Cover Image
close this bookCare and Nutrition - Concepts and Measurement (IFPRI, 1997, 56 p.)
View the document(introduction...)
View the documentForeword
View the documentAcknowledgments
View the documentSummary
View the document1. Introduction
View the document2. Developments in Conceptualizing Care
View the document3. Resources for Care
View the document4. Care Practices
View the document5. Conclusions
View the documentBibliography
View the documentInternational Food Policy Research Institute

4. Care Practices

Constructs and measurement tools are discussed for two dimensions of caring behavior: time spent (quantity of care), and the nature of the activities undertaken (quality of care), or some combination of the two dimensions. Since numerous studies have looked at quantity of time spent on child care, those studies will be summarized. Although caregiving cannot occur when the caregiver is absent, there is a great variation in the activities performed by caregivers when they are present and spending time on "child care." Since the quantitative measure of "time spent on child care" is more common and possibly easier to measure, it will be considered first.

Time Spent on Child Care

This section reviews the findings from various studies, examines the validity of nonobservational measures, and discusses the usefulness of time spent on child care as an indicator. In general, the studies conducted cannot be easily compared because they use different definitions of child care and different methods of data collection (direct observation, random spot observation, and various recall periods). It concludes that if time spent on child care is to be measured, direct observation is preferable, because it increases the chances of getting accurate measures of time allocation and allows the investigator to assess degrees of involvement and quality of child care at the same time.

Studies of Time Spent on Child Care

A number of studies have attempted direct measurement of time spent by the mother on child care, and these are summarized in Table 8. Five studies use observational data and seven are based on recall data. Time spent ranges from highs of more than 7.3 hours per day in the Philippines (Blau, Guilkey, and Popkin 1996) and 6.3 hours per day in Indonesia (Gryboski 1996) to a low of 18 minutes per day in the care of others by lead females in Nepal (Paolisso and Regmi 1995; Paolisso 1994). The recent appearance of observational studies (for example, Cohen et al. 1995a) provides a standard with which to evaluate the accuracy of estimates in recall studies. McGuire and Popkin (1990a, 1990b) presented a similar table for earlier studies but did not analyze them according to the age of the child.

Of the observational studies, two used continuous observation for 12 hours (Cohen, Haddix, et al. 1995; Gryboski 1996), one used a sampling of two-hour time blocks (Ricci et al. 1994), and two used spot observations (Paolisso and Regmi 1995; Baksh et al. 1994). The spot observations were translated into total time estimates by assuming that the percentage of time the activity was observed represents the percentage of time it will take during the day. For purposes of comparison in the table, all data are presented for units of a 12-hour day. Both total child care time and time spent breast-feeding are presented. The definition of child care in the study is also shown. Most studies evaluated direct child care activities, but two looked at child-focused activities and two evaluated not only direct care, but also watching and child supervision. When supervision is included, daily means are almost five hours greater than the means in the other studies. Thus the definition of "child care time" can substantially affect estimates.

Cohen, Haddix, et al. (1995) observed that non-working urban Honduran mothers of exclusively breast-fed 6-month-old infants spent about an hour breast-feeding out of every 12-hour day. When playing and holding the infant was included, time increased to 2.9 hours per day for primiparous women and 2.6 hours per day for multiparous women. Total time in child-focused activities, including cleaning children's clothes, was approximately 4 hours per day whether a woman was caring for one or several children. These times are lower than the 24-hour diaries of U.S. breast-feeding mothers, who reported 137 minutes per day of breast-feeding at 3 months, 98 minutes per day at 6 months, 81 minutes per day at 9 months, and 53 minutes per day at 12 months (Heinig et al. 1994); the 6-month-old children were not exclusively breast-fed.

Table 8 - Estimates of time spent on child care from observation and recall

Study

Sample

Method

Definition of child care

Age of child

Results for total child care time

Results for breast-feeding

Comment

Observation Cohen et al. 1995a 1995b

Honduras N=139

Continuous observation; simultaneous activities counted separately

(1) All child-focused activities, including play and feeding
(2) Play and feeding only

6 months

(1) Primipara EBFa 4.02 hours/12 hours; primipara mixed: 4.08 hours/12 hours; multipara EBF: 3.75 hours/12 hours; multipara mixed: 4.5 hours/12 hours
Primipara: 2.9 hours/12 hours; multipara: 2.6 hours/12 hours

Primipara EBF at 4.4 months: 71 minutes/12 hours; primipara EBF at 6 months: 62 minutes/12 hours

Urban nonworking women

Heinig et al. 1994

United States N=61

Well-kept time diaries

Breast-feeding only assessed

3 months
6 months
9 months
12 months

Not applicable

3 months: 137 minutes/24 hours; 6 months: 98 minutes/24 hours;
9 months: 81 minutes/24 hours; 12 months: 53 minutes/24 hours

