Access the full text.
Sign up today, get DeepDyve free for 14 days.
S. Subramanian, Amy Chen (2013)Treatment patterns and survival among low-income medicaid patients with head and neck cancer.
JAMA otolaryngology-- head & neck surgery, 139 5
C. Bradley, C. Given, Zhehui Luo, Caralee Roberts, G. Copeland, B. Virnig (2007)Medicaid, Medicare, and the Michigan Tumor Registry: A Linkage Strategy
Medical Decision Making, 27
(2008)Kaiser Family Foundation. Health Insurance Coverage of the Total Population
M. Halpern, E. Ward, A. Pavluck, N. Schrag, J. Bian, Amy Chen (2008)Association of insurance status and ethnicity with cancer stage at diagnosis for 12 cancer sites: a retrospective analysis.
The Lancet. Oncology, 9 3
S. Koroukian (2008)Linking the Ohio Cancer Incidence Surveillance System with Medicare, Medicaid, and Clinical Data from Home Health Care and Long Term Care Assessment Instruments: Paving the Way for New Research Endeavors in Geriatric Oncology.
Journal of registry management, 35 4
Vivian Byrd, A. Dodd (2013)Assessing the usability of MAX 2008 encounter data for comprehensive managed care.
Medicare & medicaid research review, 3 1
S. Ramsey, S. Zeliadt, L. Richardson, L. Pollack, H. Linden, D. Blough, N. Anderson (2008)Disenrollment From Medicaid After Recent Cancer Diagnosis
Medical Care, 46
SEER-Medicare: how the SEER & Medicare data are linked
(1975)SEER data submission, posted to the SEER web site
M. Hassett, M. Schymura, Kun Chen, F. Boscoe, F. Gesten, D. Schrag (2016)Variation in breast cancer care quality in New York and California based on race/ethnicity and Medicaid enrollment
D. Schrag, B. Virnig, J. Warren (2009)Linking Tumor Registry and Medicaid Claims to Evaluate Cancer Care Delivery
Health Care Financing Review, 30
Connecticut and the ACA's Medicaid expansion
Xuesong Han, Shiyun Zhu, Yao Tian, B. Kohler, A. Jemal, Elizabeth Ward (2016)Insurance Status and Cancer Stage at Diagnosis Prior to the Affordable Care Act in the United States.
Journal of registry management, 41 3
Healthinsurance . org . Connecticut and the ACA ’ s Medicaid expansion Health Insurance Coverage of the Total Population . 2008 - 2013
J. Warren, C. Klabunde, D. Schrag, P. Bach, G. Riley (2002)Overview of the SEER-Medicare Data: Content, Research Applications, and Generalizability to the United States Elderly Population
Medical Care, 40
Jeremy Snyder, K. Foley (2010)Disparities in colorectal cancer stage of diagnosis among Medicaid-insured residents of North Carolina.
North Carolina medical journal, 71 3
M. Gallaway, Bin Huang, Q. Chen, T. Tucker, J. Mcdowell, E. Durbin, D. Siegel, E. Tai (2019)Identifying Smoking Status and Smoking Cessation Using a Data Linkage Between the Kentucky Cancer Registry and Health Claims Data.
JCO clinical cancer informatics, 3
MS Gallaway, B Huang, Q Chen (2019)Identifying smoking status and smoking cessation using a data linkage between the Kentucky Cancer Registry and Health Claims Data
J Clin Oncol Clin Cancer Inform, 3
CMS takes historic steps to increase public access to Medicaid and CHIP data
C. Perkins, W. Wright, M. Allen, S. Samuels, P. Romano (2001)Breast Cancer Stage at Diagnosis in Relation to Duration of Medicaid Enrollment
Medical Care, 39
M. Guth, R. Garfield, Robin Rudowitz (2017)The Effects of Medicaid Expansion under the ACA: Updated Findings from a Literature Review
(2019)Health Insurance Coverage of the Total Population
C. Bradley, J. Gardiner, C. Given, Caralee Roberts (2005)Cancer, Medicaid enrollment, and survival disparities
F. Boscoe, D. Schrag, Kun Chen, P. Roohan, M. Schymura (2011)Building capacity to assess cancer care in the Medicaid population in New York State.
Health services research, 46 3
M. Gallaway, Bin Huang, Q. Chen, T. Tucker, J. Mcdowell, E. Durbin, S. Stewart, E. Tai (2019)Smoking and Smoking Cessation Among Persons with Tobacco- and Non-tobacco-Associated Cancers
Journal of Community Health, 44
C. Bradley, C. Given, Caralee Roberts (2003)Late Stage Cancers in a Medicaid-insured Population
Medical Care, 41
P. Nadpara, S. Madhavan (2012)Linking Medicare, Medicaid, and cancer registry data to study the burden of cancers in West Virginia.
