- Email: [email protected]
Dual health care system use is associated with higher rates of hospitalization and hospital readmission among veterans with heart failure
Dual health care system use is associated with higher rates of hospitalization and hospital readmission among veterans with heart failure R. Neal Axon, MD, MSCR, a,b Mulugeta Gebregziabher, PhD, a,c Charles J. Everett, PhD, a Paul Heidenreich, MD, MS, d and Kelly J. Hunt, PhD a,c Charleston, SC and Palo Alto, CA
Background Heart failure (HF) frequently causes hospital admission and readmission. Patients receiving care from multiple providers and facilities (dual users) may risk higher health care utilization and worse health outcomes. Methods
To determine rates of emergency department (ED) visits, hospitalizations, and hospital readmissions relative to dual use among HF patients, we analyzed a retrospective cohort of 13,977 veterans with HF hospitalized at the Veterans Affairs (VA) or non-VA facilities from 2007 to 2011; we analyzed rates of acute health care utilization using zero-inflated negative binomial regression.
Results Compared to VA-only users and dual users, individuals receiving all of their ED and hospital care outside the VA tended to be older, more likely to be non-Hispanic white and married, and less likely to have high levels of service connected disability. Compared to VA-only users, dual users had significantly higher rates of ED visits for HF as a primary diagnosis (adjusted rate ratio 1.15, 95% CI 1.04-1.27), hospitalization for HF (adjusted rate ratio 1.4, 95% CI 1.26-1.56), hospital readmission after HF hospitalization (all cause) (1.46, 95% CI 1.30-1.65), and HF-specific hospital readmission after HF hospitalization (1.46, 95% CI 1.31-1.63). With the exception of hospitalization for any primary diagnosis, non–VA-only users had significantly lower rates of ED visits, hospitalization, and readmission compared to VA-only users. Conclusions
Dual use is associated with higher rates of health care utilization among patients with HF. Interventions should be devised to encourage continuity of care where possible and to improve the effectiveness and safety of dual use in instances where it is necessary or desired. (Am Heart J 2015;0:1-7.)
Heart failure (HF) is a serious medical condition afflicting N5 million Americans, causing significant morbidity, and contributing to 1 in 9 deaths among US adults. 1 Patients with HF experience N1.1 million hospitalizations, N800,000 emergency department (ED) visits, and N3 million ambulatory care visits each year at an estimated cost of $32 billion. 2-4 Among those hospitalized for HF, readmission rates and mortality are quite high. Approximately 20.4% of Medicare
From the aCharleston Health Equity and Rural Outreach Innovation Center (HEROIC), Ralph H. Johnson VA Medical Center, Charleston, SC, bDivision of General Internal Medicine, Department of Medicine, The Medical University of South Carolina, Charleston, SC, c Department of Public Health Sciences, The Medical University of South Carolina, Charleston, SC, and dDivision of Cardiology, VA Palo Alto Healthcare System, Stanford University Medical Center, Palo Alto, CA. Research funding: VA Health Services Research and Development (IIR 12-331, Axon-PI). Reprint requests: R. Neal Axon, MD, MSCR, 109 Bee Street, Mail Code 111, Charleston, SC 29401. E-mails: [email protected], [email protected] 0002-8703 © 2015 Published by Elsevier Inc. http://dx.doi.org/10.1016/j.ahj.2015.09.023
recipients hospitalized for HF are rehospitalized within 30 days, and overall 30-day mortality is approximately 8.8%. 5 Furthermore, during the interval from 2005 to 2011, overall 30-day hospital readmission rates and 30-day mortality rates for HF patients do not appear to have changed significantly. 5 Patient-specific risk factors for HF hospitalization and readmission are well described and, among others, include age, gender, New York Heart Association functional class, left ventricular ejection fraction, and comorbid medical conditions. 6-8 However, significant variability exists in rates of hospitalization and hospital readmission for HF patients, 9,10 and current predictive models display only modest precision. 8 Variation in health care utilization among HF patients is also not fully explained by variations in quality of HF care. 11-14 Thus, additional patient and health system factors likely exist that may further describe why certain patients face higher risk of hospitalization and hospital readmission. Dual health care system use (dual use), which occurs when patients receive care from multiple providers or health care facilities, may be an important additional
2 Axon et al
explanatory factor regarding HF hospitalization and readmission. First, dual use is common. Among Medicare patients rehospitalized after an HF admission, almost 20% were readmitted to different facilities the second time around. 15 Dual use is even more common in the Veterans Health Administration (VHA) (Veterans Affairs [VA]) where, in 1 series, older VA-enrolled patients hospitalized for cardiovascular disorders relied on VA hospitals for only 27% of their hospital care. 16 Prior studies have demonstrated higher hemoglobin A1C values among veterans seen for primary care in both VA and non-VA clinics. 17 Among stroke patients, dual use was associated with significantly higher rates of 12-month rehospitalization and death. 18 A study of colon cancer patients found that, at every cancer stage, hazard of death was higher for dual users as compared to single-system users. 19 Given that HF is a chronic condition prone to frequent exacerbation, it is reasonable to hypothesize that patients might seek emergency and hospital care at multiple facilities. However, little is known regarding the impact of dual use on health care utilization and outcomes for HF patients. Most studies of dual use to date have certain limitations. For instance, some have relied on indirect or self-reported measures of dual use. 20-22 Most analyses of dual use to date have focused on Medicare populations and/or VA patient populations, and none have focused specifically on HF. 17,23-27 Unfortunately, analyses of Medicare recipients exclude information on younger subjects, whereas analyses of VA-only data exclude information on care provided at non-VA facilities. Such analyses may significantly underestimate total dual use. To study the impact of dual use on HF utilization and outcomes using a more comprehensive sample, we constructed a cohort of veterans with HF by combining VA, Medicare, and a comprehensive state data set with information on hospitalizations and ED visits at all hospitals in South Carolina. We hypothesized that dual use would be associated with higher rates of ED visits, hospitalizations, and hospital readmissions.