From 6 months onward, data are not on EBF

Paolisso 1994

Nepal N=264 rural

Spot observation; only when care was primary activity

Hold, carry, breast-feed, wash, bathe, clean, give treatment

0-5 years

Adult female: 32 minutes/12 hours; lead female: 18 minutes/12 hours

16 minutes/I 2 hours if breast-feeding

Much shared child care; child care dropped after 1 year

Baksh et al. 1994

Kenya N=169 rural

Spot observation; only when care was primary

Same

0-3 months
3-6 months
6-9 months
10-12 months
12-15 months NPNLb

2.85 hours/12 hours
2.59 hours/12 hours
2.30 hours/I2 hours
2.02 hours/12 hours
1.33 hours/12 hours
0.60 hours/12 hours

16 percent of all child care time is in breast-feeding

High work demand; 84 percent of all care activities to children less than 18 months old

Ricci et al. 1994

Egypt N=107 semi-urban

10 hours of continuous observation in 2-hour blocks

All child-focused: hold, socialize, supervise, feed, wash, illness

18-23 months 24-29 months

3.8 hours/12 hours (0.4) 3.6 hours/12 hours (0.5)

32 minutes/12 hours (0.4)
19 minutes/12 hours (1.5)

Times for mothers without other infant

Gryboski 1996

Indonesia N=60 rural

18 days of records;
6 continuous observations,
12 recalls

Direct care plus supervision

3-35 months

6.3 hours/12 hours


Much shared care; 88 percent of days had shared care

Recall Popkin 1980

Philippines N=573 Laguna rural and semi-urban

Recall of specific activities over 1 week

None provided

1-71 months

Workers: 1.26 hours/day Nonworkers: 1.70 hours/day


Used predicted child care time of mother, siblings, father

Blau, Guilkey, and Popkin 1996

Philippines N= 2,876 rural and urban Cebu

Recall of hours per day pervious day

Direct care plus watching

0-24 months

7.34 hours/day


Did not vary by child's age or mother's education

Brown and Haddad 1995

Kenya rural

Recall

None given


Nonlactating: 36 minutes/day
Lactating: 63 minutes/day

27 minutes/day

Median length of breast-feeding =22.5 (DHS)c


Ghana Brong Ahafo
N=253

Recall

None given


Nonlactating: 40 minutes/day
Lactating: 55.2 minutes/day

15 minutes/day

Median length of breast-feeding =22.8 (DHS)c


Ghana Volta N=278

Recall

None given


Nonlactating: 52 minutes/day
Lactating: 50 minutes/day

-2 minutes/day

Median length of breast-feeding =21.8 (DHS)c

Bouis and Kennedy 1989

Philippines N=448 rural

Recall of 24 hours of activity

Direct care: feed, breast-feed, play, bathe

0-5 years

Nonlactating: 58 minutes/day
Lactating: 148 minutes/day

Not given

Minutes/day breast-feeding = 16.6 (DHS)c

Current analysis

Philippines N=328 Bukidnon

Same

Same

0-1 years
1-2 years
2-3 years

About 2.5 hours/day About 1.5 hours/day About 50 minutes/day

40 minutes/day
20 minutes/day
15 minutes/day

Varies by round; rounded averages presented

a EBF is exclusively breast-feeding.
b NPNL is nonpregnant/nonlactating.
c DHS is Demographic and Health Survey.

Ricci et al. (1994) report a mean of 3.8 hours in child-centered activities for children 18-23 months and 3.6 hours per day for children 24-29 months in periurban Egypt. Breast-feeding time per day dropped with increasing age: for children 18-23 months still breast-feeding, 32 minutes per 12-hour day was spent, whereas for children 24-29 months, only 19 minutes per 12-hour day was spent breast-feeding. Mothers with a younger infant were not included in these analyses.

Observed care time drops dramatically after the first year or year and a half. In the Kenya Collaborative Research Support Program (CRSP), a rural, agricultural sample in which women carried significant workloads, Baksh et al. (1994) report the equivalent of 2.95 hours per 12-hour day for all direct child care (holding, breast-feeding, health care, feeding, washing, dressing, and other activities) for children from birth through 3 months of age, which dropped to 1.3 hours per 12-hour day for children 12-15 months and 36 minutes per day for nonpregnant, nonlactating women. Eighty-four percent of all care activities were addressed to children 0-17 months, only 11 percent to children 18-59 months, and 4 percent to children 5 years and older. Only 16 percent of time, or 24 minutes per 12 hours, was spent breast-feeding across all breast-fed children (differences by age were not reported).

Depending on the setting, it will be necessary to estimate child care time by alternate caregivers as well, rather than child care time by the mother only. Gryboski (1996) in Indonesia observed care by non-mothers, even when mothers were present, on 88 percent of days. Paolisso and Regmi (1995) in Nepal reported 32 minutes per day in child care time for children 1-5 years by all females and 18 minutes per day by the lead female, using random spot observations. Only 42 percent of all child care activities observed were performed by the lead female; 25 percent were performed by a male in the household. These data are consistent with high rates of sibling care for children over one year of age among the Malian Fulani; weaned children under five years of age were in the presence of "their biological mothers (defined according to specific operational criteria) for only about 25 percent of their day. The rest of the time they spent with their older sisters, peers, or other members of the extended family or community" (Castle 1992).

