Secondary Logo

Journal Logo

Applied Methods

Comparison of SEER Treatment Data With Medicare Claims

Noone, Anne-Michelle MS; Lund, Jennifer L. PhD; Mariotto, Angela PhD; Cronin, Kathleen PhD; McNeel, Timothy BA; Deapen, Dennis DrPH; Warren, Joan L. PhD

Author Information
doi: 10.1097/MLR.0000000000000073

Abstract

The Surveillance, Epidemiology, and End Results (SEER) program, sponsored by the National Cancer Institute (NCI), is a system of population-based cancer registries that currently covers approximately 28% of the US population from geographically defined areas. In addition to reporting national cancer statistics on incidence and survival, the SEER registries serve as a platform for studies of cancer-related care and health disparities. As such there is considerable interest in treatment information for individuals diagnosed with cancer captured in the population-based registries. The SEER registries routinely collect data on the first course of cancer treatment including information on surgery, radiation therapy (RT), chemotherapy (CT), and hormone therapy (HT). Information on surgery and RT is reported in the publically available SEER dataset, although several studies have reported underascertainment of RT on the SEER data.1,2 NCI does not make data on CT and HT publically available due to concerns about the completeness of the information.

Persons in the SEER data who are Medicare eligible have been matched to their Medicare claims to create the linked SEER-Medicare data. The SEER-Medicare data include longitudinal claims, allowing for the identification of cancer treatments from Medicare claims for cancer patients appearing in the SEER data.3 Prior validation studies have shown that Medicare claims can accurately identify persons receiving RT1,4,5 and CT.6–10 Several studies have used Medicare claims to identify individuals who received HT11–13 although these claims have not been validated.

Activities are ongoing at NCI to further investigate the opportunity and benefit of using claims to supplement treatment data. The linkage of the SEER and Medicare data offers the opportunity to validate treatment reported on the SEER data, using Medicare claims as the gold standard. As an initial step, this analysis will evaluate the completeness and validity of SEER treatment data. We compared treatment data from SEER with Medicare claims for (1) CT among individuals diagnosed with bladder, female breast, colorectal, lung, ovarian, pancreas, and prostate cancer; (2) RT among individuals diagnosed with these cancers except ovarian; and (3) HT among individuals diagnosed with prostate cancer. The concordance of treatment between the 2 data sources was estimated across a number of covariates including cancer site, stage, sex, age, race/ethnicity, geographic location, and year of diagnosis. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of SEER treatment variables were also estimated across the same covariates using Medicare claims as the gold standard.

METHODS

Data Sources

The SEER-Medicare data used for this study included cases reported from 17 SEER cancer registries, with the exception of Alaska Natives. To augment available public use of SEER data, we were able to access internal NCI data that included information about CT and HT administration collected by the SEER registries. This internal information and publically available information on RT was used for comparison to Medicare treatment information.

Individuals in the SEER data have been matched to Medicare’s master enrollment file maintained by the Centers for Medicare and Medicaid Services (CMS).14 For persons reported to a SEER registry who were aged 65 or older, 94% have been linked to Medicare and their Medicare claims have been extracted.15–17 For this analysis, we included claims from inpatient hospitalizations (MEDPAR), outpatient facilities, physician claims, or durable medical equipment (DME). These files include claims for CT, RT, and injected HT for each beneficiary with fee-for-service coverage.17 All Medicare claims include codes that indicate the type of services performed. Services are reported using the International Classification of Diseases, Ninth Revision (ICD-9), Healthcare Common Procedure Coding System (HCPCS), and National Drug Codes (NDCs). In general, oral agents for CT are covered by Medicare Part D and so could not be captured in this analysis. However, oral chemotherapeutic agents that are equivalent to those intravenously administered, such as capecitabine, are covered as a part of Medicare Part B and could be identified in our study.

