7
Behavioral Indicators of Diet and Physical Activity

Chapters 5 and 6 attempted to show that, for the purpose of determining eligibility for WIC services, one cannot make a valid and reliable assessment of an individual’s diet or physical activity patterns using conventional diet and activity assessment tools. For this reason in part, it has been suggested that there may be alternative practical measures that are strongly correlated with diet and activity that could be used in determining eligibility for WIC services. One such method would be the use of behavioral indicators. This chapter explores the concept of behavioral indicators and their possible role in the assessment of an individual’s dietary and physical activity patterns for the purposes of determining failure to meet Dietary Guidelines and therefore WIC eligibility.

Despite the theoretical and practical attractiveness of a behavioral model for dietary risk assessment, only limited research has been conducted to confirm the relationships among behavioral variables, dietary adequacy or appropriateness, nutrient intake, and health outcomes. There is no one instrument with demonstrated validity and reliability that assesses the many behavioral aspects of diet. In most cases in the literature, these practices or patterns were collected as an adjunct to or coded from other more lengthy dietary assessment methods.

Taken together, the Dietary Guidelines, the current WIC eligibility criteria, and the charge of the committee place more emphasis on diet than on activity. Accordingly, this review of behavioral indicators focuses more on diet than on activity. After reviewing the literature on behavioral indicators of diet, this chapter concludes with brief examples of how behavioral indicators might also



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 93
Dietary Risk Assessment in the WIC Program 7 Behavioral Indicators of Diet and Physical Activity Chapters 5 and 6 attempted to show that, for the purpose of determining eligibility for WIC services, one cannot make a valid and reliable assessment of an individual’s diet or physical activity patterns using conventional diet and activity assessment tools. For this reason in part, it has been suggested that there may be alternative practical measures that are strongly correlated with diet and activity that could be used in determining eligibility for WIC services. One such method would be the use of behavioral indicators. This chapter explores the concept of behavioral indicators and their possible role in the assessment of an individual’s dietary and physical activity patterns for the purposes of determining failure to meet Dietary Guidelines and therefore WIC eligibility. Despite the theoretical and practical attractiveness of a behavioral model for dietary risk assessment, only limited research has been conducted to confirm the relationships among behavioral variables, dietary adequacy or appropriateness, nutrient intake, and health outcomes. There is no one instrument with demonstrated validity and reliability that assesses the many behavioral aspects of diet. In most cases in the literature, these practices or patterns were collected as an adjunct to or coded from other more lengthy dietary assessment methods. Taken together, the Dietary Guidelines, the current WIC eligibility criteria, and the charge of the committee place more emphasis on diet than on activity. Accordingly, this review of behavioral indicators focuses more on diet than on activity. After reviewing the literature on behavioral indicators of diet, this chapter concludes with brief examples of how behavioral indicators might also

OCR for page 93
Dietary Risk Assessment in the WIC Program apply to physical activity assessment, and it provides a more detailed review of one important potential indicator of activity—television viewing. THE CONCEPT OF BEHAVIORAL INDICATORS As previously discussed, diet and physical activity are both extremely complex behaviors expressed as systematic patterns that are the end result of a complex series of many decisions (Baranowski, 1997b; Campbell and Desjardins, 1989). These decisions are affected by contextual factors that can be considered behavioral indicators, in that these indicators influence or reflect diet or activity but do not attempt to directly measure diet or activity. For example, many contextual factors affect a person’s diet, such as where one eats; who else is present and why; the cost, convenience, or familiarity of certain foods; and the presence of emotional states, such as loneliness or boredom, that can serve as eating cues. Similarly, activity levels can also be affected by contextual factors like the weather, the availability of safe outdoor areas, the support or interest of family and peers, and the presence of competing sedentary activities such as television viewing. Interest in using these behavioral indicators in WIC may also be increased by the untested assumption that, in comparison to conventional tools for assessing diet and activity, these indicators may be easier to recall, less susceptible to various types of reporting bias, and therefore most appropriate targets for behavioral counseling. A distinction can be drawn between surrogate and target behavioral indicators (See Box 7-1). Surrogate indicators are those that can be used in place of usual dietary or physical activity assessment procedures. For example, the frequency of eating a meal as a family is a possible surrogate indicator because it has been shown that families who eat dinner together tend to eat better diets (Gillman et al., 2000). If the frequency of eating family meals could be assessed more reliably than what foods a person usually eats, and if family meal eating BOX 7-1 Definitions of Two Behavioral Indicators Surrogate Behavioral Indicators indicators that are correlated with one or more aspects of diet or activity and could be used to measure those aspects of diet or activity Target Behavioral Indicators indicators that determine one or more aspects of diet or activity and, if changed, would result in changes in diet or activity

OCR for page 93
Dietary Risk Assessment in the WIC Program BOX 7-2 Criteria for Establishing Surrogate or Target Behavioral Indicators Surrogate Behavioral Criteria behavior is substantially correlated with some aspect of diet or activity behavior is consistently correlated with some aspect of diet or activity behavior is more reliably assessed than corresponding aspect of diet or activity Target Behavioral Criteria behavioral indicator causes some aspect of diet or activity behavioral indicator is modifiable changes in the behavioral indicator result in substantial change in the diet or activity could consistently and substantially discriminate between the higher consumption of certain foods (e.g., fruit and vegetables) and lower consumption of other foods (e.g., low nutrient-dense foods), then assessment of frequency of eating a meal as a family could be used as a surrogate for assessment of actual food intake when determining dietary risk (see Box 7-2). In evaluating the potential of using surrogate indicators for the purposes of determining WIC eligibility, one important issue is the level of reliability and validity of these surrogate measures in comparison to the conventional food-based assessment procedures such as dietary recalls. If the validity and reliability of the surrogate indicators are not higher than those for food-based assessment procedures, then there is little advantage to using the surrogate. If the surrogate is not substantially correlated with true consumption, then misclassification error increases substantially. Target indicators are those that identify precursors of diet or activity, which, if changed, result in improved dietary intake or levels of physical activity (Nicklas et al., in press; Siega-Riz et al., 2000). If behavioral indicators are causative of the diet or activity patterns, the behaviors are modifiable, and if the changes result in improved diet or activity practices, then they could be targets for WIC-related nutrition education efforts (see Box 7-2). To continue with the prior example, if families could be easily encouraged to more frequently eat meals together, and increased family dinners resulted in improved dietary intake, then frequency of eating a meal as a family is a likely target indicator for change. The key issue in selecting a target indicator is whether the behavioral indicator is modifiable and whether a change in the indicator results in a change in diet or activity. If a change in the target indicator is possible but is not substantially related to an alteration in diet or activity, then there would be little reason to attempt to change the target indicator. Virtually any correlate of diet or