Time spent on child care estimated from recall is also shown in Table 8. The two Philippine samples show very different investments of time in child care, but in the first, the publication did not specify how child care was defined and in the second, "watched" was included as part of child care. Breast-feeding times were not disaggregated in either sample. In four samples (one from rural Kenya, two from Ghana, and one from the rural Philippines) in which child care time was disaggregated by lactation status, nonpregnant, nonlactating women reported spending between 36 and 52 minutes per day in child care. Lactating Kenyan women spent, on average, 27 more minutes a day on child care than nonlactating women, and lactating Ghanian women in one sample spent 15 minutes more than nonlactating women per day, and in the other sample, spent no additional time per day. These figures were derived from 24-hour recall, so they probably include night breast-feeding.

According to Bouis and Kennedy (1989), lactating Philippine women spend over an hour and a half more on child care (for children aged 0-5 years) than nonlactating women. However, another analysis of the same data set (Engle and Bhattarai 1997) disaggregated child care and lactation times by the age of the child, and found lactation times more similar to the other studies - about 40 minutes a day in the first year, 20 minutes a day in the second, and 15 minutes a day in the third year (breast-fed children only, eliminating women who had another infant). If ages are not disaggregated, one cannot determine lactation times by subtracting the two estimates, since the children of nonlactating mothers are likely to be older. The low times in the African sample may also be due to extended breast-feeding (Demographic and Health Survey [DHS] estimates from 1993 for median duration of any breast-feeding are between 21 and 23 months, depending on the site).10 Therefore, the samples from Kenya and Ghana may include an older group of children who are probably spending much less time breast-feeding. These data illustrate the importance of disaggregating time in child care by the age of the child.

10 The worldwide Demographic and Health Survey program is designed to collect data on fertility, family planning, and maternal and child health. The surveys are a collaborative effort of the governments of the countries surveyed and the Institute for Resource Development/Macro International, Inc., in the United States, and partially funded by USAID and UNICEF.

Validity of Recall Methods for Time Allocation

A small number of studies have evaluated the accuracy of time use data (for example, Engle and Lumpkin 1992; Ricci et al. 1995; Piwoz et al. 1995; Engle, Hurtado, and Ruel 1996). The standard approach is to have one person observe activities and a different person interview the subject and compare the number of activities correctly recalled and the accuracy of the duration of recall (for example, Bernard et al. 1984).

Engle and Lumpkin (1992) found that among middle-class U.S. and Guatemalan mothers, it was more common for activities to be forgotten than for durations to be recalled inaccurately; in other words, the largest source of inaccuracy was forgetting the event completely rather than under- or overestimating the duration. A similar conclusion emerges from a study among rural indigenous Guatemalan women (Engle, Hurtado, and Ruel forthcoming). Given the importance of recalling the activity itself, factors that influence activities to be recalled need to be documented. Error rates vary significantly by type of activity. In general, memory is superior for events that are easily coded (for example, those that have a defined start or stop time) and for nonhabitual or highly salient or important events (Best 1989). Unfortunately, child care is a frequent, nonsalient activity and is therefore less likely to be encoded and less likely to be remembered. Not surprisingly, activities such as work are recalled with reasonable accuracy; however, recall accuracy for habitual, nonsalient activities like child care is much lower (Engle and Lumpkin 1992; Ricci et al. 1995). In a pilot study to determine whether recall could be used for the Kenya CRSP, Ricci et al. (1995) found that 83 percent of child care activities were not reported and concluded that observations had to be used in the study. Instructions that facilitate recall have been found to increase the accuracy of recall (Engle and Lumpkin 1992).

A major difference in estimates of time spent in child care depends on how simultaneous activities are coded; some are coded as separate activities, resulting in a higher total time; others ignore the secondary activity. Much of child care time is a concurrent activity (watching or keeping an eye out) that is probably neither coded mentally by the caregiver nor assessed; on the other hand, it probably has an important protective role. When it is specifically included in the definition of child care, total time increases dramatically. Second, it is probably essential to assess all child care providers in the home and to interview the respondents directly, since so much child care is shared. Another person reporting on someone's child care time may be highly inaccurate. For example, Immink et al. (1994) found that Guatemalan men's reports of the number of days that women spent in farm activities were far lower than the days that women thought they had spent.

Time Spent on Child Care as an Indicator of Care

If time spent on child care is to be used as an indicator of care, it should be measured by direct observation, which, in addition to getting accurate measures of time allocation, allows the specific child care activities to be recorded. Should, however, time spent be used as an indicator of care provision? This question is raised because a number of studies do not find a significant association between energy intake or nutritional status of the child and time spent on child care by the caregiver.

