Advertisement
Research Article| Volume 50, ISSUE 7, P705-717, July 2018

An Obesity Risk Assessment Tool for Young Children: Validity With BMI and Nutrient Values

Published:March 19, 2018DOI:https://doi.org/10.1016/j.jneb.2018.01.022

      Abstract

      Objective

      Demonstrate validity and reliability for an obesity risk assessment tool for young children targeting families' modifiable home environments.

      Design

      Longitudinal design with data collected over 100 weeks.

      Setting

      Head Start and the Special Supplemental Nutrition Program for Women, Infants, and Children.

      Participants

      Parent–child pairs (n = 133) provided food behavior assessments; 3 child-modified, 24-hour dietary recalls; 3 ≥ 36-hour activity logs; and measured heights and weights.

      Main Outcome Measure

      Five measures of validity and 5 of reliability.

      Results

      Validity was excellent for the assessment tool, named Healthy Kids, demonstrating an inverse relationship with child body mass index percentile-for-age (P = .02). Scales were significantly related to hypothesized variables (P ≤ .05): fruit or vegetable cup equivalents; folate; vitamins A, C, and D; β-carotene; calcium; fiber; sugar; screen, sleep, and physical activity minutes; and parent behaviors. Measures of reliability were acceptable.

      Conclusions and Implications

      Overall, children with higher Healthy Kids scores had a more healthful profile as well as lower body mass index percentiles-for-age 1.5 years later. Healthy Kids has potential for use by nutrition professionals as a screening tool to identify young children most at risk for excess weight gain, as an evaluation to assess intervention impact, and as a counseling tool to tailor intervention efforts. Future research should include validation in other settings and with other populations.

      Key Words

      Introduction

      Parents have direct influence over their children's physical, food, and social environments.
      • American Academy of Pediatrics
      Prevention of pediatric overweight and obesity.
      Yet, many families' nutrition and parenting practices and lifestyle behaviors create home environments that set young children on trajectories for unhealthful weight gain. Among low-income preschoolers, 31% are overweight or obese in the US.
      • Ogden C.L.
      • Carroll B.K.
      • Flegal K.M.
      Prevalence of obesity and trends in body mass index among US children and adolescents, 1999–2010.
      In response to the staggering obesity rates among children, Congress authorized federal programs to include an obesity prevention focus in their education programs for families with young children. These programs include Head Start
      • US Department of Health and Human Services
      Office of Head Start.
      ; the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC)
      • USDA Food and Nutrition Service
      Supplemental nutrition program for women, infants and children (WIC).
      ; the Supplemental Nutrition and Assistance Program–Education (SNAP-Ed)
      • USDA Food and Nutrition Service
      Supplemental nutrition assistance program education guidance (SNAP-Ed). 2014b.
      ; and the Expanded Food and Nutrition Education Program (EFNEP).
      • USDA National Institute for Food and Agriculture
      Expanded Food and Nutrition Education Program (EFNEP).
      These 4 programs have a presence in all or most low-income communities in the US. Consequently, they have the potential to affect obesity prevalence among participants.
      • Townsend M.S.
      Obesity in low-income communities: prevalence, effects, a place to begin.
      Recognizing that this young age may be ideal for intervention, the Institute of Medicine and the American Academy of Pediatrics (AAP) recommend the development of assessment tools targeting families' modifiable environmental and behavioral factors associated with the risk for pediatric obesity.
      • American Academy of Pediatrics
      Prevention of pediatric overweight and obesity.
      • Institute of Medicine of the National Academies
      Preventing Childhood Obesity: Health in Balance.
      Subsequent to the Institute of Medicine and AAP recommendations, comprehensive evidence-based literature reviews guided the selection of 13 determinants of obesity for assessment tools.
      • Ontai L.
      • Ritchie L.
      • Williams S.T.
      • Young T.
      • Townsend M.S.
      Guiding family-based obesity prevention efforts in low-income children in US: part 1−What determinants do we target?.
      The review considered multiple lines of evidence including secular trend data, observational cross-sectional and longitudinal studies, interventions, and mechanistic studies. This evidence-based analysis was followed by 2 additional literature reviews: they identified behaviors (n = 23) in this population related to the 13 determinants of obesity
      • Townsend M.S.
      • Ontai L.
      • Young T.
      • Ritchie L.D.
      • Williams S.T.
      Guiding family-based obesity prevention efforts in low-income children in US: part 2—What behaviors do we measure?.
      and existing tools for assessing the 23 behaviors.
      • Townsend M.S.
      • Ontai L.
      • Young T.
      • Ritchie L.D.
      • Williams S.T.
      Guiding family-based obesity prevention efforts in low-income children in US: part 2—What behaviors do we measure?.
      No pediatric obesity risk assessment tools were identified for low-income families with 2- to 5-year-old children covering all determinants in the diet, lifestyle, and parenting behavioral domains.
      • Townsend M.S.
      • Ontai L.
      • Young T.
      • Ritchie L.D.
      • Williams S.T.
      Guiding family-based obesity prevention efforts in low-income children in US: part 2—What behaviors do we measure?.
      Subsequent to that review, Ihmels et al
      • Ihmels M.A.
      • Welk G.J.
      • Eisenmann J.C.
      • Nusser S.M.
      • Myers E.F.
      Prediction of BMI change in young children with the family nutrition and physical activity (FNPA) screening tool.
      validated a 21-item Family Nutrition and Physical Activity tool for children aged 6–12 years. Dickin et al
      • Dickin K.L.
      • Lent M.
      • Lu A.H.
      • Sequeira J.
      • Dollahite J.S.
      Developing a measure of behavior change in a program to help low-income parents prevent unhealthful weight gain in children.
      published a 15-item evaluation tool for an EFNEP intervention targeting low-income parents of children aged 3–11 years. Importantly, no pediatric obesity risk assessment tool covering the identified behaviors in the family environment for the diet, lifestyle, and parenting domains was identified or subsequently published targeting low-income families with children aged 2–5 years for these federal programs.
      In response to this identified need, the current authors created a pediatric obesity risk assessment tool using the results of these literature reviews. The content of this tool was evidenced-based with content
      • Ontai L.
      • Ritchie L.
      • Williams S.T.
      • Young T.
      • Townsend M.S.
      Guiding family-based obesity prevention efforts in low-income children in US: part 1−What determinants do we target?.
      • Townsend M.S.
      • Ontai L.
      • Young T.
      • Ritchie L.D.
      • Williams S.T.
      Guiding family-based obesity prevention efforts in low-income children in US: part 2—What behaviors do we measure?.
      and face
      • Townsend M.S.
      • Shilts M.K.
      • Sylva K.
      • et al.
      Obesity risk for young children: development and initial validation of an assessment tool for participants of USDA programs.
      validation demonstrated for 45 items representing 23 behaviors in the child's family environment associated with the broad determinants of pediatric obesity.
      The goal of the current research was to identify children aged 2–5 years who are at risk for becoming overweight or obese before their body mass index (BMI) rises above age norms. Building on the previous qualitative work,
      • Townsend M.S.
      • Shilts M.K.
      • Sylva K.
      • et al.
      Obesity risk for young children: development and initial validation of an assessment tool for participants of USDA programs.
      the current article describes the quantitative research to further establish validity and reliability of the final version of the tool. The objectives were to select appropriate items from the 45-item tool for a final parsimonious version, entitled Healthy Kids; further establish validity of the final version of the Healthy Kids tool with low-income parents of young children; and establish reliability of the final version of Healthy Kids using 5 approaches: internal consistency, an item difficulty index, an item discrimination index, coefficient of variation, and temporal stability. Specifically the hypotheses included: (1) the Healthy Kids total score would be associated with child BMI percentiles-for-age measured 88 weeks after baseline with a lower Healthy Kids score predicting a higher BMI percentile-for-age; (2) the Healthy Kids dietary scale scores would be associated with relevant cup equivalents, dietary energy density, and micronutrient intakes with a higher Healthy Kids dietary scale score predicting higher intakes of vitamins (A, C, D, folate, and ß-carotene), minerals (calcium, magnesium, and potassium), and fiber, as well as lower saturated fats, sugar, and sodium, and energy density; (3) the Healthy Kids physical activity, screen time, and sleep scale scores would be associated with physical activity, screen time, and sleep variables calculated from ≥36-hour logs; and (4) the Healthy Kids dietary scale scores would be positively related to parent food behaviors and mediators.

      Methods

      Participants

      Parent–child dyads were recruited at Head Start (n = 13) and WIC sites (n = 2) in 2 counties in northern California. They were ethnically diverse parents or caregivers aged >18 years, who understood English as a first or second language and had ≥1 child aged 2–5 years enrolled in Head Start or WIC. Respondents received $10 for each interview, $30 for the phlebotomy session, and $30 for an optional at-home family photography session. The Institutional Review Board of the University of California, Davis approved the protocol.