Medicare & medicaid research review, 2 4
Medicaid Analytic eXtract Files (MAX) User Guide, Version 2
Vivian Byrd, A. Dodd (2012)Assessing the Usability of Encounter Data for Enrollees in Comprehensive Managed Care Across MAX 20072009
Mathematica Policy Research Reports
Cancer patients receiving Medicaid have worse prognosis. Patients in 14 Surveillance, Epidemiology, and End Results (SEER) cancer registries were linked to national Medicaid enrollment ﬁles, 2006–2013, to determine enrollment status during the year before and after diagnosis. A deterministic algorithm based on Social Security number, Medicare Health Insurance Claim number, sex, and date of birth was utilized. Results were compared with an independent linkage of Kentucky-based SEER and Medicaid data. A total 559 484 cancer cases were linked to national Medicaid enrollment ﬁles, representing 15–17% of persons with cancer yearly. About 60% of these cases were a complete match on all variables. There was 99% agreement on enroll- ment status compared with the Kentucky linked data. SEER data were successfully linked to national Medicaid enrollment data. NCI will make the linked data available to researchers, allowing for more detailed assessments of the impact Medicaid enrollment has on cancer diagnosis and outcomes. Medicaid is a federal-state program to make health-care cover- states. Linkage of multiple cancer registries to Medicaid enroll- age available for low-income residents of a state who otherwise ment files for the entire United States has never occurred. The would likely be uninsured. The program is offered to children, development of a reliable and comprehensive approach to link families, pregnant women, the aged, and the disabled, although persons in cancer registries to national Medicaid enrollment the specific eligibility criteria and services covered by Medicaid would facilitate the creation of a data resource that researchers are determined by federal and state criteria. Each state differs could use to assess the impact of Medicaid eligibility on diagno- by eligibility threshold and covered medical services. Medicaid sis and treatment for cancer patients. status is increasingly important as researchers and policy mak- This article reports on a linkage of patients included in the ers attempt to assess the value of Medicaid coverage to low- National Cancer Institute’s (NCI) Surveillance, Epidemiology, income and otherwise uninsured patients. and End Results (SEER) cancer registries to Medicaid enrollment Studies have demonstrated that cancer patients receiving data from all 50 states and the District of Columbia. We describe Medicaid are diagnosed at a more advanced stage, receive less the linkage algorithm used to match patients in the two data guideline care, and have poorer survival (1–8). Prior studies have sources and the process we used to quantify the degree of cer- linked cancer registry to Medicaid data to create data resources tainty that the correct patients were linked. We assessed the to assess the impact of Medicaid eligibility on cancer diagnosis quality of the linkage by comparing the patients identified as and treatment (8–12) . These projects have been limited to single enrolled in Medicaid via our algorithm to patients identified as Received: 19 September 2019; Revised: 19 November 2019; Accepted: 10 December 2019 Published by Oxford University Press 2020. This work is written by US Government employees and is in the public domain in the US. 89 Downloaded from https://academic.oup.com/jncimono/article/2020/55/89/5837298 by DeepDyve user on 24 August 2022 90 | J Natl Cancer Inst Monogr, 2020, Vol. 2020, No. 55 Medicaid enrollees in a separate, independent linkage of on the CMS files (15). In addition, for persons enrolled in both patients in the Kentucky Cancer Registry (KCR) with the Medicare and Medicaid (ie, dual eligibles), the PS file also Kentucky Medicaid enrollment records. includes the person’s unique Medicare number, known as the As a result of this project, NCI will create a much-needed Health Insurance Claim (HIC) number. The HICs for individuals data resource: a file that includes all SEER patients found to be on the PS files are obtained by linking state Medicaid enrollment enrolled in any Medicaid program within a 25-month window of files to Medicare’s enrollment data. PS data were available for time around the date of cancer diagnosis (12 months before, this study from 2006 to 2013 for all states except Kentucky and month of, and 12 months after). The file will include monthly New Mexico, which had data through 2012. For this project, we flags indicating whether the person was eligible for Medicaid attempted to link all persons in the SEER data to Medicaid en- and the reason for entitlement. Monthly Medicaid eligibility will rollment information for all 50 states and the District of be reported for up to three states, accounting for changes in res- Columbia. To provide a clearer description of the linkages and idence for some Medicaid patients with cancer. This