The summary of 10 studies in Table 8 does not lend strong support to the hypothesis that a quantitative measure of mother's time spent on child care has a significant association with child welfare. The studies address the issue of whether the amount of time spent on child care is associated with nutritional status or energy intake in children. Of the 10, 3 studies found that spending more time on child care was positively associated with children's anthropometric measures. Two studies are from the Philippines: Popkin (1980) reports that, using recall data, weight-for-age among children 1-35 months was marginally higher for children who received more care. Blau, Guilkey, and Popkin (1996) report that in the Cebu Philippines data set recalled child care time spent by different family members was significantly associated with increased weight of children 0-24 months. Increased mother's time was also associated with greater height. However, when fixed effects were controlled for, the only remaining associations with child weight were time spent by a nonfemale relative, and for child height, time spent by the mother.

Paolisso and Regmi (1995) find a marginally significant association of child care time spent by all family members with increased weight for children 6-36 months in Nepal. The study is unique in that it uses an estimate of observed total time spent on child care by all family members. Two studies using observational data (Gryboski 1996, in Indonesia; Ricci et al. 1994, in Egypt) find no association of total time spent by the mother on child care and energy intake (kilo-calories per day). In the former study, nonbreast-milk intake was greater when the child was cared for by someone other than the mother, although breast-milk intake was greater when cared for by the mother. Ricci et al. (1994, 302) in Kenya find that in the period between 18 and 23 months of age, "maternal child caregiving behavior was critical" for energy intake (kilocalories per day). The significant measure of care was frequency of feeding, but not total time spent on child feeding (food preparation, serving food, and feeding).

Only the original Popkin study (1980) presents clear support for the importance of mother's time spent on child care for child nutritional status. The author also reports a significant negative association of child weight-for-age with care by a sibling; possibly the reason was that the sibling provided inadequate care.

This discussion leads to the conclusion that the amount of time spent on child care may not be useful in predicting whether a child's intake or its nutritional status will be adequate. As will be shown in the next section, research resources may be better invested in collecting information on specific measures of care.

Specific Care Practices

Whereas measures of the quantity of time spent on child care can only be related to child outcomes somewhat ambiguously, measures of quality of time, or of specific practices associated with good child care, are often associated with positive outcomes. For example, different ways of disposing of child wastes can affect incidence rates of diarrhea as can hygiene practices like boiling water prior to use (Cebu Study Team 1991). Positive patterns of interaction between caregiver and child and a nurturant home environment are significantly associated with later cognitive development of children in a variety of cultural and ethnic groups (see, for example, Bradley and Caldwell 1984; Bradley et al. 1989). Child growth has also been associated with observational measures of interaction between child and caregiver (Barnard et al. 1989). Many investigators recommend that the quality of the home environment and the quality of the caregiver-child interaction must be assessed through observational means, even if the observation is brief (Heffer and Kelley 1994). This section discusses specific measures of two caregiving activities - feeding and psychosocial care. The other care practices of breast-feeding, food preparation, hygiene, and health seeking and health care will not be discussed here, but these two will serve as models.

Feeding Practices

Caregiver practices that could affect the child's nutrient intake include (1) adaptation of feeding to the child's characteristics, taking into consideration psychomotor capabilities (such as use of finger foods, spoon handling ability, ability to munch or chew) and appetite; (2) responsiveness of the caregiver to feeding situations, including encouraging the child to eat, offering additional foods, providing second helpings, stimulating eating through threats, timing of feeding, responding to poor appetite, and interacting positively with the child; and (3) appropriateness of the feeding situation, including the organization and regularity of feeding, supervision and protection of the child while eating, frequency of feeding, monitoring with whom the child eats, and elimination of distractions during eating.

Adaptation of Feeding to Child's Characteristics.

Caregivers need to be sure that children are capable of self-feeding before expecting it of them. Children also have a drive for independence and may eat more if they are allowed to use newly learned finger skills to pick up foods. A child's capacity to process food by suckling, sucking, munching, or chewing increases with age. For example, by seven months of age, the "gag reflex" moves to the posterior third of the tongue, permitting the child to ingest solids more easily (Milla 1991). The time it takes for a child to eat solid and viscous foods decreases with age, but not the time it takes to consume purées. The child's ability to hold a spoon, handle a cup, or grasp a piece of solid food also increases with age.

Poor appetite plays a major role in inadequate nutrient intake of children (Piwoz et al. 1994; Bentley, Black, and Hurtado 1995). Factors that reduce a child's appetite may include a monotonous diet, lack of nutrients needed for appetite (for example, zinc), illnesses such as fever (Neumann et al. 1994), diarrhea, malaria, measles, intestinal parasites, chronic malnutrition, sores in the mouth (perhaps caused by teething), or anxiety (Dettwyler 1986, 1987). These problems are not unique to malnourished children; as noted earlier, 24 percent of parents in affluent societies report feeding difficulties with their two-year-old children. Thus the caregiver's ability to deal with child anorexia is significant for child intake. For suggested measures of feeding practices, see Tables 9 and 10.

