Risk of Advanced-Stage Breast Cancer among Older Women with Comorbidities

Data source

The BCSC is a collaborative effort between 7 geographically dispersed mammography registries (14). Details about the BCSC have been provided elsewhere (15, 16). Data were obtained from 4 BCSC mammography registries (Carolina Mammography Registry, New Hampshire Mammography Network, San Francisco Mammography Registry, and Vermont Breast Cancer Surveillance System) that participated in linking BCSC records and Medicare claims data. Registries collected demographic, risk factor, and clinical information at each mammogram, including radiologists' indication for examination and recommendations based on the American College of Radiology's Breast Imaging Reporting and Data System (BI-RADS). Data were pooled at a central Statistical Coordinating Center (SCC) at Group Health Research Institute (Seattle, WA; ref. 14, 16). BCSC registries and the SCC received Institutional Review Board approval for active or passive consenting processes or a waiver of consent to enroll participants, link data, and conduct analysis. All procedures were Health Insurance Portability and Accountability Act (HIPAA) compliant, and registries and the SCC received a Federal Certificate of Confidentiality and other protection for the identities of women, physicians, and facilities.

Women participating in these 4 mammography registries who were also enrolled in Medicare were linked to the Center for Medicare & Medicaid Services' (CMS) Medicare Program Master Enrollment file by identifiers such as name, date of birth, and social security number. Breast cancer diagnoses and tumor characteristics were obtained through linkage with state tumor registries or regional SEER programs (17) and additional linkage to hospital-based pathology services at 3 of the 4 mammography registries (16).

Study population

The study population included all women who were 67 years of age and older who had undergone mammography in the BCSC database from the 4 sites between January 1, 2000, and December 31, 2006. We limited the study to women with at least 2 years of continuous Medicare enrollment in part A and part B and who were not enrolled in Medicare Advantage for 2 years before a mammogram in the BCSC database and with no previous history of breast cancer. We required 2 years of continuous Medicare enrollment before a mammogram to ensure complete capture of claims during the period used to assess presence of comorbidities. To further ensure complete capture of Medicare claims, we excluded mammograms included in the BCSC database for which a corresponding Medicare claim for a mammogram could not be found within 7 days before or after the examination date recorded in the BCSC database. Approximately 426,295 mammograms meeting inclusion criteria were identified in the BCSC database from 2000 to 2006 and 415,078 (97.4%) had a matching Medicare mammogram claim within 7 days of BCSC examination date. Of those with matching mammogram claims, 3,316 (2.2%) had a breast cancer diagnosis within 12 months of mammography.

Measurements and definitions

We characterized each mammogram included in the study according to type of examination (screening vs. diagnostic) and time interval between that examination and a woman's most recent prior examination. A screening mammogram was defined using the standard BCSC definition as a bilateral mammogram with a physician-stated indication for routine screening and no mammography within the past 9 months. Mammography intervals were categorized as 4 to 18 months, 19 to 30 months, or 31 to 42 months (15, 16, 18). Estimates for the number of months between mammograms were calculated by using the most recent of the dates of previous mammography in the BCSC database or the self-reported date given by the participant. All mammograms occurring within a 4-month period were considered to represent a single diagnostic series, and our analyses included only the first mammogram in the series.

We applied a recently refined list of 38 comorbid conditions (7), classified by organ system and severity, using ICD-9-CM codes listed in previously published comorbidity measures developed for Medicare claims data by Klabunde and colleagues (19), Fleming and colleagues (12), and Elixhauser and colleagues (20). We updated these measures to more recent versions of ICD-9-CM and distinguished “stable” and “unstable” comorbidities on the basis of clinical significance, judgment of seriousness, and whether the condition predicted 5-year mortality in prior studies. This distinction was previously shown to be useful in understanding variation in mammography use (7). We defined comorbidities that are life-threatening or difficult to control such as severe heart failure, cardiac arrhythmias, and end-stage liver disease as “unstable,” and age-related conditions that could affect daily function, such as arthritis, osteoporosis, depression, and diabetes as “stable” (Supplementary Data S1).

Inpatient, outpatient, and physician–supplier claims were reviewed for 2 years before the most recent mammogram in the BCSC database and used to determine the prevalence of comorbidities during that period, excluding the month of diagnosis in cancer cases. If a qualifying comorbid diagnosis code appeared only once in physician claims during that period and an identical code was not present in inpatient hospital claims, then the condition was not counted as a comorbidity (21). Likewise, if a code appeared more than once in physician claims within a 30-day period but never appeared again in either inpatient hospital or physician claims, then the condition was not counted (22). This approach was based on prior studies showing poor agreement between medical records and Medicare claims when looser methods were used to capture diagnostic information (21, 22).

To estimate total comorbidity burden, we grouped stable and unstable comorbidities as absent or present. We then counted the number of stable comorbid conditions. Unstable comorbidities were collapsed into a single group, as the prevalence of multiple unstable conditions was very low.

Breast cancer stage at diagnosis was classified according to the tumor, node, metastasis (TNM) system on the basis of the criteria of the American Joint Committee on Cancer as stage 0, I, IIA, IIB, III, or IV. Invasive tumors of stages IIB, III, and IV were considered to be advanced-stage disease (23). This definition of advanced-stage disease has been used as a proxy outcome among women with breast cancer because only 5% to 12% of stage I/II patients die within 10 years after diagnosis, compared with more than 60% of stage III patients and more than 90% of stage IV patients (24).

Statistical analysis

We estimated the overall prevalence of stable and unstable comorbidities in our study population. We summarized demographic characteristics stratified by comorbidity status: no comorbid conditions, stable conditions only (aggregating women with one or more conditions, due to their similar characteristics), or unstable conditions with or without additional stable conditions. Among women diagnosed with cancer, demographic characteristics, comorbidity burden, mammography use rates, and tumor characteristics were compared between comorbidity groups using χ² tests with statistical significance defined at P < 0.05.

We calculated stratum-specific frequency distributions for previous mammography use based on a single mammogram per woman to prevent overrepresentation of women who were frequent mammography users (15). We selected the mammogram closest to the date of diagnosis for cancer cases and a randomly selected mammogram for non-cancer cases. For mammograms followed by a cancer diagnosis, we calculated the proportion of tumors that were invasive and the distribution of tumor size, stage, grade, and estrogen receptor status.

We explored differences in previous mammography use by comorbidity status using logistic regression models for the binary outcome “adequately screened,” which we defined as having 2 mammography examinations within 30 months. This model was adjusted for age, race/ethnicity, year, and BCSC registry. In this analysis, we selected one mammogram to include in the analysis per woman as described above.

We used logistic regression to estimate overall and advanced cancer rates per 1,000 mammograms by comorbidity status, adjusting for age, race/ethnicity, year of diagnosis, and BCSC registry using generalized estimating equations (GEE). In these analyses, we excluded diagnostic mammograms with no prior mammogram within 42 months. The unit of analysis in these models was the mammogram, with women potentially contributing multiple mammograms. The method of GEE accounts for clustering among mammograms from the same woman. Adjusted cancer rates were estimated using the method of indirect standardization (25, 26). Confidence intervals (CI) were estimated using the delta method. Models were also fit including interaction terms for previous mammography use and comorbidity status to examine variation in the association between comorbidity status and cancer rates across previous mammography use groups. Adjusted cancer rates from these models are reported for each comorbidity and previous mammography use group. We also computed the mean time from the most recent prior mammography examination to cancer diagnosis to evaluate possible diagnostic delays among women by comorbidity status. All statistical analyses were carried out using R statistical software.