Identifying behavioral differences between people with and without previous cancer diagnosis

Abstract

We undertake a study to determine and assess the effects of the statistically significant predictors of the behaviors and notions that are associated with a cancer diagnosis using the 2014 Health Information National Trends Survey (HINTS) data. We implemented a new and extensive logistic regression modeling using stepwise variable selection and jackknife parameter estimation that identified the best explanatory model. Our results show that age, average time spent watching TV or playing games, usage of sunscreen, fruit intake intent, and the opinion-based variables for behaviors affecting high blood pressure, as well as the participant preference of not knowing the chance of getting cancer are the optimal set of covariates impacting the chance of getting cancer. Moreover, using more sunscreen, and a higher age was associated with increases in the chances of getting cancer. Interestingly, many usually important background covariates such as race, income, gender, geographical location, and others were not significant predictors of the outcome variable of interest. The conclusions of our analysis reveal new insights in the complexity of the behaviors and “attitudes” associated with a higher chance of a cancer diagnosis and will undoubtedly have important implications on the design and success of future healthcare messages and campaigns.

Keywords

• automatic variable selection
• logistic regression
• nonconcurrences of habits