OCR for page 93
Dietary Risk Assessment in the WIC Program activity behavior can be considered for status as a surrogate or target, but must be demonstrated to meet the corresponding criteria for such status (see Box 7-2). BEHAVIORAL INDICATORS OF DIET Categories of Behavioral Indicators of Diet This literature review attempts to identify a variety of examples of behavioral diet indicators that may be considered for surrogate or target indicators for use in the WIC program. Possible surrogate or target behavioral practices were each placed within one of the following categories: indicator foods; food, eating, or dietary patterns; meal patterns; health-related behaviors; psychosocial characteristics; parent food practices; ecological factors; or alternative technology. Most of the research on behaviors as possible surrogates for measures of dietary intake was not conducted with the intent of validating surrogates. Rather, it was conducted within the framework of understanding correlates of dietary intake. There are not many such studies. All of the methods used as indicators of validity for this purpose were self-reported (subjective), including food records. Table 7-1 summarizes the categories and provides the range of validity and reliability coefficients. Table 7-2 provides examples of indicators for each category as well as the references in which the indicator was studied. Indicator foods are single foods, consumed either for specific meals (e.g., eggs for breakfast) or during the day as a whole (e.g., red meat), that are related to variations in dietary intake usually in regard to nutrients (e.g., eat more total calories). Since indicator foods reflect rather than determine diet, they have potential as surrogate, but not target, indicators. Food, eating, or dietary patterns are either groups of foods commonly eaten together (usually based on a statistical procedure called factor analysis), or groups of people who commonly eat certain types of food (usually based on a statistical procedure called cluster analysis), or from logically placing practices together related to a particular nutrient (e.g., dietary fat). Since these patterns reflect rather than determine diet, they have primary interest as surrogate indicators. Meal patterns describe some aspect of an individual’s meal behavior other than consumption of specific foods or categories of foods. An example of a meal pattern would be “not eating breakfast.” Meal patterns could be surrogate or target behavior indicators, depending on whether the practice determines the dietary intake of interest. Health-related behavior concerns the assessment of other behaviors related to health. For example, smoking is correlated with one or most aspects of dietary

OCR for page 93
Dietary Risk Assessment in the WIC Program TABLE 7-1 Categories of Behavioral Indicators of Diet Behavioral Indicator Range of Reliability Coefficients Range of Validity Coefficients Indicator foods Consumption of specific foods (either at a specific meal or for the day as a whole) Not reported Not reported Food, eating, or dietary patterns Groups of foods or categories of foods that are usually consumed or practiced together Most none α = 0.54–0.76 trt = 0.67–0.90 Substantial relationships between factors and consumption Partial correlation r = -0.29 to -0.68 from 0.10 to 0.39 Meal patterns Differences in meal or snack consumption Not reported Partial r ≤ 0.10 Health-related behaviors Some other health-related behavior Not reported Not reported Psychosocial characteristics Psychosocial variables related to food intake Clusters of such variables From < 0.10 to > 0.90 Most are ≤ 0.30 Parent food practices Parent behavior in regard to some aspect of child’s dietary behavior From < 0.10 to > 0.90 Most are ≤ 0.30 Ecological factors Aspects of the home or neighborhood From < 0.5 to 0.9 for some indicators More are ≤ 0.30 intake. It appears that health-related behaviors reflect, rather than cause, dietary behavior and thereby have more potential as surrogate indicators. Certain psychosocial characteristics, such as food preferences or self-efficacy related to altering diet, have been related to dietary intake. Likewise, aggregates or clusters of these psychosocial characteristics have also been shown to correlate with dietary intake in marketing studies. Although the causal status of these psychosocial characteristics has not been clearly demonstrated, if established, they could become target indicators.

OCR for page 93
Dietary Risk Assessment in the WIC Program TABLE 7-2 References for Data Bearing on Behavioral Nutrition Indicators by WIC Target Group and Category of Indicator Practices 25–48 months Children/Adolescents Indicator foods Eating eggs for breakfast — — Eating ready-to-eat cereal for breakfast — Nicklas et al., in press Eating coffee, soft drink, or dessert alone for breakfast — — Added sugar — Forshee and Storey, 2001 Milk consumption — Ballew et al., 2000 Juice consumption — Ballew et al., 2000 Soft drink consumption — Ballew et al., 2000 Food, eating, or dietary patterns Many factors — — Prudent diet factor — — Western diet factor — — Fat practices — — Many clusters — — Dietary adequacy — — Vegetarian — Jacobs and Dwyer, 1988 Meal patterns Meal consistency — Siega-Riz et al., 1998 Regularly eat breakfast — Sampson et al., 1995 Snacking — — Eating out of home frequently — — Eating fast food frequently — — Eating span — Berenson et al., 1980 Longest fast > 13 hours — — Related behaviors (self) Smoking — — Physical activity — Rosmond et al., 2000 Eating while watching television — Coon et al., 2001