      Biopsychosocial Framework

      The framework used for the design of this validation research was based on systems theory.
      • Engel G.L.
      The clinical application of the biopsychosocial model.
      This framework considered the health of the child in the context of the family environment (Figure 1) and in that respect is similar to the socioecological model used to guide development of the tool's content.
      • USDA Food and Nutrition Service
      The Supplemental Nutrition Assistance Program Education (SNAP-Ed) strategies and interventions: an obesity prevention toolkit for states.
      In a hierarchical fashion, the model recognizes that the parent's relevant psychosocial mediators
      • Townsend M.S.
      • Kaiser L.L.
      Development of an evaluation tool to assess psychosocial indicators of fruit and vegetable intake for two federal programs.
      and behaviors
      • Townsend M.S.
      • Kaiser L.L.
      • Allen L.H.
      • Joy A.B.
      • Murphy S.P.
      Selecting items for a food behavior checklist for a limited resource audience.
      create the environment for the young child, and that this in turn drives or influences the child's behaviors.
      • Townsend M.S.
      • Shilts M.K.
      • Sylva K.
      • et al.
      Obesity risk for young children: development and initial validation of an assessment tool for participants of USDA programs.
      The young child's eating behaviors subsequently drive his or her food intake (measured by 3 modified 24-hour diet recalls), resulting in intakes of micronutrients. The parent's control over the child's physical, screen, and sleep activities in the framework influences the child's behaviors, measured by 3 ≥36-hour activity logs. Thus, as specified by this framework, intake of nutrients and physical activity, screen, and sleep behaviors by the child can effect changes in biochemical parameters and body weight, indicators of the child's health status (Figure 1).
      Figure 1
      Figure 1Biopsychosocial framework for validation of Healthy Kids targeting low-income families with preschool-aged children. aParent mediators of fruit and vegetable behaviors collected with the University of California Fruit and Vegetable Inventory at week 1
      • Townsend M.S.
      • Kaiser L.L.
      University of California Fruit and Vegetable Inventory.
      ; bParent dietary behaviors collected with the University of California Food Behavior Checklist at week 1
      • Sylva K.
      • Townsend M.S.
      • Martin A.
      • Metz D.
      University of California Food Behavior Checklist.
      ; cChild dietary, physical activity, sleep, and screen behaviors collected with Healthy Kids at week 12
      • Townsend M.S.
      • Sylva K.
      • Shilts M.K.
      • Davidson C.
      • Leavens L.
      • Sitnick S.L.
      Healthy Kids: Pediatric obesity risk assessment tool [45 items].
      ; dChild diet estimated with 3 modified 24-hour dietary recalls at weeks 1, 6, and 12, using the University of California Child Eating and Activity Diary. Child physical, screen, and sleep activity was estimated with 3 36-hour activity logs at weeks 1, 6, and 12, using the University of California Child Eating and Activity Diary
      • Townsend M.S.
      • Shilts M.K.
      • Davidson C.
      • Leavens L.
      My Child's Activity & Food Diary.
      ; eChild height and weight, blood, blood pressure, and body temperature were collected at weeks 15 and 100.

      Data Collection and Timeline

      The researchers collected baseline data over 15 weeks in 4 unique sessions, in person (weeks 1, 12, and 15) and by phone (week 6). Informed consent was collected before data collection began. At week 1, a parent completed the demographic questionnaire, answered a parent food behavior checklist and an assessment of parent behavioral mediators, and completed the child's ≥36-hour physical activity, screen time, and bedtime log, as well as the child's 24-hour dietary recall (Figure 1). The ≥36-hour log and 24-hour diet recall were readministered at weeks 6 and 12, for a total of 3 child activity logs and 3 child dietary recalls. The 45-item version of Healthy Kids was administered to parents on week 12. On week 15, the researchers collected the child's height and weight measurements. In addition, the child provided a fasting blood sample, blood pressure, and body temperature, the results of which will be reported elsewhere. Stability reliability data were collected a second time at week 24. The child's anthropometric data were collected at another in-person session at week 100. Each session had a maximum duration of 60 minutes. The data collection team was trained in dietary, anthropometric, and biochemical methods by the principal investigator, other project researchers, and a team at the US Department of Agriculture (USDA) Western Human Nutrition Research Center.

      Selection of Best Items for Healthy Kids

      In community settings, participant literacy and demands of the intervention's education content drive the need for a tool with low respondent burden.
      • Townsend M.S.
      • Sylva K.
      • Martin A.
      • Metz D.
      • Wooten Swanson P.
      Improving readability of an evaluation tool for low-income respondents using visual information processing theories.
      Consequently, identification of risk was the goal while having a low response burden for Healthy Kids. Box plots for BMI percentile-for-age were generated for each of the original 45 items, and multiple exploratory analyses, eg, random forest, stepwise regression, and ANOVA, were used to select Healthy Kids items for retention or elimination. A version of the tool was created with 19 items.

      Validation Process

      Five approaches were used to establish validity in addition to face and content validity previously reported.
      • Ontai L.
      • Ritchie L.
      • Williams S.T.
      • Young T.
      • Townsend M.S.
      Guiding family-based obesity prevention efforts in low-income children in US: part 1−What determinants do we target?.
      • Townsend M.S.
      • Ontai L.
      • Young T.
      • Ritchie L.D.
      • Williams S.T.
      Guiding family-based obesity prevention efforts in low-income children in US: part 2—What behaviors do we measure?.
      • Townsend M.S.
      • Shilts M.K.
      • Sylva K.
      • et al.
      Obesity risk for young children: development and initial validation of an assessment tool for participants of USDA programs.

      Predictive validity with child anthropometry

      The child's height and weight were measured using a digital scale (model 810, Sega Carisma, Seca Medical Measurement Systems and Scales, Hanover, MD) and a portable stadiometer (Seca Road Rod, Seca Medical Measurement Systems and Scales), with children dressed in light clothing, and shoes, jackets, or outerwear removed. Two research assistants weighed each child once to the nearest 0.1 kg and measured height twice to the nearest 0.1 cm with the child fully erect, head in the Frankfort plane, and at the end of a deep inhalation. Body mass index (kg/m
      • Ogden C.L.
      • Carroll B.K.
      • Flegal K.M.
      Prevalence of obesity and trends in body mass index among US children and adolescents, 1999–2010.
      ) was calculated using the means of height and weight measurements. The BMI percentile-for-age was derived using the Centers for Disease Control's BMI Percentile Calculator for Children.
      • Kuczmarski R.J.
      • Ogden C.L.
      • Grummer-Strawn L.M.
      • et al.
      CDC Growth Charts for United States: Advance Data from Vital and Health Statistics.

      Convergent validity with child 24-hour dietary recalls

      The researchers collected 3 recalls using the USDA 5-step multipass method.
      • Townsend M.S.
      • Shilts M.K.
      • Davidson C.
      • Leavens L.
      My Child's Activity & Food Diary.
      • Conway J.M.
      • Ingwersen L.A.
      • Vinyard B.T.
      • Moshfegh A.J.
      Effectiveness of the USDA 5-step multiple-pass method in assessing food intake in obese and non-obese women.
      Food recall data were entered into ESHA Research food processor nutrition analysis software (ESHA Research, Salem, OR) using the best possible match to items reported by parents. Dietary variables assessed were vegetable, fruit, and dairy cup equivalents; folate; vitamins A, C, and D; β-carotene; calcium; magnesium; sodium; total fiber; total fat, saturated fat, and calories from fat; dietary energy density; and sugar. Because breakfast, lunch, and snacks were frequently provided by Head Start and day care without parents' presence and most children ate dinners with parents, the mean from 3 diet recalls for the dinner meal was used for this modified recall analysis.

      Convergent validity for physical activity, screen, and sleep behaviors with child ≥36-hour logs

      The child's behaviors were assessed using the University of California ≥36-hour activity log, My Child's Food & Activity Diary.
      • Townsend M.S.
      • Shilts M.K.
      • Davidson C.
      • Leavens L.
      My Child's Activity & Food Diary.
      Variables collected or calculated included times of day, duration in minutes, and location for bedtime, wake time, and nap times; television, video, and other screen activities; and physical activity. One advantage of the ≥36-hour log was support for increased accuracy, ie, validity, by way of random and systematic error reduction. For example, if the parent reported the child was watching television at 3 PM, the parent could not concurrently report physical activity at the park. The incongruous report by the parent would be obvious to the data collection staff, who would then ask the parent for clarification. A second advantage of the ≥36-hour log was inclusion of a second bedtime and wakeup per log. With the 3 logs over the 15 weeks, 6 bedtimes and 6 wakeup times were reported by parents.