Methods Study population A state-level cohort of veterans with HF was created by linking multiple patient and administrative files including the VHA corporate data warehouse available from VA National Data Systems, VA/Medicare files available from the Veterans Information Resource Center, and the South Carolina Office of Revenue and Fiscal Affairs. The VA corporate data warehouse contains information including dates of service, demographic information, and diagnosis/procedure codes for outpatient visits and inpatient hospitalizations occurring in the VA system. Medicare files including inpatient, outpatient, and carrier files provide similar information for episodes of care reimbursed by the Centers for Medicare and Medicaid
American Heart Journal Month Year
Services at non-VA facilities. Under South Carolina law (§44-6-170), the SC-RFA collects and maintains uniformed billing data on all inpatient discharges, ED visits, outpatient surgery, imaging, radiation therapy, and other outpatient services from all short-term acute care hospitals and licensed freestanding medical centers in the state. 28 This material is based upon work supported by the Department of Veterans Affairs, VHA, Office of Research and Development, Health Service Research and Development (IIR 12-331-2). This project was approved by the VA Central Institutional Review Board as well as the Research and Development Committee at the Ralph H. Johnson VA Medical Center. Subjects were included in a larger cohort (n = 203,959) if they were enrolled for care in VHA, if they attended at least 1 qualifying primary care appointment at 1 of the 3 VA medical centers serving the state of South Carolina during the study period (2007-2011) and had a primary residence in the state of South Carolina (n = 136,244). Subjects were categorized as having HF if they had ≥1 inpatient or outpatient diagnoses for HF in a given year (International Classification of Diseases, Ninth Revision, 402.01, 402.11, 402.91, 429.3x, 425.xx, and 428.xx). Patients with HF were included in a final analytic data set (n = 13,977) if they had at least 1 episode of care for an ED visit or hospitalization at a VA or non-VA facility during the study period. Subjects were further categorized based on where they received acute ED and/or hospital care as VA-only users, non–VA-only users, and dual users. Subjects were followed up until death, loss to follow-up, or until December 2011.
Outcome measures Primary outcomes for this analysis were adjusted rate ratios for ED visits for any primary diagnosis and for HF as a primary diagnosis; hospitalizations for any primary diagnosis and for HF as a primary diagnosis; 30-day all-cause readmission for any primary diagnosis during the index hospitalization and for HF as primary diagnosis for the index hospitalization; and 30-day hospital readmission for HF as the primary diagnosis at index and readmission. We also report unadjusted event rates for ED visits, hospitalizations, and hospital readmissions. Covariates The primary covariates of interest were (1) age treated as a continuous variable; (2) gender categorized as male or female; (3) race/ethnicity, classified as non-Hispanic white (NHW), non-Hispanic black, Hispanic, and other/ missing/unknown; and (4) marital status categorized as married, divorced, widowed, never married, and unknown; location of residence categorized as urban, large town, small town, isolated rural, or missing; service connected disability classified as ≥50% or b50%. Service connected disability is a marker for disease burden, has implications for copayments within VHA, and has been
American Heart Journal Volume 0, Number 0
used in prior investigations of veteran patients. 29,30 Comorbidities were measured using the Elixhauser classification system and classified as present or absent for each patient. 31
Statistical analysis Typically, Poisson regression is used to model count outcomes where observations are assumed to be independent and the number of cases has variance equal to the mean for each level of the covariates. However, in our data, the equal mean and variance assumption required by the Poisson model is violated partly due to the zero inflation leading to overdispersion (ie, when the variance is greater than the conditional mean). 32 Instead, we used zero-inflated negative binomial (ZINB) models to study the association between dual use and HF-related emergency department visits, hospitalization, and hospital readmission. 33,34 The ZINB model is a mixture of a negative binomial model for the count outcome (including some zeros) combined with a logit model to determine the probability for excess zeros. Thus, the negative binomial model handles the problem of overdispersion, and the zero-inflated model handles the excess zeros. The parameters in the ZINB model have conditional or latent class interpretations, which correspond to a susceptible subpopulation at risk for the condition (in our case ED visit, hospitalization, or readmission) with counts generated from a negative binomial distribution and a nonsusceptible subpopulation that provides the extra or excess zeros. 35 We used a stepwise variable selection method based on statistical information criterion. 36 All analysis is done in SAS PROC COUNTREG 9.4 (SAS Institute, Inc, Cary, NC). The authors are solely responsible for the design and conduct of this study, all study analyses, and the drafting and editing of the manuscript, and its final contents.