OCR for page 93
Dietary Risk Assessment in the WIC Program Pregnant Lactating Postpartum Adults   — — — Siega-Riz et al., 2000 — — — Siega-Riz et al., 2000 — — — Siega-Riz et al., 2000 — — — — — — — — — — — — — — — —   — — — Randall et al., 1990, 1991b; Wolff and Wolff, 1995 — — — Fung et al., 2001 — — — Fung et al., 2001 — — — Kristal et al., 1990 — — — Huijbregts et al., 1995; Millen et al., 2001; Slattery et al., 1998; Wirfalt and Jeffery, 1997 — — — Knol and Haughton, 1998 — — — Donovan and Gibson, 1996; Janelle and Barr, 1995   — — — — — — — Nicklas et al., 1998; Siega-Riz et al., 2000 — — — Zizza et al., 2001 — — — Clemens et al., 1999; McCrory et al., 1999 — — — French et al., 2000 — — — — — — — —   Haste et al., 1990 — — Huijbregts et al., 1995; Ma et al., 2000; Randall et al., 1991a; Tucker et al., 1992 — — — Matthews et al., 1997; Rogers et al., 1995; Slattery et al., 1998 — — — —

OCR for page 93
Dietary Risk Assessment in the WIC Program Practices 25–48 months Children/Adolescents Psychosocial characteristics and clusters Psychosocial variables — Baranowski et al., 1999 Psychosocial clusters — — Parent food practices — Baranowski, 1997a; Nicklas et al., 2001 Ecological factors Availability of whole fruit, 100% juice, and vegetable (FJV) at home — Hearn et al., 1998; Kratt et al., 2000 Availability of FJV in local restaurants — Edmonds et al., 2001 Socioeconomic status — — Household food insecurity — — Parent food practices concern parent behaviors in regard to their child’s food consumption (e.g., an authoritative parenting style in which both emotional support and limit-seeking occur together). Parent food practices could be responses to child food behavior or could be causative. If shown to be causative, they could be surrogate or target indicators. Ecological factors are aspects of the family or home environment related to food intake (e.g., home availability or accessibility of certain foods). If ecological factors are demonstrated to be causative of dietary behavior, they could become surrogate or target indicators. Review of Literature on Behavioral Indicators of Diet Published studies on behavioral indicators of diet in WIC target groups are very limited. While studies on correlates of diet have been conducted with children older than 5 years of age, there are very few addressing children under 5 years. In addition, much of the work on correlates of diet in adults has occurred among adults in general, not usually among adults in the WIC targeted categories (see Table 7-2). For this reason, the following literature review will cover children older than those served by WIC and covers women in general rather than low-income women specifically. Since many of these behavioral indicators of diet have been abstracted from other assessment instruments (e.g., 24-hour dietary recalls, food frequency questionnaires), few indicators of reliability have been reported in the literature. Additionally, very few estimates of the strength of relationships of these

OCR for page 93
Dietary Risk Assessment in the WIC Program Pregnant Lactating Postpartum Adults   — — — Baranowski et al., 1999 — — — Dutta and Youn, 1999; Glanz et al., 1998 — — — Baranowski, 1997a; Nicklas et al., 2001   — — — — — — — — — — — Haste et al., 1990; Hupkens et al., 1997 — — — Casey et al., 2001; Derrickson et al., 2001; Lee and Fongillo, 2001; Tarasuk and Beaton, 1999 indicators with diet have been reported, and those that have been reported used different indicators of strength (e.g., correlation coefficients, F test values). Thus, no attempt was made to systematically report reliability coefficients or the strengths of relationships. Indicator Foods Indicator foods have been used to understand variations in what people eat. In general, consumption of an indicator food has been associated with intakes of nutrients, food groups, or energy. Thus, briefly assessing consumption of an indicator food could provide an index of consumption of a WIC nutrient or food group. For example, data from 24-hour recalls from a nationally representative sample of adults in the 1994–1996 Continuing Survey of Food Intakes by Individuals (CSFII) (n = 15,641) indicated that adults ate the following selected breakfast items: eggs (15 percent); ready-to-eat cereal (17 percent); bread only (22 percent); cooked cereal (4 percent); fruit only (6 percent); coffee, soft drink, and/or high-fat dessert (15 percent); or other (3 percent) (Siega-Riz et al., 2000). Those eating an egg-based breakfast consumed more total calories and had a higher percentage of calories from fat for breakfast, but lower carbohydrates, calcium, folate, and iron. Those consuming coffee, soft drink, and/or dessert consumed the fewest breakfast calories, lower daily intakes of protein, fiber, and folate, and the highest intake of saturated fat (Siega-Riz et al., 2000). The consumption of ready-to-eat cereal for breakfast was associated with consumption of lower total fat, and more folic acid, iron, niacin, vitamin A,

OCR for page 93
Dietary Risk Assessment in the WIC Program vitamin C, and zinc per dollar spent than those who ate fast food or other breakfasts (Nicklas et al., in press). Combining all children (> 2 years old) and adults in the 1994–1996 CSFII data set, and after controlling for age, gender, and consumption of other macronutrients, people who ate more added sugar also tended to consume more grains and lean meat, but less vegetables, fruit, and dairy, and more vitamin C and iron, but less vitamin A, calcium, and folate (Forshee and Storey, 2001). These relationships varied, however, among children by age of the child from 6 to 11 years versus 12 to 19 years. All relationships were weak. Children in the CSFII data set who drank more milk were more likely to have significantly higher intakes of vitamin A, folate, vitamin B12, calcium, and magnesium (in all age groups: 2–5 years, 6–11 years, and 12–17 years). Similarly, children in all age groups who drank more 100 percent juice were more likely to consume more vitamin C and folate (Forshee and Storey, 2001). However, children who drank more 100 percent juice were less likely to consume vitamin B12 among 12- to 17-year-olds, but not among other age groups. Children who drank more carbonated beverages were significantly less likely to consume vitamin A among all age groups. In those 2 to 11 years old, but not those 12 to 17 years old, carbonated beverage drinkers had lower vitamin C and calcium intakes (Ballew et al., 2000). For all age groups, children who drank more carbonated beverages were less likely to consume milk and 100 percent fruit juice (Ballew et al., 2000). Food, Eating, or Dietary Patterns Food, eating, or dietary patterns are consistent groupings of foods, usually determined by statistical techniques. Using food frequency data from 2,255 adults in the Western New York Diet Study (1975–1986), Randall and colleagues (1991b) demonstrated substantial intercorrelations in consumption among food groups and nutrients. Using the same data set, they reported a principal components analysis across 110 foods. Also called factor analysis, this technique identifies common patterns in foods consumed. Nine factors were extracted, which they interpreted as (1) salad, (2) Southern European/healthful, (3) fruit, (4) low cost, (5) dessert, (6) staple vegetables, (7) costly, (8) health foods, and (9) nonuse (Randall et al., 1990). Each factor was significantly correlated with several key macro- and micronutrients (e.g., energy, dietary fat, dietary fiber, and vitamins A and C). Some differences in factor structures were determined between males and females (Randall et al., 1991b). These factors correlated in expected directions with National Cancer Institute-specified consumption of dietary fat; dietary fiber density; vegetable diversity; fruit diversity; alcoholic beverage; cured, pickled, and smoked meats and fish and charbroiled meat and poultry; and sodium (Randall et al., 1991b). Using tertiles

