DOI

10.17077/etd.gfydqwgo

Document Type

Dissertation

Date of Degree

Summer 2018

Degree Name

PhD (Doctor of Philosophy)

Degree In

Geography

First Advisor

Margaret Carrel

Second Advisor

Gerard Rushton

First Committee Member

Caglar Koylu

Second Committee Member

Barcey T. Levy

Third Committee Member

Charles Lynch

Abstract

This study of the spatiotemporal patterns of colorectal cancer (CRC) in Iowa introduced the importance of examining the geographic patterns of four epidemiological measures (incidence, late-stage incidence, mortality, and survival) as inter-related phases in the natural history of the disease rather than as independent measures. To conduct such an analysis required the development of a framework for conducting spatiotemporal correlation analysis involving two or more measures across two or more periods.

This framework is based on geographic units called spatially adaptive filter areas.which effectively address the small number problem. This common spatial epidemiology problem occurs when the population in a unit of geography is too small to calculate a reliable disease rate. The spatially adaptive filter areas are created by aggregating smaller geographic units which, by themselves, do not have sufficiently large populations to calculate statistically reliable disease rates.

Conducting spatiotemporal analysis magnifies the small number problem because stratifying disease data by time further reduces the sample sizes in each period, thus increasing the potential for unreliable disease rates. This spatiotemporal framework satisfies two conditions: 1) the rates of each measure in all small areas in the study region meet a minimum level of statistical reliability in all periods, and 2) the disease outcomes measured for one period relies on the same geographic units as the rates calculated for all other periods and measures.

We applied the spatiotemporal framework to colorectal cancer data collected in the state of Iowa between 1999 and 2010. Using rates calculated in spatial filter areas, we found that the proportion of cases diagnosed at a late-stage among the population at risk for CRC is more correlated with CRC mortality than when late-stage is measured as the proportion of late-stage cases among the CRC cases diagnosed at any stage. We also found that, when measured in the context of the statewide change, an observed decline in the rate of CRC mortality in a small area does not necessarily mean that its level of mortality is improving. We also found that the correlation between rates of CRC mortality and the survival time among patients diagnosed with CRC varied across Iowa.

The results described in this dissertation could potentially reduce the burden of colorectal cancer and improve the health of communities by providing public health professionals and cancer control specialists with evidence to enhance their decision-making processes.

Keywords

cancer, disease, geographic information system, mapping, spatial, spatiotemporal

Pages

xiv, 106 pages

Bibliography

Includes bibliographical references (pages 96-106).

Copyright

Copyright © 2018 Kevin Andrew Matthews

Included in

Geography Commons

Share

COinS