To estimate the direct health-related costs attributable to food insecurity in 2014, we reviewed empirical research literature published in peer-reviewed journals from approximately 2005 to 2015, searching for quantitative findings of associations between food insecurity and health outcomes. We specifically searched for quantitative findings that involved either odds ratios (most often), likelihood ratios, or relative risk ratios expressing the differences in likelihood of a person living in a food insecure household having a disease or disease condition compared to a person living in a food-secure household (food security status is the exposure variable).
Those probability ratios were then translated into population attributable fractions (PAFs) expressing the proportion of the total prevalence of the disease in the population attributable to food insecurity (i.e., the excess fraction attributable to food insecurity). As noted above, this process requires the assumption that food insecurity is causally related to the disease conditions.
In case-control studies, if adjusted odds ratios (ORs) are available, they can be transformed into relative risk ratios using formula 1 below: 1. RR = OR/[(1-Po)+(Po*OR)], where RR is the relative risk ratio, OR is the odds ratio, and Po is the proportion of the unexposed (food secure) who develop the outcome, or become cases. This adjustment is desirable since, though the OR is an acceptable estimate of the Relative Risk ratio (RR) in case-control studies, and approaches RR in the situation of rare diseases in which very few of the unexposed develop the disease, the higher the prevalence of the disease in the unexposed population (e.g., the food-secure population), the greater the deviation of the RR from the OR.With the relative risk ratios thus calculated (or if they are available), they can be used to calculate estimates of the excess population attributable fractions (PAF) of the diseases arising due to exposure to the predictor, food insecurity, using formula 2 below: 1. PAF = Pe (RR - 1) / [Pe (RR - 1) + 1] * 100%, where PAF is the excess population attributable fraction of disease in the population considered to result from the presence of the exposure variable or condition (i.e., food insecurity), RR is the relative risk ratio calculated as above, and Pe is the proportion of controls (those who do not have the outcome or disease) who were exposed (live in a food-insecure household).
A complete table of all the conditions for which we found new studies providing the information needed to calculate attributable fractions can be found in Appendix Exhibit A1 in the PDF version of this report. For most of the health conditions, the attributable fraction (AF) is relatively small, 10 percent or less. For a few conditions we found research results leading to more than one AF for a condition. In those cases, we either used the average of the AFs, or used the one which was more reliable for the specific age group and condition under consideration. And for a few conditions, we were either unable to find data on the prevalence and number of people in the relevant sub-population with the condition, or data on the cost of treating cases of the condition. In those few instances, we were unable to estimate the disease burden or the costs. This was particularly true when the condition was failure to receive recommended or prescribed treatment, or treatment foregone due to inability to pay as a result of food insecurity.
For a couple of conditions (e.g., PEDS concerns; parents report of developmental concerns about their child), we had to add an additional link to the chain of logic such as obtaining positive predictive value of the indicator (PEDS concerns) and the outcome (special education). With a few conditions for which we could not find needed prevalence data, we relied on data from the U.S. Census Bureau on relationships between reported health status and health services utilization.
Using the information in Exhibit 1A, together with data from the Agency for Healthcare Research and Quality's Medical Expenditure Panel Survey (MEPS, or other national survey data) on the number of cases of each disease condition in the population in 2014 (when available), we estimated the fraction (proportion) of cases of each health condition attributable to food insecurity. Combining the results of these calculations with data on annual expenditures for treatment of individuals with the condition (from MEPS or other national health surveys), we estimated the total annual direct costs of treatment for all individuals with the condition.
Data on numbers of hospitalizations, and average costs of hospital stays were obtained from the Agency for Healthcare Research & Quality's Healthcare Cost & Utilization Project public access data obtained via the HCUPnet online query system (http://hcupnet.ahrq. gov/). Data were obtained from both the HCUP National Inpatient Database and the HCUP Kids' Inpatient Database. Several price index series were used to adjust the price of various healthcare services. These price indices were taken from the Bureau of Labor Statistics' online databases (http://www.bls.gov/cpi/). Resulting estimated costs for each condition are presented in Appendix Exhibit 2 of the PDF version of this report.
The Brandeis researchers estimated the cost of the private food assistance system at $17.8 billion in 2010 ($19.52 billion in 2014 dollars), and we calculated the total cost of the public food assistance system to be $103.55 billion in 2014. However discussions with healthcare colleagues and others led us to the position that the costs of these two complementary food assistance systems are more accurately viewed as the costs of prevention of food insecurity, not as a cost of food insecurity itself. The costs of these two food assistance systems are the costs of the vaccine that prevents food insecurity and hunger from occurring in the nation's households, families and children. Thus the costs of these two systems are not included as costs attributable to food insecurity.
For more information on methods, download the