The methods of this study are reported in detail elsewhere [36
]. Briefly, these data are drawn from the baseline assessment of the Multiplex Initiative, a multi-center prospective observational study. The sampling frame for the Multiplex Initiative was drawn from a pool of 350,000 members of a Midwestern health management organization. Sampling strategies have been described previously [37
]. Briefly, the sample included members who, as identified in the Master Patient Index, were: enrolled two years or more, aged 25 to 40, and self-identified as either black or white race. We selected our age range to best capture young, healthy adults who would not currently be included in population-based screening for the diseases in question. Diagnoses codes from claims data were used to exclude members who had been diagnosed with the eight health conditions on the Multiplex Genetic Test (i.e., diabetes, osteoporosis, heart disease, high cholesterol, hypertension, lung, colon or skin cancers). The eight selected health conditions are adult onset and “preventable,” meaning that there are widely accepted evidence-based prevention recommendations for these conditions, and can impact men and women. A random sample of the members meeting these criteria was drawn, oversampling for: male gender, black race, and lower educational status based on census track information associated with their current address (“lower” being ≤ high school). Recruitment occurred from February 2007 to May 2008.
Members in this random sample were sent an advance letter explaining that they would be contacted by telephone to complete a survey about their health-related attitudes and beliefs about factors that contribute to health outcomes unless they called a toll free number to decline participation. 22 called to decline our initial phone contact. Telephone contact for baseline screening was attempted for all individuals who did not call to decline. Importantly, the intent of the telephone survey was described as to assess “what people believe to be the causes of common health conditions.” As reported previously [36
], baseline surveys were attempted with 6,348 members of Henry Ford Health System. 1,292 refused the survey, 2,614 were unreachable despite repeated attempts, and 326 were ineligible. 2,116 completed the baseline. 157 of these were ineligible due to presence of a health condition, leaving 1,959 for analysis. (See Hensley-Alford et al. [37
] for a more detailed discussion of the recruitment approach).
Family history was assessed with the item, To your knowledge, do any of the following diseases run in your family? and then queried for each condition with a yes/no response. Participants responded to family histories for cancers, heart disease, osteoporosis, hypertension, and high cholesterol.
Behavioral risk factors are described below. We assessed seven risk factors in a way that would determine present behavior and therefore, current risks that could be changed. For each of the behaviors, a variable was created to indicate whether the level reported constituted a risk factor (0 – no, 1 – yes) for one of the health conditions. The risk factors, seen in , were then summed to create a behavioral risk factor score for each disease.
Behavioral risk factors for eight health conditions.
was measured with an item used in previous clinical trials of physical activity [38
], During the last 12 months, on a scale from 1 to 7 where 1 is Never and 7 is Daily, how many days per week did you do each of the following for at least 15 minutes at a time? (a) Walking for exercise; (b) Hiking; (c) Bicycling or exercycle; (d) Aerobic and calisthenics; (e) Swimming; (f) Water aerobics; (g) Weight training or strengthening; (h) Other exercise. Participants who reported being active fewer than 5 times per week were coded as having the behavioral risk factor [40
was computed using a validated 7-item nutrition screener designed for primary care settings [41
] (range = 0-14). Items include How many times in a typical week do you eat fast food meals or snacks? How many servings of fruit or vegetables do you usually eat per day? How many servings of regular sodas or sweet tea do you usually drink each day? How many times in a typical week do you eat beans (like pinto or black beans), chicken or fish? How many times in a typical week do you eat regular snack chips or crackers (not low-fat)? and How many times in a typical week do you eat desserts and other sweets?, with response options of 1, 2, or 3 or more times/week. In addition, respondents were asked How much margarine, butter or meat fat do you usually use to season vegetables or put on potatoes, bread, or corn? (Very little, some, a lot). The cutoff was the mean score for high risk patients (6.43) in the validation study.
