ISSN: 2639-2038
Using BRFSS Data to Estimate County-Level Colorectal Cancer Screening Prevalence in Missouri
Authors:
Jiang Du, Dongchu Sun*, Chester Lee Schmaltz and Jeannette Jackson Thompson
Colorectal cancer (CRC) is the third most commonly diagnosed cancer and the third leading cause of cancer death in both men and women in the US. Colorectal cancer screening (CRCS) serves an important role in the early detection of colorectal cancer and reduces the mortality rate. It is recommended that people aged 50 and older should take CRCS regularly. However, not all people follow the guideline. The state-level CRCS prevalence can be estimated from the Behavioral Risk Factor Surveillance System (BRFSS). Efforts to advocate CRCS are often conducted locally, often at the level of county or county equivalent. Knowing county-level CRCS prevalence can be important to making relevant policies. However, BRFSS does not provide county-level CRCS prevalence estimates. We examined the possibilities of using BRFSS data for county-level estimates with small area estimation (SAE) techniques. Demographic information from both BRFSS and U.S. Census population file were used in our models. In addition, county attributes related to education levels and house incomes were used to improve the estimates. A random spatial effect was also added to capture other county attributes not included in the model. We took the 2012 Missouri BRFSS (MO-BRFSS) data as an example to get county-level CRCS prevalence estimates. To evaluate the results, estimates from 2011 Missouri County Level Study (MO-CLS), which is a BRFSS-like survey but collected hundreds of responses for each county in Missouri, was used as “true†values. The evaluation results indicated the inclusion of county attributes improved the estimates significantly, but not the random spatial effect. The estimates from MO-BRFSS showed similar patterns as those from MO-CLS but less accurate.
Keywords:
Small area estimation; Bayesian hierarchical models; Survey estimates; BRFSS
