CCox - Writing Sample

11
Excerpted from: Cranston, Catherine R. (2003). Prevalence and Correlates of Drive for Thinness Among a Cohort of Black and White Adolescent Females (Master’s thesis). Emory University, Rollins School of Public Health. Methods Procedures The National Heart, Lung, and Blood Institute Growth and Health Study (NGHS) is a 10-year longitudinal cohort study to assess factors associated with the onset and development of obesity and eating disorders in a biracial cohort of girls. Black and White preadolescent girls (ages 9-10) were recruited by three clinical centers: University of California at Berkeley, University of Cincinnati/Cincinnati Children's Hospital Medical Center, and Westat, Inc./Group Health Association in Washington, DC. These study sites were chosen to ensure a wide range of income and education levels, and varied types of residential areas (inner city, urban, and suburban areas). For a detailed description of the NGHS, please refer to The National Heart, Lung, and Blood Institute Growth and Health Research Group (1992). Sample/Participants. Only those girls recruited by Westat, Inc. in the Washington, DC are included in the following analyses.

Transcript of CCox - Writing Sample

Page 1: CCox - Writing Sample

Excerpted from:Cranston, Catherine R. (2003). Prevalence and Correlates of Drive for Thinness Among a Cohort of Black and White Adolescent Females (Master’s thesis). Emory University, Rollins School of Public Health.

Methods

Procedures

The National Heart, Lung, and Blood Institute Growth and Health Study (NGHS) is a 10-

year longitudinal cohort study to assess factors associated with the onset and development of

obesity and eating disorders in a biracial cohort of girls. Black and White preadolescent girls

(ages 9-10) were recruited by three clinical centers: University of California at Berkeley,

University of Cincinnati/Cincinnati Children's Hospital Medical Center, and Westat, Inc./Group

Health Association in Washington, DC. These study sites were chosen to ensure a wide range of

income and education levels, and varied types of residential areas (inner city, urban, and

suburban areas). For a detailed description of the NGHS, please refer to The National Heart,

Lung, and Blood Institute Growth and Health Research Group (1992).

Sample/Participants. Only those girls recruited by Westat, Inc. in the Washington, DC

are included in the following analyses. Participants for the Westat study were randomly selected

from a membership roster of a Washington, DC based health maintenance organization (Group

Health Association) from eligible families who had girls aged 9 or 10 and who were enrolled in

the prepaid medical program. Because of an insufficient number of age eligible White girls in the

HMO, the cohort was augmented with white girls recruited from a troop of Girl Scouts from the

same geographic area of a predominantly White Group Health Association clinic. To be eligible

to participate, the girls had to: 1) declare themselves as being either Black or White; 2) be

between the ages of 9 and 10 within two weeks of the first clinical visit; 3) have parents or

guardians who declared themselves as the same race of the child; and 4) have their parents or

guardians give consent and complete a household demographic information sheet (National

Heart Lung and Blood Institute Growth and Health Study Research Group, 1992). The final

Page 2: CCox - Writing Sample

Catherine (Cranston) Cox

cohort of girls recruited by Westat consisted of 316 Black and 305 White girls.

Measures. Anthropometric data, including blood pressure, skin-fold thickness, height,

weight, and blood samples were collected by trained field staff at baseline and at each annual

visit. BMI was calculated using weight (in kilograms) divided by height (in meters) squared.

Further, the girls were evaluated for stage of pubertal maturation by physical exam, and a

medical history intake was performed at baseline. Demographic data were collected from the

parent or guardian of each participant at baseline and again during Year 7. Demographic

information was assessed using a self-report questionnaire and included: race, highest level of

education achieved by the parent or guardian, and household income.

A number of psychosocial instruments were administered over the course of the NGHS.

However, the Body Unhappiness Scale (adapted from Franzoi & Shields, 1984), the Figure

Rating Scale (Stunkard, 1983), two subscales (Drive for Thinness and Ineffectiveness) of the

EDI (Garner, 1991) and a measure of weight loss behavior are the focus of this study.