      Concurrent validity with parent eating behaviors

      The parents' food behaviors were assessed using the University of California Food Behavior Checklist,
      • Sylva K.
      • Townsend M.S.
      • Martin A.
      • Metz D.
      University of California Food Behavior Checklist.
      with validity previously demonstrated using serum carotenoids (r = .44; P ≤ .001)
      • Townsend M.S.
      • Kaiser L.L.
      • Allen L.H.
      • Joy A.B.
      • Murphy S.P.
      Selecting items for a food behavior checklist for a limited resource audience.
      and 3 24-hour dietary recalls.
      • Murphy S.
      • Kaiser L.L.
      • Townsend M.S.
      • Allen L.
      Evaluation of validity of items in a food behavior checklist.
      The 16-item behavior checklist included vegetable quantity, as snacks, at main meal, and types; fruit quantity, citrus; perception of diet quality; food insecurity; milk intake, milk on cereal; sugar-sweetened beverages; fat on chicken and meat; fish; and food labels. Parent responses were coded using 4 response options per item for a maximum of 4 points per item.
      • Sylva K.
      • Townsend M.S.
      • Martin A.
      • Metz D.
      University of California Food Behavior Checklist.
      The 16 items were scaled and summed for a maximum of 64 points. Seven fruit and vegetable items formed a subscale with a maximum of 28 points.

      Concurrent validity with parent psychosocial fruit and vegetable constructs

      The psychosocial variables, also referred to as mediators of parent fruit and vegetable intake, were measured by the University of California Fruit and Vegetable Inventory, with validity previously demonstrated.
      • Townsend M.S.
      • Kaiser L.L.
      Development of an evaluation tool to assess psychosocial indicators of fruit and vegetable intake for two federal programs.
      • Townsend M.S.
      • Kaiser L.L.
      Brief psychosocial fruit and vegetable tool is sensitive for United States Department of Agriculture's nutrition education programs.
      The 4 constructs were the perception of parents' diet quality; self-efficacy for planning, preparing, and eating fruits and vegetables; readiness to eat more fruit; and readiness to eat more vegetables.
      • Townsend M.S.
      • Kaiser L.L.
      University of California Fruit and Vegetable Inventory.
      The constructs were given equal weighting and summed for a total score using a maximum of 5 points/variable for a maximum of 20 total points.

      Reliability

      To give the nutrition professional additional information about the psychometric properties of Healthy Kids, 5 approaches were assessed: (1) internal consistency, the proportion of variance in the Healthy Kids score attributable to the true score, was reported as Cronbach α
      • Litwin M.S.
      How to Measure Survey Reliability and Validity.
      ; (2) the item difficulty index,
      • Litwin M.S.
      How to Measure Survey Reliability and Validity.
      the ratio of each item's mean to its maximum score, was generated by dividing the mean value of responses to the item by 5; (3) the item discrimination index, the ability of an item to discriminate between those who did well on Healthy Kids and those who did not, was the corrected item-total correlation, ie, item score divided by the total scale score without the item
      • Townsend M.S.
      • Sylva K.
      • Shilts M.K.
      • Davidson C.
      • Leavens L.
      • Sitnick S.L.
      Healthy Kids: Pediatric obesity risk assessment tool [45 items].
      • Nunnally J.C.
      • Bernstein I.H.
      Psychometric Theory.
      ; (4) temporal stability, also known as test-retest reliability, was the within-person correlation at 2 time points 12 weeks apart with no intervention
      • Litwin M.S.
      How to Measure Survey Reliability and Validity.
      ; and (5) coefficient of variation, expressed as a percentage, was calculated as item SD divided by the item mean times 100.
      • Traub R.E.

      Statistics

      Pearson chi-square test was used to compare demographic data between dropout and final samples with statistical program R (version 2014, https://cran.cnr.berkeley.edu). Significance level was set at P ≤ .05 for all analyses. Box plots of BMI percentiles-for-age were created across each of the 45 Healthy Kids items to illustrate the direction of the findings using SAS for Windows (version 9.4, SAS Institute, Cary, NC, 2017). After exploratory analysis, the 19-item Healthy Kids scores were discretized using tertiles to define low, medium, and high Healthy Kids behavior scores. The scores were compared with BMI percentiles-for-age across the Healthy Kids groups using ANOVA with the same SAS program. Because of the skewness, presenting medians was appropriate to summarize subgroup results. Linear regression is a more powerful analysis with a continuous variable. If the relationship is monotone but not linear, regression can overestimate or underestimate the mean BMI outcome. Regression will detect the trend but it predicts with less accuracy. Both analyses are reported. Nonparametric Spearman rank-order correlation and ANOVA were used to assess the relationship between dietary and other variables and Healthy Kids scores. Analyses were repeated using Kruskal-Wallace; no differences in significance were found. Consequently, Spearman correlations and P from ANOVA are reported. Using SPSS (version 21.0, SPSS, Inc, Chicago, IL, 2015), psychometric analyses were performed using the final version of Healthy Kids. With at least 130 children, the study is powered at 80% to detect a partial multivariate coefficient (R
      • Ogden C.L.
      • Carroll B.K.
      • Flegal K.M.
      Prevalence of obesity and trends in body mass index among US children and adolescents, 1999–2010.
      ) of ≥.08 owing to Healthy Kids scores in 3 categories. Power calculations showed >80% power with n = 95 using this SAS for Windows version.

      Results

      Demographics

      Parents self-reported race and ethnicity as 20% Hispanic, 2% Latino, 34% African American, 20% white, 6% Asian, 1% American Indian, and 17% multiethnic. Families enrolled in the study participated in at least 1 USDA assistance program; WIC (79%), Head Start (70%), and SNAP (61%) were reported most frequently. Parents (92% female) were aged 31 ± 8 years and enrolled children (42% female) were aged 3.3 ± 1.6 years on average and living in a household of 4.2 ± 1.0 persons. About a third of parents were married (28%); a majority reported some college education (56%) or a high school diploma (33%). Parents (65%) reported earning a household income of ≤$20,000 annually.

      Sample Size and Attrition

      Parent–child pairs (n = 176) were enrolled in the study (week 1) with 133 pairs completing the 5 sessions of data collection during year 1. By the end of year 2, 98 parent–child pairs remained in the study. The primary reason for dropout was loss of contact with the family despite numerous efforts by the researcher team. Those without all data were dropped from the analyses. The 176, 133, and 98 pairs were compared with the 43 and 35 pairs who dropped out at various stages. Tests for selective attrition between the retained and dropped families revealed no significant demographic differences for child's gender, parent's gender, marital status, parent education, household income, and ethnicity or race of child, although other unmeasured differences may have existed. With 95% certainty, initial dropouts, or those at the end of baseline data collection year 1 or by the end of year 2, were not different with regard to these measured variables compared with subjects who remained in the study.

