Background: Many veterans have dual Veterans Administration (VA) and Medicare healthcare coverage. We compared 3-year overall and cancer event-free survival (EFS) among patients with nonmetastatic colon cancer who obtained substantial portions of their care in both systems and those whose care was obtained predominantly in the VA or in the Medicare fee-for-service system.

Methods: We conducted a retrospective observational cohort study of patients older than 65 years with stages I to III colon cancer diagnosed from 1999 to 2001 in VA and non-VA facilities. Dual use of VA and non-VA colon cancer care was categorized as predominantly VA use, dual use, or predominantly non-VA use. Extended Cox regression models evaluated associations between survival and dual use.

Results: VA and non-VA users (all stages) had reduced hazard of dying compared with dual users [e.g., for stage I, VA HR 0.40, 95% confidence interval (CI): 0.28–0.56; non-VA HR 0.54, 95% CI: 0.38–0.78). For EFS, stage I findings were similar (VA HR 0.47, 95% CI: 0.35–0.62; non-VA HR 0.64, 95% CI: 0.47–0.86). Stage II and III VA users, but not non-VA users, had improved EFS (stage II: VA HR 0.74, 95% CI: 0.56–0.97; non-VA HR 0.92, 95% CI: 0.69–1.22; stage III: VA HR 0.73, 95% CI: 0.56–0.94; non-VA HR 0.81, 95% CI: 0.62–1.06).

Conclusions: Improved survival among VA and non-VA compared with dual users raises questions about coordination of care and unmet needs.

Impact: Additional study is needed to understand why these differences exist, why patients use both systems, and how systems may be improved to yield better outcomes in this population. Cancer Epidemiol Biomarkers Prev; 21(12); 2231–41. ©2012 AACR.

Most US military veterans older than 65 who use Veterans Administration (VA) healthcare services are also enrolled in the Medicare program. The majority of those use both VA and Medicare benefits (1, 2). A growing body of research has documented various aspects of dual VA-Medicare use including patient populations more likely to seek care in both systems, the types of services sought by patients using both systems, and geographic patterns of dual use (1, 3–9). However, relatively little is known about the comparative health outcomes of veterans who receive their care predominantly or entirely in one system—VA or the private sector—and those who receive substantial portions of their care in both systems (10–14).

For cancer patients, the potential consequences of dual care may be particularly hazardous. The challenges that cancer patients and their families face in navigating a complex healthcare system to access timely and appropriate care are well recognized (15–17). There are multiple barriers to coordination of care across settings and few automated systems established for doing so. Particularly for older patients with complex chronic care needs, this lack of coordination could result in delays in care, gaps in care, and duplication of services.

Whether dually covered patients who obtain their cancer care in both the VA and the private sector have different outcomes than those who remain predominantly in one system is not known. In this study, we compared 3-year survival among patients with nonmetastatic colon cancer who obtained substantial portions of their care in both systems and those whose care was obtained predominantly in the VA or in the private sector. We examined overall survival and colon cancer event-free survival (EFS) separately for each cancer stage, I to III.

Study design and population

This study used a retrospective observational cohort design. The study sample comprised veterans with stages I to III colon cancer identified from VA and National Cancer Institute Surveillance Epidemiology and End Results (SEER) Program registry records (18). VA Central Cancer Registry (VACCR) cases were reported from VA facilities nationwide, whereas SEER-reported cases resided in 1 of 8 registry coverage areas with which we had data-sharing agreements (California, Metropolitan Detroit, Georgia, Hawaii, Iowa, Louisiana, New Jersey, and Western Washington). Patients diagnosed with colon cancer between July 1, 1999 and December 31, 2001 at 66 years of age or older who were both Medicare-enrolled and eligible for VA care, and who lived at least 30 days after diagnosis, were eligible for inclusion. We excluded cases for which we had no or incomplete healthcare usage data (principally owing to Medicare HMO enrollment, Part B Medicare coverage only, or non-Medicare primary payer coverage). We also excluded patients with stage I cancer who received chemotherapy and stage II and III cancer patients who did not undergo a colectomy as those patients were likely to have differed from their counterparts in unobservable characteristics that may have had a bearing on survival.

Data sources and construction of variables

We used data from the VA, SEER program registries, the Centers for Medicare and Medicaid Services (19), and the 2000 US Census to ascertain cancer clinical characteristics, healthcare services, comorbidities, demographic information, and dates of death. The VACCR receives registered case reports from certified cancer registrars at VA medical centers, which adhere to standards of the Commission on Cancer (20). The regional SEER registries are responsible for complete case ascertainment for patients residing in their coverage areas. They adhere to rigorous SEER standards and all received North American Association of Central Cancer Registries silver or gold certification for 1999 to 2001 incidence data (20–22). The VACCR and SEER data provided information on cancer date of diagnosis, anatomic site, staging elements and AJCC-5 stage (23), and marital status.

We obtained information on chemo- and radiotherapy administration, healthcare usage, and comorbidities from national VA databases, which comprise extracts from the VA's electronic health record system, and Medicare claims (24–26). The combination of registry, medical records, and claims data provides valid and more complete ascertainment of cancer treatment over any one source alone (27–30). We obtained demographics (age, gender, and race) from the Medicare Vital Status File (19, 31).

The study endpoint for overall survival (OS) was death from any cause. For EFS, the endpoint was a colon event signaling cancer progression or death from any cause. We observed patients for 36 months; those who survived without a colon event were censored. We ascertained vital status from the VHA Vital Status File, a central source for veterans' date of death information from VA, Medicare, and Social Security Administration databases (32). The absence of a death date was accepted as evidence of survival to the 3-year endpoint.

To identify colon events, we began by defining a colon surveillance period. We defined the beginning of the colon surveillance period differently depending on initial treatment, that is, none; colectomy alone; or colectomy plus chemo- or radiotherapy. (Codes used to identify colon cancer treatment are provided in the Supplementary Data). For patients who received no initial treatment or colectomy alone, the surveillance period began 12 months from the diagnosis date. For patients who had chemo- or radiotherapy in addition to a colectomy, the surveillance period began at the later of 2 time points: 12 months from the diagnosis date or 4 months after the discontinuation of therapy. Chemo- and radiotherapy discontinuation dates were defined as the date of service on the last healthcare record with a relevant code after which there were no records with those codes for at least 30 (for chemotherapy) or 14 (for radiotherapy) days. We defined a colon event as any of the following, when it occurred during the colon surveillance period: initiation of chemotherapy or radiotherapy; a secondary malignant neoplasm diagnosis; and a lung, liver, or colon resection. We identified these events in VA pharmacy and encounter records and in Medicare claims using ICD-9 diagnosis, ICD-9 procedure, Current Procedural Terminology (CPT-4), Healthcare Common Procedure Coding System, revenue center, and diagnostic-related group codes (see Supplementary Data; ref. 33). The earliest of either the date of service on a healthcare record that satisfied an event definition or a date of death was taken as the failure event.

