Getting rurality right: Investigating cancer across the rural-urban interface.

Abstract

e19063 Background: Trends in oncology examine the influence of rural characteristics on cancer patient (pt) treatment and outcomes. Current definitions of rurality are broad and varied, with multiple standardized definitions. Few analyses exist to compare definitions and quality in assessing oncologic outcomes. We aim to determine which index is most suitable to define rurality in cancer research. Methods: 1,567 pancreatic cancer pts from the University of Wisconsin Cancer Registry, representing 84 Midwestern counties and 452 zip codes, were assigned rurality codes based on three indices spanning 1983-2013: Rural-Urban Continuum Code (RUCC), Rural-Urban Commuting Areas (RUCA), and Index of Relative Rurality (IRR). RUCC and IRR were assigned to pts at the county level and RUCA at the zip level. Pt rurality was compared across the three indices and over time via the median and interquartile range and inspected visually with violin plots. We compared indices with Spearman’s Rank Order Correlation (SROC). Results: RUCC 2003, RUCA 2004 (zip), and IRR 2000 were concordant in terms of metropolitan, micropolitan, and rural designations for 66.9% (1,049) of pancreatic cancer registry pts. The rural designation for almost one-third (489, 31.2%) of pts was discordant by one or two levels across the three indices (i.e. classified as metropolitan in one index and micropolitan or rural in another). SROC was 0.73 between RUCC 2003 and RUCA 2004 (zip) indices, 0.82 between IRR 2000 and RUCA 2004 (zip), and 0.85 between RUCC 2003 and IRR 2000. Across the 84 counties of registry pts’ residence, the median and interquartile range of RUCC decreased from 6 (3-7) in 1983 to 4 (2.25-6) in 2013 and of IRR decreased from 0.49 (0.44-0.53) in 2000 to 0.49 (0.43-0.53) in 2010. Across the 452 zip codes, RUCA decreased from 4 (1-10) in 1998 to 3 (1-8) in 2004. Pts’ median RUCC decreased from 3 (2-6) in 1983 to 3 (2-4) in 2013, median RUCA (zip) decreased from 2 (1-7) in 1990 to 1 (1-6) in 2000, and IRR remained constant at 0.42 (0.38-0.49) in 2000 and 2010. Conclusions: RUCC is preferable for state-level cancer studies incorporating rurality as it best distributes pts across the rural-urban interface compared to RUCA (skews urban) and IRR (skews central). County boundaries (RUCC) are consistent over time versus zip (RUCA) and census tract (RUCA). Our findings suggest that while the extremes of rural and urban are well-defined, rurality as a continuum is inconsistently measured. Researchers should continue to incorporate other measures of vulnerability to achieve health equity.