file will be the results, we report findings using Medicaid information only available to researchers for approved studies. from the SEER state that reported the patient’s cancer. SEER-Medicare Patient Identifiers and Crosswalk File Methods Persons in the SEER data are linked to Medicare enrollment data biennially using a deterministic match based on personal infor- Data Sources mation (SSN, sex, date of birth) included in both data sources. The primary sources were data from 14 SEER cancer registries The match results in a linkage of more than 93% of persons for persons diagnosed with cancer, 2006–2013, and enrollment aged 65 years or older in the SEER data to their Medicare data, and their HIC number is obtained (16). For persons included in data from Medicaid’s Personal Summary (PS) file for each state from 2006 to 2013. We utilized personal information for SEER the SEER-Medicare data, all SEER case numbers and HIC num- bers are extracted and stored in the SEER-Medicare Crosswalk patients augmented with information from the SEER-Medicare linkage (13). Data from the KCR linked to Kentucky Medicaid en- file. The registries have obtained IRB approval to release to NCI those personal identifiers needed to link the SEER and Medicare rollment data (KCR-KM) were used to assess the quality of the SEER-Medicaid linkage. data. The creation and maintenance of the SEER-Medicare Crosswalk file have been approved by each registry’s IRB and the NCI IRB. SEER Cancer Registry Data The SEER registries, funded by NCI, collect information for all persons with incident cancer occurring within defined geo- KCR Data The KCR has participated in the SEER program since 2000. In graphic areas. The registries available for our analysis included those in California (Northern California, Greater California, and 2015, the KCR undertook a project to link persons with lung, fe- Los Angeles), Connecticut, Georgia, Hawaii, Iowa, Kentucky, male breast, colorectal, pancreatic, ovary, and prostate cancer Louisiana, New Jersey, New Mexico, and Utah, and metropolitan diagnosed in 2000–2011 in the KCR files with health claims, Detroit and Seattle. These registries include 30% of the US popu- 2000–2011, from private and public payors, including Kentucky lation (14). New patients in the SEER registries are assigned a Medicaid enrollment data provided by the Department of unique random case number. Registries collect data about each Medicaid Services. A probabilistic approach was used to link the patient’s demographics, primary tumor site, histology and stage KCR and Medicaid data using SSN, sex, date of birth, first name, at diagnosis, initial cancer treatment, number of prior cancers, last name, and middle name (17,18). All potential matches were and follow-up for vital status. manually reviewed to confirm that they were true matches. The The registries have legal authority to obtain personal infor- final KCR-KM data were limited to cases diagnosed in 2007–2011 mation for each patient in their data. The personal information because these cases were also linked with other claims sources. collected includes a Social Security number (SSN), sex, and date of birth. This information is used to track patients over time Linkage of Persons in the SEER Data to Medicaid PS File and consolidate multiple reports for the same patient. Personal information is not released by the registries to the public but is Two files were created to link persons in the SEER data to the provided to NCI to permit the linkage of the SEER data to Medicaid PS files. The first file, the “SEER Match file,” included Medicare files following approval by each registry’s Institutional cancers occurring 2006–2013. For patients with multiple primary Review Board (IRB). NCI obtained permission from each registry cancers, the first cancer occurring in 2006–2013 was included. to reuse the personal identifiers provided for the SEER-Medicare For each patient, we retained their SEER case number, SSN, sex, linkage for the new match to Medicaid enrollment data. All per- and date of birth (Figure 1). In addition, we used the SEER case sonal information was destroyed after completion of the SEER- number to link to the SEER-Medicare Crosswalk file to obtain Medicare and SEER-Medicaid linkages. the HIC numbers for SEER patients who were also eligible for Medicare. The SEER Match file included only one record per State-Level Medicaid PS Files person. The Medicaid Analytic eXtracts (MAX) Personal Summary (PS) The second file consisted of multiple “Medicaid Match files,” files are extracted from the Medicaid Statistical Information which were created from the PS files for each state and year System (MSIS) and contain information about Medicaid enroll- (2006–2012 or 2013). For each year, persons were identified as a ment submitted from all 50 states and the District of Columbia Medicaid beneficiary if they had at least 1 month of Medicaid to the Center for Medicare and Medicaid Services (CMS). Each enrollment during the year. Persons could appear in more than