Our independent variable of primary interest was dual use of cancer care. The following services were considered colon cancer care: a colectomy; chemotherapy; and an acute inpatient stay or outpatient evaluation-and-management visit for which a diagnosis code for colon cancer was entered. We constructed a time-varying measure of percentage of all colon cancer care received in VA from the date of diagnosis to a colon event, death, or 36 months. We calculated the cumulative percentage of care received in the VA on each day of the follow-up period. We then created a 3-level categorical variable: 86% to 100% VA use (predominantly VA user), 15% to 85% VA use (dual user), and 0% to 14% VA use (predominantly non-VA user). Each patient was assigned to a category based on the cumulative proportion of VA use on the date of an event (colon event or death) or, if none, the end of the observation period (36 months). The proportion of patients' colon cancer care received in VA averaged 98.9% among predominantly VA users, 50.6% among dual users, and 1.3% among predominantly non-VA users.

Covariates included tumor grade; number of lymph nodes examined; colectomy (stage I) and chemotherapy (stages II and III) received as part of initial treatment plan (defined for colectomy as occurring in the data 1 month before or not later than 6 months after the diagnosis date, and for chemotherapy as initiated not later than 9 months after the diagnosis date); comorbidities; distance to nearest VA outpatient facility; geographic region; and demographic characteristics. Specific data sources and codes used to ascertain cancer treatment and comorbidities are shown in the Supplementary Data. To establish each case's baseline health status and patterns of healthcare usage, we ascertained comorbidities during the 6 months before diagnosis. We calculated a modified Deyo–Charlson comorbidity score with Romano and Klabunde adaptations and constructed 4 categories: 0 (no comorbidities), 1, 2 or 3, and 4 or higher (34–37).

To account for the many ways in which treatment approaches and preferences differ across geographic regions (38, 39), we identified the US Census Division in which each patient resided at the time of diagnosis but, because of sparse numbers, collapsed the Pacific and Mountain Divisions and Middle Atlantic and New England Divisions to the West and Northeast Census Regions, respectively. Patient demographic characteristics included age at diagnosis, gender, race (40, 41), ethnicity, marital status, and a census-based measure of educational attainment (census tract proportion with a high school diploma or more education among adults age 65 and older in the patient's gender group) as a proxy for socioeconomic resources.

Statistical analysis

Analyses were conducted separately for each cancer stage. Frequency distributions of patient characteristics were computed for each dual use category and the association of each characteristic with dual use was evaluated using χ2 or Fisher's exact tests for categorical variables and t tests for continuous variables. We plotted Kaplan–Meier survival curves and used log rank tests to compare the survival distributions of the dual use groups over the 3-year follow-up period. We used separate extended Cox models to evaluate associations between days to event (OS and EFS) and dual use. Models controlled for all covariates described above. We evaluated the sensitivity of our estimates of the impact of dual use on survival to a different specification of our dual use measure. Specifically, rather than assigning dual use category based on the cumulative percentage of VA use on the final day of observation (i.e., date of death, colon event, or 36 months) as we did initially, we based the assignment on the cumulative percentage of VA use as of 30 days before the final observation day. In doing so, we reduced the potential influence of reverse causation, that is, declining health leading to dual use (e.g., a dying VA user who in the last month of life needed hospice care close to home and therefore sought that care through Medicare). Proportional hazards assumptions were evaluated using log–log survival curves, observed Kaplan–Meier survival curves versus predicted Cox survival curves, Goodness-of-Fit tests (Cox–Snell residuals, Schoenfeld Residuals), and time-dependent variables (42). We judged statistical significance at the P < 0.05 level and report HR and 95% confidence intervals (CI). All tests were 2-sided. Statistical analyses were conducted using Stata MP software version 12.1. Our protocol was approved by the Hines VA Hospital Institutional Review Board and granted a waiver of informed consent.

We identified 3,014 cancer patients meeting study inclusion criteria, 1,855 from VACCR records and 1,472 in SEER registry records (313 patients had records in both sources).

Cohort characteristics

Stage I (n = 1,028), II (n = 1,129), and III (n = 857) sample characteristics broken down by dual use category are shown in Table 1. Among patients with stage I, the dual and predominantly VA user groups had similar age distributions but dual users were older than predominantly non-VA users (≥76 years: Dual 45.0% vs. VA 44.3% and Dual 45.0% vs. non-VA 61.0%). Dual users were substantially less likely than VA users but twice as likely as non-VA users to be African American (14.1% vs. 20.6% and 14.1% vs. 7.0%). Dual users were slightly less likely than VA and more likely than non-VA users to have had 5 or fewer regional lymph nodes examined (52.1% vs. 55.5% and 52.1% vs. 46.3%). More dual users than either VA or non-VA users had coexisting chronic conditions (comorbidity score ≥1: 48.8% vs. 41.7% and 48.8% vs. 44.3%). Stage I dual use groups did not differ in tumor grade, proportion who underwent colectomy, or distance to VA outpatient center.

Table 1.