OCR for page 93
Dietary Risk Assessment in the WIC Program on the factor scores, some of these relationships appeared to be strong, for example, high fat factor with dietary fat (F = 183.1); fruit factor with dietary fiber density (F = 108.4); salad factor (F = 262.4), healthful factor (F = 357.6), and traditional factor (F = 206.3) with vegetable diversity; fruit factor with fruit diversity (F = 758.8); and high fat factor with total alcohol consumption (F = 133.3) (Randall et al., 1991b). Using food frequency questionnaires (FFQs) from the Hispanic Health and Nutrition Examination Survey (Hispanic HANES), eating patterns were identified using factor analysis in 49 Mexican-American mothers (Wolff and Wolff, 1995). Seven eating pattern factors were extracted: nutrient-dense, traditional, transitional, nutrient-dilute, protein-rich, high-fat dairy, and mixed dishes. After controlling for demographic and related variables, the nutrient-dense (fruit, vegetables, low-fat dairy) and protein-rich eating patterns were associated with increased birth weight, while the transitional eating pattern was associated with decreased birth weight (Wolff and Wolff, 1995). The relationships, however, were weak. Based on the anthropological theory of core foods, an 18-item questionnaire was developed to assess aspects of reduced dietary fat practices (Kristal et al., 1990). A confirmatory factor analysis of these items revealed five factors: (1) avoiding fat as a seasoning, (2) avoiding meat, (3) modifying high fat foods, (4) substituting high-fat foods with specially manufactured low-fat foods, and (5) replacing high-fat foods with low-fat alternatives (Kristal et al., 1990). The five scales had modest Cronbach alpha reliabilities ( = 0.54–0.76) and higher test– retest reliabilities (trt = 0.67–0.90). The validity correlations with percent energy as fat varied from -0.29 to -0.68 (Kristal et al., 1990). At least one study found that variables from such food behavior questions correlated with breast cancer rates, while nutrients from FFQs did not (Byrne et al., 1996). Other investigators have employed cluster analysis as a statistical technique for identifying dietary patterns. Cluster analysis groups people into relatively homogeneous categories of consumption. One group found four cluster-determined dietary groups among noninstitutionalized senior citizens: (1) alcohol, (2) milk, cereals, and fruit, (3) bread and poultry, and (4) meat and potatoes (Tucker et al., 1992). Those in the milk, cereal, and fruit cluster had the highest intake of micronutrients and the best hematologic profile. Those in the meat and potatoes cluster had the lowest intake of micronutrients, while those in the bread and poultry cluster had the lowest reported energy intake, but the highest body mass index (BMI) (Tucker et al., 1992). Pooling FFQ data from three studies conducted with adults in the upper Midwest, six food cluster groups were determined: (1) high intake of soft drinks, (2) high intake of pastries, (3) high intake of skim milk, (4) high intake of meat, (5) high intake of meat and cheese, and (6) high intake of white bread (Wirfalt and Jeffery, 1997). Participants in the high soft drink consumption cluster had low intake of protein, fiber, and calcium and higher BMI among men, but not women. Those in the

OCR for page 93
Dietary Risk Assessment in the WIC Program high meat and high pastry clusters had higher dietary fat intake (Wirfalt and Jeffery, 1997). As part of the Framingham Offspring-Spouse study of 1,828 adult women, five eating pattern clusters were determined: (1) heart healthy, (2) light eating, (3) wine and moderate eating, (4) high-fat, and (5) empty calories (Millen et al., 2001). Statistically significant differences across eating patterns were detected for a broad variety of macro- and micronutrients and cardiovascular disease risk factors. Some of these relationships were strong, but most were not (Millen et al., 2001). One dietary pattern cluster that is easy to recognize is vegetarianism. There are several forms of vegetarianism (e.g., vegans use no animal products, lacto-vegetarians use milk products in addition to plant products, lacto-ovo-vegetarians use milk and egg products in addition to plant products, and various more restrictive practices [Jacobs and Dwyer, 1988]) that potentially could be easily identified through one self-report question. Among women in western Canada with better health practices in general, vegetarians tended to have lower BMI, and especially lower percent body fat (Janelle and Barr, 1995). Among female adolescents in southern Ontario, there were no statistically significant differences in energy consumed across lacto-ovo-vegetarians, semi-vegetarians, and omnivorous groups (Donovan and Gibson, 1996). Lacto-ovo-vegetarians tended to consume less protein and niacin, but more dietary fiber, copper, and manganese than omnivorous groups, but these differences were small. Semi-vegetarians had the greatest risk of inadequate calorie, protein, iron, zinc, and vitamin C intake (Donovan and Gibson, 1996). Food Patterns Food patterns were developed to identify naturally occurring groupings in foods consumed. Interest in food patterns has arisen, in part, because data on the intake of single nutrients are very limited in their ability to predict the development of those chronic diseases that are suspected to be related to diet (Jacques and Tucker, 2001). Food patterns related to health issues of concern to WIC (e.g., obesity) may provide important targets because they are the foods of interest. Because they would be dependent on tools similar to FFQs, it appears unlikely that food patterns could be valid or reliable enough to be used for eligibility determination. Meal Patterns Meal patterns characterize aspects of meals such as whether specific meals or snacks were consumed, where the meals were consumed, or over what time