Study Sample

We identified 433,415 individuals in the SEER-Medicare data who were diagnosed with bladder, female breast, colorectal, lung, ovary, pancreas, or prostate cancer from 2000 to 2006 for analysis. For inclusion in our study, individuals had to be at least 65 years of age at diagnosis, have continuous Part A, B, and fee-for-service coverage and not be enrolled in a Health Maintenance Organization during that year, and only have 1 cancer according to SEER data. These criteria were implemented to ensure complete Medicare claims for all individuals and so that any treatment identified using claims would be attributed to the correct cancer. In addition, those diagnosed on autopsy or death certificate (n=9276) or without a known month of diagnosis (n=4984) were excluded. A small number of individuals were part of the Patterns of Care (POC) studies performed by the SEER registries which involve obtaining more detailed information about CT and HT verified by the treating physician. These individuals were excluded as we were interested in the routine collection of SEER data (n=2671).18

Identification of Treatment

Treatment information was identified independently from the SEER data and from Medicare claims. SEER collects information on the first course of therapy. As individuals initiate therapy at different times postdiagnosis, cancer registrars update treatment information as it becomes available. For analysis, SEER treatment data for CT, RT, and HT were classified as received, not received, or unknown.

Treatment information for CT, RT, and HT was identified from Medicare claims data by reviewing the MEDPAR, outpatient, physician, and DME claims for 12 months after diagnosis. The first of the month and year of diagnosis reported by SEER registries was used as the diagnosis date. An individual was considered to have received CT, RT, or HT if at least 1 Medicare claim included a code for each specific treatment and to not have received treatment if no claims for the specific treatments were found. Thus, no individual had an unknown treatment status from the Medicare data. Codes used to identify treatment are listed in Table (Supplemental Digital Content 1, http://links.lww.com/MLR/A665, which lists codes).

Comparison of SEER and Medicare Treatment Data

Sensitivity, specificity, PPV, and NPV and their 95% confidence intervals were calculated to quantify the validity of SEER data to identify treatment receipt using Medicare claims as the gold standard. Although we calculated sensitivity, specificity, PPV, and NPV, we focused our analyses on sensitivity and PPV with the other measures reported in the Supplemental Tables (Supplemental Digital Content 2–4, http://links.lww.com/MLR/A666, http://links.lww.com/MLR/A667, http://links.lww.com/MLR/A668, which shows additional statistics).19 The analysis was performed separately for each treatment and by patient characteristics such as cancer site, stage, sex, age, race/ethnicity, geographic location, and year of diagnosis from the SEER data. As SEER does not define a fixed interval for the collection of treatment information, the overall sensitivity and PPV were also evaluated using a 4- and 8-month postdiagnosis Medicare claims window to provide a comparison to the 12-month claims window length. Concordance between the 2 data sources was quantified with the κ statistic and the percent agreement which was defined as the proportion of individuals with the same reported treatment receipt from SEER and Medicare data (ie, the number of individuals where SEER data and Medicare claims agree about treatment receipt/total number of individuals with treatment recorded in SEER). Neither of these quantities assumes a gold standard but the κ statistic does account for agreement occurring by chance alone.

RESULTS

Our analysis included a total of 433,415 individuals diagnosed with the selected cancers. Twenty-seven percent of individuals were diagnosed with prostate cancer followed by lung (24%), female breast (18%), and colorectal (18%) (Table 1). The remaining individuals were diagnosed with bladder, pancreatic, or ovarian cancer (7%, 5%, and 2%, respectively). The majority of individuals were diagnosed at localized or regional stage except for individuals with cancer of the lung, pancreas, or ovary who were primarily diagnosed with distant stage disease. Our study included individuals who were predominantly non-Hispanic whites (82%) and diagnosed at ages 65–79 years.

TABLE 1
TABLE 1:
Characteristics of 433,415 Individuals Diagnosed With Selected Cancers and Included in Analysis by Cancer Site

CT

A total of 415,341 individuals were included in the comparison of CT after excluding 4.2% with unknown CT status from SEER. For all cancers combined, there was a high level of agreement between the Medicare claims and the SEER data (90%, Fig. 1), attributable to the fact that most individuals did not receive CT. The majority of discordant findings were individuals identified as receiving CT using Medicare claims but not receiving CT using SEER data.