state’s file includes persons enrolled in Medicaid during a given one Medicaid Match file (eg, more than 1 year and/or state); year, along with their MSIS number, SSN, sex, date of birth, and each Medicaid Match file included the state-specific MSIS iden- category of Medicaid eligibility. Patient name was not available tifier, SSN, HIC (if available), sex, and date of birth (Figure 1). Downloaded from https://academic.oup.com/jncimono/article/2020/55/89/5837298 by DeepDyve user on 24 August 2022 J. L. Warren et al. |91 SEER data MEDICAID data Preparaon of a “SEER Match File” using SEER Preparaon of a “Medicaid Match File” using paent idenﬁers and SEER-Medicare Crosswalk ﬁle the records from each state’s Medicaid PS File For each person in the SEER data diagnosed with For each person, we created an annual ﬁle from cancer 2006–2013, we used paent idenﬁers sent each state’s Medicaid PS File from 2006 through by each registry that include: 2013, or the latest year available. Each year’s ﬁle included a person’s: � SEER case number � SSN � MSIS Idenﬁer � Sex � SSN � Date of birth � HIC number (if available) � Sex Using the SEER case number, we matched to the � Date of birth SEER-Medicare Crosswalk ﬁle to determine which paents were eligible for Medicare and obtain their: � HIC Medicaid Match File SEER Match File Match SEER Match File and Medicaid Match Files using � SSN SEER and Medicaid data had to agree exactly � HIC (if any) on SSN and/or HIC to be considered a match � Sex � Date of birth Calculate match score (see Figure 2) Figure 1. Process of linking persons in the Surveillance, Epidemiology, and End Results (SEER) data diagnosed with cancer, 2007 to 2013, to persons in each State’s Medicaid Personal Summary (PS) ﬁle. HIC ¼ Health Insurance Claim; MSIS ¼ Medicaid Statistical Information System; SSN ¼ Social Security number. We used a deterministic matching process. To be considered required with the SSN match because the SSN on the file may a match, a patient’s HIC or SSN, as listed in the SEER Match file, be that of the spouse rather than the Medicaid recipient. Unlike had to exactly match a record in one of the Medicaid Match the SSN, the HIC, by design, includes information about files. If a patient’s HIC or SSN was not matched to any of the whether the individual is the husband or wife. An exact match Medicaid Match files, they were considered not enrolled in on SSN but without agreement on sex was given a lower value Medicaid. The number of matched patients was tallied by SEER (three points). A match on sex was given one point. For date of registry and year. The percentage of patients that matched was birth, we used a hierarchical approach: exact match on (day/ month/year, three points), month and year (two points), or year determined by dividing the number of matched patients by the total number of patients for each registry by year. only (one point). The points were summed; the maximum pos- To quantify the strength of the match, we assigned points sible score was 12, the minimum was 3 (all cancer cases were for the agreement between the SEER and the Medicaid data on required to match at least on SSN to be included). We expected three tiers of variables: 1) HIC and SSN, 2) sex, and 3) date of that the accuracy of personal information on the Medicaid data birth (Figure 2). We first assessed the match on HIC and/or SSN. could vary over multiple years of within-state enrollment. If there was an exact match on all digits in both the HIC and Therefore, we reported the patient’s highest state-specific SSN, eight points were given. Four points were given if there match score in any year and applied that score to all years. For was an exact match on HIC, but not the SSN or if there was a ease of reporting, we consolidated match scores into four match on SSN and sex but not on the HIC. A match on sex was groups: complete match (score: 12), strong match (score: 10–11), Downloaded from https://academic.oup.com/jncimono/article/2020/55/89/5837298 by DeepDyve user on 24 August 2022 92 | J Natl Cancer Inst Monogr, 2020, Vol. 2020, No. 55 where the percent increased from 10.9% in 2006 to 21.5% in SEER Match File and Medicaid Match File agreed on: Points 2013, resulting from the state’s early adoption of Medicaid HIC number and SSN 8 expansion (19). OR Table 2 reports the distribution of match scores by SEER reg- HIC number only or (SSN and sex) 4 istry. In every registry except Greater California, approximately OR 60% of the SEER patients who were linked to Medicaid data had a complete match on all variables, with almost all remaining SSN only 3 matches having a moderate score. For all registries, except Sex 1 Greater California, less than 2% of all matches had a weak Exact DOB: day, month, and year 3 match score. In the Greater California registry, almost no per- OR sons had a complete match on variables. We assessed the Year and month of birth 2 reporting of specific variables in Greater California and found OR that 1% of persons matched on day of birth. The match rate on Year