Comparing Two Populations using Standardized Rates Section
Rate ratios are useful for comparing the rate in one population to the rate in another population. Confidence intervals can be calculated for rate ratios,thereby giving an estimate of the uncertainty in the measurement of the rate. The following example uses surveillance data to calculate adjusted rates and ratios.
Example 4-2: Cancer Incidence in Appalachia Section
Is cancer incidence in Appalachia, a large rural area with increased poverty and reduced access to health as compared to the rest of the United States, higher than the remainder of the country? The Appalachia Cancer Network (ACN) is a network of academic and community organizations that conducts surveillance, intervention, and dissemination research related to cancer in Appalachia. In Figure 1 below, Appalachian counties are dark green. The counties that are striped have been designated as rural counties by a different designation.
In the United States, cancer registries collect information about cases of cancer within a region. This study compares incidence rates calculated from WV, PA, and KY cancer registry data during 1994-1998 to national rates calculated from SEER data.
239,670 cancer cases were extracted from the state data. The cases were further categorized as to type and stage. (Table 1 ). Two groupings were used: rural Appalachian counties only and all Appalachian counties.
Table 1: Categorization of Cancer Cases:
Incidence rates were calculated for the entire Appalachian region and for rural and nonrural counties separately. Rates were directly adjusted to the US 2000 population. These rates are compared to the U.S. as a whole, using rates calculated from 1994-1998 SEER data, also adjusted to the US 2000 population. Consider Table 2 below.
Compare the SEER sample with the Appalachian rates. For example, using the SEER U.S. data, the age-adjusted rate for all new cancer cases was 471.2. The accompanying 95% confidence interval is fairly precise, 470.1 to 472.2. The corresponding age-adjusted rate for all of the counties in Appalachia was 471.7. The rate ratio was 1.0, meaning that there was basically no difference between overall occurrence of cancer in Appalachia as compared to SEER.
What about the rural counties? For all-site cancer, the rate is 468.3 while for non-rural counties, the rate is 473.1. Is this what you expected? The rate ratio between rural areas and SEER is 0.99. with a 95% confidence interval that includes 1. Is there truly no difference in cancer incidence between rural counties and the rest of the country? Or are there differences, not in the overall incidence rate for all sites of cancer, but in the distribution by stage or type of cancer? Comparing the age-adjusted incidence rates, using a confidence interval for a rate ratio provides insight. In this situation, the results are unexpected, raising additional questions. For example, what is the impact of ‘unknown stage of cancer’? What are the reasons for unknown stage? Patient request? Lack of access to care? Why is stage missing in the registry?
Come up with an answer to this question by yourself and then click on the icon to the left to reveal the solution.
What do you observe? Did any types or stages of cancer occur more frequently in Appalachia than the rest of the U.S.?
Lung cancer, in all stages, is much more frequent in rural Appalachia as compared to SEER. Unknown stage and regional cancers seem to occur more frequently in Appalachia as well. Competing risks?
The same rate ratios are presented using bar charts in the two figures below. The different format may help in focusing on an individual disease, across stages, or by cancer site. What stands out from these graphs?
Ratios of Incidence Rates - Appalachia as a Whole vs. SEER by Stage at Diagnosis and Cancer Site
Ratios of Incidence Rates - Rural Appalachia vs. SEER by Stage at Diagnosis and Cancer Site
(From: Lengerich et al, 2005. Journal of Rural Health.)
Do you see the greater rates of unstaged cancer in the rural areas compared to SEER? The rate ratio is above 1.20 for all types of cancer in rural Appalachia while the rate ratio for distant cancer is similar between the two populations.
The conclusions of this study were the following:
- Overall cancer incidence is not higher in Appalachia compared to SEER (U.S.).
- However, specific cancer incidence compared to SEER varies:
- Unstaged at diagnosis vs. SEER