Results The study cohort consisted of 13,977 veterans with HF with ED visits or hospitalizations during the study time interval (Table I). Compared to VA-only users and dual users, individuals receiving all of their ED and hospital care outside the VA tended to be older (73.5 vs 67.0 years [only VA users] and 68.2 years [dual users]), more likely to be NHW (82.2% vs 60.2% [only VA users] and 62.0% [dual users]) and married (69.9% vs 57.6% [only VA users] and 55.0% [dual users]). Individuals receiving all their ED and hospital care outside the VA were also less likely to have high levels of service connected disability than the other 2 groups (14.6% vs 22.8% [only VA users] and 21.7% [dual users]). Comorbidity burden varied across the 3 exposure categories with the burden of disease depending upon the comorbidity in question (Table I). Unadjusted rates of ED visits, hospitalizations, and hospital
Axon et al 3
readmissions are summarized in Table II. Considering visits or admissions from any primary cause, dual users had the highest rate of ED visits, hospitalizations, and 30-day hospital readmissions, whereas only non-VA users had the lowest rates. Hospitalization rates were 53.7 per 10,000 patient-days in dual users and 26.0 per 10,000 patient-days in only non-VA users. When limited to visits or admissions with HF as the primary diagnosis, ED visits were again highest in dual users (7.0 per 10,000 patient-days) and lowest in only non-VA users (3.6 per 10,000 patient-days) as were HF hospitalizations (7.5 vs 2.8 per 10,000 patient-days). Thirty-day hospital readmission rates after a HF hospitalization follow the same pattern regardless of whether the primary diagnosis on readmission was HF. Table III shows the adjusted rate ratio estimates of the association between count outcomes and dual use after adjusting for age, race, gender, year of visit, and comorbidities that were found to be significant. Among those who were at risk for having an ED visit, the rate of ED visits from any cause was 1.18 (1.14-1.22) times higher in dual users than in VA-only users, whereas the rate for non-VA users was 0.62 (0.60-0.64) times lower than VA-only users. The adjusted hospitalization rate from any cause was 1.93 (1.85-2.01) times higher in dual users than VA-only users and similar in non–VA-only users and VA-only users. The adjusted 30-day hospital readmissions rate from any cause was 1.82 (1.74-1.90) times higher in dual users than VA-only users. As depicted in Supplementary Table I, anemia, depression, fluid and electrolyte disorders, and psychoses were comorbidities selected for inclusion in models focused on health care utilization for all diagnoses. Focusing on primary diagnosis of HF, the rate of ED visits for HF were 1.15 (1.04-1.27) times higher and the rate of hospitalizations for HF were 1.40 (1.26-1.56) times higher in dual users than VA-only users. Moreover, after a hospitalization for HF, 30-day hospital readmission rates for any cause and HF specifically as the primary diagnosis are 1.46 (1.30-1.65) times higher and 1.46 (1.31-1.63) times higher, respectively, in dual users when compared to VA only users. As depicted in Supplementary Table II, fluid and electrolyte disorders and lung conditions were most frequently included in models focused on health care utilization where HF was the primary diagnosis at index encounter.
Discussion We examined a cohort of veterans hospitalized for HF and found that dual users had significantly higher rates of acute health care utilization as compared to VA-only users and non–VA-only users. This was true for ED visits, hospitalizations, and hospital readmissions for any diagnosis and for HF-specific health care visits. To our knowledge, this is the first published report of the impact of dual use on HF-related utilization and outcomes.
