IMPACT OF AREA-POVERTY RATE ON LATE-STAGE COLORECTAL CANCER INCIDENCE IN INDIANA, 2010-2014 NAACCR JUNE 22, 2017 Aaron Cocke, Amanda K. Raftery, Timothy McFarlane
SECTION 1 OVERVIEW OF PROJECT
Purposes of Project 1. Part of an academic assignment for the Fairbanks School of Public Health 2. Gain experience working with real epidemiological data 3. Assist with the Indiana State Department of Health s (ISDH) ongoing projects and goals
Summary of Project Objectives 1. Identify measures of socioeconomic status that may be significantly related to late-stage colorectal cancer incidence 2. Contribute to current understanding between colorectal cancer incidence and socioeconomic status in Indiana 3. Provide direction for future studies 4. Potentially identify other significant relationships between incidence and included variables
Colorectal Cancer Incidence in Indiana 1. Colorectal cancer incidence rate in Indiana is higher than the national rate. 2009-2013 age-adjusted rate for IN: 44.0 per 100,000 National estimate: 40.6 per 100,000 2. Five-year survival rate diagnosed at local stage is 90%, but decreases to 70% and 13% when diagnosed at regional and distant stages 3. The ISDH is participating in efforts to increase colon cancer screening rate to 80% by 2018
Socioeconomic Status and Colorectal Cancer Colorectal cancer incidence was formerly associated with higher socioeconomic status, but has changed to be associated with lower status Unhealthy lifestyle habits more commonly associated with low socioeconomic status (SES) are estimated to account for 70% of cases Few studies attempt to precisely analyze the impact of SES on colorectal cancer incidence Ongoing surveillance and studies are needed to understand the dynamic relationship between incidence and SES
SECTION 2 MATERIALS AND METHODS
Publication: The Relationship Between Area Poverty Rate and Site-Specific Cancer Incidence in the United States 2014 Boscoe et al. Sought to demonstrate the importance of including measures of SES in cancer incidence surveillance in the United States Analyzed census tract area-poverty data from 2005-2009 data across16 states by assigning individuals to one of four poverty rate categories Calculated site-specific and sex-specific rates of incidence by poverty rate Found that numerous cancers were strongly associated with low SES
Differences between Boscoe study and this project Scale of study reduced from 16 states to solely Indiana and multiple types of cancer to solely colorectal cancer Included additional measures of SES for analysis Colorectal cancer was selected for study to assist with ISDH efforts to improve colon cancer screening
Commission on Cancer Late Stage Diagnosis 1. An outcome variable was created using SEER Summary Stage 2000 in the North American Association of Central Cancer Registries data dictionary 2. Codes were collapsed to create a dichotomous categorical variable Stage of Diagnosis during years 2010-2014 Early: In situ / localized Late: Regional / distant
Hypotheses 1. Later stages of colorectal cancer are significantly associated with areas of poverty in Indiana 2. Later stages of colorectal cancer are significantly associated with areas lacking insurance coverage in Indiana 3. Later stages of colorectal cancer are significantly associated in areas with larger drive times to Commission on Cancer-accredited hospitals
Drive Time Map ArcGIS allowed us to: Plot colorectal cancer cases by address Assign cases to subcategories for regression analysis 15 minutes 16 to 30 minutes Layer data obtained from ISDH, US Census Bureau, and Indiana Department of Transportation
Variables Selected for Analysis Independent Variables 1. Area-poverty rate by census tract (source: American Community Survey) <10%, 10-20%, 20-30%, >30% 2. Uninsured rate by census tract (source: American Community Survey) <8.0%, 8.0-10.0%, 10.0-12.0%, 12.0%-14.0%, >14.0% 3. Drive time to nearest Commission on Cancer-accredited hospital (created internally) <15 minutes, 15-30 minutes, >30 minutes 4. Age group (source: Indiana State Cancer Registry) 50-64, 65+ years 5. Sex (source: Indiana State Cancer Registry) Female, Male 6. Ethnicity (source: Indiana State Cancer Registry) Non-Spanish or Non-Hispanic, Spanish or Hispanic 7. Race (source: Indiana State Cancer Registry) -Black, White, Other
Testing Methods Utilized SAS 9.4 to create and run the following: Descriptive statistics of data Cochran-Mantel-Haenszel Chi-square test of association Backward multiple regression to determine significant predictors of outcome variable All tests used an alpha of 0.05 and 95% confidence intervals
SECTION 3 RESULTS AND DISCUSSION
Chi-Square Test Results Chi-Square test shows significant associations between all variables and outcome variable Demographic variables are highly significant and may warrant further analysis Variable Chi-Square Value P-Value Age Group 66.0484 <.001 Sex 122.8181 <.001 Race 45.3872 <.001 Ethnicity 213.7739 <.001 Drive Time 6.2081 0.0449 Poverty Group 14.6065 0.0022 Uninsured Group 12.9710 0.0114
Logistic Regression Analysis of Maximum Effects Significant relationships are seen between incidence and poverty group Drive time and uninsured group are not significant and are removed from model All demographic variables are significantly related Effect Chi-Square P-Value Age Group 64.8745 <.0001 Sex 38.5242 <.0001 Race 7.9870 0.0184 Ethnicity 113.3730 <.0001 Drive Time* 5.9512 0.0510 Poverty Group 11.4769 0.0094 Uninsured Group* 4.9766 0.2897 *Removed from final model due to non-significance.
Age 65+ 50-64 Independent Variable Sex Male Female Ethnicity Spanish, Hispanic Non-Spanish, Hispanic Unknown Odds Ratio 1.0 1.12 1.0 1.09 1.0 0.90 0.29 95% CI - 1.09 1.16-1.06 1.12-0.80 1.03 0.23 0.37 Logistic Regression Odds Ratios and 95% Confidence Limits Greater association in 50-64 than 65+ Greater association in women than men Several poverty category comparisons show greater association in areas of high poverty compared to areas of low poverty Race White Black Other 1.0 1.02 0.80-0.97 1.08 0.69 0.94 Poverty Group <10% 10-20% 20-30% >30% 1.0 1.01 1.07 1.05-0.98 1.04 1.02 1.12 1.01 1.10
Potential Improvements for the Future 1. Include more measures of socioeconomic status for study 2. Potentially use same methodology to examine how the selected measures are associated with other cancers 3. Re-examine category definitions for uninsured and poverty groups
Conclusions 1. Area-poverty rate is significantly associated with late-stage colorectal cancer incidence in Indiana 2. Areas of higher poverty are more closely associated with late-stage incidence than areas of lower poverty 3. Future studies may wish to examine the impact of these socioeconomic measures on other types of cancer and/or include more measures for study and look at individual-level measures of SES 4. Future screening interventions may be better suited for areas with high area-poverty rates than areas with low insurance coverage or are far from hospitals
Acknowledgements ISDH: Amanda Raftery Project Preceptor Keylee Wright and Laura Ruppert Proofreading and additional feedback Indiana University-Purdue University, Indianapolis: Aaron Cocke MPH candidate, Fairbanks School of Public Health Sarah Johnson Project Coordinator Timothy McFarlane and Thomas Duszynski Academic Advisors