FIGURE 1
FIGURE 1:
Agreement between chemotherapy, radiation, and hormone therapy collected by SEER and treatment identified using Medicare claimsa. aA 12-month postdiagnosis window was used to identify treatment using Medicare claims (gold standard). SEER indicates Surveillance, Epidemiology, and End Results.

The sensitivity of SEER data to identify individuals who received CT was 68% overall and varied by patient and tumor characteristics (Table 2). The sensitivity of SEER data was highest for ovarian cancer and lowest for bladder and prostate cancers (84%, 22%, and 7%, respectively). In addition, the sensitivity decreased with age for all cancers combined, specifically 74% for individuals aged 65–69 compared with 47% for those aged 85 and older. The sensitivity was low for detecting CT in individuals diagnosed with in situ or localized disease. In addition, there was some variation in sensitivity by registry (range, 60%–78%). The sensitivity, however, did not vary greatly based on year of diagnosis (data not shown), sex, or race with the exception of individuals with unknown race.

TABLE 2
TABLE 2:
Sensitivity and PPV of SEER Data to Identify Chemotherapy Receipt and κ Statistics by Patient and Tumor Characteristics*

The poor results for prostate cancer did not, however, greatly affect the overall results because so few men receive CT as first-course therapy. Specifically, among men with all cancer sites combined, the sensitivity was 62.3% and modestly increased to 66.3% when prostate cancer was excluded. Finally, we further investigated potential misclassification of men with prostate cancer who had a claim for CT administration that may have been used for delivery of HT. We considered additional CT administration claims found on the same day as HT but the number of men with these additional codes was small and hence their impact on the overall results would be minimal.

Overall the PPV was high indicating that among those identified in the SEER data as receiving CT, the vast majority also had Medicare claims for CT (Table 2). The κ statistics showed a moderate level of agreement overall and also followed a similar pattern to sensitivity and PPV. The specificity (range, 93%–100%) and NPV (range, 80%–99% excluding bladder) were high across all characteristics (Table, Supplemental Digital Content 2, http://links.lww.com/MLR/A666, which shows additional statistics for CT). Finally, as an addition to the sensitivity analysis we evaluated the individuals excluded because of unknown CT in SEER data and found them to have similar rate of CT as those with known CT status (72% vs. 76%, respectively).

RT

Overall 2.8% of individuals had unknown RT status reported in the SEER data, thus 412,350 patients were available for comparison. The overall agreement between Medicare claims and the SEER data for identifying RT was high (91%, Fig. 1). The majority of discordant findings were individuals identified as receiving RT using Medicare claims but were not identified using SEER data (7.4%).

The sensitivity of SEER data to identify individuals who received RT was 80% overall and varied by patient and tumor characteristics (Table 3). The sensitivity of SEER data was highest for prostate cancer (85%), lowest for bladder (54%), and ranged between 66% and 80% for the remaining cancer sites. The sensitivity decreased by year of diagnosis (data not shown, range, 82%–78%) and increasing age (range, 81%–65%). The sensitivity did not vary greatly by stage or race with the exception of low sensitivity of individuals in the unstaged or unknown race categories.

TABLE 3
TABLE 3:
Sensitivity and PPV of SEER Data to Identify Radiation Therapy and κ Statistics by Patient and Tumor Characteristics*

For all cancer sites combined, the PPV of RT on the SEER data was 95%. Although there was some variation in PPV by patient and tumor characteristics, the PPV was high for all cancer sites. The overall κ statistics were moderate to strong and followed a similar pattern as sensitivity and PPV. In addition, the overall specificity and NPV were high. The specificity overall was 97% and ranged between 94% and 100% by characteristic and the NPV was 89% overall and ranged from 82% to 96% (Table, Supplemental Digital Content 3, http://links.lww.com/MLR/A667, which shows additional statistics for RT). For all cancer sites combined, 73% of those with unknown RT in the SEER data did not have claims, ranging from 90% for colorectal cancer to 60% for breast cancer.