OCR for page 93
Dietary Risk Assessment in the WIC Program period during the day one ate. For example, assessing whether a child consistently ate breakfast could be an easy way to assess total caloric intake. Using three 24-hour diet recalls from the 1989–1999 CSFII, meal patterns were analyzed among a nationally representative sample of 1,310 adolescents (11–18 years of age) (Siega-Riz et al., 1998). On any one day, about 58 percent of adolescents ate three meals plus snacks. The second most common meal pattern was breakfast, dinner, and snacks (about 15 percent). About 3 percent ate only one meal or only snacks on any particular day. Adolescents were categorized into consistent, moderately consistent, and inconsistent meal patterns based on how many meals were consumed across the 3 days of assessment. Adolescents with consistent meal plans ate more total calories, fiber, calcium, iron, vitamin E, fruit and vegetable servings, grain and legume servings, and sodium, but had a lower diet quality index (Siega-Riz et al., 1998). Skipping Breakfast. Whether a person skipped breakfast or not could be assessed by a single question. In the 1994–1996 CSFII data with a nationally representative sample of 15,641 adults, 17 percent did not consume breakfast, but this percentage was higher in 18- to 40-year-olds (23 percent) than in 41- to 65-year-olds (13 percent) (Siega-Riz et al., 2000). Among 509 young adults in Bogalusa, Louisiana (studied from 1988–1991), using 24-hour diet recalls, 37 percent skipped breakfast the previous day (Nicklas et al., 1998). Those not eating breakfast consumed 568 fewer calories per day, less protein, less saturated fat, but 121 mg more cholesterol. Men who skipped breakfast consumed less total fat than all females or males eating breakfast (Nicklas et al., 1998). Those not eating breakfast were less likely to consume two-thirds or more of the Recommended Dietary Allowance (RDA) for a variety of micronutrients, but even among those eating breakfast, those achieving two-thirds or more of the RDA varied from approximately 40 to 90 percent (Nicklas et al., 1998). Among 1,151 mostly lower-income African-American second to fifth grade children in New Jersey who completed a 24-hour diet recall and 4 days of food surveys (Sampson et al., 1995), children not eating breakfast before school varied from 22 to 26 percent per day. Across all 4 days, 71 percent reported eating breakfast each day and 4 percent reported eating breakfast for none of the days, with no differences by gender. Children who skipped breakfast had significantly lower daily intakes of calories and micronutrients. While those who skipped breakfast consumed a higher percentage of calories from fat, they had lower intakes of cholesterol and sodium. There was no difference between groups in BMI (Sampson et al., 1995). Snacking Patterns. Using 24-hour diet recall data from three CSFII data sets (1977–1978, 1989–1991, 1994–1996), snacking patterns were assessed in nationally representative samples of young adults (19–29 years of age) (Zizza et al., 2001). The percentage of young adults who did not snack across the multiple days of assessment changed from 23.4, to 25.8, to 15.6 percent across the three

OCR for page 93
Dietary Risk Assessment in the WIC Program time intervals. The average number of snacking occasions per day increased from 1.70, to 1.69, to 1.92. The calories consumed per snacking occasion increased from 247, to 265, to 313. This increase in calories was accounted for in part by the caloric density per gram of snack food, which increased from 1.05, to 1.30, to 1.32. The caloric density per gram of food at meals remained a near constant 1.11 to 1.13 across the three time intervals. In each year, those who snacked consumed more energy, more carbohydrates (but not as a percentage of calories), more fat (but not as a percentage of calories), and more saturated fat (Zizza et al., 2001). Meals Away from Home. Using a 7-day food record from 129 young women (average age 30 years), the sample was divided into more frequent events (6–13 times per week), or less frequent events (≤ 5 times per week) of eating meals outside the home (Clemens et al., 1999). Women who ate out more frequently consumed more total calories, fat, carbohydrates, protein, and sodium, but not fiber or calcium. Using an FFQ with 73 healthy men and women, there was substantial variability in the rate of eating outside the home: 7.5± 8.5 times per month (McCrory et al., 1999). After statistically controlling for demographic variables, people who ate outside the home more frequently had higher BMIs (r = 0.36). After also controlling for amount of physical activity, the relationship increased (r = 0.42). People who ate outside the home more frequently consumed more energy (r = 0.59), more dietary fat (r = 0.28), and less fiber (r = -0.45) (McCrory et al., 1999). Focusing more specifically on fast food consumption, FFQs were given to 891 women (20 to 45 years of age) enrolled in a weight gain prevention study. Results indicated that 21 percent of the women had three or more fast food-eating events in a week. Additionally, 16 percent reported two events, 39 percent reported one event, and 24 percent reported none (French et al., 2000). Women who reported the highest number of fast food-eating events (highest tertile or X = 3.3 events) consumed significantly more total calories, fat as a percentage of calories, hamburgers, french fries, and soft drinks. Additionally they consumed less dietary fiber, vegetables, and fruit (French et al., 2000). Women who increased the frequency of eating out over 3 years consumed more total calories, percent of energy as fat, hamburgers, french fries, and soft drinks, and less vegetables. Weight increased with increased fast food consumption (French et al., 2000). Eating Span. Using 24-hour diet recalls among 10-year-old children in the Bogalusa study, eating span was defined as the number of hours from first food or beverage consumed to last consumed (Berenson et al., 1980). Three groups were identified: short span (10 hours or less, n = 19); moderate span (10 to 13 hours, n = 95), and long span (13 hours or more, n = 71). There were no gender