American Heart Journal Month Year
4 Axon et al
Table I. Subject characteristics by dual use status Characteristic Age Continuous (mean, y) ≥65 y (%) Gender (%) Male Female Race/ethnicity (%) NHW Non-Hispanic black Hispanic Other Missing Marital status (%) Married Divorced Widowed Never married Unknown Service connected disability (%) N50% Rurality (%) Urban Large town Small town Isolated rural Missing Comorbidities (%) Anemia Cancer Cardiovascular disease Cerebrovascular disease Depression Diabetes Fluid electrolyte disorder Hypertension Hypothyroidism Liver disease Lung conditions Obesity Peripheral vascular disease Psychoses Substance abuse Other⁎
Only VA users, n = 2242
Only non-VA users, n = 8825
Dual users, n = 2910
Total, N = 13,977
P
67.0 52.5
73.5 78.4
68.12 56.0
71.3 69.6
b.0001 b.0001
97.5 2.5
98.2 1.8
97.9 2.1
98.0 2.0
.0801
60.2 37.2 0.4 1.2 1.0
82.2 17.0 0.2 0.7 0.0
62.0 36.6 0.3 1.2 0.0
74.4 24.3 0.3 0.8 0.2
b.0001
57.6 20.0 10.2 10.8 1.4
69.9 9.2 12.7 4.8 3.4
55.0 19.2 12.2 11.3 2.2
64.8 13.0 12.2 7.1 2.8
b.0001
22.8
14.6
21.7
17.4
b.0001
88.5 5.7 1.2 1.1 3.6
88.2 7.0 1.4 0.9 2.5
88.8 5.6 1.2 1.4 3.0
88.4 6.5 1.3 1.0 2.8
.0027
27.7 26.2 74.0 16.6 42.9 61.7 61.4 97.5 13.4 12.6 19.0 41.2 32.0 12.9 25.2 84.7
30.0 28.2 79.4 25.4 35.2 57.8 67.3 97.7 20.9 11.4 21.8 27.9 44.2 10.1 13.1 85.4
40.6 27.9 81.6 33.6 50.4 65.1 81.1 98.4 17.9 17.6 27.1 36.2 44.8 19.6 28.0 91.5
31.8 27.8 79.0 25.7 39.6 60.0 69.2 97.8 19.1 12.9 22.4 31.8 42.4 12.5 18.2 86.6
b.0001 .1646 b.0001 b.0001 b.0001 b.0001 b.0001 .0658 b.0001 b.0001 b.0001 b.0001 b.0001 b.0001 b.0001 b.0001
⁎ Includes AIDS, chronic pulmonary disease, coagulopathy, peptic ulcer disease, renal failure, and rheumatoid arthritis.
Notably, our data sources included information on younger veterans and on episodes of care paid for by carriers other than VA and Medicare. These results are important because they highlight a novel system-level risk factor for HF hospitalization and readmission. These findings also add to the building body of evidence suggesting that dual health care system use may be less efficient and less safe. Previous work has demonstrated that many veterans receive care from both VA and non-VA sources. For example, Weeks et al 16 analyzed patterns of hospital utilization for veterans residing in New York State and found that veterans of all ages relied more heavily on VA for mental health and substance use disorders and less
heavily for circulatory, respiratory, or digestive disorders. In the outpatient realm, rural veterans have been observed to rely on Medicare more for primary care and on VA services for specialty and mental health care. 37 We observed proportions of non-VA hospital utilization that were similar to these prior reports. For instance, Jia et al 18 examined 12-month hospital readmission rates for veterans after stroke, and they found adjusted odds ratios of 1.5 (95% CI 1.1-1.9) for readmission among VA/ Medicare dual users and 2.3 (95% CI 1.2-4.5) among VA/ Medicaid dual users as compared to VA-only users. 18 There are several potential reasons why dual users may face higher hospitalization and readmission rates. First, dual use may be systematically associated with greater
American Heart Journal Volume 0, Number 0
Axon et al 5
Table II. Unadjusted rates of emergency department visits, hospitalization, and hospital readmission per 10,000 patient-days Only VA users, Only non-VA n = 2242 users, n = 8825
Dual users, n = 2910
Table III. Adjusted rate ratios of emergency department visits, hospitalizations, and hospital readmissions⁎ Only VA users, n = 2242
Mean (SD)⁎ All diagnoses ED visits 72.6 Hospitalizations 31.5 (total) 30.6 30-d hospital readmissions (all cause) HF as primary diagnosis ED visits 6.2 HF hospitalizations 6.5 (total) 30-d hospital 7.4 readmissions (all cause) 6.4 30-d hospital readmissions (principle diagnosis HF)
Only non-VA users, n = 8825
Dual users, n = 2910
Rate ratio (95% CI)
(79.0)† (56.7)
45.0 (67.5) 26.0 (34.5)
94.3 (135.4) 53.7 (64.8)
(54.6)
23.2 (32.4)
50.2 (59.5)
(22.8) (27.1)
3.6 (12.6) 2.8 (12.4)
7.0 (17.7) 7.5 (26.8)
(36.5)
2.6 (11.4)
8.8 (31.8)
(27.0)
2.2 (10.2)
7.0 (24.4)
All diagnoses ED visits Hospitalizations (total) 30-d hospital readmissions (all cause) HF as primary diagnosis ED visits HF hospitalizations (total) 30-d hospital readmissions (all cause) 30-d hospital readmissions (principle diagnosis HF)
1.00 1.00
0.62 (0.60-0.64) 1.18 (1.14-1.22) 0.98 (0.95-1.02) 1.93 (1.85-2.01)
1.00
0.87 (0.83-0.90) 1.82 (1.74-1.90)
1.00 1.00
0.60 (0.55-0.66) 1.15 (1.04-1.27) 0.61 (0.55-0.68) 1.40 (1.26-1.56)
1.00
0.51 (0.45-0.57) 1.46 (1.30-1.65)
1.00
0.51 (0.46-0.57) 1.46 (1.31-1.63)
⁎ Count outcomes for ED visits, hospitalizations, and hospital readmissions demonstrated heavy zero inflation and skewness. Thus, SDs are wide. † Event rates per 10,000 patient-days.