      Item Reduction

      Items were removed from the 45-item tool because of (1) inadequate box plot analysis with BMI percentile-for-age data, (2) duplication of content, and/or (3) an item difficulty index >90. For example, parent responses to My child eats breakfast were universally positive. All children in the sample ate breakfast at Head Start. As written, this item did not differentiate or offer the opportunity for positive behavior change. Nineteen items were retained and are presented in Table 1 by category and domain. Items with 1–5 points each generated a possible scale score range of 19–95 points. Subsequent validation and reliability analyses were conducted using this 19-item final version.
      Table 1Behavioral Domain and Construct, Item Text, and Item Visual Content for Final 19 Items for Healthy Kids Assessment Tool
      Behavioral Domain and ConstructItem TextItem VisualResponse
      Each item has a minimum of 1 and a maximum of 5 points.
      (mean ± SD)
      Vegetables
      Vegetable availability1. I buy vegetables.
      • Left: woman getting ready to buy broccoli with her daughter
      • Right: bag of frozen mixed vegetables, box of frozen green beans, can of diced tomatoes, can of corn, and can of mixed vegetables
      4.5 ± 0.8
      Vegetable as snack2. My child eats snack foods such as apples, bananas, or carrots.
      • Left: boy eating carrot stick
      • Right: girl eating banana
      3.7 ± 1.1
      Vegetables as main meal3. My child eats ___ vegetables at her main meal.
      • Left: girl eating dinner
      • Right: plate with spaghetti, broccoli, bowl of fruit, and glass of milk
      3.3 ± 1.0
      Vegetable variety4. My child eats >1 kind of vegetable a day.
      • Top left: boy biting into a celery stick
      • Bottom left: the same boy biting into a carrot
      • Right: range of vegetables including cob of corn, carrots, cucumbers, radishes, peppers, potato, fresh tomatoes, broccoli, onion, can of salsa, can of tomato sauce, bottle of tomato juice, and head of lettuce
      2.9 ± 1.2
      Fruit
      Fruit intake5. My child eats fruit.
      • Left: boy biting into an apple
      • Top right: 0.5 cup of canned fruit cocktail
      • Bottom right: 0.5 cup of fresh grapes
      3.8 ± 1.0
      Fruit availability6. I buy fruit.
      • Left: woman picking up an apple at a grocery store to put in her shopping cart
      • Right: examples of fruit include bag of frozen strawberries, orange, apple, can of peaches, jar of applesauce, and box of raisins
      4.4 ± 0.9
      Fruit accessibility7. I keep fruit ready for my child to eat.
      • Left: shelf in refrigerator with apple slices in bag, orange slices in bag, and bowl of washed cherries Right: bowl of washed fruit with a girl biting into an apple
      4.0 ± 1.0
      Fruit as snack2. My child eats snack foods such as apples, bananas, or carrots.
      • Left: boy eating a carrot stick
      • Right: girl eating a banana
      3.7 ± 1.1
      Dairy
      Milk frequency8. My child drinks milk ___ times a day.
      • Left: boy drinking a cup of milk for snack
      • Right: girl drinking a cup of milk with a meal
      3.3 ± 1.0
      Sugar-sweetened beverages
      Soda frequency9. My child drinks soda or sugared drinks.
      • Left: boy drinking soda from a can
      • Right: girl drinking soda from a cup while eating a meal
      4.1 ± 0.7
      Sports and sugared drinks frequency10. My child drinks sports or sugared drinks ___ times a day.
      • Left: boy drinking Kool-Aid from a pouch
      • Right: SunnyD, Hawaiian Punch, Propel Fitness Water, Gatorade, and Kool-Aid
      4.2 ± 0.8
      Saturated fat
      Energy-dense foods11. My child eats candy, cake, or cookies ___ times a day.
      • Left: boy eating a cookie
      • Right: girl eating a cupcake
      4.4 ± 0.5
      Dairy fat12. My child drinks milk.
      • Whole milk (4% fat)
      • Reduced-fat milk (2% fat)
      • Low-fat milk (1% fat)
      • • Fat-free milk (0%)
      • Soy milk
      3.0 ± 0.8
      Energy-dense snacks13. My child eats chips ___ times a day.
      • Left: boy eating Cheetos from a bag
      • Top right: bags of chips (Funyuns, Cheetos, Doritos) and plate of chips
      • Bottom right: plate of Doritos, Funyuns, and Cheetos chips
      4.4 ± 0.7
      Meat fat14. I trim fat before eating meat.
      • Left: hands of a person trimming fat off raw meat on a cutting board
      • Right: hands of person trimming fat off cooked meat on a plate
      3.6 ± 1.5
      Parenting
      Modeling at mealtime15. I sit and eat a meal with my child.
      • Father and daughter are about to eat their meal
      4.1 ± 1.1
      Screen time
      Television16. My child watches television ___ hours a day.
      • Boy watching television in his bedroom
      3.5 ± 0.8
      Other screen use17. My child plays video or computer games ___hours a day.
      • Boy playing game on an iPad
      4.5 ± 0.6
      Physical activity
      Play, sedentary time18. My child likes playing instead of watching television.
      • Left: girl playing a board game
      • Right: boy playing with toy cars
      3.4 ± 1.1
      Sleep
      Bedtime19. My child goes to bed around ___ nighttime
      • Mother tucking a child into bed in the child's dark bedroom
      2.9 ± 0.9
      a Each item has a minimum of 1 and a maximum of 5 points.

      Body Mass Index

      Table 2 shows BMI percentile-for-age distribution over the 19-item Healthy Kids scale scores. There was a significant inverse relationship between the child's Healthy Kids scale scores and BMI percentiles measured at week 100 (P = .02). The higher the child's Healthy Kids scale score, ie, healthier behaviors, the lower the child's BMI percentile (R2 = .09). The unadjusted effect of Healthy Kids was 9% of variability. If the baseline BMI variable were included, P was .04 and R2 would increase to .60, showing the variability in the BMI variable explained by the baseline BMI and Healthy Kids tool. The R values are not reported because of the lack of linearity between the Healthy Kids scores and BMI percentile-for-age.
      Table 2Relationship of 19-Item Healthy Kids Assessment Tool With BMI: Medians and Means of BMI Percentile-for-Age by Tertile Categories of Healthy Kids Scale Scores
      Healthy KidsnChild BMI Percentile-for-Age
      Baseline(wk 12)Baseline (wk 15)Wk 100
      MedianInterquartile RangeMeanSDMedian
      Kruskal-Wallis test with 19-item Healthy Kids as tertile categories for Healthy Kids scores (tertile 1 = 0–67.2; tertile 2 = 67.3–76.0; tertile 3 = 76.1–95; 19 items × 5 points maximum each) for BMI percentile-for-age as a continuous variable (P = .02).
      Interquartile RangeMean
      ANOVA using 19-item Healthy Kids scores as continuous variable testing BMI percentile-for-age adjusted for baseline BMI percentile-for-age as outcome variable (P = .04);
      SD
      Tertile 1306559573381.045.069.1
      Means of BMI percentile-for-age with different superscripts are significantly different from one another.
      29.2
      Tertile 2276544632872.036.072.4
      Means of BMI percentile-for-age with different superscripts are significantly different from one another.
      22.7
      Tertile 3345347543159.050.053.3
      Means of BMI percentile-for-age with different superscripts are significantly different from one another.
      31.6
      BMI indicates body mass index.
      a Kruskal-Wallis test with 19-item Healthy Kids as tertile categories for Healthy Kids scores (tertile 1 = 0–67.2; tertile 2 = 67.3–76.0; tertile 3 = 76.1–95; 19 items × 5 points maximum each) for BMI percentile-for-age as a continuous variable (P = .02).
      b ANOVA using 19-item Healthy Kids scores as continuous variable testing BMI percentile-for-age adjusted for baseline BMI percentile-for-age as outcome variable (P= .04);
      1,2 Means of BMI percentile-for-age with different superscripts are significantly different from one another.