of birth 1 month and year of birth (not specific day) exceeded 97% for per- Sum of all points - maximum possible score sons in the Greater California data. The data file from the (HIC and SSN) + sex + exact DOB Greater California registry may have had an anomaly, but we could not effectively assess the data because, per the agreement Figure 2. Scoring weights assigned to each variable used to calculate the with the SEER registries, NCI destroyed the file with the per- strength of the match between persons in the SEER data and the Medicaid sonal information from each of the registries after the linkage Personal Summary ﬁle. DOB ¼ date of birth; HIC ¼ Health Insurance Claim; SSN was completed. ¼ Social Security number. In the comparison of the SEER-Medicaid data from Kentucky with the independent KCR-KM linkage, there were approxi- moderate match (score: 7–9), and weak match (score: 6 or less). mately 14 000 KCR patients diagnosed with incident lung, These scores are reported by cancer registry. breast, colorectal, pancreatic, ovary, and prostate cases annu- ally (Table 3). For each year, the SEER-Medicaid and the KCR-KM Assessment of the Linkage linkages agreed that approximately 20% of patients were eligible We assessed the quality of our linkage process by comparing a for Medicaid and 79% of patients were not eligible, a percent subset of cancer patients from the Kentucky registry included in agreement in excess of 99% per year. The two data sources had our SEER-Medicaid linkage to patients included in the indepen- 97% or greater agreement on the number of months patients dently linked KCR-KM data, as described above. Because the were eligible each year of diagnosis. KCR is part of the SEER program, the Kentucky patients in the SEER-Medicaid and the KCR-KM data had the same SEER case numbers, allowing us to directly match patients with incident Discussion diagnoses in the two data sources for the years of overlapping Our analysis matched 3 493 820 incident cancer cases in the data, 2007–2011. The SEER-Medicaid data were limited to per- SEER data to national Medicaid enrollment files and identified sons with lung, female breast, colorectal, pancreatic, ovary, and 559 484 cases that were eligible for Medicaid. The percent of prostate cancer cases to correspond with the cancers included cancer patients that were matched to Medicaid in our study, 15– in the KCR-KM data. The level of agreement about Medicaid en- 17% per year, is slightly higher than national Medicaid enroll- rollment between the SEER-Medicaid and KCR-KM data was de- ment rates of 11–13% reported between 2008 and 2013 (20). The termined for the year of the cancer diagnosis and classified in higher rates of Medicaid enrollment that we observed may be four categories: both sources agreed that the person was en- the result of newly diagnosed cancer patients enrolling in rolled in Kentucky Medicaid; both sources agreed that the per- Medicaid during the peri-diagnostic period, as reported in prior son was not enrolled in Kentucky Medicaid; or only one source, studies (2,21). In our data, the state ranking by percent of cancer KCR-KM or SEER-Medicaid, reported enrollment in Kentucky patients enrolled in Medicaid was consistent with national Medicaid. In both data sources, patients were classified as reports for those years (20). Medicaid enrolled if they were enrolled for at least 1 month of We required that to be considered a match, the SEER data the year. In addition to determining if there was agreement be- and Medicaid files must agree exactly on the patient’s HIC or tween the SEER-Medicaid and the KCR-KM data as to Medicaid SSN. Prior linkages of cancer registries to state Medicaid data enrollment, we also assessed the agreement between the two have used deterministic and probabilistic approaches using sources on the number of months that the cancer patients were identifiers such as SSN, patient name, sex, and date of birth (8– enrolled in Medicaid. 11). One study of Michigan cancer patients age 65 years and older also included a HIC number in the matching process (12). Likewise, we included the HIC number as a match variable. We Results believe that HIC number is a strong match variable because it is There were more than 420 000 patients with incident cancer unique to each individual and is used to process Medicare reported each year by the SEER registries between 2006 and 2013 claims. A HIC number was found in both the SEER Match file (Table 1). Of these patients, 15–17% were matched to Medicaid and the Medicaid Match files for 67.5% of cancer patients in our each year. The percent of cancer patients enrolled in Medicaid data. This means that most SEER cases matched to the varied by registry, with Louisiana having the highest percent of Medicaid data were dual-eligible, reflecting the fact that cancer patients with over 20%. Utah had the lowest percent of cancer is a disease of the elderly. Except for the Greater California reg- patients enrolled in Medicaid at 10% or less each year. The per- istry, more