⁎ Calculated using ZINB regression. All models adjusted for age, race, gender, year of visit, dual use category, and comorbidities that were found to be significant using a stepwise selection procedure. Zero model also adjusted for year of visit.
variation in care quality leading to overtreatment or undertreatment of HF patients. Higher HF care quality, as measured by Centers for Medicare and Medicaid Services core measures compliance at the hospital level, has been associated with modestly but significantly lower 30-day hospital admission rates for HF patients. 38 Second, dual use may be a marker for poorer actual or perceived access to HF care either in the primary care setting or in a cardiology clinic. Before a hospitalization, poorer care continuity and less timely access to care for patients with acute HF symptoms could lead to less effective management and/or delays in treatment seeking. Heart failure is considered by the Agency for Healthcare Quality and Research to be an ambulatory care sensitive condition, and several prior studies have demonstrated that lower access to timely primary care is associated with higher hospital admission and readmission rates. 27,39,40 It may also be true that patients seeking care outside their usual system of care (ie, VA) may be more likely to be admitted to the hospital than those for whom medical records and follow-up options are more readily available. Although direct evidence of this phenomenon is lacking, hospital admission rates clearly vary widely by area and are a key determinant of hospital readmission rates. 9,10,41 This study should be evaluated in light of certain limitations. First, this is an observational study using administrative data sets, which is subject to the inherent limitations of this analytic approach. Although we attempted to control for variations in known likely confounders, there appears to be significant baseline
differences across the patient groups. Specifically, non– VA-only subjects tended to be older, more likely to be NHW and married, and less likely to have high levels of service connected disability. Demographically, they seem less similar to the other 2 groups. In addition, a higher proportion of dual users had several comorbidities including depression, diabetes, fluid/electrolyte disorders, lung conditions, and liver disease. Although we accounted for these differences in our models, our analysis may be subject to some residual confounding due to unmeasured comorbidities or overall illness burden. Although our research team had access to information on mortality for censoring purposes, we did not analyze this as a specific outcome in our analyses. Finally, we were not able to include information on the severity of HF in this analysis, such as New York Heart Associate class or ejection fraction. If this were available, it might allow for adjustment based on HF severity. Despite these limitations, our results add to a body of evidence suggesting that dual use is associated with higher rates of health care utilization and worse health care outcomes. Our analyses are strengthened by the use of robust statistical methods to account for heavy zero inflation and overdispersion. The VA and other integrated health delivery systems have reason for concern about fractured cross-system care, and care transitions measures such as hospital readmission have received increasing scrutiny in the era of the Accountable Care Act. 42 From a health systems perspective, even if outcomes for dual
American Heart Journal Month Year
6 Axon et al
users are influenced by severity of illness, dual use remains a marker for a patient population that may require more intensive clinical management and care coordination. Additional research is needed to better understand the factors leading to dual use for HF and other conditions in the veteran population and to measure the impacts of dual use in the general population of HF patients.
14.
15. 16.
Appendix. Supplementary data Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.ahj.2015.09.023.
17.
18.
References 1. Go AS, Mozaffarian D, Roger VL, et al. Heart disease and stroke statistics—2013 update: a report from the American Heart Association. Circulation 2013;127:e6-245. 2. Lloyd-Jones D, Adams R, M. C. Heart Disease and Stroke Statistics 2009 Update. A report from the American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Circulation 2008;108: 191261. 3. Blecker S, Ladapo JA, Doran KM, et al. Emergency department visits for heart failure and subsequent hospitalization or observation unit admission. Am Heart J 2014;168:901-8. [e901]. 4. Heidenreich PA, Trogdon JG, Khavjou OA, et al. Forecasting the future of cardiovascular disease in the United States: a policy statement from the American Heart Association. Circulation 2011;123:933-44. 5. Cheng RK, Cox M, Neely ML, et al. Outcomes in patients with heart failure with preserved, borderline, and reduced ejection fraction in the Medicare population. Am Heart J 2014;168:721-30. 6. Chaudhry SI, McAvay G, Chen S, et al. Risk factors for hospital admission among older persons with newly diagnosed heart failure: findings from the Cardiovascular Health Study. J Am Coll Cardiol 2013;61:635-42. 7. Lupon J, Vila J, Bayes-Genis A. Risk prediction tools in patients with heart failure. JACC Heart fail 2015;3:267. 8. Ross JS, Mulvey GK, Stauffer B, et al. Statistical models and patient predictors of readmission for heart failure: a systematic review. Arch Intern Med 2008;168:1371-86. 9. Goodman DC, Fisher ES, Chang C-H. After hospitalization: a Dartmouth Atlas Report on post-acute care for Medicare beneficiaries. In: Bronner KK, ed. The Dartmouth Atlas of Healthcare; 2011. 10. Goodman DC, Fisher ES, Chiang-Hua C. The revolving door: a report on U.S. hospital readmissons. Robert Wood Johnson Foundation. 2013. 11. Adams Jr KF, Fonarow GC, Emerman CL, et al. Characteristics and outcomes of patients hospitalized for heart failure in the United States: rationale, design, and preliminary observations from the first 100,000 cases in the Acute Decompensated Heart Failure National Registry (ADHERE). Am Heart J 2005;149:209-16. 12. Heidenreich PA, Hernandez AF, Yancy CW, et al. Get With The Guidelines program participation, process of care, and outcome for Medicare patients hospitalized with heart failure. Circ Cardiovasc Qual Outcomes 2012;5:37-43. 13. Heidenreich PA, Lewis WR, LaBresh KA, et al. Hospital performance recognition with the Get With The Guidelines Program and mortality
19.