Body Unhappiness Scale. The Body Unhappiness (BH) scale is a 7-item assessment of

the level of unhappiness with 7 individual body parts and two general body unhappiness

questions. Items on the Body Unhappiness scale include, "How happy or unhappy are you with

your present weight," "How happy or unhappy are you with the way your body looks," and

"How happy or unhappy are you with these parts of your body," including waist, stomach, hips,

legs, and behind. These items are rated on a 4-point Likert type scale ranging from 1 ("Very

Happy") to 4 ("Very Unhappy"). This scale was adapted from the Body Esteem Scale (BES)

developed by Franzoi and Shields (1984). Responses are totaled; higher scores reflect a greater

level of unhappiness or dissatisfaction. A measure of internal consistent (Cronbach's alpha) for

the original subscale of the BES was .87 (n=633) (Franzoi, 1984).

Figure Rating Scale. The Figure Rating (FD) scale is a rating instrument that presents

nine silhouettes gradually progressing in weight. Individuals are asked, "Right now I look like,"

2

Page 3: CCox - Writing Sample

Catherine (Cranston) Cox

and "I would like it best if I now looked like." Individuals select the silhouette that best

represents how they feel about each of the two statements. Figure dissatisfaction is calculated by

subtracting the number corresponding with the "right now" figure from the "like best" figure.

The larger the discrepancy, the greater the body-image dissatisfaction (Stunkard, 1983). Figures

used were patterned after those of Stunkard but redrawn to represent age appropriate growth.

These figures were also rated in focus groups as being racially negative.

Ineffectiveness. The Ineffectiveness (IN) scale of the EDI-2 (Garner, 1991) is a 10-item

scale which assesses, "the feelings of general inadequacy, insecurity, worthlessness, emptiness,

and lack of control over one's life" (Garner, 1991). Items consist of statements rated on a 6-point

Likert scale ranging from "Always" to "Never." Scores are calculated from 0-3. Responses in the

non-symptomatic direction do not contribute to the total subscale score reflecting symptomatic

behavior, thus responses of "Never," "Rarely," and "Sometimes" are scored as 0, and the

responses of "Often, " "Usually," and "Always" are scored as 1, 2, and 3 respectively (unless the

item is reverse scored). High scores on this scale indicate decreased self-esteem owing to

overwhelming feelings of inadequacy. The Ineffectiveness subscale of the EDI has well-

established validity and reliability. For non=ED populations, the coefficient of internal

consistency (Cronbach's alpha) is .88 (n=271) (Garner, 1991).

Drive For Thinness. The Drive For Thinness (DT) scale of the EDI-2(Garner, 1991) is a

7-item scale which assesses, "excessive concern with dieting, preoccupation with weight, and

fear of weight gain" (Garner, 1991). Items consist of statements rated on a 6-point Likert scale

ranging from "Always" to "Never." As with the Ineffectiveness scale, scores from 0-3 are

assigned for each item and the sum of all items totaled. Higher scores reflect an increased desire

to be thin and fear of weight gain. For non-patient populations, alpha is .87 (n=271) (Garner,

1991). Both the DT and IN subscales were reworded to be comprehendible to 9-10 year olds. In

Year 10, the scales were administered with their original wording.

3

Page 4: CCox - Writing Sample

Catherine (Cranston) Cox

Weight Loss Behavior. Weight loss behavior (WLB) was measured with one item,

"During the past 30 days, which of the following did you do to lose weight or to keep from

gaining weight?" A list of 10 items corresponding to various weight loss behaviors followed, and

respondents were asked to check all that applied. These 10 items included dieting, not eating for

one or more days, skipping meals, exercising, taking diet pills, smoking, taking laxatives,

vomiting, using diet drinks, and "other." Positive responses were totaled, and a possible score

ranged from 0 to 10.

Analyses

Descriptive statistics including maximum parental education, household income, mean

BMI, and mean scores on each instrument by race were calculated. Multiple logistic-regression

was used to determine the odds ratio of having an increased DT score (defined as a score of 3 or

greater) adjusting for selected variables, including: race, BMI, maximum parental education,

household income, and score on the BH, FD, and IN scales for both Years 1 and 10. Multiple

logistic-regression was also used to determine the odds ratio of having an extremely high DT

score (defined as those in the highest 15% (score of >10) vs. all others) by those variables found

to be significant in the initial regression analysis. Logistic regression analysis was used to predict

weight loss behavior given DT in Year 10. SPSS v. 11.01 was used for all analyses.