Patient characteristics

Stage IStage IIStage III
Predominantly VADual userPredominantly non-VAPredominantly VADual userPredominantly non-VAPredominantly VADual userPredominantly non-VA
(n = 528)(n = 213)(n = 287)(n = 534)(n = 174)(n = 421)(n = 403)(n = 120)(n = 334)
%%%Pa%%%Pa%%%Pa
Age at diagnosis  <0.01    <0.01    <0.01 
  Mean (Std.) 74.9 (5.9) 75.1 (5.1) 76.7 (5.5)  75.1 (5.6) 76.9 (5.7) 76.3 (5.5)  74.1 (5.5) 76.1 (6.4) 76.3 (5.9)  
 66–75 55.7 54.9 39.0  53.6 38.5 44.9  61.3 41.7 43.1  
 76–85 39.0 42.7 55.1  43.3 55.2 50.4  35.7 48.3 50.6  
 ≥ 86 5.3 2.3 5.9  3.2 6.3 4.8  3.0 10.0 6.3  
Gender  <0.01    <0.01    <0.01 
 Male 98.7 98.1 93.0  97.9 98.3 92.6  99.0 95.8 93.1  
Race  <0.01    0.04    0.24 
 African American 20.6 14.1 7.0  16.7 14.4 10.9  15.1 18.3 12.3  
Marital status  <0.01    <0.01    <0.01 
 Not married 38.4 31.9 26.8  44.2 41.4 32.5  38.0 49.2 28.4  
 Married 54.9 54.0 69.7  55.4 55.7 66.0  61.5 50.8 68.3  
 Unknown 6.6 14.1 3.5  0.4 2.9 1.4  0.5 0.0 3.3  
High school graduation (area rate)  <0.01    <0.01    <0.01 
 Mean Percent (Std.) 58.8 (16.9) 64.0 (17.5) 67.8 (18.7)  59.6 (17.5) 62.2 (17.6) 66.7 (17.6)  60.8 (17.3) 62.9 (16.8) 67.5 (18.8)  
Tumor grade  0.06    0.04    0.32 
 Grade I 17.6 14.6 16.7  8.1 12.6 9.7  6.9 5.0 5.1  
 Grade II 59.1 61.0 63.4  75.8 70.7 69.1  70.0 71.7 67.7  
 Grade III 4.4 3.8 7.7  11.2 12.6 16.9  17.1 20.8 23.4  
 Grade IV 0.2 0.5 0.0  0.2 0.0 1.2  0.7 0.8 0.9  
 Unknown 18.8 20.2 12.2  4.7 4.0 3.1  5.2 1.7 3.0  
Regional lymph nodes examined    0.01    0.25    0.18 
 0–5 55.5 52.1 46.3  24.7 20.1 23.8  14.6 25.0 14.7  
 6–11 22.7 23.0 29.6  33.5 36.8 34.0  38.0 32.5 39.8  
 12–90 20.6 19.7 20.2  40.4 40.8 38.5  46.7 42.5 44.6  
 Other/Unknown 1.1 5.2 3.8  1.3 2.3 3.8  0.7 0.0 0.9  
Colectomy  0.46    <0.01    <0.01 
 90.9 90.1 88.2  18.9 7.5 30.9  70.2 40.0 57.5  
Comorbidity scoreb  0.03    0.22    0.03 
 0 58.3 51.2 55.7  53.9 44.8 48.2  55.1 50.8 47.0  
 1 22.9 27.2 18.1  26.0 32.2 26.4  27.3 22.5 27.2  
 2 or 3 15.5 18.3 19.2  16.1 18.4 22.1  14.9 18.3 21.6  
 4 or higher 3.2 3.3 7.0  3.9 4.6 4.3  2.7 8.3 4.2  
Distance to VA outpatient center (miles)  0.25    0.37    0.12 
  Mean (Std.) 13.4 (14.3) 14.1 (16.4) 14.8 (15.6)  14.0 (13.9) 15.4 (14.1) 14.1 (14.7)  14.6 (13.7) 14.1 (15.2) 13.2 (14.0)  
 Bottom tertile 2.3 (1.2) 2.4 (1.1) 2.6 (1.1)  2.3 (1.3) 2.1 (1.2) 2.7 (1.2)  2.6 (1.4) 2.3 (1.4) 2.4 (1.3)  
 Middle tertile 8.3 (2.8) 7.9 (2.3) 8.6 (2.7)  9.3 (3.1) 9.5 (3.3) 8.6 (2.9)  9.2 (2.8) 8.0 (2.8) 8.2 (2.4)  
 Top tertile 29.7 (14.1) 32.3 (16.3) 32.5 (14.8)  31.1 (11.2) 29.9 (11.4) 32.2 (12.5)  30.0 (11.1) 33.2 (13.0) 30.8 (12.2)  
US Census Division/Region    <0.01    <0.01    <0.01 
 Northeast 18.8 24.4 20.9  17.0 14.4 23.5  17.6 11.7 22.5  
 East North Central 11.6 11.7 8.4  11.4 12.1 5.9  10.9 4.2 7.2  
 West North Central 9.7 10.8 11.8  8.4 8.6 11.9  12.4 19.2 10.8  
 South Atlantic 23.5 18.3 13.9  21.3 20.7 14.5  19.9 22.5 16.2  
 East South Central 10.4 7.0 0.7  10.1 5.7 0.5  7.2 9.2 1.2  
 West South Central 14.0 8.0 9.8  15.0 13.2 10.5  16.1 15.8 9.9  
 West 12.1 19.7 34.5  16.7 25.3 33.3  15.9 17.5 32.3  
Stage IStage IIStage III
Predominantly VADual userPredominantly non-VAPredominantly VADual userPredominantly non-VAPredominantly VADual userPredominantly non-VA
(n = 528)(n = 213)(n = 287)(n = 534)(n = 174)(n = 421)(n = 403)(n = 120)(n = 334)
%%%Pa%%%Pa%%%Pa
Age at diagnosis  <0.01    <0.01    <0.01 
  Mean (Std.) 74.9 (5.9) 75.1 (5.1) 76.7 (5.5)  75.1 (5.6) 76.9 (5.7) 76.3 (5.5)  74.1 (5.5) 76.1 (6.4) 76.3 (5.9)  
 66–75 55.7 54.9 39.0  53.6 38.5 44.9  61.3 41.7 43.1  
 76–85 39.0 42.7 55.1  43.3 55.2 50.4  35.7 48.3 50.6  
 ≥ 86 5.3 2.3 5.9  3.2 6.3 4.8  3.0 10.0 6.3  
Gender  <0.01    <0.01    <0.01 
 Male 98.7 98.1 93.0  97.9 98.3 92.6  99.0 95.8 93.1  
Race  <0.01    0.04    0.24 
 African American 20.6 14.1 7.0  16.7 14.4 10.9  15.1 18.3 12.3  
Marital status  <0.01    <0.01    <0.01 
 Not married 38.4 31.9 26.8  44.2 41.4 32.5  38.0 49.2 28.4  
 Married 54.9 54.0 69.7  55.4 55.7 66.0  61.5 50.8 68.3  
 Unknown 6.6 14.1 3.5  0.4 2.9 1.4  0.5 0.0 3.3  
High school graduation (area rate)  <0.01    <0.01    <0.01 
 Mean Percent (Std.) 58.8 (16.9) 64.0 (17.5) 67.8 (18.7)  59.6 (17.5) 62.2 (17.6) 66.7 (17.6)  60.8 (17.3) 62.9 (16.8) 67.5 (18.8)  
Tumor grade  0.06    0.04    0.32 
 Grade I 17.6 14.6 16.7  8.1 12.6 9.7  6.9 5.0 5.1  
 Grade II 59.1 61.0 63.4  75.8 70.7 69.1  70.0 71.7 67.7  
 Grade III 4.4 3.8 7.7  11.2 12.6 16.9  17.1 20.8 23.4  
 Grade IV 0.2 0.5 0.0  0.2 0.0 1.2  0.7 0.8 0.9  
 Unknown 18.8 20.2 12.2  4.7 4.0 3.1  5.2 1.7 3.0  
Regional lymph nodes examined    0.01    0.25    0.18 
 0–5 55.5 52.1 46.3  24.7 20.1 23.8  14.6 25.0 14.7  
 6–11 22.7 23.0 29.6  33.5 36.8 34.0  38.0 32.5 39.8  
 12–90 20.6 19.7 20.2  40.4 40.8 38.5  46.7 42.5 44.6  
 Other/Unknown 1.1 5.2 3.8  1.3 2.3 3.8  0.7 0.0 0.9  
Colectomy  0.46    <0.01    <0.01 
 90.9 90.1 88.2  18.9 7.5 30.9  70.2 40.0 57.5  
Comorbidity scoreb  0.03    0.22    0.03 
 0 58.3 51.2 55.7  53.9 44.8 48.2  55.1 50.8 47.0  
 1 22.9 27.2 18.1  26.0 32.2 26.4  27.3 22.5 27.2  
 2 or 3 15.5 18.3 19.2  16.1 18.4 22.1  14.9 18.3 21.6  
 4 or higher 3.2 3.3 7.0  3.9 4.6 4.3  2.7 8.3 4.2  
Distance to VA outpatient center (miles)  0.25    0.37    0.12 
  Mean (Std.) 13.4 (14.3) 14.1 (16.4) 14.8 (15.6)  14.0 (13.9) 15.4 (14.1) 14.1 (14.7)  14.6 (13.7) 14.1 (15.2) 13.2 (14.0)  
 Bottom tertile 2.3 (1.2) 2.4 (1.1) 2.6 (1.1)  2.3 (1.3) 2.1 (1.2) 2.7 (1.2)  2.6 (1.4) 2.3 (1.4) 2.4 (1.3)  
 Middle tertile 8.3 (2.8) 7.9 (2.3) 8.6 (2.7)  9.3 (3.1) 9.5 (3.3) 8.6 (2.9)  9.2 (2.8) 8.0 (2.8) 8.2 (2.4)  
 Top tertile 29.7 (14.1) 32.3 (16.3) 32.5 (14.8)  31.1 (11.2) 29.9 (11.4) 32.2 (12.5)  30.0 (11.1) 33.2 (13.0) 30.8 (12.2)  
US Census Division/Region    <0.01    <0.01    <0.01 
 Northeast 18.8 24.4 20.9  17.0 14.4 23.5  17.6 11.7 22.5  
 East North Central 11.6 11.7 8.4  11.4 12.1 5.9  10.9 4.2 7.2  
 West North Central 9.7 10.8 11.8  8.4 8.6 11.9  12.4 19.2 10.8  
 South Atlantic 23.5 18.3 13.9  21.3 20.7 14.5  19.9 22.5 16.2  
 East South Central 10.4 7.0 0.7  10.1 5.7 0.5  7.2 9.2 1.2  
 West South Central 14.0 8.0 9.8  15.0 13.2 10.5  16.1 15.8 9.9  
 West 12.1 19.7 34.5  16.7 25.3 33.3  15.9 17.5 32.3  

