Exploring the Relationship between a Cancer Diagnosis and Household Food Insecurity: The Concern of Selection Bias

Background: Due to advancements in prevention, detection, and treatment, cancer is now considered a chronic condition, and approximately 5% of the US population has been diagnosed with cancer during their lifetime.  Cancer impacts not just an individual’s physical, but also his/her financial well-being.  Research suggests that cancer patients may face a higher financial burden, are at increased risk for bankruptcy, and regularly experience decreased income.  Due to financial struggles and other physical limitations, cancer patients may be forced to choose between competing demands such as food and medical treatment.  This suggests cancer patients may be at increased risk for experiencing food insecurity.  However, few studies have empirically tested this hypothesis.  This study aims to investigate the association between a cancer diagnosis and household food insecurity. 

Methods: We use data of a nationally representative sample (N=9,029) from the 2015 Panel Study of Income Dynamics (PSID).  Household food insecurity is indicated by an 18-item standardized food security scale created by the USDA.  The sum of these items (0-18) is considered a continuous dependent variable.  According to the USDA, we also create a dichotomous indicator (Food Secure=0 and Food Insecure=1) based on the continuous scale.  The major independent variable is household heads’ cancer diagnosis (No=0 and Yes=1).  We use OLS and logit model to predict continuous and dichotomous measures of food insecurity, respectively.  All analyses are addressed for the survey features of the PSID and extensive demographic and socioeconomic variables are included (eg. age, gender, race, education, employment, marital status, number of children, household income, nutrition assistance program participation, and medical expenditure). 

Results: Nearly 6% of the household heads reported being diagnosed with cancer, and about 14% experienced food insecurity in the previous year.  Bivariate analyses suggested that household heads with cancers actually had better socioeconomic status than those without, in terms of education, household income, health insurance, and nutrition assistance program participation, while these differences between the two groups were not statistically significant.  Household heads with cancer had a slightly lower likelihood to be food insecure than those without cancer.  Multivariate analyses also revealed the household heads’ cancer diagnosis was not a statistically significant predictor of household food insecurity. 

Conclusion and Implications: While it is hypothesized that patients with cancer may suffer from a higher risk of household food insecurity due to their financial stress and physical limitations, analyses on the PSID do not support this hypothesis.  Heads of households diagnosed with cancer in this study were of higher socioeconomic status in comparison to the sample not diagnosed with cancer, which might imply that the current design and analyses do not address the potential for selection bias sufficiently. For example, household heads who report a cancer diagnosis might be a group of participants who had better socioeconomic status and greater access to healthcare, leading to earlier diagnosis and treatment.  This may result in lower food security scores among cancer patients than the comparison sample.  It raises an important consideration for future research to carefully identify the comparison group of cancer patients.