Table 9 - Feeding practices: Caregiver-child interactions

Construct

Measurement tools

Comment

Presence, absence of caregiver feeding practices

Observation of one or more eating episodes; can adapt existing scales (for example, the Nursing Child Assessment Feeding Scale) or develop list of behaviors related to caregiver responsiveness and the feeding environment

Caregiver feeding practices will depend on child behavior, which should be recorded. May compensate for child behavior problems.

Frequency of behaviors

Quantitative assessment of behaviors related to feeding, number of spoonfuls, number of touches

Must have careful training of observers

Overall affective quality of interaction

Rate child and caregiver separately on scales representing domains of behavior with 1-5 point scales

Distinguish failure to thrive from normal U.S. inner-city children

Caregiver behavior in a structured situation

Present a challenge to caregiver and observe what she does with the child, or ask her to report the results at a later date

More often used in qualitative research; could be a quantitative technique

Caregiver's Ability to Feed Responsively. Particularly with young children, feeding can be an active process: caregivers can encourage, cajole, offer more helpings, talk to children while eating, model eating behavior, and monitor how much the child eats. In many societies, caregivers are passive feeders, leaving the initiative to eat to children (the child controls the feeding) (Dettwyler 1987; Bentley et al. 1991; Engle and Zeitlin 1996). At the other extreme are cultural patterns that support caregiver control of eating, characterized by forced feeding and continued and even intrusive pressure on children to eat (Brown et al. 1988; Launer and Habicht 1989). In this case, rather than providing an opportunity for interaction and educational enhancement, feeding can become a time of conflict with intrusive but ineffective caregiver strategies and high levels of child refusal.

Passive feeding may be due to lack of time and energy or to beliefs that children should not be pressured to eat - that "the stomach knows its limits" (Bentley, Black, and Hurtado 1995). Although this belief may seem reasonable, if a child has anorexia or a poor appetite, extra encouragement may be necessary for adequate nutrient intake. Anorexic children are difficult to feed. When anorexia is a problem, care-givers need to actively encourage food consumption. But this means having the time, knowledge, resources, self-confidence, and support to encourage anorexic children to eat (Griffiths 1988).

Table 10 - Feeding practices: Child variables

Construct

Measurement tools

Comment

Appetite and hunger

Observe whether food is completely eaten or interest level during eating

May depend on specific food or initial or subsequent contact with food

Adaptive food preferences

Observation of child's interest in standard foods; survey of caregiver's observations

Children who reject major food are more undernourished; high incidence of "picky eaters" in failure-to-thrive children

Child has characteristics preferred (or not) by parents

Use qualitative methods to identify variables - may be gender, parentage, physical attributes, birth order

These preferences are hard to assess directly; may use frequency of care as an indicator

Physical difficulties in self-feeding

Low birth weight (LBW); oral/motor dysfunction as diagnosed by physician; developmental delays in skills related to self-feeding assessed by clinician

LBW associated with poorer suckling ability, oral/motor dysfunction fairly common in failure-to-thrive children

Where feeding encouragement is normally low, increased encouragement of eating has been observed when children are ill (Bentley et al. 1991) or refuse food (Engle and Zeitlin 1996). These findings suggest that active feeding may have a compensatory rather than an enhancement role. In other words, the caregiver may feed more intensively if she perceives that the child is not eating. Caregiver understanding of and response to children's hunger cues may be critical for adequate food intake. For example, if caregivers perceive a child's tongue thrust, a typical mouthing response to new food sensations at a particular age, as a food refusal and cease to feed, a child will receive less food.

The person who is doing the feeding may influence the child's willingness to eat; often children will refuse food if the preferred caregiver is not present. Patience and understanding, plus recognizing the child's need to gain familiarity with the caregiver, will increase the chances of successful feeding. Caregiver beliefs about the appropriate level of demand for food by children can result in the shaping of children's behavior to reduce demand for food. If the caregiver feels that a child should learn not to ask for food, or that immediate responses to children's requests for food will "spoil" or inappropriately indulge a child, particularly after infancy, the chances of the child achieving adequate intake are lowered, since child demand plays a large role in the amount of food ingested (Garcia, Kaiser, and Dewey 1990).

