Effects of Scanning (Routine Health Information Exposure) on Cancer Screening and Prevention Behaviors in the General Population

Abstract

Research on health information exposure focuses primarily on deliberate information-seeking behavior and its effects on health. By contrast, this study explores the complementary and perhaps more influential role of health information acquired through exposure to routinely used sources, called scanning. The authors hypothesized that scanning from nonmedical sources, both mediated and interpersonal, affects cancer screening and prevention decisions. The authors used a nationally representative longitudinal survey of 2,489 adults 40 to 70 years of age to analyze the effects of scanning on 3 cancer screening behaviors (mammography, prostate-specific antigen [PSA], and colonoscopy) and 3 prevention behaviors (exercising, eating fruits and vegetables, and dieting to lose weight). After adjustment for baseline behaviors and covariates, scanning at baseline predicted weekly exercise days 1 year later as well as daily fruit and vegetable servings 1 year later for those whose consumption of fruits and vegetables was already higher at baseline. Also, among those reporting timely screening mammogram behavior at baseline, scanning predicted repeat mammography. Scanning was marginally predictive of PSA uptake among those not reporting a PSA at baseline. Although there were strong cross-sectional associations, scanning did not predict dieting or colonoscopy uptake in longitudinal analyses. These analyses provide substantial support for a claim that routine exposure to health content from nonmedical sources affects specific health behaviors.

Acknowledgments

The authors acknowledge the support of the National Cancer Institute–funded Center of Excellence in Cancer Communication, located at the Annenberg School for Communication, University of Pennsylvania (P50-CA095856-05&P20-CA-095856).

The authors are grateful to the following members of the Seeking and Scanning Behavior research group who were important contributors to this research program: Katrina Armstrong, M.D., Angel Bourgoin, Ph.D., Angela DeMichele, M.D., Taressa Fraze, Ph.D., Martin Fishbein, Ph.D., Stacy Gray, M.D., Nehama Lewis, Ph.D., Lourdes Martinez, Ph.D., Rebekah Nagler, Ph.D., Jeff Niederdeppe, Ph.D., Susana Ramirez, Ph.D., Anca Romantan, Ph.D., and Andy Tan, Ph.D.

Notes

¹Colorectal cancer screening can be accomplished ordinarily in one of three ways: colonoscopy every 10 years, sigmoidoscopy every 5 years, or fecal occult blood test every year. Our screening measure assessed colonoscopy only, partly because at the time of this research it was considered the gold standard, and partly because we were constrained by questionnaire space. Thus, although colonoscopy is the predominant form of screening for colorectal cancer in the United States, we may have risked underestimating the degree of colorectal screening, overall. However, since our scanning measures also were about colonoscopy only, they matched the behavioral measures. We then are likely to underestimate any effects of colonoscopy scanning on colorectal screening other than colonoscopy. Thus, our tests of effects, in all probability, are conservative.

²Some guidelines for PSA and mammography screening changed since the Time 1 and Time 2 surveys took place. To accurately represent recommended behavior during that period, analyses herein are consistent with screening guidelines that existed between 2004 and 2006.

³Age and age squared; education; gender; race/ethnicity; marital status; employment status; income and dual income; having kids younger than 18 years of age; type, size, head, and ownership status of household; smoking behavior; overall health status; owning web television; living in a large city; use of newspaper, national news, local news, email, Internet; body mass index and body mass index squared; history of breast, prostate, colon, and/or other cancers among friends or family; personal history of cancer; religious attendance.

⁴Knowledge of cholesterol levels; knowledge of heart disease risk among men; belief in genetic disease inevitability; belief in cancer myths about surgical treatments and cures; health orientation; cancer fatalism. To reduce loss of degrees of freedom, some potential confounders were excluded from particular models if there was not any observed bivariate relation with the specific scanning or behavioral measure. Perceived relative risk for breast/prostate cancer was included in the mammography, colonoscopy, and PSA models only; perceived relative risk for colon cancer was not included in the PSA or exercise models; perceived relative risk for other cancers was not included in the PSA model; locus of control in treatment decisions was included in the fruit/vegetable, dieting, and exercise models only; locus of control in lifestyle and cancer screening decisions were not included in the mammography or PSA models; frequency of doctor consultations was not included in the mammography model.

Note. Data are unweighted.

^a Using Time 1 survey data.

Note. Cases (numbers and percentages) represent nonmissing data and are weighted to the population or subpopulation size. PSA = prostate-specific antigen.

^a The scanning measure is an 8-point index of media and interpersonal sources.

^b Weighted according to the Time 1 sample.

^c Weighted according to the Time 2 sample.

Note. All predictors are measured at Time 1. N refers to the weighted size of the population or subpopulation. PSA = prostate-specific antigen.

^a Effect of Time 1 scanning on Time 1 screening/prevention behavior, conditional on 33–34 applicable confounders.

^b Effect of Time 1 scanning on Time 1 screening/prevention behavior at Time 2, conditional on Time 1 screening/prevention behavior and its interaction with scanning.

^c Effect of scanning at Time 1 on prevention behaviors at Time 2, conditional on Time 1 screening/prevention behavior, its interaction with (if significant in Longitudinal Model 1), and 45–49 applicable confounders.

^d Longitudinal models for Colonoscopy included only those cases that reported not having had a colonoscopy at Time 1 and were eligible for screening. This led to no variation in Time 1 behavior. As a result, interactions with Time 1 behavior could not be examined.

^e If the interaction term was not significant in the Longitudinal 1 model, the term was removed from the full model adjusted for confounders.

*p ≤ .05. **p ≤ .01. ***p ≤ .001.

⁵A reduced model including the interaction and only those confounders that added some predictive power to the equation (utilizing backwards stepwise regression and eliminating variables that did not have a p < .10) exhibited a significant lagged simple main effect of scanning [OR = 1.18, SE = .09, p = .036].