HT

A total of 115,724 men with prostate cancer were in the study cohort and 111,466 had known HT status recorded in the SEER data and were included in the analysis of HT. The overall agreement between SEER and Medicare was 81% (Fig. 1).

The sensitivity of the SEER data to identify HT was 69% and followed a similar pattern to CT. Specifically, the sensitivity was lower for localized/regional stage or unknown stage disease and decreased with increasing age (Table 4). The sensitivity did not vary much by race with the exception of unknown race and, finally, the sensitivity varied by registry (range, 47%–76%).

TABLE 4
TABLE 4:
Sensitivity and PPV of the SEER Data to Identify Hormone Therapy and κ Statistics by Patient and Tumor Characteristics Among Men With Prostate Cancer*

The overall PPV was 86% and did not vary substantially by characteristic. American Indian/Alaska Natives and those with distant-stage disease had lower PPVs (66% and 75%, respectively). The overall κ showed a low level of agreement. Similar to CT and RT, the specificity was high except for distant stage; however, the NPV was only 78% overall (Table, Supplemental Digital Content 4, http://links.lww.com/MLR/A668, which shows additional statistics for HT). The NPV was lower for individuals with distant (61%) or unstaged (66%) cancers compared with local/regional stage (79%) or for older patients (87% for age 65–69 vs. 62% for age 85+ ). The NPV did not vary substantially by race or registry.

DISCUSSION

In order for the treatment data recorded by the SEER registries to be useful to assess how patients are treated, the information must have good sensitivity and PPV. We found that the overall utility of the SEER data to identify cancer treatment was limited. The sensitivity of the SEER data varied by treatment type, cancer site, stage, age at diagnosis, and registry. For CT and HT, the sensitivity was low, whereas the sensitivity was better for reporting of RT. We investigated whether SEER data may be appropriate for use among specific site/stage strata. However, for the cancer sites included in this analysis the sensitivity was <90% for all stages and treatment types (See Table, Supplemental Digital Content 5, http://links.lww.com/MLR/A672 which shows sensitivity and PPV by cancer site and stage). These results indicate that using SEER data to assess treatment will miss a number of patients who have been treated, resulting in misclassification and inaccurate estimates of the proportion of individuals treated.

Although the sensitivity was poor, the PPV was high among all treatment types and did not vary greatly by cancer site or characteristic with the exception of CT for prostate cancer. As within many site/stage strata SEER reliably identified individuals who truly received treatment, SEER treatment data may have a limited role in research projects to identify a cohort of treated individuals within these strata (See Table, Supplemental Digital Content 5, http://links.lww.com/MLR/A672 which shows sensitivity and PPV by cancer site and stage). SEER data, however, will not capture all individuals who received treatment.

This is the first study to compare CT reported in the SEER data with Medicare claims. Prior studies from 2 SEER registries compared CT data collected by medical records to registry data for women of all ages with breast cancer. These earlier studies had similar rates of concordance as the current study.7,20 As treatment is more often received in settings outside of the hospital system, either a physician’s office or an outpatient clinic, identification of treatment is difficult and may be a major contributing factor to the low capture of CT in the SEER data. For example, the low sensitivity of CT for bladder cancer is likely explained by more patients being treated primarily in the physician office. Other challenges include a trend away from intravenous CT drugs to oral CT and identification of prescription drug use is difficult for cancer registrars to find.