For the purpose of analyses, the education variable was reduced from thirteen to four

groups as follows: high school degree/equivalent or less, some college/post-high school

education, 4 years of college, or more than 4 years of college. Annual household income

included four groups: less than $10,000, $10,000 to $19,999, $20,000 to $39,999, and greater

than or equal to $40,000. These variables were only assessed in Year 1. BMI was divided into

quartiles in both Years 1 and 10. Finally, all remaining continuous variables (DT, BH, FD, and

IN) were categorized for both years, as none were normally distributed. For both DT and IN, a

median split was used to categorize participants into "High" or "Low" risk groups. For DT in

4

Page 5: CCox - Writing Sample

Catherine (Cranston) Cox

Year 1, a score of <2 (51.9%) was categorized as "Low," while any score >3 (48.1%) was

considered "High." These categories remained the same in Year 10. In Year 1, an IN score of <2

(50.2%) was considered "Low," and anything greater than 2 (49.8%) was considered "High." In

year 10, this split was made between the scores of 0 and 1, where 0+"Low" (57%) and

>1="High" (43%), as the mean score for this scale dropped considerably between Years 1 and

10. BH and FD scores were categorized into three groups of roughly 1/3 each, given the

distribution of responses observed. In Year 1, participants' Body Unhappiness score was

categorized into Happy (score <15, 32.5%), Neutral, meaning neither happy nor unhappy (score

16-18, 35%) and Unhappy (score >19, 32.5%). Figure Dissatisfaction was also categorized into

three groups: those girls wanting to be thinner (score <0) those girls wanting to stay the same

(score=0), and those girls wanting to be heavier (score >1). In Year 10, participants' FD scores

were grouped into the same categories as Year 1. The Body Unhappiness groups changed as

follows: Happy (score <18, 34.7%), Neutral (19-22, 33%), and Unhappy (score >23, 32.3%), to

maintain approximately the same percentage distribution in each category.

Initially for both Years 1 and 10, each independent variable was assessed separately to

determine if they were significantly associated with DT. The "Enter" method was used for all

regressions. This method allows all variables to be retained in the model, regardless of

significance. With the exception of race, each regression was run using this method, each time

removing those variables found not to be significant at the α=.05 level, until all variables

remaining in the model were significant. These unadjusted values are found in Table 4 (Year 1)

and Table 6 (Year 10). Race was forced into the model regardless of significance since it is the

basis for much of the between group comparisons. Participants with missing variables were

excluded from all analyses.

Another regression was run for both Years 1 and 10 to determine what factors were

associated with having the highest DT scores. The highest DT scores were taken from the highest

5

Page 6: CCox - Writing Sample

Catherine (Cranston) Cox

15% of all scores (>10). This level of drive for thinness is consistent with the literature

accompanying the EDI highest 15% drive for thinness scores found in a normative population of

college females (Garner, 1991). While in Year 1 the NGHS population was younger than the

normative EDI population, scores for girls with the highest drive for thinness were similar and

remained the same in Year 10.

To test the Theory of Reasoned Action (TRA), DT was regressed against a derived

Weight Loss Behavior (WLB) score in Year 10 from summing the number of dietary behaviors

practiced. The WLB score was calculated by adding all "yes" responses to any of the 10 weight

loss behaviors engaged in over the thirty days prior to the completion of the survey. Possible

activities included dieting, not eating for more or more days, skipping meals, exercising,

vomiting, taking diet pills, using laxatives, using diet drinks, smoking, and "other." DT was

regressed on this score to determine whether higher DT scores meant these girls were more

likely to engage in any WLB. Data were not available for Year 1 since the review group felt it

was inappropriate to ask dieting behaviors of young girls for fear of being suggestive.

A second WLB score was created only for adverse weight loss behaviors, those that may

be an indication of disordered eating behavior. Behaviors constituting adverse behaviors and

included in this calculation were vomiting, taking diet pills, not eating, and using laxatives. DT

was regressed on this score to determine whether higher DT scores meant these girls were more

likely to engage in potentially adverse WLB's. Finally, a third WLB score was created to look at

the remaining dieting behaviors seen as normal, non-clinical weight loss behaviors. DT was

regressed on this score to determine whether higher DT scores were associated with normal

weight loss behaviors.

6