The study was approved by the Institutional Review Board at the University of Hawai‘i at Manoa, and informed consent and assent were obtained. Health center personnel made the initial contact with eligible child-caregiver pairs, and scheduled study visits were in compliance with Health Insurance Portability and Accountability Act privacy rules. Through the fall of 2005 and summer of 2006, interviews were conducted at the health center or at the respondent's home, according to their preference and availability.
Participants were drawn from two rural sites on O‘ahu and two sites on the Big Island of Hawai‘i in the state of Hawai‘i, selected for rural location and high proportion of Native Hawaiian population. On the island of O‘ahu, child-caregiver pairs were selected from local health center patient databases. In these communities, eligibility criteria included: (1) Child had an encounter with the health care facility during the past 2 years, and (2) Child was 8–12 years old at the time of the pre-intervention survey interview.
For the two sites on the Big Island of Hawai‘i, sampling was done according to census tracts.5
A random cluster sampling technique was used to identify how many children from each census block would be recruited. Within the selected census block cluster, data collectors went systematically from door to door to obtain the specified proportion of respondents, until the desired sample was achieved. The overall response rate was 80%. In total, 183 child - caregiver pairs were interviewed.
Questionnaires were administered to both the caregiver and the child. There is a high level of ethnic mixing in Hawai‘i. Thus, care-givers were asked the standard ethnic identity question, “What ethnic or racial group do you most identify with?” However, they were permitted to identify more than one group. The following ethnic groups were provided on the questionnaire: Native American, African American/Black, Chinese, Filipino, Japanese, Korean, Hispanic, Hawaiian/Part-Hawaiian, Other Pacific Islander, and White. The following algorithm was used to define ethnicity for the presented analyses:
- Native Hawaiian - if the subject checked Hawaiian/Part-Hawaiian. (However, this group could be mixed with other ethnic groups, which were White, Chinese, Filipino, Japanese, Korean, Hispanic, Pacific Islander, Native American or African American. Only Hawaiian ethnicity was reported by 85 of 110 individuals, while 18/110 reported one other ethnicity, six reported two other ethnicities and three reported three other ethnicities).
- Filipino - if the subject checked Filipino and did not check Hawaiian/Part-Hawaiian. (This group could be mixed with other ethnic groups which included White, Chinese, Japanese, Korean, Hispanic, Pacific Islander, Native American or African American. Filipino ethnicity only was reported by 22 of 28, while 6 of 28 reported one other ethnicity).
- White - if the subject checked White and did not select Hawaiian/Part-Hawaiian or Filipino. (This group could be mixed with the Chinese, Japanese, Korean, Hispanic, other Pacific Islander, Native American or African American. However, no other ethnic group was reported for any White participant).
Twenty-seven child-caregiver pairs who did not meet the ethnic criteria for this analysis were excluded due to small sample size of that ethnic group (those who were either 100% Chinese, Japanese, Korean, Hispanic, other Pacific Islander, Native American, African American, or who they said they were mixed but were unable to specify the mixture). Thus, 156 child-caregiver pairs were included in the analysis (18 White, 28 Filipino, 110 Native Hawaiian). Among those who had specified the type of caregiver (144 out of the 156 caregivers), 82% (118 out of 144) were a parent.
A single 24-hour dietary recall was collected from each child and then from their caregiver using a modified USDA five-step multiple pass methodology.6
Children were interviewed directly, but the assistance of caregivers was accepted when the child could not remember. To decrease the length of time for the food recall, time and occasion of food and beverage consumption were collected in combination with either the first or third step, yielding a four-step method. A simple set of measuring instruments and plastic food models was used to aid in quantification of amount consumed and a list of commonly “forgotten foods” was used as a prompt during the last stage of the recall. One-sixth of recalls were collected on Mondays, in order to include (recalled) weekend intake. All other recalls reflected weekdays. Data collectors were trained staff from the Healthy Living in the Pacific Islands, Healthy Pacific Child Project, and Healthy Foods Hawai‘i Project at the University of Hawai‘i at Manoa (Novotny R, PI). Healthy Eating Index (HEI) component scores were calculated for nine HEI components: grain, vegetables, fruits, dairy, meat, total fat, saturated fat, cholesterol, and sodium, with a score of 1–10 for each component.7
In addition, data for education and socio-demographic characteristics of the caregiver (birth date, sex, marital status, and years of education) were collected. The completed number of years of school and the employment status were recorded. If employed, the caregiver was asked for hours/week worked, which was classified as: 1 = Seasonally/Occasionally employed; 2 = Part time (less than 30 hrs/wk); 3 = 30–40 hrs/wk; 4 = More than 40 hrs/wk. Due to the high number of unemployed persons and the prevalence of bartering for goods in the community, a material style of living score developed for use in rural, semi-subsistence economies was computed to estimate socio-economic status based on ownership of various items (TV, car, microwave, etc) in working condition available in their household.8
To determine if the study samples were homogeneous, differences between the community of residence, material style of living, and caregiver education among the three ethnic groups were tested using Fisher's exact test in SAS (version 9.1.3, 2002–2003, SAS Institute Inc, Cary, NC) PROC FREQ.
Before performing statistical analysis, the data for dietary intakes were processed using the University of Hawai‘i Cancer Center's food composition database, which contains local foods and recipes not found in the USDA standard reference.9
Food group data were compared to Food Guide Pyramid reference values that were used at the time of analysis.10
To determine if there were ethnic differences in the average dietary intakes, first the overall ethnic effect on any of the food groupings (or the nutrients, or the HEI scores) was simultaneously tested using multivariate analysis of variance (MANOVA). Statistical analyses were conducted using SAS PROC GLM. For both children and caregivers, three independent MANOVAs were performed for food groups, nutrients, and HEI scores. If the results of MANOVA indicated that there was a significant overall ethnic effect, then individual ANOVA were performed to test if the individual elements of the food/nutrient/HEI score was influenced by ethnic group. Multiple comparisons were adjusted for at p<0.05 using the methods of Benjamini and Hochberg.11
Following the individual ANOVA tests, pairwise comparisons (Tukey's honest significance test) were performed on those that showed significant ethnic effect to determine magnitude of effect. All models were adjusted for the potential confounders of age, sex, and “food season.” To adjust for expected “seasonal” (holiday) variation in dietary intake, “food season” variables were created and defined as follows:
- Halloween (yes, if measured between October 31, 2005 and November 10, 2005; or no for any other date)
- Winter holiday (yes, if measured between November 20, 2005 and January 7, 2006; or no for any other date)
- Summer vacation (yes, if measured during June, July, or August in 2006; Or no for any other date)
- School year (yes, if measured between January 8 and January 31, February, March, April, May, September October, and between November 11 and November 19 in 2005 or 2006).