OCR for page 93
Dietary Risk Assessment in the WIC Program differences in span. Children with a longer eating span consistently consumed 40 percent more than short span and 20 percent more than moderate span children of calories, protein, fat, carbohydrate, and sodium. Children with the longer eating span also had higher total serum cholesterol (Berenson et al., 1980). Pregnant women who regularly fasted more than 13 hours (overnight) were more likely to have premature and small-for-gestational-growth babies (Siega-Riz et al., 2000). Although this should be a relatively easily measured phenomenon, little has been published about it or its measurement. There are likely other behavioral phenomena that have similar health implications. While research needs to identify the physiological and behavioral processes that account for this adverse outcome, behavior change programs can be initiated to attempt to change these fasting patterns and in turn assess the extent to which change in the fasting pattern results in improved pregnancy outcome. Related Health Behaviors Research has revealed that there are patterns across several health behaviors. For example, people who smoke have less healthy diets (Ma et al., 2000). Since regular smoking can be relatively easily assessed, it is possible that this could provide important surrogate or target behavioral indicators. Among the previously described dietary pattern studies using FFQs, smoking was negatively correlated with the salad, fruit, healthful, and whole grain factors, and positively correlated with the high-fat factor among men, and with similar patterns among women (Randall et al., 1991a). Nonsmokers were more common in the healthy diet cluster even after controlling for possible confounders (Huijbregts et al., 1995). Smoking was most common in the alcohol cluster (Tucker et al., 1992). Using two 24-hour diet recalls in the 1994–1996 CSFII, whether a person smoked (current, former, or non) or drank alcohol (abstainers, occasional, moderate, or liberal drinkers) were assessed among a nationally representative sample of 6,745 adults (19 years of age and older) (Ma et al., 2000). The smoking and drinking subgroups differed in age, income, education, BMI, exercise, and other behavior, which could confound other relationships. Current and former smokers drank more alcohol than nonsmokers. Men and women who smoked the most cigarettes reported the lowest consumption of fruit, carotenes, and vitamin C. Men and women who drank the most alcohol reported the lowest consumption of fruit; grain; carbohydrates, fat and protein as a percent of calories; carotenes; and dietary fiber (Ma et al., 2000). Using a 7-day weighed dietary intake record of women in London at 28 (n = 206) and at 36 weeks (n = 178) of gestation, intake was compared between smokers (≥ 15 cigarettes per day, n = 83) and nonsmokers (n = 101) (Haste et al., 1990). Smokers consumed less macro- and micronutrients, whether

OCR for page 93
Dietary Risk Assessment in the WIC Program expressed as nutrients or nutrient density, even after controlling for social class (Haste et al., 1990). In one of the diet pattern studies described earlier, leisure time physical activity was higher in those consuming a prudent diet (Slattery et al., 1998). Using an FFQ with 211 less well-educated adult patients in the General Medicine Clinic in Syracuse, New York, diet and physical activity were assessed (Rogers et al., 1995). Patients with less physical activity were less likely to consume vegetables, fruit (especially those 20 to 49 years of age), and high-fiber grains (Rogers et al., 1995). Using a 7-day diet recall and one 24-hour diet recall with 919 adults in the WATCH hyperlipidemia trial, physical activity was assessed using a new questionnaire that assessed type, duration, and intensity of regular activity (Matthews et al., 1997). Those reporting more minutes of physical activity reported consuming less fried foods, but there were nonlinear patterns of consumption for other food groups. The nonlinear patterns were also obtained for nutrients consumed (Matthews et al., 1997). In a randomly selected sample of 40-year-old women (n = 1,464) in Sweden, reports of fighting and playing with boys in childhood were positively related, and playing with girl toys and other girls during childhood were negatively related, to the probability of being overweight (Rosmond et al., 2000). Whether the television was on while 10-year-old children ate breakfast, afternoon snacks, or dinner was assessed in a sample of 91 parent–child pairs, along with three 24-hour diet recalls (Coon et al., 2001). Children for whom the television was on for two or three of the eating occasions consumed less fruit, vegetables, and juice and more meat, pizza, snacks, and carbonated beverages (Coon et al., 2001). Psychosocial Correlates A recent review of psychosocial correlates (e.g., self efficacy or preference) of intake of dietary fat, fruit, juice, and vegetables revealed a large body of research employing many different psychosocial constructs. However, most of these relationships were weak to moderate (Baranowski et al., 1999). As a result, there appears to be no advantage to using these as surrogate diet indicators. Alternatively, psychosocial variables have been proposed as the most likely mediators of dietary change interventions (Baranowski et al., 1997, 1998). Thus, better understanding of these relationships may lead to more effective dietary change interventions, which suggests they could be target indicators. A particularly promising avenue of research comes from the realm of social marketing. Market segmentation is a marketing technique that divides the population into relatively homogeneous groups that are related to diet in various ways (Dutta and Youn, 1999; Glanz et al., 1998). The clustering often uses

OCR for page 93
Dietary Risk Assessment in the WIC Program psychosocial variables, and is referred to as “psychographics” (Dutta and Youn, 1999). Thus, empirically identifying profiles of groups of people based on several psychosocial correlates of diet may lead to different, and hopefully more effective, interventions for each group. At the present time, it is not yet clear how one would design interventions using such profiles. Parent Food Practices Reviews of family correlates of dietary intake have also appeared recently (Baranowski, 1997a; Nicklas et al., 2001). In a vein similar to the psychosocial variables, the documented relationships have been mostly weak to modest, thereby precluding their use as surrogate diet indicators. Better understanding of these relationships, however, also could suggest improved targets for promoting dietary change in the families of children and in spousal pairs (Baranowski and Hearn, 1997). It appears likely that, since dietary behaviors and practices are established early in life, they have a long-term influence on diet (Baranowski et al., 2000; Costanzo and Woody, 1984). Thus, research on parent food practices as targets for intervention has particular promise for having long-term impacts on the health of both adults and children. (Baranowski et al., 2000). Ecological or Environmental Correlates Other possible sources of behavioral indicators of diet are ecological or environmental variables. For example, it could be possible that assessing whether certain foods are available in the home could provide surrogate indicators of consumption of these foods. Bronfenbrenner (1993) has been a leading advocate of ecological approaches to understanding behavior, but it has only recently been applied to diet behavior (Black, 1999). One decision in the cascade of decisions that result in food consumed at home is what foods are purchased and kept in the home (Baranowski, 1997b; Campbell and Desjardins, 1989), otherwise called availability (e.g., carrots in the refrigerator vegetable bin) (Hearn et al., 1998; Kratt et al., 2000). Accessibility includes whether foods kept in the home are in a form that encourages their consumption at an appropriate time (e.g., clean, scraped, sliced carrots in a plastic bag on a child-accessible shelf next to the child’s favorite dip at 3:00 p.m. on a school day). In a large sample of third grade children, whole fruit, 100 percent juice, and vegetable (FJV) availability and accessibility were related to consumption (Hearn et al., 1998). FJV availability and accessibility moderated the relationship of psychosocial variables to consumption (Kratt et al., 2000). School lunch FJV availability was related to child school lunch FJV consumption (Hearn et al., 1998), and availability in restaurants in the same census tract in which participating adolescent males lived was related to adolescent male FJV consumption