Current smoking status
was based on self-report of ever having smoked and having smoked in the last 7 days [42
]. Current smokers were at-risk.
was assessed using three items. In the past 30 days, how often did you drink any alcoholic beverages (every day, almost every day, 3-4 time/week; once per week, 2-3 times/month, 1/month, never); In the past 30 days, on the days when you drank alcoholic beverages, how many drinks did you have each day, on the average? We also asked In the past 30 days, how many times did you have 5 or more drinks on any one occasion? Drinking more than an average of one (women) or two (men) alcoholic beverages a day or consuming more than 5 drinks in any one sitting was considered to indicate a risk factor [43
Sun exposure was measured with one item: How many times in the last 12 months did you get a sunburn that blistered or peeled from the sun? Having one or more sunburns was considered to indicate a risk factor.
was assessed using one item: On average, how many days a week do you take multivitamins or folic acid? Less frequently than four times per week was considered to indicate a risk factor [45
Body mass index (BMI; BMI=kg/m2)
was calculated from self-reported height and weight. BMI ≥ 30 (obese) was considered to indicate a risk factor for diabetes, high cholesterol, hypertension, and colon cancer [46
]. BMI< 18.5 (underweight) was considered a risk factor for osteoporosis [47
Educational status was based on a single item, What is the highest grade or year in school you completed? This was trichotomized as high school or less, some college/vocational school, and college degree or more.
Causal attributions related to behavioral risk factors were assessed using the item On a scale from 1 to 7 where 1 is Not at all and 7 is Completely, how much do you think health habits such as diet, exercise, and smoking determine whether or not a person will develop each of the following conditions? Attributions related to genetic make-up were assessed using the item On a scale from 1 to 7 where 1 is Not at all and 7 is Completely, how much do you think genetic make-up, that is characteristics that are passed from one generation to the next, determines whether or not a person will develop the following conditions?
These items were combined to compute a new variable ranging from 0 to 1 indicating a participant's perceptions of the relative contribution of genetic to behavior as causes for each of the health conditions. At the extreme value of 0, the person attributes the health condition entirely to behavior. At the extreme value of 1, the person attributes it entirely to genes. At the mid-point (0.50), the person attributes the condition equally to genes and behavior.
Information Preference Outcomes, based on genetic and behavioral risk factors, were assessed using two items. Preferences for behaviorally-based information was assessed using the item On a scale from 1 to 7 where 1 is Not at all important and 7 is Very important, how important is it to you to learn more about how your health habits affect your chance of getting a certain health condition? Preferences for information related to genetic make-up were assessed using the item On a scale from 1 to 7 where 1 is Not at all important and 7 is Very important, how important is it for you to learn more about how your genes, that is the characteristics that are passed from one generation to the next, affect your chance of getting certain health conditions? Due to the skewed distribution of these items, each was dichotomized (1-5 = Less important, 6-7 = More important).
Covariates, including age, gender, marital status (married/partnered v. other), and race (White, African American, Other) were assessed with standard items.
We generated frequency distributions and descriptive statistics to determine the participants’ sociodemographics and to determine variable frequencies and distributions. We conducted bivariate analyses to determine the relationships between covariates and outcomes. Significant covariates were entered into our multivariate models.
Mediation analyses would have been the preferred method for understanding whether behavioral risk and family history impacted information outcomes through the causal attributions held by those at risk. We did not pursue specific mediation analyses given our cross-sectional data [48
]. However, we did follow the steps of mediation analyses, without formally testing statistical significance of mediation. This approach enabled an orderly approach to assessing the interrelationships among our variables prior to entering predictors into our final multivariate models. These steps included testing the associations between (1) the predictors (behavioral risk and family history) and the outcome (information preferences), (2) the predictors and the mediator (attributions), and (3) the mediator and the outcome, as well as (4) the final model, which incorporates the predictors and the mediator to predict the outcome [49
Hypothesis 1, that participants would place greater importance on learning about how genes affect risk for health conditions than about how health habits affect risk for health conditions, was tested by using McNemar's exact test to assess the difference between these two categorical variables.
Hypothesis 2, that reporting more behavioral risk factors and a family history of the condition would be positively associated with the tendency to hold causal attributions that favor genetic over behavioral explanations as the cause of health conditions, was tested using bivariate correlations to examine the associations among behavioral risk factors, family history, and causal attributions (See ).
Theoretical Model and Hypothesized Effects
We ran logistic regressions using PROC LOGISTIC from SAS 9.2 (SAS Institute, Inc., Cary, NC) to test hypotheses 3 and 4, that reporting more behavioral risk factors, a positive family history, and holding causal beliefs favoring genetic explanations would be associated with expressing less interest in seeking information about how health habits influence risk but more interest in seeking information about how genetics influence risk. Significance levels of p < .05 were used for all analyses.