Studies comparing failure-to-thrive children with normally growing children have found differences in the feeding style of the two groups. In failure-to-thrive groups, an authoritarian disciplinary approach may override children's internal regulatory system for hunger, and there may be low maternal responsiveness and sensitivity to cues. This style may be combined with family isolation and possibly with difficult temperaments or subtle oral/motor feeding problems in the children, leading to a breakdown of the caregiver-child relationship (MacPhee et al. 1993; Black 1995). Interventions to modify these relationships through increasing family support have met with only modest success (for example, Drotar et al. 1990), although one carefully executed experimental investigation showed significant effects on cognitive development, although not growth, among children when the intervention began prior to 12 months (Black et al. 1995). Strategies that use behavior modification, including shaping of parental behavior and presentation of positive role models, have resulted in changes in feeding practices. For example, among U.S. African-American adolescent mothers, a videotape of positive feeding practices that was culturally appropriate and relied on social learning theory resulted in significant changes in the mother's attitude toward child feeding and in observed maternal mealtime behavior such as maternal communication, amount of verbalization, and quality of verbalization (Black and Teti 1996). For suggested measures, see Tables 9 and 10.

The Feeding Situation. Children can be fed on a regular basis each day, sitting in a prescribed place with food easily accessible, or feeding can occur while children wander around, or at a time that the caregiver finds convenient. Children can be easily distracted, particularly if food is difficult to eat (for example, soup with a spoon that is beyond the child's ability to handle) or not particularly tasty. If supervision of feeding is not adequate, other siblings or even animals may take advantage of a young child's vulnerability and take food away, or food may be spilled on the ground. Feeding from a common pot may reduce the chances of a younger child getting enough food and may make it harder for a caregiver to be sure food has been allocated to the youngest child.

Studies in developing countries have found associations between specific feeding behaviors, such as location of feeding, organization of the feeding event, and use of spoon, with mother's education (Guldan et al. 1993). The authors conclude that more educated mothers had more labor-intensive child care strategies, particularly in selecting a clean and protected location for feeding. Linkages with child nutritional status were not made.

Caregivers may not be aware of how much their children eat; one project found that when mothers paid more attention to the quantity children ate, they were surprised by the small amounts and were willing to increase amounts fed (Dickin, Griffiths, and Piwoz 1996). When children are fed from a common pot, the amount eaten is not easy to determine. Having a separate bowl for each child can help the caregiver evaluate the quantities eaten.

Measurement Tools to Assess Care Behaviors and Practices in Feeding. Five types of observational measures have been used to examine child and parent behavior in feeding situations: (1) observations of time spent on feeding, (2) presence or absence of specific feeding practices, (3) quantitative assessments (frequency of specific feeding behaviors), (4) behavioral ratings (to measure the quality of the caregiver-child interaction), and (5) structured situation challenge (the caregiver is presented with a new food and her reactions are observed). Recall of child or caregiver time spent on feeding, discussed in a previous section, will not be considered here.

Observation of specific practices. The most commonly used assessment technique is to code the presence or absence of child or caregiver behaviors on a series of items (Barnard et al. 1989; MacPhee et al. 1993; Engle and Zeitlin 1996; Guldan et al. 1993). The most widely used instrument in clinical settings in the United States has been the Nursing Child Assessment Feeding Scale (NCAFS) (Barnard et al. 1989), in which the observer watches a single instance of child feeding and rates the behavior according to carefully defined operational criteria on 76 items. The authors have defined a threshold (a score of 50 or less) which has been shown to distinguish high- and low-risk infant feeding behaviors (Farel et al. 1991). The scale is valid through 12 months of age. Use of the scale requires that the observer be trained to achieve a level of concordance with a previously certified trainer. Other scales from the United States include the MacPhee (25 items), the Crittendon (81 items), and the Chatoor (46 items) (MacPhee et al. 1993). The more items, the better the scale discriminates between those with and without feeding difficulties.

In developing countries, a similar approach has been used. Guldan et al. (1993) in Bangladesh found a number of caregiver practices that differed according to the mother's education, adjusting for household education, wealth, child age, birth order, and gender. Variables associated with maternal education included some from each of the three categories defined earlier: (1) adaptation to the child's psychomotor skills (number of finger food feedings per hour, child less likely to feed self, percent of time using cup, percent of time using bottle), (2) "responsiveness of the feeder" (noticing when food was dropped, whether mother initiated the feeding, caregiver less likely to be doing something else at the same time as breast-feeding), and (3) the feeding situation (location, cleanliness, absence of distraction). Guldan et al. (1993, 925) conclude that more educated mothers adopted "more attentive feeding practices" and "more labor-intensive child care," However, child behavior was not assessed.

The Guldan et al. (1993) study found frequency of feeding to be associated with maternal education. This variable has appeared to be associated with child anthropometric status in several studies. Frequency of feeding, observed during continuous observations, was associated with child nutritional status for children 18-23 months of age in Kenya (Ricci et al. 1994).

Engle and Zeitlin (1996) observed 37 different items per eating event in Managua, Nicaragua, and constructed a scale for active feeding behavior of the caregiver and child demand from a subset of the items. Whereas child demand was positively associated with nutritional status, active feeding was not; rather, active feeding was associated with a child's lack of interest in food, suggesting that caregivers in this situation may feed actively in response to child refusal.