20.
21.
22. 23.
24. 25.
26.
27.
28. 29.
30.
31.
32.
33. 34.
for acute myocardial infarction and heart failure. Am Heart J 2009;158:546-53. Hernandez AF, Fonarow GC, Liang L, et al. The need for multiple measures of hospital quality: results from the Get with the Guidelines– Heart Failure Registry of the American Heart Association. Circulation 2011;124:712-9. Nasir K, Lin Z, Bueno H, et al. Is same-hospital readmission rate a good surrogate for all-hospital readmission rate? Med Care 2010;48:477-81. Weeks WB, West AN, Wallace AE, et al. Comparing the characteristics, utilization, efficiency, and outcomes of VA and non-VA inpatient care provided to VA enrollees: a case study in New York. Med Care 2008;46:863-71. Helmer D, Sambamoorthi U, Shen Y, 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. Jia H, Zheng Y, Reker DM, et al. Multiple system utilization and mortality for veterans with stroke. Stroke 2007;38:355-60. Tarlov E, Lee TA, Weichle TW, et al. Reduced overall and event-free survival among colon cancer patients using dual system care. Cancer Epidemiol Biomarkers Prev 2012;21:2231-41. Wolinsky FD, An H, Liu L, et al. 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. Wolinsky FD, Miller TR, An H, et al. Dual use of Medicare and the Veterans Health Administration: are there adverse health outcomes? BMC Health Serv Res 2006;6:131. Ross JS, Keyhani S, Keenan PS, et al. Dual use of Veterans Affairs services and use of recommended ambulatory care. Med Care 2008;46:309-16. Carey K, Montez-Rath ME, Rosen AK, et al. Use of VA and Medicare services by dually eligible veterans with psychiatric problems. Health Serv Res 2008;43:1164-83. Liu CF, Bolkan C, Chan D, et al. Dual use of VA and non-VA services among primary care patients with depression. J Gen Intern Med 2009;24:305-11. Petersen LA, Byrne MM, Daw CN, et al. Relationship between clinical conditions and use of Veterans Affairs health care among Medicare-enrolled veterans. Health Serv Res 2010;45:762-91. Zhu CW, Penrod JD, Ross JS, et al. 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. 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. South Carolina Revenue and Fiscal Affairs Office. website accessed at, http://rfa.sc.gov/healthcare 2015. [on July 16, 2015.]. Egede LE, Gebregziabher M, Hunt KJ, et al. Regional, geographic, and racial/ethnic variation in glycemic control in a national sample of veterans with diabetes. Diabetes Care 2011;34:938-43. Egede LE, Gebregziabher M, Hunt KJ, et al. Regional, geographic, and ethnic differences in medication adherence among adults with type 2 diabetes (February). Ann Pharmacother 2011;45:169-78. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;43:1130-9. Payne E, Hardin JW, Egede LE, et al. Approaches for dealing with various sources of overdispersion in modeling count data: scale adjustment versus modeling. Stat Methods Med Res May 31, 2015, http://dx.doi.org/10. 1177/0962280215588569. [Epub ahead of print]. Lambert D. Zero-inflated Poisson regression, with an application to defects in manufacturing. Technometrics 1992;34:1-14. Greene WH. Accounting for excess zeros and sample selection in Poisson and negative binomial regression models. New York University Stern School of Business Working Papers. 1994.
American Heart Journal Volume 0, Number 0
35. Atkins DC, Gallop RJ. Re-thinking how family researchers model infrequent outcomes: a tutorial on count regression and zero-inflated models. J Fam Psychol 2007;21:726-35. 36. Cameron AC, Trivedi PK. Regression analysis of count data. Cambridge, UK: Cambridge University Press. 1998. 37. Weeks WB, Bott DM, Lamkin RP, et al. Veterans Health Administration and Medicare outpatient health care utilization by older rural and urban New England veterans. J Rural Health 2005;21:167-71. 38. Heidenreich PA, Zhao X, Hernandez AF, et al. Patient and hospital characteristics associated with traditional measures of inpatient quality of care for patients with heart failure. Am Heart J 2012;163: 239-45. [e233].