      Dietary Scales

      The 14 eating behavior items were grouped by content for validation with corresponding nutrients and food group (Table 3). With regard to the fruit and vegetable Healthy Kids scale, key variables were significant: vegetable cup equivalents (r = .36; P ≤ .01), fruit cup equivalents (r = .22; P ≤ .05), vitamin C (r = .29; P ≤ .001), vitamin A (r = .17; P ≤ .05), ß-carotene (r = .27; P ≤ .01), folate (r = .28; P ≤ .01), and total fiber (r = .29; P ≤ .001). This scale was negatively related to sodium (r = –.22; P ≤ .05). The dairy item My child drinks milk was related to calcium (r = .41; P ≤ .001) and vitamin D (r = .47; P ≤ .001), which offers support for the validity of this dairy item. The 2-item scale regarding soda, juice drinks, and sports drinks was inversely related to sugar intake (r = –.18; P ≤ .05). The saturated fat scale was negatively related to total fat (r = –.18; P ≤ .05) and dietary energy density (r = –.38; P ≤ .001).
      Table 3Validation Using Relationship of Healthy Kids Subscales and Categories With Hypothesized Variables From the Mean of 3 24-h Diet Recalls, 3 ≥ 36-h Activity and Sleep Logs, the University of California Food Behavior Checklist, and the University of California Fruit and Vegetable Inventory for Low-Income Mother–Child Pairs (n = 133)
      Categories/Subscales From 19-Item Healthy KidsItems, nConvergent validity (n = 133)r, P
      Dietary subscale or category
      Fruit/vegetable
      Spearman rank-order correlation coefficient r estimated using 3 modified 24-hour dietary recalls showing parent report of child's diet. Results expressed per 1,000 kcal.
      7Food groups
      Cup equivalents, vegetables.36***
      Cup equivalents, fruit.22**
      Nutrients
      Vitamin C, mg.29****
      Vitamin A, IU.17**
      ß-Carotene, µg.27***
      Folate, µg.28***
      Total fiber, g.29****
      Potassium, mg.16*
      Sodium, mg–.22**
      Parent
      Fruit and Vegetable Inventory (fruit/vegetable mediators)
      Parent perception of diet quality, vegetable readiness to eat more, fruit readiness to eat more, vegetable and fruit self-efficacy.
      .36****
      Food Behavior Checklist (7 fruit/vegetable items)
      Sixteen items on the University of California Food Behavior Checklist include parent vegetable quantity, as snacks and at main meal, and types; fruit quantity, types, as citrus; perception of diet quality; food insecurity, milk intake, on cereal; sugar-sweetened beverages; fat on chicken and meat; fish; and food labels.
      .54*****
       Dairy/milk
      Spearman rank-order correlation coefficient r estimated using 3 modified 24-hour dietary recalls showing parent report of child's diet. Results expressed per 1,000 kcal.
      1Food group
      Cup equivalents, dairy.15*
      Nutrients
      Calcium, mg.41*****
      Vitamin D, IU.47*****
      Sugar-sweetened beverages
      Spearman rank-order correlation coefficient r estimated using 3 modified 24-hour dietary recalls showing parent report of child's diet. Results expressed per 1,000 kcal.
      2Food item
      Sugar, g–.18**
       Saturated fat
      Spearman rank-order correlation coefficient r estimated using 3 modified 24-hour dietary recalls showing parent report of child's diet. Results expressed per 1,000 kcal.
      4Nutrients
      Saturated fat, g–.16*
      Total fat, g–.18**
      Calories from fat, kcal–.15*
      Dietary energy density, kcal/g–.38*****
      Healthy Kids dietary scale
      Of the 19 items in the final version of Healthy Kids, 14 are specific to dietary behaviors; *P ≤ .10, **P ≤ .05, ***P ≤ .01, ****P ≤ .001, *****P ≤ .0001.
      14Parent
      Food Behavior Checklist (16 dietary items)
      Sixteen items on the University of California Food Behavior Checklist include parent vegetable quantity, as snacks and at main meal, and types; fruit quantity, types, as citrus; perception of diet quality; food insecurity, milk intake, on cereal; sugar-sweetened beverages; fat on chicken and meat; fish; and food labels.
      .47*****
      Other subscales or category
      Television
      Spearman rank-order correlation coefficient r estimated using 3 days of 36-hour logs showing parent report of television time, and computer time. Average sleep from logs with 6 bedtimes and 6 wakeup times, and average physical activity from 3 logs are shown.
      1Average total television, min–.53*****
      Other screen time
      Spearman rank-order correlation coefficient r estimated using 3 days of 36-hour logs showing parent report of television time, and computer time. Average sleep from logs with 6 bedtimes and 6 wakeup times, and average physical activity from 3 logs are shown.
      1Average total video/computer, min–.50*****
      Physical activity
      Spearman rank-order correlation coefficient r estimated using 3 days of 36-hour logs showing parent report of television time, and computer time. Average sleep from logs with 6 bedtimes and 6 wakeup times, and average physical activity from 3 logs are shown.
      1Child physical activity, average of 3 d, min.21***
      Bedtime
      Spearman rank-order correlation coefficient r estimated using 3 days of 36-hour logs showing parent report of television time, and computer time. Average sleep from logs with 6 bedtimes and 6 wakeup times, and average physical activity from 3 logs are shown.
      1Sleep, no nap, average of 3 d, min.22***
      Average sleep with nap, min.22***
      Parenting1No data currently available to validate
      a Spearman rank-order correlation coefficient r estimated using 3 modified 24-hour dietary recalls showing parent report of child's diet. Results expressed per 1,000 kcal.
      b Spearman rank-order correlation coefficient r estimated using 3 days of 36-hour logs showing parent report of television time, and computer time. Average sleep from logs with 6 bedtimes and 6 wakeup times, and average physical activity from 3 logs are shown.
      c Parent perception of diet quality, vegetable readiness to eat more, fruit readiness to eat more, vegetable and fruit self-efficacy.
      d Sixteen items on the University of California Food Behavior Checklist include parent vegetable quantity, as snacks and at main meal, and types; fruit quantity, types, as citrus; perception of diet quality; food insecurity, milk intake, on cereal; sugar-sweetened beverages; fat on chicken and meat; fish; and food labels.
      e Of the 19 items in the final version of Healthy Kids, 14 are specific to dietary behaviors; *P ≤ .10, **P ≤ .05, ***P ≤ .01, ****P ≤ .001, *****P ≤ .0001.

      Other Categories and Scales

      The television item was related to the television screen time reported on the mean of 3 ≥ 36-hour logs (r = –.53; P ≤ .001), as shown in Table 3. The computer/video game screen item was related to the computer/video screen time (r = –.50; P ≤ .001). For the sleep variable, bedtime was related to average sleep time with naps (r = .22; P = .01). The analysis was repeated without naps, with similar results (r = .22; P = .01). No data were collected to validate the parenting item. The single physical activity item was related to activity time, in minutes, averaged for the 3 ≥ 36-hour logs (r = .21; P = .01). The parent behaviors assessed by the University of California Food Behavior Checklist were significantly related to the child's score on Healthy Kids (r = .47; P ≤ .001) on the 14 items composing the child dietary scale. In addition, a scale score composed of 4 parent mediators of fruit and vegetable behaviors assessed by the University of California Fruit and Vegetable Inventory was positively related to the child's fruit and vegetable score on Healthy Kids (r = .30; P ≤ .001).

      Reliability

      Item difficulty ranged from a low of .51 for 3 items about fruits or vegetables to a high of .90 for Child plays computer games (Table 4). Item discrimination ranged from a low of .08 for My child drinks milk to a high of .60 for I buy fruit. Item and scale stability examined reliability longitudinally. Item stability ranged from r = .31 for I buy vegetables to r = .69 for Type of milk child drinks (Table 4). The tool of 19 items produced a test-retest reliability coefficient of .74 (P ≤ .01). Internal consistency was an acceptable α, .76, for the 19 items. Coefficient of variation (CV), the interpersonal variation among participants, ranged from a low of 15% for My child eats candy, cake, and cookies and My child plays computer games to a high of 42% for I trim fat from meat.
      Table 4Healthy Kids Assessment Tool With Corresponding Test for Each Item and Overall Scale: Item Difficulty Index, Item Discrimination Index, Test Retest Reliability for Items and Scale, Coefficient of Variation, and Internal Consistency
      No.Item ContentItem Difficulty Index
      Item difficulty scores were calculated by dividing the item's mean by 5, the maximum possible score on a 5-item scale (n = 174). Desired cut points = .20 ≥ n ≤ .90.
      Item Discrimination Index
      Discrimination index is the corrected item-total correlation (item score divided by total scale score without item) (n = 174). Desired cut point ≥ .20.
      Reliability/Stability
      Nonparametric Spearman rank order correlation, 1-tailed, at 2 time points 8 weeks apart with no intervention, coefficient r (n = 98). Desired cut point ≥ .30.
      Coefficient of Variation
      Coefficient of variation was calculated as the SD divided by the item mean × 100 (n = 172).
      (%)
      1Parent buys vegetables.84.54.3119
      2Child snacks on apples and carrots.71.54.5731
      3Child eats vegetables at main meal.51.28.4730
      4Child ≥1 kind of vegetable.51.54.6041
      5Child eats fruit.67.55.5728
      6Parent buys fruit.87.60.4521
      7Fruit ready for child to eat.51.59.6027
      8My child drinks milk.65.08.4732
      9Child drinks soda.82.28.4618
      10Child drinks sports drinks.83.30.5222
      11Child eats candy, cake, and cookies.79.23.5515
      12Milk type.63.21.6927
      13Child eats chips.87.25.5918
      14Parent trims fat from meat.70.29.4942
      15Parent eats with child.83.30.6025
      16Child watches x hours of television.69.24.4424
      17Computer games.90.12.3415
      18Child plays instead of television.67.22.3231
      19Bedtime.62.1938
      19-item Healthy Kids.75NA.74NA
      NA indicates not available.
      a Item difficulty scores were calculated by dividing the item's mean by 5, the maximum possible score on a 5-item scale (n = 174). Desired cut points = .20 ≥ n ≤ .90.
      b Discrimination index is the corrected item-total correlation (item score divided by total scale score without item) (n = 174). Desired cut point ≥ .20.
      c Nonparametric Spearman rank order correlation, 1-tailed, at 2 time points 8 weeks apart with no intervention, coefficient r (n = 98). Desired cut point ≥ .30.
      d Coefficient of variation was calculated as the SD divided by the item mean × 100 (n = 172).

      Other psychometric properties

      The addition of photographs portraying each item allowed the number of words for an item to be substantially reduced, resulting in an improved readability index score. The Flesch-Kincaid readability index produced an overall score of grades 1–2 (MS Word 2013, Microsoft Corporation, Redmond, WA). For administration of the 19-item Healthy Kids in a group community setting, respondents took ≤10 minutes to complete on average, meeting the researchers' criteria for rapid assessment. Items with 1- and 2-syllable words and the pictorial format met the criteria for minimal respondent burden for a self-administered tool. The 19-item Healthy Kids is arranged on 1 11 × 17 folded, 2-sided, sheet, and is shown in Figure 2.
      • Townsend M.S.
      • Shilts M.K.
      • Styne D.
      • Lanoue L.
      • Ontai L.
      Healthy Kids.
      Figure 2
      Figure 2Healthy Kids assessment tool with 19 items, printed in color on 1 11 × 17-in page, folded in booklet format.

      Discussion

      This study provides support for the validation of a risk assessment questionnaire with low-literate parents. Using the Biopsychosocial Framework (Figure 1), these data show that the Healthy Kids tool is valid to use for concurrent as well as predictive risk assessment in families with 2- to 5-year-old children.
      The relationship was significant between Healthy Kids scores and BMI percentiles-for-age measured at nearly 1.5 years after baseline but was not significant between them at baseline. Children with higher Healthy Kids scores, ie, healthier, at baseline as a group tended to have lower BMI percentiles-for-age. If 2 children had the same BMI percentile-for-age at baseline, the child with a higher Healthy Kids score at baseline more likely had a lower BMI 88 weeks later. These results suggest that the consequences of a suboptimal family environment may not have affected 2- to 3-year-old children, as assessed by BMI, when they entered this study, but the family environment affected them more 1.5 years later, as indicated by the children's less than healthful weight gain.
      An incremental decrease between BMI percentile-for-age and the Healthy Kids scores was not consistently observed in the current study. Using the Family Nutrition and Physical Activity tool to identify home environmental factors to identify children at risk for obesity, Peyer and Welk
      • Peyer K.L.
      • Welk G.J.
      Construct validity of an obesity risk screening tool in two age groups.
      and Yee et al
      • Yee K.E.
      • Pfeiffer K.A.
      • Turek K.
      • et al.
      Association of the family nutrition and physical activity screening tool with weight status, percent body fat, and acanthosis nigricans in children from a low socioeconomic, urban community.
      also observed no or low correlation between the scores and changes in BMI percentiles in first-through eighth-graders and 10th-graders. However, results from both populations showed that children with scores in the lowest tertiles had increased odds for obesity compared with children in the highest tertiles. With this Healthy Kids tool, scores in the highest tertile associated with the most healthful behaviors had significantly lower BMI-for-age percentiles (Table 2). For a linear regression model, this would indicate a nearly flat slope over the lower range of scores with a change in slope at the higher tertile.
      • Draper N.
      • Smith H.
      Applied Regression Analysis.
      Hence, although the BMI variable decreased monotonically with improved Healthy Kids scores, the rate of decrease was not constant. This lack of linearity in the low-range Healthy Kids scores and BMI-for-age percentile could be explained by a social desirability bias, eg, parents may have been hesitant to admit providing chips at snack time or soda on most days.
      • Draper N.
      • Smith H.
      Applied Regression Analysis.
      • Hebert J.R.
      • Clemow L.
      • Pbert L.
      • Ockene I.S.
      • Ockene J.K.
      Social desirability bias in dietary self-report may compromise the validity of dietary intake measures.
      Thus, for these 2 examples, social desirability bias would suggest overreporting of healthful snacking behavior and underreporting of the frequency of soda consumption. This bias would have the effect of reducing the level of statistical significance when examining relationships between Healthy Kids scores and outcome variables, ie, underreporting the true effect.
      • Hebert J.R.
      • Clemow L.
      • Pbert L.
      • Ockene I.S.
      • Ockene J.K.
      Social desirability bias in dietary self-report may compromise the validity of dietary intake measures.
      • Rothman K.
      • Greenland S.
      • Lash T.
      Modern Epidemiology.
      Despite these potential sources of misreporting,
      • Hebert J.R.
      • Clemow L.
      • Pbert L.
      • Ockene I.S.
      • Ockene J.K.
      Social desirability bias in dietary self-report may compromise the validity of dietary intake measures.
      • Rothman K.
      • Greenland S.
      • Lash T.
      Modern Epidemiology.
      the medians of BMI percentile-for-age showed a significant inverse association with Healthy Kids scores expressed as tertiles; the majority of children in highest tertile had a lower BMI percentile than the majority of children in the lowest tertile. Thus, the median data showed the robustness of the 19-item tool to predict BMI percentiles-for-age changes in this longitudinal study.
      The item difficulty index varied for each of the 19 items of the Healthy Kids tool. One interpretation is that some of the items were easier to assess or remember by parents than others.
      • Townsend M.S.
      • Sylva K.
      • Shilts M.K.
      • Davidson C.
      • Leavens L.
      • Sitnick S.L.
      Healthy Kids: Pediatric obesity risk assessment tool [45 items].
      Another is that some behaviors were easier to implement than others.
      • Townsend M.S.
      • Sylva K.
      • Shilts M.K.
      • Davidson C.
      • Leavens L.
      • Sitnick S.L.
      Healthy Kids: Pediatric obesity risk assessment tool [45 items].
      For example, almost all parents (90%) consistently reported that their child did not play computer games whereas 51% of parents responded that their child ate vegetables. Item discrimination ranged from a low of .08 for My child drinks milk to a high of .60 for I buy fruit (Table 4). Parents reporting that they bought fruit responded similarly to other Healthy Kids items on the response continuum; hence, there was a high score of .60.
      • Townsend M.S.
      • Sylva K.
      • Shilts M.K.
      • Davidson C.
      • Leavens L.
      • Sitnick S.L.
      Healthy Kids: Pediatric obesity risk assessment tool [45 items].
      • Nunnally J.C.
      • Bernstein I.H.
      Psychometric Theory.
      However, parents did not respond as expected to My child drinks milk. Parents who reported healthful behaviors for their children sometimes did not report milk consumption as 1 of them. The interpretation is that frequency of milk drinking was not related to other healthful behaviors such as eating vegetables, although it was related to BMI.
      • Townsend M.S.
      • Sylva K.
      • Shilts M.K.
      • Davidson C.
      • Leavens L.
      • Sitnick S.L.
      Healthy Kids: Pediatric obesity risk assessment tool [45 items].
      Perhaps this is the result of a misinterpretation of the advisory from the American Academic of Pediatrics recommending that children aged ≥2 years transition to drinking nonfat or 1% milk from whole milk; this could have led to misclassification and/or underreporting by parents. An item that discriminates as expected would score similarly to the others in the scale, as indicated by a desired value of r ≥ .20.
      • Nunnally J.C.
      • Bernstein I.H.
      Psychometric Theory.
      Three items did not. These results showed that low-income families practiced the behaviors related to obesity differentially; in other words, a parent practicing 1 healthy behavior may not have practiced other healthy behaviors.
      • Nunnally J.C.
      • Bernstein I.H.
      Psychometric Theory.
      Despite acknowledging this information, the item discrimination index provides some insights and is presented.
      The interpretation for the item stability results of r = .31–.69 is that the I buy vegetables item was the least reliable temporally, meaning that there was more variability in responses from the first administration of the tool to the second 12 weeks later.
      • Townsend M.S.
      • Sylva K.
      • Shilts M.K.
      • Davidson C.
      • Leavens L.
      • Sitnick S.L.
      Healthy Kids: Pediatric obesity risk assessment tool [45 items].
      • Litwin M.S.
      How to Measure Survey Reliability and Validity.
      The type of milk child drinks item was the most stable, meaning that parents perceived they served the same milk type consistently to their child, whereas the vegetable item was answered less consistently. The tool of 19 items was considered to have good test-retest reliability (r = .74; P ≤ .01) (Table 4), because this value exceeded the desired minimum of .70 for good consistency.
      • Litwin M.S.
      How to Measure Survey Reliability and Validity.
      Internal consistency was an acceptable α, .76, for the 19 items.
      • Litwin M.S.
      How to Measure Survey Reliability and Validity.
      The CV of the Healthy Kids ranged from 15% to 42%. The 2 items with the lowest CV, My child eats candy, cake, and cookies and My child plays computer games, indicated that parents responded similarly, ie, with minimal variation, to the candy, cake, and cookies item and also to the computer item.
      • Traub R.E.
      The item with the highest CV of 42% was I trim fat from meat, which suggested that parents responded with high variation to trimming fat from meat. Higher CV scores are more appropriate for risk assessment.
      Test-retest reliability for Healthy Kids indicated good consistency in parent responses repeated 12 weeks later,
      • Litwin M.S.
      How to Measure Survey Reliability and Validity.
      which suggested that variation in Healthy Kids scores reflected real change in underlying constructs rather than the random measurement error. With a similar low-income audience, Dickin et al
      • Dickin K.L.
      • Lent M.
      • Lu A.H.
      • Sequeira J.
      • Dollahite J.S.
      Developing a measure of behavior change in a program to help low-income parents prevent unhealthful weight gain in children.
      reported a larger reliability coefficient for a behavioral checklist but with a short 2-week test-retest period. At the same time, Dickin et al did not report child dietary recall data or child heights and weights for a comparison with this study. Alternately, the results from Ihmels et al
      • Ihmels M.A.
      • Welk G.J.
      • Eisenmann J.C.
      • Nusser S.M.
      • Myers E.F.
      Prediction of BMI change in young children with the family nutrition and physical activity (FNPA) screening tool.
      with middle- and low-income parents completing an obesity risk survey for children aged 6–12 years were comparable to those of this study. In addition, the survey by Ihmels et al predicted changes in BMI percentiles-for-age at 1-year follow-up (P = .05; n = 704), a result similar to that observed in this study with low-income parents and a smaller sample size (P = .02; n = 98).
      Healthy Kids diet scores correlated well with parent behaviors (r = .47; P ≤ .001). This result was essentially identical to those of Dickin et al
      • Dickin K.L.
      • Lent M.
      • Lu A.H.
      • Sequeira J.
      • Dollahite J.S.
      Developing a measure of behavior change in a program to help low-income parents prevent unhealthful weight gain in children.
      (r = .48; P ≤ .001), using the University of California food behavior checklist.
      • Sylva K.
      • Townsend M.S.
      • Martin A.
      • Metz D.
      University of California Food Behavior Checklist.
      The Healthy Kids fruit and vegetable scale correlated strongly with parent fruit and vegetable behaviors (r = .54; P ≤ .001), also similar to results from Dickin et al (r = .46; P ≤ .001). Convergence of Healthy Kids with the University of California food behavior checklist, previously validated against serum carotenoids,
      • Townsend M.S.
      • Kaiser L.L.
      • Allen L.H.
      • Joy A.B.
      • Murphy S.P.
      Selecting items for a food behavior checklist for a limited resource audience.
      strengthens the credibility of the Healthy Kids findings.
      The relationship of Healthy Kids subscale scores with an individual nutrient was weak (r = .16 for potassium) to strong (r = .47 for vitamin D) in statistical significance (Table 3). However, when an individual nutrient is considered in combination with other relevant nutrients and cup equivalents, evidence for a relationship between the child's nutrient intakes and the Healthy Kids scale score becomes stronger and clinically significant. Several possible factors explain some weaknesses of individual correlation coefficients. Healthy Kids items asked parents about usual behavior, whereas parents responded to the 24-hour dietary recalls for specific days, which resulted in more day-to-day variation. In addition, there was >1 food group source for most nutrients, such as dietary fiber. A high concordance cannot be expected with Healthy Kid scores from the fruit and vegetable items because foods other than fruits and vegetables, ie, whole grains, legumes, nuts, and seeds, contribute to dietary fiber.
      Federal program participants are requested to complete assessment tools for data reporting, although no federal mandate exists. Therefore, it behooves nutrition education professionals to design data collection tools to accommodate participant literacy needs and to use a tool format that motivates the participant.
      • Townsend M.S.
      • Sylva K.
      • Martin A.
      • Metz D.
      • Wooten Swanson P.
      Improving readability of an evaluation tool for low-income respondents using visual information processing theories.
      A desired outcome of this research process is a reduction in the error associated with participant misunderstanding or skipping items associated with literacy and related motivational issues.
      • Townsend M.S.
      • Ganthavorn C.
      • Neelon M.
      • Donohue S.
      • Johns M.C.
      Improving the quality of data from EFNEP participants with low-literacy skills: a participant-driven model.
      The Healthy Kids behavioral checklist format with color photographs and minimal literacy demands circumvents the limitations of traditional data collection methods
      • Townsend M.S.
      • Sylva K.
      • Martin A.
      • Metz D.
      • Wooten Swanson P.
      Improving readability of an evaluation tool for low-income respondents using visual information processing theories.
      such as the 24-hour diet recall or the invasiveness of measuring height and weight in school or community settings.
      • Ikeda J.P.
      • Crawford P.B.
      • Woodward-Lopez G.
      BMI screening in schools: helpful or harmful.
      The simplicity of the tool works for parent groups in a community setting where self-administration is the norm.
      • Townsend M.S.
      • Ganthavorn C.
      • Neelon M.
      • Donohue S.
      • Johns M.C.
      Improving the quality of data from EFNEP participants with low-literacy skills: a participant-driven model.
      The primary strength of this study was the use of multiple and robust analytical approaches to confirm the validity of Healthy Kids (Figure 1) and inform reliability. The framework for the study incorporated both subjective measures using self-report data, ie, parent reports of the child's diet, and objective measures external to the parent, ie, anthropometric markers, to determine the validity of parent responses on Healthy Kids. The external validity of these findings to other limited-resource audiences is unknown. However, these findings were consistent with theories of visual information processing
      • Townsend M.S.
      • Sylva K.
      • Martin A.
      • Metz D.
      • Wooten Swanson P.
      Improving readability of an evaluation tool for low-income respondents using visual information processing theories.
      • Levie W.H.
      • Lentz R.
      Effects of text illustrations: a review of research.
      with efforts to reduce client cognitive load for clients with minimal literacy skills.
      • Townsend M.S.
      • Ganthavorn C.
      • Neelon M.
      • Donohue S.
      • Johns M.C.
      Improving the quality of data from EFNEP participants with low-literacy skills: a participant-driven model.
      These results may be applicable to similar groups of program participants in California. Other limitations should be considered. The analysis for the 3 recalls included only dinners. The variability of parent responses in this cohort may be lower than that of the general limited-resource target population, because parents volunteered to participate in the study. Consequently, selection bias must be considered as a threat to external validity.
      • Cook T.D.
      • Campbell D.T.
      Quasi-experimental: Design and Analyses Issues for Field Studies.
      Items that can compromise internal validity are social desirability bias influencing parent responses on all parent-report tools and the Hawthorne effect from the 24-hour diet recalls and the ≥36-hour activity and sleep logs serving as motivators for healthier behaviors.
      • Draper N.
      • Smith H.
      Applied Regression Analysis.
      • Levie W.H.
      • Lentz R.
      Effects of text illustrations: a review of research.

      Implications for Research and Practice

      Healthy Kids is a work in progress with face, content, convergent, and predictive validity that has been established. The next step includes reporting validation results using children's blood values. Also under way is a research study with another target audience, Spanish-speaking parents and their young children. To provide consistency in administering the tool to reduce random error and thereby enhance reliability, an instruction guide was developed and reviewed by 4 experts in program content and 3 paraprofessional staff familiar with low-income clients.
      • Townsend M.S.
      • Shilts M.K.
      • Leavens L.
      • et al.
      Instruction guide: Healthy Kids (Child Obesity Risk Assessment Tool). Version 1.0.
      My Child at Mealtime, a parenting practices tool developed by the same research group, is intended to be used in tandem with Healthy Kids.
      • Ontai L.
      • Sitnick S.
      • Sylva K.
      • Leavens L.
      • Davidson C.
      • Townsend M.S.
      My Child at Meal Time. Version 1.
      • Ontai L.L.
      • Sitnick S.
      • Shilts M.K.
      • Townsend M.S.
      My child at mealtime: a visually enhanced self-assessment of feeding styles for low-income parents of preschoolers.
      Tools for families with preschoolers usually target 1 or 2 domains such as dietary fat, dairy, physical activity, or screen time.
      • Townsend M.S.
      • Ontai L.
      • Young T.
      • Ritchie L.D.
      • Williams S.T.
      Guiding family-based obesity prevention efforts in low-income children in US: part 2—What behaviors do we measure?.
      Instead, Healthy Kids was developed and validated as a novel multidomain risk assessment tool for obesity prevention targeting low-income families with young children and with a readability index of grades 1–2.
      This tool could serve nutrition professionals, educators, and clinicians in many capacities. First, used in tandem with My Child at Meal Time, Healthy Kids scores could contribute to an assessment of programmatic needs for agency professionals. Second, it could assist with screening to identify young children who are most at risk for excess weight gain. Third, nutrition professionals could use that information to target counseling or nutrition education classes for families of children identified for early intervention efforts as recommended by the AAP.
      • American Academy of Pediatrics
      Prevention of pediatric overweight and obesity.
      Fourth, with prevention as the goal, the tool could be a valuable health promotion opportunity for providing individualized feedback and intervention information to parents enrolled in SNAP-Ed, EFNEP, WIC, or Head Start. Tailored goals can be generated
      • Shilts M.K.
      • Townsend M.S.
      • Leavens L.L.
      • Davidson C.A.
      • Ontai L.L.
      • Reed M.L.
      Healthy Kids Website.
      based on parent Healthy Kids responses. Fifth, it could function as an evaluation to assess the impact of the intervention. Finally, in a medical setting, clinicians could use this risk assessment tool to supplement a physical examination to predict a BMI trajectory in young children and craft personalized messages for patients' families. Healthy Kids has the potential for use by nutrition professionals as a rapid and easily administered assessment of a child's family environment and risk for becoming overweight or obese.

      Acknowledgments

      This material is based on work that is supported by the National Institute of Food and Agriculture, USDA, under Award Nos. 2009-55215-05019 and 2010-85215-20658. The authors thank Professor Kathryn Sylva for her graphic design expertise; Brenda Campos and Meghan Marshall of SETA Head Start in Sacramento, CA, for their ongoing support making this study possible; Christine Davidson, Larissa Leavens, and Danielle Rehnstrom for data collection and processing; and Shaoxuan Wang for analytical support.

      References

        • American Academy of Pediatrics
        Prevention of pediatric overweight and obesity.
        Pediatrics. 2003; 112: 424-450
        • Ogden C.L.
        • Carroll B.K.
        • Flegal K.M.
        Prevalence of obesity and trends in body mass index among US children and adolescents, 1999–2010.
        JAMA. 2012; 307: 483-490
        • US Department of Health and Human Services
        Office of Head Start.
        http://www.acf.hhs.gov/programs/ohs
        Date: 2016
        Date accessed: October 30, 2017
        • USDA Food and Nutrition Service
        Supplemental nutrition program for women, infants and children (WIC).
        http://www.fns.usda.gov/wic
        Date: 2017
        Date accessed: October 30, 2017
        • USDA Food and Nutrition Service
        Supplemental nutrition assistance program education guidance (SNAP-Ed). 2014b.
        http://snap.nal.usda.gov
        Date: 2017
        Date accessed: October 30, 2017
        • USDA National Institute for Food and Agriculture
        Expanded Food and Nutrition Education Program (EFNEP).
        • Townsend M.S.
        Obesity in low-income communities: prevalence, effects, a place to begin.
        J Am Diet Assoc. 2006; 106: 34-37
        • Institute of Medicine of the National Academies
        Preventing Childhood Obesity: Health in Balance.
        National Academies Press, Washington, DC2005
        • Ontai L.
        • Ritchie L.
        • Williams S.T.
        • Young T.
        • Townsend M.S.
        Guiding family-based obesity prevention efforts in low-income children in US: part 1−What determinants do we target?.
        Int J Child Adolesc Health. 2009; 2: 19-30
        • Townsend M.S.
        • Ontai L.
        • Young T.
        • Ritchie L.D.
        • Williams S.T.
        Guiding family-based obesity prevention efforts in low-income children in US: part 2—What behaviors do we measure?.
        Int J Child Adolesc Health. 2009; 2: 31-48
        • Ihmels M.A.
        • Welk G.J.
        • Eisenmann J.C.
        • Nusser S.M.
        • Myers E.F.
        Prediction of BMI change in young children with the family nutrition and physical activity (FNPA) screening tool.
        Annal Behav Med. 2009; 38: 60-68
        • Dickin K.L.
        • Lent M.
        • Lu A.H.
        • Sequeira J.
        • Dollahite J.S.
        Developing a measure of behavior change in a program to help low-income parents prevent unhealthful weight gain in children.
        J Nutr Educ Behav. 2012; 44: 12-21
        • Townsend M.S.
        • Shilts M.K.
        • Sylva K.
        • et al.
        Obesity risk for young children: development and initial validation of an assessment tool for participants of USDA programs.
        Forum Fam Consum Issues. 2014; 3
        • Engel G.L.
        The clinical application of the biopsychosocial model.
        Am J Psychiatry. 1980; 137: 535-544
        • Townsend M.S.
        • Kaiser L.L.
        University of California Fruit and Vegetable Inventory.
        • Sylva K.
        • Townsend M.S.
        • Martin A.
        • Metz D.
        University of California Food Behavior Checklist.
        • Townsend M.S.
        • Sylva K.
        • Shilts M.K.
        • Davidson C.
        • Leavens L.
        • Sitnick S.L.
        Healthy Kids: Pediatric obesity risk assessment tool [45 items].
        University of California Cooperative Extension, Davis, CA2011
        • Townsend M.S.
        • Shilts M.K.
        • Davidson C.
        • Leavens L.
        My Child's Activity & Food Diary.
        • USDA Food and Nutrition Service
        The Supplemental Nutrition Assistance Program Education (SNAP-Ed) strategies and interventions: an obesity prevention toolkit for states.
        • Townsend M.S.
        • Kaiser L.L.
        Development of an evaluation tool to assess psychosocial indicators of fruit and vegetable intake for two federal programs.
        J Nutr Educ Behav. 2005; 37: 170-184
        • Townsend M.S.
        • Kaiser L.L.
        • Allen L.H.
        • Joy A.B.
        • Murphy S.P.
        Selecting items for a food behavior checklist for a limited resource audience.
        J Nutr Educ Behav. 2003; 35: 69-82
        • Townsend M.S.
        • Sylva K.
        • Martin A.
        • Metz D.
        • Wooten Swanson P.
        Improving readability of an evaluation tool for low-income respondents using visual information processing theories.
        J Nutr Educ Behav. 2008; 40: 181-186
        • Kuczmarski R.J.
        • Ogden C.L.
        • Grummer-Strawn L.M.
        • et al.
        CDC Growth Charts for United States: Advance Data from Vital and Health Statistics.
        National Center for Health Statistics, Hyattsville, MD2000: 314 (Report)
        • Conway J.M.
        • Ingwersen L.A.
        • Vinyard B.T.
        • Moshfegh A.J.
        Effectiveness of the USDA 5-step multiple-pass method in assessing food intake in obese and non-obese women.
        Am J Clin Nutr. 2003; 77: 1171-1178
        • Murphy S.
        • Kaiser L.L.
        • Townsend M.S.
        • Allen L.
        Evaluation of validity of items in a food behavior checklist.
        J Am Diet Assoc. 2001; 101: 751-761
        • Townsend M.S.
        • Kaiser L.L.
        Brief psychosocial fruit and vegetable tool is sensitive for United States Department of Agriculture's nutrition education programs.
        J Am Diet Assoc. 2007; 107: 2120-2124
        • Litwin M.S.
        How to Measure Survey Reliability and Validity.
        Sage, Thousand Oaks, CA1995
        • Nunnally J.C.
        • Bernstein I.H.
        Psychometric Theory.
        3rd ed. McGraw-Hill, New York, NY1994
        • Traub R.E.
        Reliability for the Social Sciences: Theory and Applications. Vol. 3. Sage, Thousand Oaks, CA1994
        • Townsend M.S.
        • Shilts M.K.
        • Styne D.
        • Lanoue L.
        • Ontai L.
        Healthy Kids.
        • Peyer K.L.
        • Welk G.J.
        Construct validity of an obesity risk screening tool in two age groups.
        Int J Environ Res Public Health. 2017; 14: 419
        • Yee K.E.
        • Pfeiffer K.A.
        • Turek K.
        • et al.
        Association of the family nutrition and physical activity screening tool with weight status, percent body fat, and acanthosis nigricans in children from a low socioeconomic, urban community.
        Ethn Dis. 2015; 25: 339-404
        • Draper N.
        • Smith H.
        Applied Regression Analysis.
        3rd ed. John Wiley & Sons, New York1998
        • Hebert J.R.
        • Clemow L.
        • Pbert L.
        • Ockene I.S.
        • Ockene J.K.
        Social desirability bias in dietary self-report may compromise the validity of dietary intake measures.
        Int J Epidemiol. 1995; 24: 389-398
        • Rothman K.
        • Greenland S.
        • Lash T.
        Modern Epidemiology.
        3rd ed. Lippincott, Williams & Wilkins, Philadelphia, PA2008
        • Townsend M.S.
        • Ganthavorn C.
        • Neelon M.
        • Donohue S.
        • Johns M.C.
        Improving the quality of data from EFNEP participants with low-literacy skills: a participant-driven model.
        J Nutr Educ Behav. 2014; 46: 309-314
        • Ikeda J.P.
        • Crawford P.B.
        • Woodward-Lopez G.
        BMI screening in schools: helpful or harmful.
        Health Educ Res. 2006; 21: 761-769
        • Levie W.H.
        • Lentz R.
        Effects of text illustrations: a review of research.
        Educ Commun Technol J. 1982; 30: 195-232
        • Cook T.D.
        • Campbell D.T.
        Quasi-experimental: Design and Analyses Issues for Field Studies.
        Houghton Mifflin, Boston, MA1979
        • Townsend M.S.
        • Shilts M.K.
        • Leavens L.
        • et al.
        Instruction guide: Healthy Kids (Child Obesity Risk Assessment Tool). Version 1.0.
        https://ucdavis.app.box.com/s/3ux4l1bp2vzvp6np01m9zadr82phecpn
        Date: 2011
        Date accessed: November 10, 2017
        • Ontai L.
        • Sitnick S.
        • Sylva K.
        • Leavens L.
        • Davidson C.
        • Townsend M.S.
        My Child at Meal Time. Version 1.
        Townsendlab.ucdavis.edu
        Date: 2011
        Date accessed: November 6, 2017
        • Ontai L.L.
        • Sitnick S.
        • Shilts M.K.
        • Townsend M.S.
        My child at mealtime: a visually enhanced self-assessment of feeding styles for low-income parents of preschoolers.
        Appetite. 2016; 99: 76-81
        • Shilts M.K.
        • Townsend M.S.
        • Leavens L.L.
        • Davidson C.A.
        • Ontai L.L.
        • Reed M.L.
        Healthy Kids Website.
        Regents of the University of California, Davis, CA2014
        http://healthykids.ucdavis.edu/
        Date accessed: November 10, 2017

      CHORUS Manuscript

      View Open Manuscript