Among men with prostate cancer, the sensitivity of CT was poor, even among men with distant-stage disease (13.6% vs. 4.9% for localized/regional stage). These findings could be influenced by the challenges of distinguishing men who received CT, HT, or both using Medicare claims. Among men with prostate cancer included in this analysis who received CT according to Medicare claims, 67% also had a claim for HT within the same year. For these men, we cannot definitively determine if all claims were related to administration of HT or if some were for CT. A recent study compared CT and HT data compiled by the Cancer Research Network’s (CRN) Virtual Data Warehouse, which included encounter, claims, and electronic medical record data, to gold standard tumor registry data abstracted from medical charts.21 The PPV for identifying CT for prostate cancer was only 6%, similar to our current findings. These results reflect the fact that registry data is limited to the first course of therapy and the difficulty in distinguishing CT and HT using claims for prostate cancer. In contrast to poor reporting of CT in the CRN data, the authors found that among men with prostate cancer identified as receiving HT in the CRN data, 92% also had HT captured in the tumor registry data. The success in identifying HT from the CRN claims data compared with Medicare claims may be due to availability of pharmacy information in the CRN data.

We found moderate sensitivity and agreement between SEER data and Medicare claims for receipt of RT. In previous comparisons of reported RT use between SEER and Medicare data, breast, endometrial, lung, prostate, and rectal cancer had strong agreement when comparing RT use for individuals diagnosed between 1991 and 1996.4 A more recent study compared receipt of RT as reported in the SEER data for individuals diagnosed with breast cancer aged 20 to 79 and diagnosed between 2005 and 2007 with self-reported treatment information obtained through patient survey.2 This analysis found overall agreement of 83%, and 21% of individuals who self-reported receiving RT were recorded as not receiving RT in the SEER data. Underascertainment of RT was higher in women under the age of 65 who would not be eligible for Medicare. We also found more underreporting of RT in the SEER data, about 21%, which is consistent with the more recent study.1,2,5 The poorer performance with more recent diagnoses could be due to the increasing difficulty for registries to collect treatment information delivered in the outpatient setting. Underascertainment of RT may also result if individuals receive surgical treatment in a large cancer center but receive RT in a community setting.

There are several limitations to this study. One caveat is the potential for underreporting of treatment by both Medicare claims and the SEER data. Underreporting of treatment in the Medicare claims would result from individuals receiving care outside of the Medicare system, such as from the Veterans Health Administration, or if individuals are still working and have employer-sponsored insurance. Whereas, underreporting of treatment in the SEER data may be caused by care not captured by the registry, such as prescription drugs or outpatient treatment, or individuals leaving the registry catchment area for treatment. The sensitivity among those over the age of 85 is lower which may be due to these individuals leaving home while receiving treatment. In addition, the sensitivity varied by registry. This may be due to differences in how the data are collected at the registry and the proportion of patients receiving inpatient treatment.

Apparent differences in the reporting of treatment between the SEER and Medicare data may be caused by misalignment of the treatment window captured by SEER and Medicare files. In this evaluation, a 12-month postdiagnosis window was used to identify claims from Medicare. This timeframe was chosen as it aligned with guidelines from the SEER Program Coding and Staging Manual followed by SEER registrars stating that in the absence of additional documentation the first course of therapy ends after 1 year postdiagnosis.22 However, SEER is only capturing first-course treatment, whereas Medicare claims would also capture secondary treatment due to an inability of individuals to tolerate the initial treatment or failure of the initial treatment to produce the desired outcome. The overall sensitivity of the SEER data increased and the PPV decreased using a shorter window postdiagnosis to identify claims but these differences were modest (Tables 2–4).

In addition, there may have been some missed opportunity for the SEER registries to capture treatment. We reviewed the claims for individuals identified as having received treatment in Medicare but not in the SEER data to determine the number of encounters the individual had with the health care system. The number of encounters was quantified by the number of days in the first year postdiagnosis a claim for treatment was found. Thirty-two percent of individuals with any claims for CT and 55% of those with any claims for RT had at least 10 encounters. As most of the reporting was from inpatient claims (Table 1) this may imply that SEER may have missed the opportunity to capture treatment information. Finally, treatment status for individuals in managed care systems may be more easily identified by the registries as there is a unified record of incident cancer and treatment but only individuals with fee-for-service were included in this analysis.