Bentley et al. (1991) and Bentley et al. (1992) developed a scoring system to measure child and caregiver behaviors for each food rather than for each eating event. Rather than assessing the presence or absence of specific behaviors, they constructed a Guttman scale for child behaviors and for caregiver behaviors. The assumption behind a Guttman scale is that there is a logical order among dichotomously coded items and that they tend to always appear in that order. Theoretically a scale based on a correct implicit ordering will be more predictive than a scale based on a simple sum, as in the Engle and Zeitlin (1996) study. For the child, the three-point scale was food refusal, food appetite, and food request. For the caregiver, the scale was no response, verbal encouragement, verbal pressure, and physical force. The scale illustrated that caregivers were more active feeders when the children were convalescing from diarrhea than when they were healthy (Bentley et al. 1991).

Quantitative assessments of feeding behaviors. This approach counts the number of instances, rather than the presence or absence, of a behavior during a feeding episode (Polan and Ward 1994; Sanders et al. 1993; Klesges et al. 1986; Zeitlin, Houser, and Johnson 1989). If behaviors are relatively discrete, the number of instances of that behavior during a feeding episode, such as the number of times the mother touches the child, can be counted (Polan and Ward 1994). Sanders et al. (1993) rated the frequency of 14 parent and 17 child behaviors using the Mealtime Observation Schedule in Australia. For behaviors varying in duration, one can code whether or not the behavior is occurring after a fixed interval (for example, every five minutes). Zeitlin, Houser, and Johnson (1989) coded the presence of active feeding behaviors and child feeding behaviors every five seconds for Mexican infants. In each case, the measurement was able to differentiate between children who were growing well and those who were growing poorly.

Behavioral ratings. This technique rates the overall quality of the child-caregiver interaction. Black et al. (1994) have used it based on the coding scheme from the Parent Child Early Relation Assessment (PCERA) (dark 1985). The technique of behavioral ratings is widely used for the assessment of child behavior (for example, the Behavior Rating Scale from the Bayley Scales for Infant Development) (Bayley 1993) in the United States and has been used in developing countries (Engle et al. 1996). A domain of behavior is defined (for example, "Parent reads child's cues and responds sensitively and appropriately"), it is carefully described in behavioral terms (for example, "This variable is composed of parent's ability to accurately observe the child's cues, to understand what the child needs and wants, and to demonstrate the capacity to respond appropriately"), with some behavioral descriptions ("for example, if an infant squirms, or shows discomfort in the way he or she is held, a parent adjusts holding position; if an older infant tugs at mother's skirt, she responds to the need for attention by touching, holding, etc.") (Black et al. 1994). Finally, different levels of the item are defined, and the coder has to decide which represents the overall behavior (for example, 1 is insensitive to child, oblivious, indifferent, or unresponsive to child's cues; consistently misreads or misinterprets child's cues; 5 is very empathic, characteristically reads child's cues and responds sensitively and appropriately).

In order to use the scale, coders must have experience with the codes and know how to interpret them. Normally, they will be trained and must reach agreement with a previously trained coder before being able to rate these behaviors. If interactions are videotaped, coders should rate the child behavior separately from the caregiver's behavior (Black et al. 1995). Ratings from trained coders have been found to discriminate between children growing well and not growing well in the United States (Black et al. 1994). One of the reasons for using a rating system rather than predefined codes is that the caregiver's overall style, rather than specific behaviors, may be a better predictor of child outcomes. Black (personal communication) has compared the rating scale with a quantitative assessment of feeding behaviors and found no significant difference between the two; most of the variance in the quantitative measurement was captured by the rating scale.

Structured situation challenge. Under the rubric of "social marketing," the Manoff Group, a consulting corporation in Washington, D.C., has employed a strategy of presenting the caregiver with a challenge, such as a new food, and observing the caregiver's behavior to assess acceptability and potential problems with the introduction. Dickin, Griffiths, and Pivoz (1996) explain the technique in great detail and the way it should be used. Although one could code each observation systematically, the authors have used it in a more qualitative manner, assessing whether the innovation was acceptable and what kinds of problems appeared to arise in the situation. Validity of the method was not assessed.

These measurement techniques are, of course, prone to methodological problems. Does the caregiver being observed exhibit typical behavior? One can argue that even if the caregiver's behavior is altered, she is probably demonstrating her notion of ideal behavior, which can be revealing. For example, Black et al. (1994) found that 40 percent of mothers of children with failure to thrive were observed to neglect their children (interacting rarely) in spite of being videotaped in a clinical setting, a behavior they might have felt was not ideal. There does appear to be evidence for the validity of these measures, since a number of studies have indicated that they can discriminate between children with feeding difficulties and normally growing children (Heffer and Kelley 1994). A second concern with these rather brief observations is their reliability over time (would the child and caregiver behave the same way on a different day?). Barnard et al. (1989) found consistencies in maternal feeding behaviors during the first year of life, but more research is needed in this area. A third concern is the generalizability of these behaviors to a different kind of behavior (for example, is a caregiver who is responsive to the child during feeding also responsive to the child during play?). Black et al. (1995) found that there was more variability in maternal control in the play situation than the feeding situation, probably because feeding behaviors are more constrained, but maternal nurturance was similar in the two settings.