aP values from χ2 tests or Fisher's exact tests (gender, marital status, tumor grade, regional lymph nodes, and comorbidity score).

bWe constructed a modified Deyo–Charlson Score with Romano and Klabunde Adaptations.

Among patients with stage II, dual users were older; 61.5% were 76 years or older compared with 46.5% (VA) and 55.2% (non-VA). African Americans comprised larger percentages of the dual (14.4%) and VA (16.7%) than non-VA users (10.9%). Fewer dual and VA users than non-VA users had a grade III or IV tumor (Dual 12.6%; VA 11.4%; non-VA 18.1%). Dual users were much less likely than either VA or non-VA users to have received chemotherapy (7.5% vs. 18.9% and 7.5% vs. 30.9%). There were no differences in number of lymph nodes examined, comorbidity score, or distance to VA across dual use categories for patients with stage II.

Among patients with stage III, dual users were substantially older than predominantly VA but not predominantly non-VA users (≥75 years: 58.3% vs. 38.7% and 58.3% vs. 56.9%). Substantially fewer dual users than either VA or non-VA users received chemotherapy services (40.0% vs. 70.2% and 40.0% vs. 57.5%). Dual users' comorbidity burden was similar to that of non-VA users (comorbidity score 0 or 1: 73.3% vs. 74.2%) but lower than that of VA users (comorbidity score 0 or 1: 73.3% vs. 82.4%). There were no differences across dual use categories in proportion African American, tumor grade, number of nodes examined, or distance to VA outpatient center.

For all stages, some of the differences in characteristics, although statistically significant, were small and of questionable clinical importance. Notably, across all stages, distance to VA outpatient center was not associated with dual use category; dual users tended to live in areas with higher percentages of high-school graduates compared with VA users but lower percentages compared with non-VA users; and predominantly non-VA patients were concentrated in the West region of the US.

Overall survival

Table 2 presents numbers of deaths per 100 person-years of follow-up. Dual users experienced more deaths than both VA and non-VA users in stages II and III but not I. VA users experienced fewest deaths in all stages. Figure 1 shows Kaplan–Meier curves for overall survival, comparing survival distributions across user groups within each stage. The survival distributions were statistically different for stages I and III (P = 0.032 and P = 0.003, respectively), for stage I showing lower survival in the dual compared with the VA group but higher compared with the non-VA group and for stage III showing lowest survival among the dual group across nearly the entire distribution.

Figure 1.

Overall survival Kaplan–Meier curves for predominantly VA users (), dual users (), and predominantly non-VA users (). A, stage I. B, stage II. C, stage III. X-axis, analysis time (months); Y-axis, survival probability. Note: Kaplan–Meier estimates are unadjusted.

Figure 1.

Overall survival Kaplan–Meier curves for predominantly VA users (), dual users (), and predominantly non-VA users (). A, stage I. B, stage II. C, stage III. X-axis, analysis time (months); Y-axis, survival probability. Note: Kaplan–Meier estimates are unadjusted.

Close modal
Table 2.

Number of events per 100 person-years of follow-up

Stage IStage IIStage III
Deaths    
 VA 7.7 11.7 19.0 
 Dual 10.2 14.9 30.9 
 Non-VA 11.2 13.0 22.6 
Colon eventsa    
 VA 12.0 19.2 34.9 
 Dual 15.7 19.6 46.1 
 Non-VA 17.3 23.9 38.3 
Stage IStage IIStage III
Deaths    
 VA 7.7 11.7 19.0 
 Dual 10.2 14.9 30.9 
 Non-VA 11.2 13.0 22.6 
Colon eventsa    
 VA 12.0 19.2 34.9 
 Dual 15.7 19.6 46.1 
 Non-VA 17.3 23.9 38.3 

aEvents include deaths.

HR estimated from extended Cox models controlling for age, comorbidity score, marital status, race, area percentage of high school graduates, distance to VA outpatient center, colectomy (stage I), chemotherapy (stages II and III), tumor grade, number of regional lymph nodes examined, and geographic region are shown in Table 3. Among patients with stages I, II, and III, predominantly VA and predominantly non-VA users had better overall survival. For example, compared with dual users with stage I cancer, predominantly VA users had 60% and predominantly non-VA users had 46% reduced hazard of dying from any cause (VA HR 0.40, 95% CI: 0.28–0.56; non-VA HR 0.54, 95% CI: 0.38–0.78). Other HR ranged from 0.59 for stage II and III VA users (stage II—95% CI: 0.43–0.80; stage III—95% CI: 0.44–0.78) to 0.70 for stage II non-VA users (95% CI: 0.52–0.93).

Table 3.

