Appropriate considerations of “rural” in National Cancer Data Base analyses

Abstract

I would like to commend Virostko and his colleagues for addressing the important topic of early-onset colorectal cancer trends in their recent article entitled “Recent Trends in the Age at Diagnosis of Colorectal Cancer in the US National Cancer Data Base, 2004-2015,” which was published in the November 1st edition of Cancer. The authors examined data from the National Cancer Data Base (NCDB), a high-quality clinical surveillance database of cancers diagnosed at Commission on Cancer (CoC)–accredited hospitals that is inclusive of nearly 70% of all cancer cases diagnosed in the United States. They found that the proportion of colorectal cancers diagnosed in individuals under the age of 50 years is increasing overall, within certain subgroups such as non-Hispanic white men and women and Hispanic women, and among urban populations. However, they identified no increases among rural populations under the age of 50 years old. The authors are appropriately forthcoming about the limitations of using trends in proportions rather than incidence rates, which can be calculated only from population-based data. Furthermore, a corresponding editorial by Doubeni also rightly noted the limitations of NCDB data with respect to the representation of different age groups and other population groups as well as the exclusion of non–CoCaccredited and military hospitals. Beyond these nuances and limitations noted by both Virostko et al and Doubeni, I would like to highlight 2 additional important points that should be considered when one is using NCDB data to examine rural-urban differences. First, a more appropriate definition of rural should be used. Second, greater attention should be paid to the limitations of using NCDB data to study cancer in rural areas. In Virostko et al’s article, rural is defined in a figure footnote as referring to those living in counties with fewer than 2500 people. The NCDB public use file data dictionary notes that the rural-urban variable available in its participant use file is the Rural-Urban Continuum Code (RUCC), which is a county-level code that takes into consideration population size and proximity to a metropolitan area. The American College of Surgeons data dictionary for the NCDB data indicates that counties with fewer than 2500 people are rural counties, and this is the definition that the authors have applied. This definition corresponds to RUCCs 8 and 9, which are inclusive of just 1.5% of the US population. However, this is not indicative of how rural is commonly defined in most studies. Most often, RUCCs 4 to 9 (ie, nonmetropolitan counties) are considered rural, and they are representative of 15% of the US population. Indeed, defining rural is a source of ongoing debate. However, using a categorization that is inclusive of such a small proportion of the population greatly underrepresents the generally agreedupon proportions of the US population and subsequently may cause incorrect conclusions to be drawn regarding rural trends in early-onset colorectal cancers. An additional point of necessary nuance is that CoC-accredited hospitals are far less likely to be in rural areas. Thus, even if a less restrictive definition of rural is employed, rural patients may not be captured by NCDB data. To be sure, rural patients may seek care at a CoCaccredited hospital in either a rural setting or an urban setting, but it is important to recognize this important limitation when one is making geographically based interpretations from clinical surveillance data.