The Missouri Cancer Registry needs to measure female breast cancer survival proportions to be able to evaluate the impact of Missouri’s breast cancer control program and the burden of female breast cancer in Missouri. The measured and visualized survival data will transform our registry from being an incidence registry to becoming a survival registry for breast cancer.

Survival data mirrors female breast cancer prediction in a specific period [25]. We used the Missouri Cancer Registry records because Missouri Cancer Registry is a nationally recognized, population-based registry with data that originates from diverse sources including hospitals, ambulatory surgical centers, freestanding cancer treatment centers, pathology laboratories, long-term care facilities, and physician offices. It also contains cases obtained through case-sharing agreements with 19 states. The Missouri Cancer Registry data undergo a strict quality control process and the data are evaluated following specific national measures [26]. Several studies have revealed the significance of linking NDI data to a central cancer registry’s data to obtain more accurate population-based cancer survival data [3-5].

From this study’s results, as shown in Tables 1-4, we can create female breast cancer survival profiles for the 34 Missouri senate districts. By creating these profiles, we can compare each district’s results to the state and to other districts’ results and give more detailed information to public health practitioners and decision makers about female breast cancer in their district.

Cancer incidence and mortality data have traditionally been presented in tabular and descriptive statistics formats; these are easily understood by health professionals with specific knowledge and experience in statistics and epidemiology. At the Missouri Cancer Registry and Research Center, we strive to present our data in formats that meet the needs of a wide range of potential data users. That is why we chose to combine our survival data with geographical data to produce interactive mapping reports at the Missouri senate district level. InstantAtlas is an interactive, internet-based mapping tool licensed to the Missouri Cancer Registry and Research Center that allows users to visually display data gathered from the registry and other databases. Use of interactive data visualization and mapping software allows users to interact with the datasets. We built two interactive mapping reports that include our senate district-level female breast cancer survival data [23,24]. The two maps, the area profile map and the double map, have not yet been published on the Missouri Cancer Registry and Research Center’s website. The area profile report shows a single map and focuses on displaying many indicators for a selected state senate district and compares the district’s findings to other districts and to Missouri. The double map focuses on exploring the relationships between selected indicators; it displays two indicators simultaneously along with a scatterplot or a table.

The InstantAtlas reports can facilitate communication between collaborators from different fields related to breast cancer, enhance female breast cancer research and policy, and inform public health professionals and policy makers. These maps can be used as educational tools at the community level for women at risk and the public about the distribution of female breast cancer in Missouri by age, race, stage and grade at diagnosis, and by senatorial district. These data could be used as a knowledge base at Missouri oncology facilities to assess management plan decisions taken by providers and by female breast cancer cases.

During the matching processes, some cases did not have a social security number, which is the best available unique identifier. Also, some identifiers, such as date of birth and last and/or first name, showed differences when the NDI database and the registry database were compared, possibly due to data entry errors or changed last name. Such cases were manually reviewed. Manual review of all partial matches was done by more than one Missouri Cancer Registry and Research Center staff member, including at least one certified tumor registrar, to reduce possible mistakes.

Survival was measured using cause-specific survival rather than relative survival (another common net measure of survival) to avoid the need of having detailed population lifetables by senatorial district. Potential disadvantages of using cause-specific survival is that, unlike relative survival, it relies on additionally having the cause of death rather than just the fact and date of death and on accurate coding of the cause of death [18]. To decrease the number of known decedents with unknown cause of death in the Missouri Cancer Registry database, these cases were included in the NDI linkage to try to obtain their cause of death. To lessen the impact of miscoded cause of death (eg, a breast cancer death being misattributed to the location of a metastatic site), the data used here was defined “breast cancer death” according to the SEER cause-specific death classification recode variable [18]. It should be noted that this will miss indirect deaths originating from a diagnosis of breast cancer, such as toxic effects of chemotherapy. Moreover, the use of this death classification variable is limited to first primary tumors only as used in these analyses and cannot be used to analyze second and subsequent tumors.

Due to aggregating the cases to areal units, this study is subject to the modifiable areal unit problem [27]. State senate areal units were selected for this project because they would be relevant to policy makers making decisions at the senate district level and to constituents within those districts. It should be noted that the modifiable areal unit problem implies that differing conclusions can potentially be drawn from the same data had different areal units been used.

The survival rates presented in these mapping reports are the observed percentages rather than rates that have been spatially smoothed. Observed percentages may be more directly interpretable and relevant to the residents of each of the individual senate districts; however, observed percentages have the disadvantage of being less stable and more prone to spurious high and low values than spatially smoothed survival rates. This instability is mitigated somewhat by the fact that, with the notable exception of African Americans in many districts, survival was calculated with fairly large sample sizes: always more than 100 and generally at least 200 or more.

For the selected cases, only approximately 1% had the district imputed. Due to the relatively small number of cases, a sensitivity analysis was not performed.

In the future, by combining mortality and incidence data in the survival profiles, we will be able to inform every district’s decision makers about the full picture of female breast cancer burden by district and we could help them assess female breast cancer interventions and policies on geographical bases. Due to small sample sizes, we do not have county-level results from the Behavioral Risk Factor Surveillance System, a state-based health survey that annually gathers data on health events, behaviors, preventive practices, and access to health care. A similar Behavioral Risk Factor Surveillance System-based survey known as the “County-Level Study” has been conducted at the county level in Missouri [28]. In the future, we hope to combine these results with female breast cancer survival data and create InstantAtlas mapping reports at the senatorial district level that include survival and other measured contextual indicators (eg , demographic, environment, and socioeconomic), similar to the currently published county-level maps. This kind of mapping report could be used to explore the relationship between female breast cancer and other measured contextual indicators all over Missouri.

In this paper, we measured 5-year cause-specific survival proportions of female breast cancer for the 34 senate districts in Missouri. In the future, we will consider the feasibility of measuring the same data for all 163 Missouri legislative districts [29,30]. We will also consider measuring 5-year cause-specific survival for other screening-amenable cancers (eg, colorectal cancers) and for cancers that impact many residents (eg, lung cancer).

Before we publish senate district maps on our website, we aim to test the usability of the survival maps using a pilot sample of actual users, similar to one we conducted with our previously published maps [31], in order to make them more user friendly.

Net measures of survival factor out other causes of death and are useful from a policy-based perspective. These measures enable comparisons of cancer survival across geographical regions and between groups of patients without differences in background mortality rates of other causes impacting the results.

Cancer registry data are very rich and can be used in the exploration of many scientific theories and models. Registry data are a valuable source for survival data on breast cancer by race, age, and stage at diagnosis. Using cancer registry data supplemented by SSDI and NDI information will be beneficial and can improve accuracy of breast cancer survival data by age, stage, or race, as well as by geographic area (counties and senatorial districts).