Estimated HR comparing healthcare system user groupsa

HR (95% CI)
Overall survival
Stage IStage IIStage III
Healthcare system use(N = 1,028)(N = 1,129)(N = 857)
 Predominantly VA 0.40 (0.28–0.56) 0.59 (0.43–0.80) 0.59 (0.44–0.78) 
 Dual User Ref Ref Ref 
 Predominantly Non-VA 0.54 (0.38–0.78) 0.69 (0.50–0.94) 0.70 (0.52–0.93) 
 EFS 
 Stage I Stage II Stage III 
Healthcare system use (352 events) (524 events) (575 events) 
 Predominantly VA 0.47 (0.35–0.62) 0.74 (0.56–0.97) 0.73 (0.56–0.94) 
 Dual User Ref Ref Ref 
 Predominantly Non-VA 0.64 (0.47–0.86) 0.92 (0.69–1.22) 0.81 (0.62–1.06) 
HR (95% CI)
Overall survival
Stage IStage IIStage III
Healthcare system use(N = 1,028)(N = 1,129)(N = 857)
 Predominantly VA 0.40 (0.28–0.56) 0.59 (0.43–0.80) 0.59 (0.44–0.78) 
 Dual User Ref Ref Ref 
 Predominantly Non-VA 0.54 (0.38–0.78) 0.69 (0.50–0.94) 0.70 (0.52–0.93) 
 EFS 
 Stage I Stage II Stage III 
Healthcare system use (352 events) (524 events) (575 events) 
 Predominantly VA 0.47 (0.35–0.62) 0.74 (0.56–0.97) 0.73 (0.56–0.94) 
 Dual User Ref Ref Ref 
 Predominantly Non-VA 0.64 (0.47–0.86) 0.92 (0.69–1.22) 0.81 (0.62–1.06) 

aHR were obtained from Extended Cox survival models adjusted for age, race, comorbidity score, marital status, area high school graduation rate, tumor grade, number of regional lymph nodes examined, colectomy (stage I), chemotherapy (stages II and III), distance to VA outpatient center, and geographic region.

When we ascertained patients' dual use status at 30 days before death and did not allow it to change after that, the results were very similar (Table 4). While we found that the magnitude of the HRs were somewhat closer to 1.0 and in some models lost statistical significance, the direction of the effects was unchanged. For example, among patients with stage III, HRs were 0.71 (95% CI: 0.53–0.96) for VA and 0.80 (95% CI: 0.59–1.09) for non-VA users compared with 0.59 and 0.69, respectively, when dual use was allowed to vary up to the time of death.

Table 4.

Estimated HR comparing healthcare system user groups: revised dual use measurea

HR (95% CI)
Overall survival
Stage IStage IIStage III
Healthcare system use(N = 1,028)(N = 1,129)(N = 857)
 Predominantly VA 0.50 (0.35–0.71) 0.72 (0.52–0.99) 0.71 (0.53–0.96) 
 Dual User Ref Ref Ref 
 Predominantly Non-VA 0.63 (0.43–0.91) 0.79 (0.57–1.11) 0.80 (0.59–1.09) 
 EFS 
 Stage I Stage II Stage III 
Healthcare system use (352 events) (524 events) (575 events) 
 Predominantly VA 0.55 (0.41–0.74) 0.91 (0.68–1.23) 0.84 (0.64–1.09) 
 Dual User Ref Ref Ref 
 Predominantly Non-VA 0.69 (0.50–0.94) 1.12 (0.83–1.51) 0.90 (0.68–1.18) 
HR (95% CI)
Overall survival
Stage IStage IIStage III
Healthcare system use(N = 1,028)(N = 1,129)(N = 857)
 Predominantly VA 0.50 (0.35–0.71) 0.72 (0.52–0.99) 0.71 (0.53–0.96) 
 Dual User Ref Ref Ref 
 Predominantly Non-VA 0.63 (0.43–0.91) 0.79 (0.57–1.11) 0.80 (0.59–1.09) 
 EFS 
 Stage I Stage II Stage III 
Healthcare system use (352 events) (524 events) (575 events) 
 Predominantly VA 0.55 (0.41–0.74) 0.91 (0.68–1.23) 0.84 (0.64–1.09) 
 Dual User Ref Ref Ref 
 Predominantly Non-VA 0.69 (0.50–0.94) 1.12 (0.83–1.51) 0.90 (0.68–1.18) 

aHR were obtained from Extended Cox survival models adjusted for age, race, comorbidity score, marital status, area high school graduation rate, tumor grade, number of regional lymph nodes examined, colectomy (stage I), chemotherapy (stages II and III), distance to VA outpatient center, and geographic region. Dual use status was held static at its value 30 days before the event.

Event-free survival

Table 2 presents numbers of events per 100 person-years of follow-up. Again, in all 3 stage groups, VA users experienced the fewest deaths. Stage III dual users had substantially higher event rates than their VA and non-VA counterparts (46 compared with 35 and 38 per 100 person-years, respectively). Figure 2 shows Kaplan–Meier curves for EFS. Survival distributions differed across dual use groups only in patients with stage I (P = 0.008). In stage I, survival in the dual group was between that of VA (highest) and non-VA (lowest) users across the entire follow-up period.

Figure 2.

Event-free survival Kaplan–Meier curves for predominantly VA users (), dual users (), and predominantly non-VA users (). A, stage I. B, stage II. C. stage III. X-axis, analysis time (months); Y-axis, survival probability. Note: Kaplan–Meier estimates are unadjusted.

Figure 2.

Event-free survival Kaplan–Meier curves for predominantly VA users (), dual users (), and predominantly non-VA users (). A, stage I. B, stage II. C. stage III. X-axis, analysis time (months); Y-axis, survival probability. Note: Kaplan–Meier estimates are unadjusted.

Close modal

Among patients with stage I, both predominantly VA and predominantly non-VA users had better EFS (Table 3). Compared with dual users, VA users had 53% and non-VA users had 36% reduced hazard of a progression event (VA HR 0.47, 95% CI: 0.35–0.62; non-VA HR 0.64, 95% CI: 0.47–0.86). Among patients with stage II, VA users had a reduced hazard of a progression event (HR 0.74, 95% CI: 0.56–0.97) but there was no difference in risk of a progression event among non-VA users compared with dual users (non-VA HR 0.92, 95% CI: 0.69–1.22). Similarly, among patients with stage III, VA users had a 27% reduced hazard of a progression event (HR 0.73, 95% CI: 0.56–0.94), whereas the non-VA HR of 0.81 (95% CI: 0.62–1.06) did not reach statistical significance.

When we ascertained patients' dual use status at 30 days before an event and did not allow it to change after that, results were similar (Table 4). As we found for OS, HRs obtained with the revised dual use measure were somewhat attenuated. For example, HRs among patients with stage I were 0.55 (95% CI: 0.41–0.74) for VA users and 0.69 (95% CI: 0.50–0.94) for non-VA users, compared with 0.47 and 0.64, respectively, when dual use was allowed to vary up to the time of the event. In patients with stage III, while HRs continued to suggest a greater hazard of progression among dual users, CI for both VA and non-VA crossed 1.0. The same was true for stage II VA users. As before, the hazard of an event among stage II non-VA users was not different from that among dual users.

In this cohort of veterans with nonmetastatic colon cancer who were dually eligible for VA and Medicare benefits, we found associations between receiving substantial portions of a patient's cancer care in 2 systems—both VA and private sector non-VA—rather than one, and reduced OS and EFS. Relative differences were largest and most consistent for care received predominantly in the VA. Regardless of stage and accounting for multiple demographic and clinical factors, patients who received their cancer care predominantly in the VA had substantially better OS and EFS than dual users. Patients who received predominantly non-VA cancer care had better OS than dual users regardless of stage and, among patients with stage I, improved EFS as well. None of our analyses suggested that dual system use was associated with a survival advantage. Our sensitivity analyses show that while unobserved declining health status among dual users may have contributed to their outcomes, this cannot be the only explanation for poorer survival in that group.