Other problems include deciding on which behaviors to code, establishing the reliability of coding of the behaviors, defining the categories operationally, and deciding on the appropriate setting in which to evaluate the behavior. It is essential that observers be well trained and standardized in order for the assessment to have validity (meaning). For suggested measures, see Tables 9 and 10.

Psychosocial Care

Psychosocial care refers to the provision of affection and warmth, responsiveness to the child, and the encouragement of autonomy and exploration. In this area, as in no other, is the importance of culture central. As noted earlier in the section on culture, a universalist perspective would suggest that the same function (for example, showing affection) may be demonstrated differently from one culture to another, and careful work with members of the cultural group is required to be sure that correct interpretations of behavior are made. For example, affection can be shown through physical, visual, and verbal contact with children. However, the patterns of expression may vary by culture. Becker and Becker (1994, 192) warn that "any given indicator of maternal attachment as derived in one racial group may not necessarily be valid in other racial groups."

Measurement of Psychosocial Care. Three categories of measures of psychosocial care can be described: (1) direct measurement of child-caregiver interaction, (2) assessment of the home environment, and (3) assessment of the child's appearance.

Observations of child-caregiver interaction. Typically, the frequency of caregiver behaviors, child behaviors, and child-caregiver interaction patterns are coded (Rahmanifar et al. 1992). Behaviors most often assessed are verbalizations of child and caregiver and looking and touching, although these behaviors will vary with the age of the child. A second technique is to present the caregiver with a task, such as asking her to play with a specific toy with the child, and then rate her behavior on a checklist (Barnard et al. 1989) or count the number of times an event occurs. Several studies suggest two child risk factors: a child is inactive ("doing nothing") a high percentage of the time, and a child is carried and held excessively after 18 months. High rates of holding and carrying of children beyond 18 months has been negatively associated with cognitive development and social interactions with other children (Wachs et al. 1991; Sigman et al. 1989), both because the holding and carrying restricts the opportunities for learning and because children who choose to be held tend to be lighter and are more likely to be ill (Sigman et al. 1989).

Assessment of the home environment. The most commonly used global assessment of the living situation is the Home Observation for Measurement of the Environment (HOME) scale. Using this scale in an interview with the caregiver, both the environment and incidence of positive effect between caregiver and child are rated (Bradley and Caldwell 1984). The HOME scale assesses both the emotional responsive-ness of the caregiver and the characteristics of the environment that are supportive of autonomy and exploration by the child (including avoidance of punishment, provision of appropriate play materials or location, opportunities for variety in daily stimulation, and organization of the physical and temporal environment). It has been adapted and used in many countries and has had consistent positive associations with cognitive development in a variety of settings (Bradley and Caldwell 1984; Bradley et al. 1989), in addition to controlling for the effects of socioeconomic status.

Child appearance. Several studies have suggested that the appearance of a child, either rated in a public place, such as the public health clinic (Joshi 1994), or rated over a number of instances in the home (Alien et al. 1992) may be an indicator of care (or lack thereof). Having developed a checklist to rate child appearance, Alien et al. were able to distinguish "poor but clean and well cared for" children from "wealthier but unkempt" ones. They found highly significant associations between child appearance during the second and third years of life and cognitive development at 30 months. Using multiple regressions, these associations were significant even when controlling for socioeconomic status, nutrient intake of animal fats, and body length at 18 months, which would usually account for most of the variance in cognitive development. For suggested measures, see Tables 11 and 12.

Table 11 - Psychosocial care: Child and caregiver interactions

Construct

Measurement tools

Comment

Caregiver-child interaction

Naturalistic observation of caregiver and child for a short period; code variables such as delay in responding, type of response, and level of vocalization by caregiver and child

Depends on age of child and context; reliable and valid measures have been developed; key variables are time without interaction, "doing nothing"

Caregiver-child interaction rating scale

Rating of caregiver behavior in structured teaching or play situation

Allows valid measures in a brief time; limited if structured situation is too distinct from child's life

Overall rating of environment

HOME scale (standardized rating scale)

Has been used in a variety of cultures; needs to be adapted to each setting carefully

Caregiver's understanding of developmental milestones

Assessment of caregiver's judgments about stages of development

May be linked to parental stimulation of cognitive development

Table 12 - Psychosocial care: Child variables

Construct

Measurement tools

Comment

Alertness or drowsiness

Brazelton Assessment Scale for newborns; observations of child behavior

More caregiver interaction with more alert children

Verbalization

Count verbalizations during observation

Measurement depends on age of child

Rate of motor and mental development

Test, observation

Earlier achievement of motor development associated with cognitive development

Physical health or disability status

Survey, observation, clinical observation

Depends on particular disability