than 90% of SEER patients matched to Medicaid had cent of cancer patients enrolled in Medicaid in each state a score that was a complete, strong, or moderate match. The in- remained stable from 2006 to 2012 or 2013 except Connecticut, clusion of additional variables used to calculate match scores Downloaded from https://academic.oup.com/jncimono/article/2020/55/89/5837298 by DeepDyve user on 24 August 2022 J. L. Warren et al. |93 Table 1. Number of people with incident cancers in the 2006–2013 SEER data* and percent who matched the MAX PS ﬁle during the year follow- ing cancer diagnosis by SEER registry and year 2006 2007 2008 2009 2010 2011 2012 2013 SEER registry No. (%) No. (%) No. (%) No. (%) No. (%) No. (%) No. (%) No. (%) California Los Angeles 39 727 (23.9) 41 031 (23.8) 40 458 (23.8) 40 621 (24.4) 40 011 (23.2) 39 414 (24.1) 38 356 (23.4) 38 531 (23.2) Northern California 32 375 (15.5) 33 050 (15.5) 33 126 (15.4) 33 513 (16.1) 33 532 (15.8) 33 015 (16.2) 32 818 (16.6) 32 540 (16.0) Greater California 90 049 (15.6) 92 350 (15.6) 93 445 (16.5) 93 296 (16.7) 92 820 (16.5) 91 948 (17.5) 91 739 (17.6) 92 337 (17.8) Connecticut 22 348 (10.9) 22 047 (11.6) 21 770 (11.7) 21 972 (12.0) 21 408 (17.3) 20 928 (19.4) 20 624 (21.0) 20 406 (21.5) Detroit 23 554 (14.2) 23 873 (13.3) 23 297 (14.3) 23 280 (15.2) 22 884 (16.4) 23 253 (16.6) 22 218 (16.4) 22 042 (15.5) Georgia 42 620 (15.6) 44 801 (15.6) 45 233 (16.9) 45 700 (17.2) 44 775 (16.5) 46 236 (16.9) 47 256 (17.6) 46 773 (17.0) Hawaii 6730 (13.2) 6837 (12.7) 7038 (13.4) 7030 (14.1) 6903 (14.81) 6834 (15.4) 6830 (15.4) 6996 (15.5) Iowa 176 57 (11.9) 17 450 (12.4) 17 464 (13.0) 17 911 (13.1) 17 594 (13.2) 17 481 (13.9) 16 983 (14.8) 16 753 (14.3) Kentucky 26 413 (19.8) 27 078 (19.7) 27 359 (19.9) 27 680 (20.1) 28 427 (20.3) 28 685 (20.4) 29 481 (19.6) 29 432† Louisiana 23 506 (22.5) 24 469 (22.8) 24 955 (23.0) 25 972 (23.5) 25 869 (24.1) 26 660 (24.9) 26 699 (24.4) 26 823 (23.2) New Jersey 54 328 (9.5) 54 943 (9.3) 55 077 (9.7) 55 840 (9.8) 55 438 (10.1) 56 296 (10.1) 56 252 (10.5) 57 213 (10.8) New Mexico 9245 (14.1) 9678 (14.5) 9910 (15.4) 9985 (18.0) 9992 (17.8) 9962 (17.6) 9880 (18.7) 9893† Seattle 25 217 (10.9) 26 744 (10.7) 27 031 (10.9) 27 976 (11.7) 28 094 (12.6) 28 840 (12.5) 28 570 (12.3) 29 119 (12.6) Utah 9456 (8.7) 9763 (8.43) 10 231 (8.7) 10 465 (8.7) 10 886 (9.0) 11 228 (9.5) 11 332 (10.0) 11 537 (9.6) Total 423 225 (15.2) 434 114 (15.2) 436 394 (15.7) 441 241 (16.2) 438 633 (16.4) 440 780 (17.0) 439 038 (17.1) 440 395 (15.2) *For patients with multiple primary cancers, includes the ﬁrst cancer diagnosed during 2006–2013. MAX ¼ Medicaid Analytic eXtracts; PS ¼ Personal Summary; SEER ¼ Surveillance, Epidemiology, and End Results. †2013 PS ﬁle data not available. Table 2. Number of cancer patients in the SEER data who matched in the Medicaid Enrollment Data and the score for the strength of the match by individual cancer registry* Medicaid match score (12¼ highest; 1¼ lowest) Complete match Strong match Moderate match Weak match Score 12 10–11 7–9 6 SEER registry % % % % Total cases, No. California Los Angeles 63.7 1.8 32.4 2.1 75 462 Northern California 65.5 1.2 31.4 2.0 41 976 Greater California 0.7 62.9 2.8 33.7 123 368 Connecticut 68.5 1.4 29.1 1.0 26 678 Detroit 59.8 1.2 37.5 1.5 28 067 Georgia 62.1 2.2 33.9 1.8 60 610 Hawaii 58.8 0.7 39.6 1.0 7901 Iowa 65.2 0.8 33.5 0.4 18 525 Kentucky 66.7 2.0 29.7 1.6 38 946 Louisiana 63.0 2.1 33.4 1.5 48 298 New Jersey 65.9 2.2 29.9 2.0 44 396 New Mexico 58.4 1.7 38.0 2.0 11 398 Seattle 64.4 0.6 34.2 0.9 26 122 Utah 57.5 0.6 40.9 1.1 7737 *SEER ¼ Surveillance, Epidemiology, and End Results. Table 3. Agreement between SEER-Medicaid and KCR-KM data on the percent of cancer patients who were eligible for Medicaid and months of Medicaid enrollment, 2007–2011* Agreement on Medicaid eligibility Agreement on months enrolled Ineligible in Eligible in In SEER-Medicaid In KCR-KM Percent Percent No. both files both files data only data only agreement No. agreement 2007 14 130 79.4 20.4 0.1 0.1 99.8 2885 97.5 2008 14 030 78.5 21.1 0.1 0.2 99.6 2963 98.2 2009 14 024 79.2 20.5 0.1 0.2 99.7 2869 98.1 2010 14 099 77.9 21.6 0.1 0.4 99.5 3050 96.7 2011 14 200 78.3 21.0 0.3 0.4 99.3 2983 98.3 *KCR-KM ¼ Kentucky Cancer Registry-Kentucky Medicaid; SEER ¼ Surveillance, Epidemiology, and End Results. Downloaded from https://academic.oup.com/jncimono/article/2020/55/89/5837298 by DeepDyve user on 24 August 2022 94 | J Natl Cancer Inst Monogr, 2020, Vol. 2020, No. 55 allows researchers to assess the quality of the match and per- available to researchers once the data release process has been finalized. form sensitivity analyses, including and excluding patients with lower match scores. This study has several important strengths. This is the first Notes linkage of all patients in the SEER data to Medicaid enrollment information from all states. The state-based data provide a Affiliations of authors: Healthcare Assessment Research unique opportunity to compare differences in cancer stage and