To the Editor: Prior studies have shown that tumor registry and administrative databases, including the National Cancer Institute Surveillance Epidemiology End Results (SEER) registry and Medicare data, provide valid, accurate, and complete information on various treatment modalities for women with breast cancer.1–3 Another alternative to obtain this information is patient self-report. Three studies have shown that breast cancer survivors can accurately recall important treatment information,4–6 but these studies were small and included largely younger women. Validation of self-report of breast cancer treatments specifically in elderly women is important, because more than 40% of women with breast cancer are diagnosed at age 65 and older. The current study sought to determine whether a population-based cohort of older breast cancer survivors could accurately recall important treatment information.

The study cohort consisted of 3,083 community-dwelling women aged 65 to 89 residing in four states (California, Florida, Illinois, New York) who underwent initial breast cancer surgery in 2003 and completed a telephone survey approximately 30 months after the surgery.7 Self-reported data regarding treatment (breast surgery, radiation therapy, chemotherapy, and reconstructive surgery) were validated against Medicare claims data from 1 month before diagnosis until the time of the survey.8 Agreement was assessed using the Cohen kappa statistic.9

The mean age of the women when they completed the survey was 76.1±5.5 (range 67–93); 94% were Caucasian, and the majority were healthy (65% with no comorbidities). Half were married, and 92% had at least a high school degree. Median annual household income in 2004 was $29,000. According to Medicare claims, 66% underwent breast-conserving surgery (BCS), 67% underwent radiation therapy, and 21% received chemotherapy. Only 4.6% (n=143) underwent breast reconstruction surgery.

Assuming Medicare claims as the criterion standard, the validation of each item on the self-report questionnaire is summarized in Table 1. Overall, agreement was excellent for the four treatments examined. Kappa values varied between 0.83 for type of breast surgery and 0.95 for receipt of radiation therapy. Sensitivity (80–99%), specificity (98–99%), positive and negative predictive values (90–99%) of self-report were high for all four treatments. Proxy response, patient age group, and educational status did not significantly affect accuracy of self-report (data not shown).

Table 1. Agreement Between Self-Report and Medicare Claims for Initial Treatment of Breast Cancer
TreatmentCorrect, %Kappa (95% CI)%
SensitivitySpecificityPPVNegative Predictive Value
  • Excluding missing values: radiation therapy (n=3), chemotherapy (n=6), type of breast surgery (n=25), and reconstructive surgery (n=5).

  • *

    “Best case” scenario for type of breast surgery (see text for explanation): proportion correct=98.2%, kappa=0.96 (95% confidence interval (CI)=0.95–0.97), sensitivity for mastectomy=95.9%, positive predictive value (PPV) for mastectomy=98.9, PPV for breast-conserving surgery (BCS)=97.9.

Receipt of radiation therapy97.90.95 (0.94–0.96)97.698.699.395.3
Receipt of chemotherapy96.80.90 (0.88–0.92)90.098.694.597.4
Type of surgery*92.80.83 (0.81–0.85)    
 BCS  99.579.790.398.7
 Mastectomy  79.799.598.790.3
Receipt of breast reconstructive surgery99.10.90 (0.86–0.94)90.299.690.899.5

Agreement for type of breast surgery was probably higher than what is reported. Of the 2,207 women who self-reported undergoing BCS, 214 had Medicare claims for mastectomy; 171 (80%) of these 214 women also had claims for BCS. Therefore, these 171 women presumably underwent initial BCS followed by mastectomy at a later date and were identified by the Medicare claims algorithm8 as total mastectomy cases. The survey did not capture these 171 women as undergoing eventual mastectomy, because a positive response to the BCS survey item was erroneously programmed to skip the next item about mastectomy. Therefore, these 171 women were placed in the BCS self-report group. If these 171 women are recategorized as self-reported total mastectomy cases, then the sensitivity of self-report for mastectomy increases from 79.7% to 95.9%, and the kappa improves from 0.83 to 0.96 (95% confidence interval=0.95–0.97). These values would represent the “best case” scenario, and the true agreement probably falls somewhere between those reported in Table 1 and this “best case” scenario.

These results confirm the validity of self-report described in previous smaller studies in elderly breast cancer survivors6 and younger women.4,5 Several factors may explain the disagreement between self-report data and Medicare claims in the current study. From a survey standpoint, errors may be attributed to biased self-report, recall, or unclear wording of the survey items. It was attempted to limit this bias by providing brief descriptions of each type of treatment, referencing the time frame of treatment, and wording questions at the eighth-grade level. Although it was assumed that the Medicare claims data were correct, there are limitations of Medicare data regarding accuracy of coding, completeness of claims, and the potential for underreporting due to care provided by other agencies.1,3

In summary, this study demonstrates that a population-based cohort of more 3,000 elderly breast cancer survivors accurately reported general information regarding several key cancer treatments 2 to 3 years later. Details regarding stage of disease and treatment would probably still need to be obtained from other sources, but for investigators performing survey studies in breast cancer survivors that require only broad key treatment information (quality-of-life or patient satisfaction studies), patient self-report appears to be an excellent, lower-cost alternative for obtaining this.


  1. Top of page

Conflict of Interest: The editor in chief has reviewed the conflict of interest checklist provided by the authors and has determined that the authors have no financial or any other kind of personal conflicts with this paper. This research was supported by Grants K07CA125586 (TWY), R21CA131643 and K08AG021631 (JMN), and R01CA81379 (ABN) from the National Institutes of Health, National Cancer Institute and National Institute on Aging. These grant funds are not related to products, technology, or methodologies involved in this manuscript submission.

Author Contributions: Study concept and design: Laud, Nattinger, Yen. Acquisition of subjects and data: Nattinger, Sparapani. Analysis and interpretation of data: Guo, Sparapani, Laud, Nattinger, Neuner, Yen. Preparation of manuscript: Laud, Nattinger, Neuner, Sparapani, Yen.

Sponsor's Role: None.


  1. Top of page