Axon et al 7
39. Ambulatory care sensitive conditions: age-standardized acute care hospitalization rate for conditions where appropriate ambulatory care prevents or reduces the need for admission to the hospital, per 100,000 population younger than age 75 years. Available at: http://www.qualitymeasures.ahrq.gov/content.aspx?id=35186 2014. Accessed 3/21/2014. 40. Axon RN. Frequency and costs of hospital transfers for ambulatory care sensitive conditions. Am J Manag Care 2015;21:11-9. 41. Pines JM, Mutter RL, Zocchi MS. Variation in emergency department admission rates across the United States. Med Care Res Rev 2013;70: 218-31. 42. Axon RN, Williams MV. Hospital readmission as an accountability measure. J Am Med Assoc 2011;305:504-5.
Background Heart failure (HF) frequently causes hospital admission and readmission. Patients receiving care from multiple providers and facilities (dual users) may risk higher health care utilization and worse health outcomes. Methods
To determine rates of emergency department (ED) visits, hospitalizations, and hospital readmissions relative to dual use among HF patients, we analyzed a retrospective cohort of 13,977 veterans with HF hospitalized at the Veterans Affairs (VA) or non-VA facilities from 2007 to 2011; we analyzed rates of acute health care utilization using zero-inflated negative binomial regression.
Results Compared to VA-only users and dual users, individuals receiving all of their ED and hospital care outside the VA tended to be older, more likely to be non-Hispanic white and married, and less likely to have high levels of service connected disability. Compared to VA-only users, dual users had significantly higher rates of ED visits for HF as a primary diagnosis (adjusted rate ratio 1.15, 95% CI 1.04-1.27), hospitalization for HF (adjusted rate ratio 1.4, 95% CI 1.26-1.56), hospital readmission after HF hospitalization (all cause) (1.46, 95% CI 1.30-1.65), and HF-specific hospital readmission after HF hospitalization (1.46, 95% CI 1.31-1.63). With the exception of hospitalization for any primary diagnosis, non–VA-only users had significantly lower rates of ED visits, hospitalization, and readmission compared to VA-only users. Conclusions
Dual use is associated with higher rates of health care utilization among patients with HF. Interventions should be devised to encourage continuity of care where possible and to improve the effectiveness and safety of dual use in instances where it is necessary or desired. (Am Heart J 2015;0:1-7.)
Heart failure (HF) is a serious medical condition afflicting N5 million Americans, causing significant morbidity, and contributing to 1 in 9 deaths among US adults. 1 Patients with HF experience N1.1 million hospitalizations, N800,000 emergency department (ED) visits, and N3 million ambulatory care visits each year at an estimated cost of $32 billion. 2-4 Among those hospitalized for HF, readmission rates and mortality are quite high. Approximately 20.4% of Medicare
From the aCharleston Health Equity and Rural Outreach Innovation Center (HEROIC), Ralph H. Johnson VA Medical Center, Charleston, SC, bDivision of General Internal Medicine, Department of Medicine, The Medical University of South Carolina, Charleston, SC, c Department of Public Health Sciences, The Medical University of South Carolina, Charleston, SC, and dDivision of Cardiology, VA Palo Alto Healthcare System, Stanford University Medical Center, Palo Alto, CA. Research funding: VA Health Services Research and Development (IIR 12-331, Axon-PI). Reprint requests: R. Neal Axon, MD, MSCR, 109 Bee Street, Mail Code 111, Charleston, SC 29401. E-mails: [email protected], [email protected] 0002-8703 © 2015 Published by Elsevier Inc. http://dx.doi.org/10.1016/j.ahj.2015.09.023
recipients hospitalized for HF are rehospitalized within 30 days, and overall 30-day mortality is approximately 8.8%. 5 Furthermore, during the interval from 2005 to 2011, overall 30-day hospital readmission rates and 30-day mortality rates for HF patients do not appear to have changed significantly. 5 Patient-specific risk factors for HF hospitalization and readmission are well described and, among others, include age, gender, New York Heart Association functional class, left ventricular ejection fraction, and comorbid medical conditions. 6-8 However, significant variability exists in rates of hospitalization and hospital readmission for HF patients, 9,10 and current predictive models display only modest precision. 8 Variation in health care utilization among HF patients is also not fully explained by variations in quality of HF care. 11-14 Thus, additional patient and health system factors likely exist that may further describe why certain patients face higher risk of hospitalization and hospital readmission. Dual health care system use (dual use), which occurs when patients receive care from multiple providers or health care facilities, may be an important additional
2 Axon et al
explanatory factor regarding HF hospitalization and readmission. First, dual use is common. Among Medicare patients rehospitalized after an HF admission, almost 20% were readmitted to different facilities the second time around. 15 Dual use is even more common in the Veterans Health Administration (VHA) (Veterans Affairs [VA]) where, in 1 series, older VA-enrolled patients hospitalized for cardiovascular disorders relied on VA hospitals for only 27% of their hospital care. 16 Prior studies have demonstrated higher hemoglobin A1C values among veterans seen for primary care in both VA and non-VA clinics. 17 Among stroke patients, dual use was associated with significantly higher rates of 12-month rehospitalization and death. 18 A study of colon cancer patients found that, at every cancer stage, hazard of death was higher for dual users as compared to single-system users. 19 Given that HF is a chronic condition prone to frequent exacerbation, it is reasonable to hypothesize that patients might seek emergency and hospital care at multiple facilities. However, little is known regarding the impact of dual use on health care utilization and outcomes for HF patients. Most studies of dual use to date have certain limitations. For instance, some have relied on indirect or self-reported measures of dual use. 20-22 Most analyses of dual use to date have focused on Medicare populations and/or VA patient populations, and none have focused specifically on HF. 17,23-27 Unfortunately, analyses of Medicare recipients exclude information on younger subjects, whereas analyses of VA-only data exclude information on care provided at non-VA facilities. Such analyses may significantly underestimate total dual use. To study the impact of dual use on HF utilization and outcomes using a more comprehensive sample, we constructed a cohort of veterans with HF by combining VA, Medicare, and a comprehensive state data set with information on hospitalizations and ED visits at all hospitals in South Carolina. We hypothesized that dual use would be associated with higher rates of ED visits, hospitalizations, and hospital readmissions.