OCR for page 93
Dietary Risk Assessment in the WIC Program (Edmonds et al., 2001). Availability and accessibility were identified as mediators of dietary change in a community intervention project with children (Baranowski et al., in press). A limitation of this research for purposes of behavioral indicators has been the weak relationships obtained (0.1 ≤ r ≤ 0.25). A pervasive aspect of the environment is socioeconomic status (SES). In concept, SES indicates the resources available to purchase foods and knowledge of the best foods to buy. In addition, certain values, beliefs, and behaviors may be held in common at different levels of SES. SES is often measured by how much education and/or income a person or family has, and/or their type of occupation. Using the FFQs and an indicator of educational attainment with 849 women from families with school children in three communities in the Netherlands, people at lower levels of SES ate more meat, oils, fats, bread, and potatoes but less cheese, dietary fiber, and vegetables, than other SES levels (Hupkens et al., 1997). In a health-related behavior study using a 7-day weighed dietary intake and a measure of SES involving coding of occupation of male partner, SES was a significant predictor of consumption of total calories and every macro- and micronutrient (Haste et al., 1990). SES was a significant predictor even after statistically controlling for smoking, which was also shown to be a predictor of consumption of most nutrients (Haste et al., 1990). One reason low SES may lead to poor diets is the lack of money to purchase food. This has been called “food insufficiency.” Using the 1994–1996 CSFII data, when including a single item statement of how often the household did not have enough food to eat in the past 3 months, only 5.9 percent of low-income households (at or below 130 percent of the poverty guideline) reported inadequate food, but 7.9 percent of low-income households with children reported food insufficiency (Casey et al., 2001). No statistically significant differences in nutrient consumption were detected between food-insufficient and food-sufficient, low-income families. Food-insufficient, low-income families consumed fewer servings of dark green leafy vegetables, other vegetables, nuts and seeds, and added sugar, but more eggs, than food-sufficient, low-income families. Only relatively small differences were detected between groups (Casey et al., 2001). In a sample of 153 women (19–48 years of age) seeking food assistance in Toronto, 15.1 percent reported food insecurity with severe hunger and 35.3 percent reported food insecurity with moderate hunger (Tarasuk and Beaton, 1999). Women reporting food insecurity with severe hunger consumed 1,486 fewer kJ/d than the no food insecurity group and 468 less kJ/d than the food insecure with moderate hunger group. The food insecure with severe hunger group reported lower protein, iron, magnesium, and zinc intake than other groups (Tarasuk and Beaton, 1999).

OCR for page 93
Dietary Risk Assessment in the WIC Program Evaluation of the Behavioral Indicators of Diet in Relationship to the Suggested Criteria for Dietary Assessment Tools The committee was asked to consider the possible role of behavior-based indicators in dietary risk assessment. Any candidate behavioral indicator of diet that might be used for the purpose of determining the eligibility of individuals for the WIC program would need to be considered against the eight suggested criteria for evaluating dietary assessment tools described in Chapter 4. In the categories of behavioral indicators of diet that have been reviewed, the committee was unable to identify any indicators that would satisfy the eight evaluation criteria. As with all the food-based, dietary assessment tools, there is a major difficulty with the criteria of reliability and/or validity (criterion 4) for all behavioral indicators where individual assessment is concerned. Indeed, the categories of indicator foods, dietary patterns, and meal patterns are derived from traditional, food-based assessment techniques such as 24-hour dietary recall and FFQs. Furthermore, very little of the research on behavioral indicators has been performed in the populations served by WIC. Finally, there are no randomized trials that attempt to change any target behavioral indicators and measure the impact on diet. Perhaps a reasonable goal for behavioral measures is to achieve the status of a target indicator and thereby provide a focus for WIC nutrition education. To be a target indicator, the behavioral measure must be demonstrated to correlate with, and be causative of, an important dietary behavior at some reasonably high level; it should be amenable to change through some demonstrated method; and the change in the targeted variable should be demonstrated to be related to changes in the dietary practices of interest. Change in the targeted practice should be relatively easy to accomplish considering the operational constraints of WIC. Although substantial error exists in dietary and behavioral indicator assessment that precludes reasonable accuracy for determining WIC eligibility for individuals, there is merit in doing research on correlates of diet in groups representative of WIC populations. A variety of statistical procedures are available that can correct for known sources of error (Traub, 1994) and thereby provide reasonable tests of relationships. Thus, while a relationship between a behavioral indicator and diet may not be true of any specific individual, it would be true of the group assessed. WIC might then use the findings from such research to make inferences about the behavior of groups (e.g., pregnant Hispanic teenagers) rather than individuals, and to make decisions about the content of dietary counseling that is targeted to the specific groups that have been studied.

