.76, for children of 19 months or more it was 0.84 (overall correlation = 0.79).

In attempting to measure a possible training effect, analysis of covariance (ANCOVA) was used because of group differences in W/A (children of the control group being significantly better nourished), analysis of variance produced between group difference (P < .001), and the high correlation between successive W/A measurements. As there were only small differences between treatment groups (see table 2), the comparison made (controlled for the children's previous W/A) was between the W/A of control group children and of all children in groups who received a training input, and also between W/A of younger (18 months or less) and older (19 months or more) children. Applying the F test, the training effect was insignificant (P < .90), as was the age effect (P < .22). There was no significant interaction between these variables.

FIG. 1. Mean Changes in Children's Proportion of Standard Weight for Age over Four Months between First and Second Interview

TABLE 2. Changes in Mean Standard Percentage of Weight for Age for Children of Two Age Cohorts - 18 Months or Less, and 19 Months or More - according to Training Provided to Their Mothers or to Health Centre Staff at Clinics Attended by Their Mothers

Mean percentages are rounded off. Figures in parentheses are standard deviations.

TABLE 3. Mean Accuracy in Identification of Weight Charts by Mothers of Children 18 Months or Less and 19 Months or More during First and Second Interviews

Score is number of answers correct out of a total of four.

These results indicate that, despite changes in the mean weights of children, there was not a significant difference in weight loss between younger and older children. Children's W/A was consistent, although close to the minimum standard for W/A (80 per cent of the norm). The educational strategies attempted had no nutritional effect over the four-month period.

Mothers' Accuracy in Identifying Test Charts

Mothers' accuracy in correctly labelling four test weight charts was measured in the two interviews. Means of group scores (table 3) do not show any treatment effect; the only trend noticeable is for group means to regress to a score slightly below three (of the maximum of four). There was no consistency in mothers' scores over the period between interviews (correlation between scores = .06). ANCOVA of accuracy scores (controlling for previous score) enabled comparison between the combined training groups and the untrained group and between mothers of younger (18 months or less) or older children.

The training effect was insignificant (P < .80) as the effect of children's age more nearly approached significance (P < .10). There was no significant interaction between these variables. These results indicate that the test graphs were not a reliable measure and that they were insensitive to the effect, if any, of training, but that mothers of younger children showed a trend towards a relative increase in their accuracy in labelling the test charts.

Changes in mothers' accuracy in correctly labelling development trends in their own child's weight chart were compared for training groups and the control group. Mothers in the training groups were 82 per cent accurate at first testing and 83 per cent accurate at the second test. The control group mothers showed a drop in accuracy from first to second test of 94 per cent to 76 per cent (a significant difference at the 1.6 per cent level, binomial test [5] ). It appears that the reduced accuracy of untrained mothers over the follow-up period is the result of a regression of this measure to the mean. Comparisons of mothers' accuracy in correctly identifying their child's weight chart between the first and second assessments were made for mothers of younger (18 months or less) and older children (19 months or more). Neither group showed changes in accuracy in identifying their child's weight chart over the period.

It seemed possible that changes in children's W/A motivated mothers to learn about weight charts. Cross tabulations were made between changes in children's W/A (improved or unchanged contrasted with worsened) and mothers' accuracy in correctly identifying test weight charts (mothers achieving a 100 per cent score [4/4] contrasted with the others) both at the first and at the second interview, At the first interview mothers of younger children (18 months or less) whose children would show reduced W/A over the following four months made up 53 per cent of the total group of mothers of younger children. The chi-square test showed that the proportion of women who could correctly identify ail four test charts (91 per cent) was significantly better (P < .05) than the proportion (76 per cent) of mothers whose children's W/A would not deteriorate. Perplexingly, at the follow-up test the difference between the groups had narrowed; mothers of W/A-reduced children were now 87 per cent accurate in contrast to 78 per cent for the mothers of children whose W/A did not decrease - a non-significant difference. There was no relation between this assessment of knowledge of weight charts for the mothers of older (19 months or more) children in terms of changes in children's W/A.

Mothers' accuracy in identifying their child's weight chart trends was cross-tabulated against changes in the child's W/A (improved or unchanged contrasted with worsened). No significant differences in accuracy were shown before or after changes in children's W/A.

To summarize, attempts to teach mothers about weight charts, either directly or indirectly, produced no changes in mothers' understanding of weight charts. Such teaching produced no changes in the children's nutritional status.

There was a tendency for mothers of younger children to become more accurate in identifying test weight charts. It was hypothesized that this was because these mothers were more sensitive to the meaning of weight charts because many of them had children who had worsened nutritionally over the follow-up period. However, comparisons of mothers whose children's W/A worsened with mothers whose children remained at a constant W/A or improved did not reveal any tong-term increase in the accuracy of mothers of at-risk children in identifying test weight charts or the weight charts of their own children.

These results indicate that weight charts achieve little in the way of motivating urban Papua New Guinea mothers to provide nourishment for their children. Despite the wide popularity of weight charts in nutrition education (Austin et al. [6] estimated that between 31 and 74 per cent of nutrition projects in Third World countries used weight charts), their role in involving mothers in assisting child development has not been demonstrated. It is possible that more participatory and group-oriented teaching methods than those described here would help mothers to become knowledgeable about, and active agents in, child development (7, 8). Education that encourages a group identity should also promote regular clinic attendance. A participatory orientation could result in mothers reinforcing each other's understanding of nutrition and generating practical solutions to malnourishment.

ACKNOWLEDGEMENTS

I wish to thank Ms. Rone Keloup for her contribution to the training of mothers, and also to thank Dr. Terry Fairclough for his encouragement and suggestions in the data analysis. Dr. Peter Poore and Mr. John Pomat gave help and advice.

REFERENCES

1. J. E. Austin, Nutrition Programs in the Third World (Oelgeschlager, Gunn, and Hain, Cambridge, Mass., USA, 1981).

2. S. S. Basta, "Nutrition and Health in Low Income Urban Areas of the Third World," Ecol.. food Nutr., 6: 113 (1977).

3. S. J. Forsyth, "Assessment of Nutrition Education: Urban Mother Understanding of Weight for Age Graphs," Papua New Guinea Med. J., 25: 239 ( 1982).

4. D. T. Campbell and J. C. Stanley, Experimental and Ouasi-Experimental Designs for Research (Rand McNally, Chicago, 1963).

5. S. Seigel, Nonparametric Statistics for the Behavioural Sciences (McGraw-Hill Kogakusha, Tokyo, 1956).

6. J. Austin, M. Mahin, D. Pyle, and M. Zeitlin, Annotated Dictionary of Nutrition Programs in Developing Countries (Harvard Institute for International Development, Cambridge, Mass., USA, 1978).

7. A. Praun, "Nutrition Education: Development or Alienation? "Human Nutr.: Applied Nutr., 36A: 28 (

1982).

8. D. Werner and B. Bower, Helping Health Workers Learn (Hesperian Foundation, Palo Alto, Calif., USA, 1982).

Continue

Contents - Previous - Next