Registries’ ability to capture accurate treatment information is becoming more difficult over time because of trends in cancer treatment and also privacy concerns. Although the Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule includes provisions to allow mandated cancer reporting without patient consent for public health surveillance and research, it does create confusion on the disclosure of protected health information by health care providers.23 A survey of North American Association of Central Cancer Registries (NAACCR) found that two thirds of registries cited HIPAA as interfering with nonresearch operations. Among those, 75% reported that HIPAA was identified as the reason for either ceased case reporting from at least 1 source, prevented case finding or auditing, or interfered with collection of follow-up data.24

Although there are significant limitations found with the SEER treatment data, there are several possible uses. For example, researchers used SEER treatment data to select a cohort of treated individuals.25 Furthermore, SEER treatment data could be improved by correcting the bias using the POC data, as done in the study by Mariotto et al,26 and also by potentially augmenting SEER treatment data using other larger scale data sources. Linkages between the SEER data and data sources such as the CRN, state-employee data, electronic medical records, and insurance billing claims data, such as Medicare, may provide an opportunity to enhance the quality of treatment data for individuals with cancer and possibly provide an opportunity to estimate rates of treatment receipt in the population. For example, data from CRN’s Virtual Data Warehouse have been shown to reliably identify CT as treatment for breast,21,27 colorectal,21 lung,21 and ovarian28 cancer. Additional benefits to these data are that they include pharmacy data, individuals under 65 years of age, and those receiving treatment in HMOs.

In conclusion, caution should be taken when using treatment information from the SEER data. As the sensitivity of SEER data to identify treatment was consistently low, these data cannot be used to accurately describe the proportion of individuals in the population who received treatment. In addition, comparisons between individuals who received and did not receive treatment based on the SEER data would be biased. In contrast, the PPV was high for most cancer sites suggesting that SEER treatment data could be used reliably to identify a cohort of individuals with cancer who received treatment. RT data are currently available in the public-use SEER research data file. CT and HT data are not publically available but approval may be granted upon special data request to the SEER program. Our findings can inform investigators about potential uses of the CT and HT data, albeit limited. Finally, NCI is currently investigating linkages to the SEER data, such as Medicare data, that could be used to supplement SEER treatment information leading to more accurate and complete information about treatment for individuals with cancer.