Branch, Healthcare Delivery Research Program, Division of survival based on varying Medicaid coverage across individual Cancer Control and Population Science, National Cancer states. The quality of the SEER-Medicaid linkage was confirmed Institute, Bethesda, MD (JLW, LE); Fu Associates, Ltd, Arlington, by comparison with an independent linkage performed by the VA (SB); Information Management Services, Calverton, MD (JS); KCR. Finally, NCI’s creation of a file that includes Medicaid en- Department of Biostatistics, College of Public Health, Markey rollment information for cancer patients eliminates the admin- Cancer Center, University of Kentucky, Lexington, KY (BH); istrative and logistical challenges that researchers encounter Division of Data, Research, and Analytic Methods, Center for when linking cancer registry data to Medicaid files. Medicare & Medicaid Innovation, Centers for Medicare and There are also several limitations to our study. The data in Medicaid Services, Baltimore, MD (LZ); Department of Health our analysis are from 2006 to 2013, before the Affordable Care Systems, Management and Policy, School of Public Health, Act’s 2014 provisions for Medicaid expansion were adopted (22). University of Colorado, Aurora, CO (CJB). The data we used were the most recent national Medicaid en- The contents of this article are solely the responsibility of rollment data available from CMS at the time of this linkage. In the authors and do not necessarily represent the official views 2013, CMS began to transition the Medicaid data to the of the US Department of Health and Human Services or any of Transformed MSIS (T-MSIS); however, data quality challenges its agencies. The authors have no disclosures. have delayed T-MSIS implementation (23). We believe that the approaches we developed to link persons in the SEER data to References the Medicaid PS file could be useful for more current SEER- Medicaid linkages once the T-MSIS system is fully operational. 1. Bradley CJ, Given CW, Roberts C. Late stage cancers in a Medicaid-insured population. Med Care. 2003;41(6):722–728. NCI is committed to updating the linkage of SEER patients to 2. Bradley CJ, Gardiner J, Given CW, Roberts C. Cancer, Medicaid enrollment, more recent Medicaid enrollment information. An additional and survival disparities. Cancer. 2005;103(8):1712–1718. limitation is that we only evaluated the linkage of cancer 3. Halpern MT, Ward EM, Pavluck AL, Schrag NM, Bian J, Chen AY. Association of insurance status and ethnicity with cancer stage at diagnosis for 12 cancer patients to the Medicaid data for the state where they lived dur- 1sites: a retrospective analysis. Lancet Oncol. 2008;9(3):222–231. ing the year of their cancer diagnosis. We did not assess the 4. Han X, Zhu S, Tian Y, Kohler BA, Jemal A, Ward E. Insurance status and can- cer stage at diagnosis prior to the Affordable Care Act in the United States. J quality of the linkage for patients who moved to another state Registry Manag. 2016;41(3):143–151. after their cancer diagnosis. Although the SEER registries in- 5. Hassett MJ, Schymura MJ, Chen K, Boscoe FP, Gesten FC, Schrag D. Variation clude almost one-third of the US population, states where regis- in breast cancer care quality in New York and California based on race/eth- tries are not part of the SEER program are not included in the nicity and Medicaid enrollment. Cancer. 2016;122(3):420–431. 6. Snyder JW, Foley KL. Disparities in colorectal cancer stage of diagnosis linkage. This limits the generalizability of analyses using the among Medicaid-insured residents of North Carolina. N C Med J. 2010;71(3): linked SEER-Medicaid enrollment data, especially for states that 206–212. have not expanded their Medicaid coverage. Finally, the focus of 7. Subramanian S, Chen A. Treatment patterns and survival among low- income Medicaid patients with head and neck cancer. JAMA Otolaryngol Head our project was to develop a method to link SEER cancer Neck Surg. 2013;139(5):489–495. patients to Medicaid enrollment information. We opted not to 8. Perkins CI, Wright WE, Allen M, Samuels SJ, Romano PS. Breast cancer stage at diagnosis in relation to duration of Medicaid enrollment. Med Care. 2001; obtain Medicaid claims for health-care utilization because the 39(11):1224–1233. MAX claims are complex, with variable coverage and quality by 9. Boscoe FP, Schrag D, Chen K, Roohan PJ, Schymura MJ. Building capacity to state and year. The MAX claims are to be replaced with the T- assess cancer care in the Medicaid population in New York state. Health Serv MSIS Analytic File (TAF) data. CMS announced on November 7, Res. 2011;46(3):805–820. 10. Koroukian SM. Linking the Ohio Cancer Incidence Surveillance System with 2019, that the TAF research identifiable files will be made avail- Medicare, Medicaid, and clinical data from Home Health Care and long term able to researchers (24). With the availability of these data, NCI care assessment instruments: paving the way for new research endeavors in will assess the possibility of obtaining Medicaid claims for SEER geriatric oncology. J Registry Manag. 