What might account for poorer survival among those receiving substantial care in 2 systems or, conversely, improved outcomes among cancer patients receiving care in a single system? Some dual users may have failed to receive timely or appropriate diagnosis and treatment as a result of fragmented care. We note that stage II and III dual users were substantially less likely than VA and non-VA users to receive chemotherapy. Perhaps for that reason, dual users also had fewer total provider visits for colon cancer care (not shown), leading us to wonder whether these patients were lost to follow-up. Stage I dual users, in contrast, had more provider visits for cancer care than their single-system counterparts and yet they, too, had poorer survival.

Regardless of whether dual users received more care, the effectiveness of that care in prolonging survival seems to have been diminished. Previous research suggests that care fragmentation has deleterious effects on health, including increased emergency department (ED) visits among patients with diabetes and greater probability of hospitalization for ambulatory care sensitive conditions in a VA population, and conversely, that greater continuity is associated with lower probability of hospitalization, fewer ED visits, and higher patient satisfaction (43–48). Fragmentation of care associated with dual use may influence outcomes through reduced continuity and associated coordination of care. Cancer care in particular is complex, multifaceted, and often prolonged. Coordination of the spectrum of care needed by cancer patients is a challenge for providers, patients, and families (17, 49). Even within one system such as the VA that has an electronic health record system, imperfect communication between multiple providers and between providers and patients, and imperfect systems for accomplishing administrative tasks involved in coordinating care can present barriers to timely care and optimal treatment (50). When care is sought across unrelated networks of providers, those problems are surely magnified.

To our knowledge, ours is the first study to examine impacts of dual healthcare system use in colon cancer. Recently, Landrum and colleagues (51) used VA and SEER-Medicare data to compare survival of older veterans with colon cancer in VA care to men in a general population sample using fee-for-service (FFS) Medicare. They found better all-cause and cancer-specific survival rates in VA, partly explained by more favorable stage at diagnosis. The authors did not examine dual use or account for varying proportions of VA patients' care in versus outside the VA. Because of those and other differences in the studies, direct comparisons of study results are not possible although our findings of better survival among predominantly VA users are not inconsistent with the results of that study.

Our study has limitations, particularly those related to the use of administrative data. It is possible, for example, that our finding of poorer survival among dual users is a result of the effect of unmeasured factors in our analysis. In that case, the seeming association between dual use and survival could have been confounded by unmeasured health status, preferences for treatment, quality of noncolon cancer care, or other factors. We accounted for comorbidities at baseline and the magnitude of the comorbidity HRs diminished as stage progressed (not shown), consistent with the expectation that underlying health is less influential in determining survival in patients with stage III cancer, for example, than those with stage I cancer and offering reassurance that our comorbidity measure captured important aspects of health. Nonetheless, our ability to ascertain and control for other potentially determinative conditions—frailty, for example—or changes in noncancer-related health status after diagnosis was limited.

Measurement error in our colon progression event outcome, if substantial and unequally distributed between VA and Medicare data sources, could bias results. For example, an ICD-9 code for “Secondary malignant neoplasm” accounted for a higher percentage of progression events in VA records than in Medicare claims. Differences in ICD-9 coding practices, if they existed, could have contributed to the fewer and smaller survival differences we found between predominantly non-VA and dual users than between predominantly VA and dual users. In addition, some events signaling cancer progression are more measurable in administrative data than others, regardless of the source. Chemotherapy, for example, is readily ascertained, whereas a positive biopsy result is not (28, 30, 33). If patients receiving care outside the VA are more likely to receive treatment services that are measurable (e.g., chemotherapy among patients with stage II), predominantly VA users could appear to have lower rates of progression than patients who use both systems. We note, however, that these potential limitations in outcome measurement would not have impacted our OS analysis in which we used a single data source for death ascertainment and found larger and more consistent deleterious effects associated with dual use. Thus, we conclude that findings of greater propensity toward death and progression among dual users are unlikely to be attributable to measurement error alone. Finally, because we did not have information on cause of death, we are unable to draw conclusions about the association of dual use with cancer death per se.

Our findings signal potential problems among patients with colon cancer who use both VA and Medicare FFS healthcare. Providing inducements to patients for single system care and mechanisms to facilitate improved care coordination across systems may be two potential avenues to achieve improved outcomes. To be successful, interventions targeting dual use will need to address the removal of barriers to single system care that may confront patients, alignment of financial incentives on the private sector side, and advancement of the electronic health record and health information exchange (2, 50).

While this study has revealed some important differences in outcomes according to dual versus single system use, further study using prospectively collected data is needed. Future research should confirm these results and learn more about why the differences exist, why patients use both systems, what problems occur as a result, and how dual users who have Medicare managed care coverage fare. Together with a better understanding of barriers to better provider communication and coordination, this information will lead to potential implementable solutions and better patient outcomes in colon cancer care.

No potential conflicts of interest were disclosed. The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs or the National Cancer Institute's Surveillance, Epidemiology, and End Results Program.

Conception and design: D.M. Hynes, E. Tarlov

Development of methodology: T.A. Lee, R. Durazo-Arvizu, D.M. Hynes, E. Tarlov

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Tarlov, T.A. Lee, T. Weichle, R. Durazo-Arvizu, Q. Zhang, R. Perrin, D.J. Bentrem, D.M. Hynes

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Perrin, D.M. Hynes

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Durazo-Arvizu, Q. Zhang, R. Perrin, D.M. Hynes

Writing, review, and/or revision of the manuscript: E. Tarlov, T.A. Lee, T. Weichle, R. Durazo-Arvizu, Q. Zhang, R. Perrin, D.J. Bentrem, D.M. Hynes

Study supervision: E. Tarlov, D.M. Hynes

The authors thank members of our Clinical Advisory Committee: Al B. Benson, MD, FACP, Feinberg School of Medicine, Northwestern University; Nirmala Bhoopalam, MD, Hines VA Hospital; Vivien W. Chen, PhD, Louisiana Tumor Registry and Louisiana State University Health Science Center School of Public Health; Marc T. Goodman, PhD, Cancer Research Center, University of Hawaii; M. Rosario Ferreira, MD, Feinberg School of Medicine; Dawn Provenzale, MD, Division of Gastroenterology, Department of Medicine, Durham VA Medical Center and Durham Epidemiologic Research and Information Center, GI Outcomes Research Group, and Division of Gastroenterology, Duke University Medical School; Beth Virnig, PhD, University of Minnesota School of Public Health; and Min-Woong Sohn, PhD, Center for Management of Complex Chronic Care, Hines VA Hospital and Feinberg School of Medicine, who advised on design and measurement issues in our project.

Supported by the Department of Veterans Affairs (VA), Health Services Research, and Development Service Project No. IIR 03-196, “Quality and Costs of Colon Cancer Care in VA and Medicare” and Project No. SDR 02-237 “VA CMS Data for Research” (principal investigator, D.M. Hynes). Also supported by a VA Health Services Research and Development Service Research Career Scientist Award (D.M. Hynes). Medicare data (claims and enrollment) obtained through the VA-Medicare Data Merge Initiative of the VA Information Resource Center within the VA Health Services Research Service. The collection of the California cancer incidence data used in this study was supported by the California Department of Public Health and by the Northern California Cancer Center under contract from the National Cancer Institute's Surveillance, Epidemiology and End Results Program.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Hynes
DM
,
Koelling
K
,
Stroupe
K
,
Arnold
N
,
Mallin
K
,
Sohn
MW
, et al
Veterans' access to and use of Medicare and Veterans Affairs health care
.
Med Care
2007
;
45
:
214
23
.
2.
Shen
Y
,
Hendricks
A
,
Zhang
S
,
Kazis
LE
. 
VHA enrollees' health care coverage and use of care
.
Med Care Res Rev
2003
;
60
:
253
67
.
3.
Bean-Mayberry
B
,
Chang
CC
,
McNeil
M
,
Hayes
P
,
Scholle
SH
. 
Comprehensive care for women veterans: indicators of dual use of VA and non-VA providers
.
J Am Med Womens Assoc
2004
;
59
:
192
7
.
4.
Liu
CF
,
Bolkan
C
,
Chan
D
,
Yano
EM
,
Rubenstein
LV
,
Chaney
EF
. 
Dual use of VA and non-VA services among primary care patients with depression
.
J Gen Intern Med
2009
;
24
:
305
11
.
5.
Liu
CF
,
Chapko
M
,
Bryson
CL
,
Burgess
JF
,
Fortney
JC
,
Perkins
M
, et al
Use of outpatient care in Veterans Health Administration and Medicare among veterans receiving primary care in community-based and hospital outpatient clinics
.
Health Serv Res
2010
;
45
(
5 Pt 1
):
1268
86
.
6.
Liu
CF
,
Bryson
CL
,
Burgess
JF
 Jr
,
Sharp
N
,
Perkins
M
,
Maciejewski
ML
. 
Use of outpatient care in VA and Medicare among disability-eligible and age-eligible veteran patients
.
BMC Health Serv Res
2012
;
12
:
51
.
Available from
: http://www.biomedcentral.com/1472-6963/12/51/abstract.
Published March 5, 2012. Accessed March 25, 2012
.
7.
Petersen
LA
,
Byrne
MM
,
Daw
CN
,
Hasche
J
,
Reis
B
,
Pietz
K
. 
Relationship between clinical conditions and use of Veterans Affairs health care among Medicare-enrolled veterans
.
Health Serv Res
2010
;
45
:
762
91
.
8.
Weeks
WB
,
Bott
DM
,
Lamkin
RP
,
Wright
SM
. 
Veterans Health Administration and Medicare outpatient health care utilization by older rural and urban New England veterans
.
J Rural Health
2005
;
21
:
167
71
.
9.
Zhu
CW
,
Penrod
JD
,
Ross
JS
,
Dellenbaugh
C
,
Sano
M
. 
Use of Medicare and Department of Veterans Affairs health care by veterans with dementia: a longitudinal analysis
.
J Am Geriatr Soc
2009
;
57
:
1908
14
.
10.
Ajmera
M
,
Wilkins
TL
,
Sambamoorthi
U
. 
Dual Medicare and Veteran Health Administration use and ambulatory care sensitive hospitalizations
.
J Gen Intern Med
2011
;
26
Suppl 2
:
669
75
.
11.
Fischer
MJ
,
Stroupe
KT
,
Kaufman
JS
,
O'Hare
AM
,
Browning
MM
,
Huo
Z
, et al
Predialysis nephrology care among older veterans using Department of Veterans Affairs or Medicare-covered services
.
Am J Manag Care
2010
;
16
:
e57
e66
.
Available from
: http://www.ajmc.com/publications/issue/2010/2010-01-vol16-n02/AJMC_10feb_FischerXclu_e57to66/.
Published February, 2010. Accessed March 29, 2012
.
12.
Helmer
D
,
Sambamoorthi
U
,
Shen
Y
,
Tseng
CL
,
Rajan
M
,
Tiwari
A
, et al
Opting out of an integrated healthcare system: dual-system use is associated with poorer glycemic control in veterans with diabetes
.
Prim Care Diabetes
2008
;
2
:
73
80
.
13.
Jia
H
,
Zheng
Y
,
Reker
DM
,
Cowper
DC
,
Wu
SS
,
Vogel
WB
, et al
Multiple system utilization and mortality for veterans with stroke
.
Stroke
2007
;
38
:
355
60
.
14.
Wolinsky
FD
,
An
H
,
Liu
L
,
Miller
TR
,
Rosenthal
GE
. 
Exploring the association of dual use of the VHA and Medicare with mortality: separating the contributions of inpatient and outpatient services
.
BMC Health Serv Res
2007
;
7
:
70
.
Available from
: http://www.biomedcentral.com/1472-6963/7/70.
Published May 9, 2007. Accessed March 25, 2012
.
15.
Grady
D
. 
Cancer patients, lost in a maze of uneven care
.
New York, NY:
New York Times
. 
2007
:
A1
.
16.
Nekhlyudov
L
,
Latosinsky
S
. 
The interface of primary and oncology specialty care: from symptoms to diagnosis
.
JNCI Monographs
2010
;
2010
:
11
7
.
17.
Sussman
J
,
Baldwin
LM
. 
The interface of primary and oncology specialty care: from diagnosis through primary treatment
.
J Natl Cancer Inst Monogr
2010
;
2010
:
18
24
.
18.
Yu
JB
,
Gross
CP
,
Wilson
LD
,
Smith
BD
. 
NCI SEER public-use data: applications and limitations in oncology research
.
Oncology (Williston Park)
2009
;
23
:
288
95
.
19.
VA/CMS Data for Research
. 
VA Information Resource Center Web site
.
Hines, IL: VA Information Resource Center, Health Services Research and Development Service, Department of Veterans Affairs. Available from
: http://www.virec.research.va.gov/DataSourcesName/VA-CMS/VACMSoverview.htm.
Updated January 30, 2012. Accessed February 12, 2012
.
20.
Commission on Cancer
. 
Facility Oncology Registry Data Standards (FORDS), 2010 revision
.
Chicago, IL
:
American College of Surgeons
2002
.
21.
SEER Quality Improvement
. 
Surveillance Research Program, National Cancer Institute Web site
.
Bethesda, MD: Surveillance Research Program, National Cancer Institute. Available from
: http://seer.cancer.gov/qi/index.html.
Updated March 28, 2012. Accessed March 29, 2012
.
22.
NAACCR Certification Levels
. 
North American Association of Central Cancer Registries Web site
.
Springfield, IL: North American Association of Central Cancer Registries. Available from
: http://www.naaccr.org/Certification/CertificationLevels.aspx.
Updated 2012. Accessed March 29, 2012
.
23.
Fleming I, Cooper JS, Henson DE, et al. (Eds.). American Joint Committee on Cancer
. 
AJCC cancer staging manual
. 5 ed.
Philadelphia
:
Lippincott-Raven
; 
1997
.
24.
VA Information Resource Center
. 
The Medical SAS® Inpatient Datasets—FY2000: A VIReC Resource Guide
.
Hines
EJ
 Jr
.
VA Hospital, Hines, IL
:
U.S. Department of Veterans Affairs, VA Health Services Research and Development Service
.
Available from
: http://www.virec.research.va.gov/References/RUG/RUG.htm. 
2001
.
25.
VA Information Resource Center
. 
The Medical SAS® Outpatient Datasets—FY2000: a VIReC Resource Guide
.
Hines
EJ
 Jr
.
VA Hospital, Hines, IL
:
U.S. Department of Veterans Affairs, VA Health Services Research and Development Service
.
Available from
: http://www.virec.research.va.gov/References/RUG/RUG.htm. 
2001
.
26.
VA Information Resource Center
. 
VIReC Research User Guide: VHA Pharmacy Prescription Data
, 2nd ed.
Hines
EJ
 Jr
.
VA Hospital, Hines, IL
:
U.S. Department of Veterans Affairs, VA Information Resource Center
.
Available from
: http://www.virec.research.va.gov/References/RUG/RUG.htm. 
2008
.
27.
Ayanian
JZ
,
Zaslavsky
AM
,
Fuchs
CS
,
Guadagnoli
E
,
Creech
CM
,
Cress
RD
, et al
Use of adjuvant chemotherapy and radiation therapy for colorectal cancer in a population-based cohort
.
J Clin Oncol
2003
;
21
:
1293
300
.
28.
Du
XL
,
Key
CR
,
Dickie
L
,
Darling
R
,
Geraci
JM
,
Zhang
D
. 
External validation of medicare claims for breast cancer chemotherapy compared with medical chart reviews
.
Med Care
2006
;
44
:
124
31
.
29.
Lamont
EB
,
Herndon
JE
,
Weeks
JC
,
Henderson
IC
,
Lilenbaum
R
,
Schilsky
RL
, 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
3
.
30.
Warren
JL
,
Harlan
LC
,
Fahey
A
,
Virnig
BA
,
Freeman
JL
,
Klabunde
CN
, et al
Utility of the SEER-Medicare data to identify chemotherapy use
.
Med Care
2002
;
40
(
8 Suppl
):
IV
61
.
31.
Medicare Data Documentation
. 
Research Data Assistance Center Web site
.
Minneapolis, MN: Research Data Assistance Center. Available from
: http://www.resdac.org/ddvh/Index.asp.
Updated March 6, 2012. Accessed March 25, 2012
.
32.
VHA Vital Status File
. 
VA Information Resource Center Web site
.
Hines, IL: VA Information Resource Center, Health Services Research and Development Service, Department of Veterans Affairs. Available from
: http://www.virec.research.va.gov/DataSourcesName/VitalStatus/VitalStatus.htm.
Updated December 30, 2011. Accessed January 15, 2012
.
33.
Earle
CC
,
Nattinger
AB
,
Potosky
AL
,
Lang
K
,
Mallick
R
,
Berger
M
, et al
Identifying cancer relapse using SEER-Medicare data
.
Med Care
2002
;
40
(
8 Suppl
):
IV
81
.
34.
Charlson
ME
,
Pompei
P
,
Ales
KL
,
MacKenzie
CR
. 
A new method of classifying prognostic comorbidity in longitudinal studies: development and validation
.
J Chronic Dis
1987
;
40
:
373
83
.
35.
Klabunde
CN
,
Potosky
AL
,
Legler
JM
,
Warren
JL
. 
Development of a comorbidity index using physician claims data
.
J Clin Epidemiol
2000
;
53
:
1258
67
.
36.
Klabunde
CN
,
Harlan
LC
,
Warren
JL
. 
Data sources for measuring comorbidity: a comparison of hospital records and medicare claims for cancer patients
.
Med Care
2006
;
44
:
921
8
.
37.
Romano
PS
,
Roos
LL
,
Jollis
JG
. 
Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives
.
J Clin Epidemiol
1993
;
46
:
1075
9
.
38.
Pritchard
RS
,
Fisher
ES
,
Teno
JM
,
Sharp
SM
,
Reding
DJ
,
Knaus
WA
, et al
Influence of patient preferences and local health system characteristics on the place of death. SUPPORT Investigators. Study to Understand Prognoses and Preferences for Risks and Outcomes of Treatment
.
J Am Geriatr Soc
1998
;
46
:
1242
50
.
39.
Sirovich
BE
,
Gottlieb
DJ
,
Welch
HG
,
Fisher
ES
. 
Regional variations in health care intensity and physician perceptions of quality of care
.
Ann Intern Med
2006
;
144
:
641
9
.
40.
Sohn
MW
,
Zhang
H
,
Arnold
N
,
Stroupe
K
,
Taylor
BC
,
Wilt
TJ
, et al
Transition to the new race/ethnicity data collection standards in the Department of Veterans Affairs
.
Popul Health Metr
2006
;
4
:
7
.
41.
Stroupe
KT
,
Tarlov
E
,
Zhang
Q
,
Haywood
T
,
Owens
A
,
Hynes
DM
. 
Use of Medicare and DOD data for improving VA race data quality
.
J Rehabil Res Dev
2010
;
47
:
781
95
.
42.
Therneau
TM
,
Grambsch
PM
. 
Modeling survival data
. 
Extending the Cox Model
.
New York
:
Springer
; 
2000
.
43.
Burge
F
,
Lawson
B
,
Johnston
G
. 
Family physician continuity of care and emergency department use in end-of-life cancer care
.
Med Care
2003
;
41
:
992
1001
.
44.
Fan
VS
,
Burman
M
,
McDonell
MB
,
Fihn
SD
. 
Continuity of care and other determinants of patient satisfaction with primary care
.
J Gen Intern Med
2005
;
20
:
226
33
.
45.
Gill
JM
,
Mainous
AG
 3rd
. 
The role of provider continuity in preventing hospitalizations
.
Arch Fam Med
1998
;
7
:
352
7
.
46.
Gill
JM
,
Mainous
AG
 3rd
,
Nsereko
M
. 
The effect of continuity of care on emergency department use
.
Arch Fam Med
2000
;
9
:
333
8
.
47.
Liu
CW
,
Einstadter
D
,
Cebul
RD
. 
Care fragmentation and emergency department use among complex patients with diabetes
.
Am J Manag Care
2010
;
16
:
413
20
.
48.
Pizer
SD
,
Gardner
JA
. 
Is fragmented financing bad for your health?
Inquiry
2011
;
48
:
109
22
.
49.
Hewitt M, Simone JV. National Cancer Policy Board, Institute of Medicine and National Research Council. Ensuring Quality Cancer Care
.
The National Academies Press
; 
1999
.
50.
Hesse
BW
,
Hanna
C
,
Massett
HA
,
Hesse
NK
. 
Outside the box: will information technology be a viable intervention to improve the quality of cancer care?
J Natl Cancer Inst Monogr
2010
;
2010
:
81
9
.
51.
Landrum
MB
,
Keating
NL
,
Lamont
EB
,
Bozeman
SR
,
Krasnow
SH
,
Shulman
L
, et al
Survival of Older Patients With Cancer in the Veterans Health Administration Versus Fee-for-Service Medicare
.
J Clin Oncol
2012
;
30
:
1072
9
.