REFERENCES

1. Cooper GS, Yuan Z, Stange KC, et al.. Agreement of Medicare claims and tumor registry data for assessment of cancer-related treatment. Med Care. 2000;38:411–421.
2. Jagsi R, Abrahamse P, Hawley ST, et al.. Underascertainment of radiotherapy receipt in Surveillance, Epidemiology, and End Results registry data. Cancer. 2012;118:333–341.
3. National Cancer Institute. SEER-Medicare Publications. Available at: http://healthservices.cancer.gov/seermedicare/overview/publications.html. Accessed March 8, 2012.
4. Virnig BA, Warren JL, Cooper GS, et al.. Studying radiation therapy using SEER-Medicare-linked data. Med Care. 2002;40supplIV-49–IV-54.
5. Du X, Freeman JL, Goodwin JS. Information on radiation treatment in patients with breast cancer: the advantages of the linked Medicare and SEER data. Surveillance, Epidemiology and End Results. J Clin Epidemiol. 1999;52:463–470.
6. Lund JL, Sturmer T, Harlan LC, et al.. Identifying specific chemotherapeutic agents in Medicare data: a validation study. Med Care. 2013;51:e27–e34.
7. Du XL, Key CR, Dickie L, et al.. External validation of Medicare claims for breast cancer chemotherapy compared with medical chart reviews. Med Care. 2006;44:124–131.
8. Lamont EB, Herndon JE, Weeks JC, et al.. Criterion validity of Medicare chemotherapy claims in Cancer and Leukemia Group B breast and lung cancer trial participants. J Natl Cancer Inst. 2005;97:1080–1083.
9. Warren JL, Harlan LC, Fahey A, et al.. Utility of the SEER-Medicare data to identify chemotherapy use. Med Care. 2002;40supplIV-55–IV-61.
10. Lamont EB, Lauderdale DS, Schilsky RL, et al.. Construct validity of Medicare chemotherapy claims: the case of 5FU. Med Care. 2002;40:201–211.
11. Elliott SP, Jarosek SL, Wilt TJ, et al.. Reduction in physician reimbursement and use of hormone therapy in prostate cancer. J Natl Cancer Inst. 2010;102:1826–1834.
12. Shahinian VB, Kuo YF, Freeman JL, et al.. Increasing use of gonadotropin-releasing hormone agonists for the treatment of localized prostate carcinoma. Cancer. 2005;103:1615–1624.
13. Williams SB, Gu X, Lipsitz SR, et al.. Utilization and expense of adjuvant cancer therapies following radical prostatectomy. Cancer. 2011;117:4846–4854.
14. National Cancer Institute. SEER-Medicare Data. Available at: http://healthservices.cancer.gov/seermedicare/. Accessed September 30, 2011.
15. Altekruse SF, Kosary CL, Krapcho M, et al.. SEER Cancer Statistics Review, 1975-2007. Bethesda, MD: National Cancer Institute; 2010. Available at http://seer.cancer.gov/csr/1975_2007/.
16. Potosky AL, Riley GF, Lubitz JD, et al.. Potential for cancer related health services research using a linked Medicare-tumor registry database. Med Care. 1993;31:732–748.
17. Warren JL, Klabunde CN, Schrag D, et al.. Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care. 2002;40supplIV-3–IV-18.
18. National Cancer Institute. Patterns of Care/Quality of Care. Available at http://appliedresearch.cancer.gov/surveys/poc/. Accessed September 30, 2011.
19. Chubak J, Pocobelli G, Weiss NS. Tradeoffs between accuracy measures for electronic health care data algorithms. J Clin Epidemiol. 2012;65:343–349.
20. Malin JL, Kahn KL, Adams J, et al.. Validity of cancer registry data for measuring the quality of breast cancer care. J Natl Cancer Inst. 2002;94:835–844.
21. Ritzwoller DP, Carroll N, Delate T, et al.. Validation of electronic data on chemotherapy and hormone therapy use in HMOs. Med Care. 2013;51:e67–e73.
22. Adamo MB, Johnson CH, Ruhl JL, et al.. 2011 SEER Program Coding and Staging Manual. NIH Publication Number 11-5581. 2011.Bethesda, MD: National Cancer Institute.
23. Deapen D. Cancer surveillance and information: balancing public health with privacy and confidentiality concerns (United States). Cancer Causes Control. 2006;17:633–637.
24. Deapen D. Negative Impact of HIPAA on Population-Based Cancer Registry Research: Update of a Brief Survey. Washington, DC: Institute of Medicine: Committee on Health Research and the Privacy of Health Information: the HIPAA Privacy Rule. Available at: http://www.iom.edu/∼/media/Files/Activity%20Files/Research/HIPAAandResearch/HIPAAIOMmeeting607rHowe.ashx.
25. Morton LM, Dores GM, Tucker MA, et al.. Evolving risk of therapy-related acute myeloid leukemia following cancer chemotherapy among adults in the United States, 1975-2008. Blood. 2013;121:2996–3004.
26. Mariotto AB, Feuer EJ, Harlan LC, et al.. Dissemination of adjuvant multiagent chemotherapy and tamoxifen for breast cancer in the United States using estrogen receptor information: 1975-1999. J Natl Cancer Inst Monogr. 2006;36:7–15.
27. Delate T, Bowles EJ, Pardee R, et al.. Validity of eight integrated healthcare delivery organizations’ administrative clinical data to capture breast cancer chemotherapy exposure. Cancer Epidemiol Biomarkers Prev. 2012;21:673–680.
28. Aiello Bowles EJ, Tuzzio L, Ritzwoller DP, et al.. Accuracy and complexities of using automated clinical data for capturing chemotherapy administrations: implications for future research. Med Care. 2009;47:1091–1097.
Keywords:

SEER; Medicare; treatment; validation; chemotherapy

Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved.