Multiple System Utilization and Mortality for Veterans With Stroke
Background and Purpose— Many Veteran Health Administration (VHA) enrollees receive health services outside the VHA system. However, limited information is available about poststroke utilization and mortality by veterans who used multiple sources of health care. This study assessed the likelihood of 12-month poststroke rehospitalization and mortality of veterans who used VHA only versus those who used multiple sources of care.
Methods— Our retrospective observational study examined veterans living in Florida and diagnosed with acute stroke. We categorized users into 4 groups: VHA-only, VHA-Medicare, VHA-Medicaid, and VHA-Medicare-Medicaid based on their use of each health care program. Logistic regression models were fitted for 12-month poststroke general rehospitalization, recurrent stroke readmission, and mortality, adjusting for sociodemographic and clinical factors.
Results— The sample consisted of 29% VHA-only users, 61% VHA-Medicare users, 3% VHA-Medicaid users, and 7% VHA-Medicare-Medicaid triple users. Compared with the VHA-only users, multiple system users were significantly more likely to be rehospitalized for any cause and for recurrent stroke 12-months postindex. Mortality outcomes depended on when the outcome was measured; at the index admission date, we found no significant difference in mortality across the user groups; at the index discharge date, the VHA-only users was less likely to die within the first 12 months than the users of the 2 dual groups (VHA-Medicare and VHA-Medicaid).
Conclusions— Multiple health care source use was common among VHA enrollees with acute stroke in Florida. Multiple system users were more likely to be rehospitalized and the mortality outcomes were dependent on when the outcome was measured.
The Department of Veterans Affairs (VA), through the Veterans Health Administration (VHA), operates the largest integrated health care delivery system in the United States. However, VHA is not the only source of care for its enrollees who receive care from both VHA and non-VHA providers. This is particularly true for those who are also eligible for multiple health care plans such as Medicare and Medicaid. According to early studies, 42% of the 2.3 million VHA enrollees, 65 years old and older during 1989, were also eligible for Medicare, and 22% of those veteran patients actually received some Medicare benefits.1,2 It appears that the proportion of the VHA-Medicare dually eligible has been increasing since these studies were conducted. In a recent study on VHA-Medicare dual use, Hynes found that half of the 6.4 million veterans of all ages were dually eligible in 1999.3 Another recent study on Medicare utilization by VHA patients in a Midwest Veterans Integrated Service Network reported that two-thirds of the Veterans Integrated Service Network’s enrollees aged 65 years and older received inpatient and/or outpatient care under the Medicare program in 1998.4
Several studies compared the characteristics and health-related outcomes between the VHA-only users and the VHA-Medicare dual users with different types of diseases. Significant differences in patient sociodemographics, health plan coverage, accessibility, efficiency of care, and patient satisfaction between the 2 user groups have been documented.2,5–12 Wright et al, for example, used VHA and Centers for Medicare and Medicaid Services’ Medicare inpatient databases to identify dually eligible veterans with acute myocardial infarction who used VHA facilities and non-VHA facilities. These researchers found that more than half of the veterans with acute myocardial infarction obtained their initial care in Medicare hospitals, dual users were significantly more likely to undergo cardiac procedures than the VHA-only users, and the 2 user groups had similar adjusted mortality rates.8
Although the VHA-Medicare dual utilization and related outcomes by VHA patients are well-documented, less is known about the VHA stroke patients who used VHA-only versus those stroke patients who used multiple sources of care. No study has compared the poststroke utilization and mortality between these different user groups. This study assessed the likelihood of 12-month postindex stroke rehospitalization (for any cause and for recurrent stroke) and mortality for veterans with acute stroke who used VHA-only versus those veterans who used VHA-Medicare, VHA-Florida Medicaid, and VHA-Medicare-Florida Medicaid. Such information is needed to understand the full spectrum of services veterans receive both inside and outside the VHA.
Materials and Methods
Data used for this study were from three major sources. First, from the VHA Austin Automation Center, the VHA data files used included the Patient Treatment Files (Main files) and the Functional Status Outcomes Database for inpatient care, the outpatient files, and the Beneficiary Identification and Records Locator Subsystem Death File for patients’ vital status along with other inpatient files. Second, from the VA Information Resource Center, veteran-Medicare merged data used for this study included: the Medicare Denominator file for Medicare beneficiary’s eligibility and sociodemographic information; the Medicare Provider Analysis and Review File for Medicare inpatient and skilled nursing facility care; the Medicare Part B’s carrier file for noninstitutional care; and the outpatient file for institutional outpatient care. Third, from Florida Agency for Health Care Administration, the Florida Medicaid data used included the recipient file for the beneficiaries’ eligibility and demographic information and the facility file for the beneficiary’s inpatient and outpatient claims information.
The VHA data are stored by federal fiscal year whereas the Medicare and Medicaid data are stored by calendar year. To make these different timeframes compatible, we consistently used the calendar year system in this study. To ensure that a patient identified in the VHA database was the same person in the Medicare and/or Medicaid databases, we conducted a dual-system matching using a modified matching criteria developed by Fleming and Fisher.2 Our initial matching met these criteria for 99% of both VHA versus Medicare and VHA versus Medicaid samples, respectively. Whenever there was missing information in one source of data, we were able to find it from other sources of data being used.
We identified 1818 veterans with acute stroke. These patients: (1) lived in the state of Florida; (2) had at least 1 inpatient stay with admission or discharge diagnoses that matched Reker’s stroke high-sensitivity International Classification of Diseases (ICD)-9 codes13 in the VHA, Medicare, or Florida Medicaid databases between 2000 and 2001; (3) had index hospital stay <365 days; (4) were discharged alive at the index hospitalization; and (5) had at least 3 health care encounters, inpatient and/or outpatient, VHA and/or non-VHA systems, during the study period. The latter criterion facilitated compatibility in utilization comparison between the 4 user groups (eg, a VHA-only user could have only 1 health care encounter within the VHA system, whereas a triple user could have experienced at least 3 health encounters under different systems). We excluded 7 patients whose health care encounters were <3 during the study period. This study was approved by the Institutional Review Board at the University of Florida and local Research and Development Subcommittee for Clinical Investigations at the VA Medical Center in Gainesville, Florida.
Rehospitalization in general was established when a patient was discharged from a facility for index care and returned to a facility within 12 months for inpatient care for any cause. Readmission for recurrent stroke refers to patients who were discharged from a facility for their index stroke event and then readmitted to a facility within 12 months for another cerebrovascular accident (eg, secondary stroke or recurrent stroke). Mortality refers to all deaths that occurred during 12 months postindex discharge day.
Patients’ rehospitalization information was extracted from each of the inpatient data files described in the data sources section in this article. Beneficiary Identification and Records Locator Subsystem death file is a commonly used source for VHA enrollees’ vital status by VA investigators. Previous studies showed that the sensitivity of the Beneficiary Identification and Records Locator Subsystem death file ranged from 80.0%14 to 96.5%,15 that the quality of the file varied over time with the more recent the better the quality,14,16 and that the file should be supplemented with other data sources.17,18 For this study, we found that the death events identified in the Beneficiary Identification and Records Locator Subsystem file were also found in other files being used and vice versa.
Independent Variables or User Groups
Patients were categorized into 4 user groups (VHA-only, VHA-Medicare, VHA-Medicaid, and triple) according to their actual use of each health care program (VHA, Medicare, Medicaid) for inpatient and/or outpatient care during the study period.
Variables of interest and for risk adjustment in our final models included patient age (continuous), race (white, black, all other), gender (male, female), marital status (yes, no), and priority for VHA medical care (high, low). Patient self-reported racial/ethnic information was extracted from Medicare’s denominator file if available; otherwise, the data were obtained from VHA or Medicaid inpatient or outpatient data files. Patient gender was from VHA inpatient or outpatient databases. Patient priority for VHA medical care was created based on the means test indicator in VHA patient treatment files main database. The means test indicator is commonly used by VHA investigators in determining a patient’s eligibility to receive care within the VHA system.
To provide a valid comparison of the outcomes across the different user groups, it is essential to risk-adjust each outcome variable. Based on previous literature reports,19–21 we chose the following 2 categories of clinical risk-adjustment measures. First, at the index hospitalization, the measures included length of stay (continuous), comorbid conditions, stroke type (hemorrhagic stroke, ICD-9 code 430, 431, and 432; ischemic stroke, ICD-9 code 433, 434, 435, and 436; and all other for multiple stroke diagnoses and/or for those confirmed stroke inpatients identified in functional status outcomes database but their stroke type was not identified), number of days for intensive care unit use, care setting (acute inpatient care, long-term or extended inpatient care), mechanical ventilation or intubation (ICD-9 procedural code 96.04, 96.05, 96.70, 96.71, and 96.72), atrial fibrillation (ICD-9 code 427.3), dysphagia (ICD-9 code787.2), and malnutrition (ICD-9 codes 263.9). A modified Charlson comorbidity index was used to assess the patients’ comorbid conditions with the higher the weighted summary score, the more severe the burden of comorbidity.22 For this study, we excluded the related stroke diagnoses (ICD-9 codes 342, 430.xx – 438.xx) when calculating the comorbidity summary score because these diagnoses, primary or secondary, were used to identify the study cohort.
Second, we also included the next 12-month pre-index measures in our analyses: number of hospitalization for any cause, transient ischemic attack (ICD-9 code 435), and stroke diagnosis (Reker’s high-sensitivity ICD-9 codes14).
All data were analyzed using SAS version 8.1 (SAS Institute, Cary, NC). First, descriptive statistics were obtained on the sociodemographic, clinical, and dependent variables. Statistical inference (χ2 test on discrete variables and F test on continuous variables) were performed to compare the demographic and clinical characteristics between the groups. Second, collinearity diagnostics (conditional indices and variance proportion) were calculated to measure degrading or harmful multicollinearity among all independent and controlling variables. Consequently, because of multicollinearity, we chose to remove only the gender variable from our final models. Third, a multivariable logistic regression model was fitted for each dependent variable, adjusting for all the demographic and clinical factors discussed.
Among the 1818 veterans in Florida with acute stroke diagnosed between 2000 and 2001, 29% were VHA-only users, 61% were VHA-Medicare dual users, 3% were VHA-Florida Medicaid dual users, and 7% were triple users. There was a significant difference in several aspects of the sociodemographic and clinical characteristics across the user groups (Table 1). Demographically, compared with the other user groups, the VHA-Medicare group was more likely to be older, white, married, and of low priority for VHA care. In contrast, the VHA-Medicaid and VHA-only users were more likely to be younger and unmarried. Clinical characteristics in Table 1 are presented in 2 categories: index and 12-month pre-index admission. During the index hospitalization, compared with the other user groups, the VHA-Medicare users were significantly more likely to have ischemic stroke, atrial fibrillation, and malnutrition diagnosed, had more days in intensive care unit, and received more acute inpatient care versus long-term or extended care. During the 12 months pre-index, the VHA-only users were significantly less likely to receive inpatient care compared with other groups.
Table 2 shows the descriptive information of the dependent variables. Compared with other groups, the VHA-only group had smaller proportion of death and less rehospitalization utilization both for any causes and for recurrent stroke during the first 12 months postindex hospitalization.
Table 3 presents the results from our multivariable logistic regression analyses for each outcome measure. The dual and triple system users were significantly more likely to be rehospitalized in general and readmitted for recurrent stroke than the VHA-only users (reference group) after adjusting for patient sociodemographic, clinical, and disease severity factors. Specifically, compared with the VHA-only: the odds of receiving rehospitalization for any cause and recurrent stroke 12-month postindex were 1.5-times and 3.0-times for VHA-Medicare users; 2.3-times and 3.4-times for VHA-Medicaid users; and 13.6-times and 5.2-times for triple users, respectively. Other factors significantly associated with the utilization outcomes included: index age, high comorbidity summary score, stroke type, care setting, dysphagia, and malnutrition diagnoses; 12-month pre-index hospitalization and stroke diagnosis; and 12-month postindex death.
Our results also show barely significant group effects in mortality outcomes. The VHA-Medicare (adjusted odds ratio, 1.6) and VHA-Medicaid (adjusted odds ratio, 2.8) users were more likely to die within 12 months after index than the VHA-only users (Table 3). Other factors significantly related to mortality outcome were: older age, high priority for VHA care, higher comorbidity summary score, hemorrhagic stroke, intubation or mechanical ventilation care, dysphagia, malnutrition at index hospitalization, and more general hospitalization 12 months after index.
We conducted several other analyses to confirm these findings. First, these mortality results only reflect the death events occurring within the 12-month postindex discharge. In a separate analysis, when we also included the deaths that occurred during the index hospitalization (n=105), no significant difference was found between the VHA-only and the other user groups after adjusting the same set of risk adjusters. The difference between the 2 results (with and without index hospitalization death) was probably attributable to the higher in hospital death rate among the VHA-only group (8.7% for VHA-only users, 7.9% for VHA-Medicare dual users, 4.9% for VHA-Medicare dual users, and 1.4% for triple users).
Second, the higher death rate at index and lower rate post index discharge among the VHA-only group may introduce a survivorship effect, resulting in a healthier VHA-only users group and leading to reduced postindex mortality and utilization for the VHA-only group compared with other groups. To test this likelihood, we have rerun the 2 rehospitalization models (for any cause and for recurrent stroke), respectively, by excluding all the patients who died during the 12-month postindex (n=212). The odds of receiving rehospitalization for any cause and recurrent stroke 12-month postindex were 1.4-times and 2.9-times for VHA-Medicare users; 2.3-times and 3.5-times for VHA-Medicaid users; and 14.9-times and 5.8-times for triple users, respectively; which are very consistent with the original analyses. That is, VHA-only group had consistently used significantly less follow-up rehospitalization than the dual and triple users.
In addition, to test the consistency of these findings in utilization and mortality outcomes between the patients of all ages and the patients aged 65 years and older, we reran the 3 multivariable logistic regression models with the same set of dependent variables and covariates with the following exceptions: (1) the sample only included patients who were 65 years and older (N=1303); and (2) the independent variable (user group) consisted of 3 categories of triple, VHA-Medicare, and VHA-only users. The VHA-Medicaid user group was excluded in the consistency test analyses because the group had only 4 patients older than 65 years of age. We found that the VHA-Medicare and triple users were more likely to be rehospitalized for any cause (adjusted odds ratio, 1.2, 21.6) and for recurrent stroke (adjusted odds ratio, 2.9, 5.4) and the VHA-Medicare dual users were more likely to die (adjusted odds ratio, 2.0) than the VHA-only users 12 months after index hospitalization. These results for the patients aged 65 years and older were consistent with the results shown in Table 3 for the patients of all ages.
This is the first study that examines and compares the adjusted health care utilization and mortality of VHA inpatients with stroke diagnosed and used services under 3 different programs (VHA, Medicare, and Medicaid). First, our findings indicate that 71% of the study sample received care, inpatient or outpatient, out of VHA system during the study period. The high proportion of multiple-source use among the VHA stroke patients in Florida is consistent with previous national study reports. Using the 1999 Large Health Survey of VHA Enrollees, Shen et al showed that 73% of the VHA enrollees had multiple health care coverage (Medicare and Medicaid)23 but did not examine the actual utilization of the non-VHA programs.
The utilization of multiple sources of care by the VHA stroke patients has important implications for measuring the care quality. Studies have shown repeatedly that appropriate, consistent, high-intensity rehabilitation, and effective secondary prevention initiated in the early postacute phase of stroke may enhance the recovery process, minimize functional disability, reduce the risk of secondary stroke, and improve the quality of life of the patients with stroke. Despite the evidence regarding rehabilitation and secondary prevention poststroke, the use of multiple sources of care by the patients may result in a discontinuity of care and lead to an incomplete medical history at VHA for the care the patients received outside the system.
After adjusting for patient demographic and clinical characteristics, we found that compared with the VHA-only users, the triple and dual users were significantly more likely to be rehospitalized both for any cause and for recurrent stroke within the 12 months after index. We also found that the mortality outcome for this study cohort was heavily dependent on when the outcome was measured: if we counted the death event from the index admission date, there was no significant difference across the user groups. However, if we counted it from the index discharge date, the VHA-only users was less likely to die within the first 12 months than the dual users, even though the group effects were barely significant based the 95% CIs. These findings suggest that when comparing the long-term mortality outcomes between multiple groups it is important to examine the baseline mortality rates between the groups and survivorship effect. Counting mortality from index admission date versus from index discharge date could lead to different results. Further studies are needed to understand why the index death rate is higher among the VHA-only group than the other user groups. It may be that the VHA-only users were sicker than other groups to start with. For example, they had more ischemic stroke than the VHA-Medicare and triple users. Generally, sicker patients are more likely to stay within VHA system even if they were also eligible for multi-health care programs because VHA provides more comprehensive and systematic care than other programs.
Whether these differences in likelihood reflect differences in quality care, access to care, severity of disease, or patient clinical characteristics pose critical issues for VHA health care and need ongoing research. Future research is necessary to compare the outcomes of different user groups and understand the factors such as accessibility, scope of services provided by each system, scope of benefits under each plan, and patient satisfaction that may impact patient preference and choice of care across the different systems.
The generalizability of our findings is limited by our study’s focus on the state of Florida. Future research needs to understand the characteristics and utilization outcomes of the stroke patients at the national level. In addition, our sample size for triple users and VHA-Florida Medicaid dual users was small. Despite these limitations, our findings indicate that VHA stroke patients in Florida commonly used multiple health care systems. The user groups are different in several aspects of sociodemographic and clinical characteristics, as well as 12-month postindex rehospitalization utilization and mortality. Our findings may help improve clinicians’ understanding of their patients’ continuum of care poststroke; demonstrate the importance of considering dual-system and/or triple-system utilization when conducting program evaluations for health care systems with a high proportion of dual or triple enrollees; and serve as a basis for future studies.
In this retrospective observational study, we assessed the likelihood of 12-month post-stroke rehospitalization and mortality of veterans with stroke who used VHA only versus those who also used Medicare and/or Medicaid. We found that using multiple health care sources was common among VHA stroke patients in Florida. Our results showed that compared with the VHA-only users, the dual- and triple-system users were more likely to be rehospitalized for any cause and for recurrent stroke within 12 months postindex hospitalization and the mortality outcome was dependent on when the outcome was measured.
The authors thank the Veterans Affairs Information Resource Center for their assistance in extracting the Medicare data needed for this study, Florida’s Agency for Health Care Administration and Florida Center for Medicaid Issues at the University of Florida for their assistance in extracting the Florida Medicaid information needed for this study.
Sources of Funding
This research was supported through the VA HSR&D (grant # IIR 03-151-1).
The views and opinions in this article reflect those of the authors and do not necessarily reflect those of the Department of Veterans Affairs.
- Received June 30, 2006.
- Revision received August 7, 2006.
- Accepted September 13, 2006.
Hynes DM. Veterans’ use of Medicare and VA health care services by VISN (abstract). Presentation at the 2002 Annual VA HSR&D Meeting. 2003.
Barnett MJ, Brooks JM, Hendel ME, Rosenthal GE. Population-based evaluation of variation in Medicare utilization by VAMC patients (abstract). Presentation at the 2002 Annual VA HSR&D Meeting; 2003.
Cowper DC, Manheim LM, Weaver FM, Pawlow AJ. Mix and match: VA and non-VA care for elderly veterans. VA Practitioner. 1993; 10: 41–44.
Schlesinger M, Wetle T. Care of the elder veteran: New directions for change. Health Aff. 1986; 5: 59–71.
Wright SM, Jackson SL, Lo S. Enrollment and utilization in Medicare among veterans age 65 (abstract). Presentation at the 20th Annual VA HSR&D Meeting; 2002.
Yano EM, Kominski G, Rubenstein LV. Characteristics associated with obtaining care outside the VA (abstract). Presentation at the 13th Annual VA HSR&D Meeting; 1995.
Boyle CA, Decoufle P. National sources of vital status information: Extent of coverage and possible selectivity in reporting. Am J Epidemiol. 1990; 131: 160–168.
Hemphill JC 3rd, Newman J, Zhao S, Johnston SC. Hospital usage of early do-not-resuscitate orders and outcome after intracerebral hemorrhage. Stroke. 2004; 35: 1130–1134.
Shen Y, Hendricks A, Zhang S, Kazis LE. VHA enrollees’ health care coverage and use of care. Med Care Res Rev. 2003; 60: 253–267.