2008;35(4):156–165. 11. Nadpara PA, Madhavan SS. Linking Medicare, Medicaid, and cancer registry patients. However, these files have different data structures and data to study the burden of cancers in West Virginia. Medicare Medicaid Res contents. The quality of these claims would need to be assessed Rev. 2012;2(4):mmrr.002.04.a01. 12. Bradley CJ, Given CW, Luo Z, Roberts C, Copeland G, Virnig BA. Medicaid, carefully, either by NCI or researchers, because prior studies Medicare, and the Michigan Tumor Registry: a linkage strategy. Med Decis have reported considerable issues in the completeness of Making. 2007;27(4):352–363. Medicaid claims data (25–27). 13. Warren JL, Klabunde CN, Schrag D, Bach PB, Riley GF. Overview of the SEER- In conclusion, this study developed and tested an approach Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care. 2002;40(8 suppl):3–18. to link persons in multiple SEER cancer registries to national 14. Howlader N, Noone AM, Krapcho M, et al. eds. SEER Cancer Statistics Review, Medicaid enrollment data to identify those cancer patients who 1975–2016, Bethesda, MD: National Cancer Institute. https://seer.cancer.gov/ were Medicaid enrollees. Knowing which patients are enrolled csr/1975_2016/, based on November 2018 SEER data submission, posted to the SEER web site, April 2019. in Medicaid and the timing of their Medicaid enrollment rela- 15. Chronic Condition Warehouse. Medicaid Analytic eXtract Files (MAX) User tive to their cancer diagnosis will allow researchers to under- Guide, Version 2.2 https://www.ccwdata.org/documents/10280/19002246/ccw- take studies across states related to stage and outcomes for max-user-guide.pdf. Published January 2019. Accessed June 19, 2019. 16. National Cancer Institute. Division of Cancer Control and Population some of the most vulnerable cancer patients. NCI will create a Sciences. SEER-Medicare: how the SEER & Medicare data are linked. https:// file for all persons in the SEER data that are matched to healthcaredelivery.cancer.gov/seermedicare/overview/linked.html. Last Medicaid in any state or multiple states. This file will be made updated May 16, 2019. Accessed June 19, 2019. Downloaded from https://academic.oup.com/jncimono/article/2020/55/89/5837298 by DeepDyve user on 24 August 2022 J. L. Warren et al. |95 17. Gallaway MS, Huang B, Chen Q, et al. Identifying smoking status and smok- 23. Center for Medicare and Medicaid Services. Letter to state health ofﬁcials ing cessation using a data linkage between the Kentucky Cancer Registry and about the T-MIS system. https://www.medicaid.gov/Federal-Policy- Health Claims Data. J Clin Oncol Clin Cancer Inform. 2019;3:1–8. Guidance/downloads/SHO18008.pdf. Published 2018. Accessed June 19, 2019. 18. Gallaway MS, Huang B, Chen Q, et al. Smoking and smoking cessation among 24. Center for Medicare and Medicaid Services. CMS takes historic steps to in- persons with tobacco- and non-tobacco-associated cancers. J Community crease public access to Medicaid and CHIP data. https://www.cms.gov/news Health. 2019;44(3):552–560. room/press-releases/cms-takes-historic-steps-increase-public-access-med 19. Healthinsurance.org. Connecticut and the ACA’s Medicaid expansion. icaid-and-chip-data. Accessed November 7, 2019. https://www.healthinsurance.org/connecticut-medicaid/. Published 25. Schrag D, Virnig BA, Warren JL. Linking tumor registry and Medicaid claims November 13, 2018. Accessed July 3, 2019. to evaluate cancer care delivery. Health Care Financ Rev. 2009;30(4):61–73. 20. Kaiser Family Foundation. Health Insurance Coverage of the Total 26. Byrd VL, Dodd AH. Assessing the Usability of Encounter Data for Enrollees in Population. 2008-2013. https://www.kff.org/other/state-indicator/total-popu Comprehensive Managed Care Across MAX 2007–2009. Washington, DC: lation/. Accessed June 19, 2019. Centers for Medicare & Medicaid Services, December 2012a. https://www. 21. Ramsey SD, Zeliadt SB, Richardson LC, et al. Disenrollment from Medicaid af- cms.gov/Research-Statistics-Data-and-Systems/Computer-Data-and-Systems/ ter recent cancer diagnosis. Med Care. 2008;46(1):49–57. MedicaidDataSourcesGenInfo/Downloads/MAX_IB_15_AssessingUsability.pdf 22. Antonisse L, Garﬁeld R, Rudowitz R, Artiga S. The effects of Medicaid expan- Accessed June 19, 2019. sion under the ACA: updated ﬁndings from a literature review. Kaiser Family 27. Byrd VL, Dodd AH. Assessing the usability of MAX 2008 encounter data for Foundation. https://www.kff.org/medicaid/issue-brief/the-effects-of-medicaid- comprehensive managed care. Medicare Medicaid Res Rev. 2013;3(1): expansion-under-the-aca-updated-ﬁndings-from-a-literature-review-march- mmrr.003.01.b01. 2018/. Published 2018. Accessed June 19, 2019.
JNCI Monographs – Oxford University Press
Published: May 1, 2020
Keywords: cancer; medicaid; seer program; cancer registries; medicare; kentucky; national cancer institute
Access the full text.
Sign up today, get DeepDyve free for 14 days.