OCR for page 93
Dietary Risk Assessment in the WIC Program BEHAVIORAL INDICATORS OF PHYSICAL ACTIVITY A literature review has recently been completed on the correlates of physical activity in children (Sallis et al., 2000). While this review did not examine studies in preschool children per se, it did review the literature on correlates in 4- to 12-year-old children. These correlates can all be considered as potential surrogate or target behavioral indicators of physical activity, according to the criteria previously discussed (see Box 7-2). Target indicators, in particular, would be precursors of physical activity. If changed, they would result in increased levels of physical activity and might also serve as the appropriate targets for behavioral counseling efforts. In the review by Sallis et al., the two strongest correlates with physical activity in children were time spent outdoors (Baranowski et al., 1993; Klesges et al., 1990; Sallis et al., 1993) and access to recreational facilities or play spaces (Garcia et al., 1995; Sallis et al., 1993; Stucky-Ropp and DiLorenzo, 1993). With regard to the factors that parents evaluate in selecting outdoor play spaces for preschoolers, perceived safety may be the most important factor (Sallis et al., 1997). Despite this finding, perceived neighborhood safety (how safe is it “for your child to play outdoors with other children in your neighborhood without adult supervision”) was not predictive of change in physical activity of fourthgraders from suburban San Diego who were followed for 20 months (Sallis et al., 1999). There are many factors that affect the perception of safety outdoors, ranging from neighborhood crime levels, to traffic patterns and sidewalk availability, to playground disrepair. Research has not been conducted to determine which aspects of neighborhood safety are most relevant in the decisions made by families in WIC about spending time outdoors. It is intuitive that parental activity would affect the activity of preschoolers, and the Dietary Guidelines emphasize that parents should be active with their children. Parent activity has been the most widely studied potential correlate of child physical activity, but the review of physical activity correlates by Sallis et al. (2000) did not find compelling evidence across 29 studies that parental activity and child activity were correlated. While this finding does not negate the many potential benefits of parents being active with their children, it does not support the idea that increasing a parent’s activity level will necessarily increase their child’s activity. Thus, there is not sufficient evidence that either the time children spend outdoors or parental physical activity levels would meet the criteria (Box 7-2) to be target behavioral indicators for physical activity in WIC-enrolled preschool children. Television Viewing as a Behavioral Indicator of Physical Activity Television viewing has also been considered a potential behavioral indicator of physical activity and, for several reasons, the committee felt that this indicator

OCR for page 93
Dietary Risk Assessment in the WIC Program merited separate discussion. One reason is that the Dietary Guidelines address television viewing in each of the two guidelines under “Aim for Fitness”: Aim for a Healthy Weight and Be Physically Active Each Day. A second reason is the well-documented association between television viewing and obesity and the increasing problem of obesity in the populations served by WIC. Although adverse affects of television viewing on fitness and fatness may not occur until school-age or beyond, the trajectories towards television viewing habits in middle childhood appear set in children by 24 months of age and are less favorable for children whose mothers are less educated (Certain and Kahn, 2001). This makes television viewing a particular concern for the WIC population. Several large studies, based on nationally representative samples, show a direct association between television viewing and obesity in children and adolescents (Andersen et al., 1998; Dietz and Gortmaker, 1985; Gordon-Larsen et al., 1999; Gortmaker et al., 1996; Pate and Ross, 1987). Recent experimental evidence has shown that reducing television viewing time in school-age children reduces the normal age-related increase in BMI (Robinson, 1999), suggesting that television viewing may cause obesity. As for adult women, several observational studies have also shown a direct association between television viewing and obesity (Crawford et al., 1999; Jeffery and French, 1998; Sidney et al., 1996; Tucker and Bagwell, 1991), but there have been no randomized trials that involve reducing television viewing in adults. Despite these studies, the association between television viewing and fatness in preschoolers is less clear (DuRant et al., 1994; Klesges et al., 1995b). This could be due to the difficulty of measuring television viewing accurately in preschoolers. For example, there are no validated measures of parent-reported hours of child television viewing for preschool children, and parent reports may be subject to social desirability bias. The short attention span of young children and their proclivity for frequent but short bouts of activity may mean that many preschoolers are active while watching television (DuRant et al., 1994). The Relationship of Television Viewing to Physical Activity and Overweight/Obesity The interrelationships between television viewing (and other sedentary behaviors), physical activity, and obesity are complex and not yet fully understood. Despite the persistent association between television viewing and obesity, television viewing is not always highly correlated to physical activity, making it an unsuitable candidate as a behavioral indicator for physical activity. This is especially true for children (DuRant et al., 1994; Sallis et al., 2000; Strauss et al., in press) and women (Jeffery and French, 1998) in relationship to moderate or low-intensity activity.

OCR for page 93
Dietary Risk Assessment in the WIC Program There are at least three possible explanations for the lack of a consistent relationship between television viewing and physical activity. The most likely explanation is that the relationship between television viewing and obesity is only partly mediated by reductions in physical activity (Robinson, 1998). Television viewing has been shown to be associated with food consumption (Coon et al., 2001; Jeffery and French, 1998). Children and adults may not only be eating while watching television, but their food consumption may be influenced by the advertising (Gorn and Goldberg, 1982; Jeffrey et al., 1982; Taras and Gage, 1995). A second possibility is that television viewing substitutes for moderate or low-level activities that are difficult to measure, and television viewing is less often substituted for the vigorous activities that are more often and more accurately measured. Finally, television is just one form of sedentary activity and may not serve equally across ages, races, or cultural groups as a proxy for inactivity. Other forms of sedentary behavior in both women and preschool children, aside from television viewing, have not been well characterized. In summary, while television viewing may be a particular form of inactivity that contributes to overweight and obesity both by reducing energy expenditure and increasing energy intake, the exact mechanism is uncertain. While television viewing may be a plausible target behavior for physical activity, more needs to be known about whether altering television viewing levels in the populations served by WIC would have a demonstrable impact on activity or fatness. CONCLUSIONS REGARDING THE USE OF BEHAVIORAL INDICATORS FOR ELIGIBILITY DETERMINATION Behavioral indicators of food intake or physical activity hold no promise of distinguishing individuals who are ineligible from those eligible for WIC based on the criterion failure to meet Dietary Guidelines, or on nutrient intake or level of physical activity. However, assessment methods and behavioral indicators do offer promise of improving the understanding of diet and activity behaviors of WIC participants as a group, which could be used to design effective nutrition education programs that target behavior change related to the Dietary Guidelines. Likewise, behavioral indicators hold promise for monitoring the dietary intake and physical activity levels of groups and thereby evaluation of program effectiveness. Research efforts should focus on determining the feasibility and validity of assessing target behavioral indicators for diet and physical activity in the population served by WIC.