Methods Study population A state-level cohort of veterans with HF was created by linking multiple patient and administrative files including the VHA corporate data warehouse available from VA National Data Systems, VA/Medicare files available from the Veterans Information Resource Center, and the South Carolina Office of Revenue and Fiscal Affairs. The VA corporate data warehouse contains information including dates of service, demographic information, and diagnosis/procedure codes for outpatient visits and inpatient hospitalizations occurring in the VA system. Medicare files including inpatient, outpatient, and carrier files provide similar information for episodes of care reimbursed by the Centers for Medicare and Medicaid
American Heart Journal Month Year
Services at non-VA facilities. Under South Carolina law (§44-6-170), the SC-RFA collects and maintains uniformed billing data on all inpatient discharges, ED visits, outpatient surgery, imaging, radiation therapy, and other outpatient services from all short-term acute care hospitals and licensed freestanding medical centers in the state. 28 This material is based upon work supported by the Department of Veterans Affairs, VHA, Office of Research and Development, Health Service Research and Development (IIR 12-331-2). This project was approved by the VA Central Institutional Review Board as well as the Research and Development Committee at the Ralph H. Johnson VA Medical Center. Subjects were included in a larger cohort (n = 203,959) if they were enrolled for care in VHA, if they attended at least 1 qualifying primary care appointment at 1 of the 3 VA medical centers serving the state of South Carolina during the study period (2007-2011) and had a primary residence in the state of South Carolina (n = 136,244). Subjects were categorized as having HF if they had ≥1 inpatient or outpatient diagnoses for HF in a given year (International Classification of Diseases, Ninth Revision, 402.01, 402.11, 402.91, 429.3x, 425.xx, and 428.xx). Patients with HF were included in a final analytic data set (n = 13,977) if they had at least 1 episode of care for an ED visit or hospitalization at a VA or non-VA facility during the study period. Subjects were further categorized based on where they received acute ED and/or hospital care as VA-only users, non–VA-only users, and dual users. Subjects were followed up until death, loss to follow-up, or until December 2011.
Outcome measures Primary outcomes for this analysis were adjusted rate ratios for ED visits for any primary diagnosis and for HF as a primary diagnosis; hospitalizations for any primary diagnosis and for HF as a primary diagnosis; 30-day all-cause readmission for any primary diagnosis during the index hospitalization and for HF as primary diagnosis for the index hospitalization; and 30-day hospital readmission for HF as the primary diagnosis at index and readmission. We also report unadjusted event rates for ED visits, hospitalizations, and hospital readmissions. Covariates The primary covariates of interest were (1) age treated as a continuous variable; (2) gender categorized as male or female; (3) race/ethnicity, classified as non-Hispanic white (NHW), non-Hispanic black, Hispanic, and other/ missing/unknown; and (4) marital status categorized as married, divorced, widowed, never married, and unknown; location of residence categorized as urban, large town, small town, isolated rural, or missing; service connected disability classified as ≥50% or b50%. Service connected disability is a marker for disease burden, has implications for copayments within VHA, and has been
American Heart Journal Volume 0, Number 0
used in prior investigations of veteran patients. 29,30 Comorbidities were measured using the Elixhauser classification system and classified as present or absent for each patient. 31
Statistical analysis Typically, Poisson regression is used to model count outcomes where observations are assumed to be independent and the number of cases has variance equal to the mean for each level of the covariates. However, in our data, the equal mean and variance assumption required by the Poisson model is violated partly due to the zero inflation leading to overdispersion (ie, when the variance is greater than the conditional mean). 32 Instead, we used zero-inflated negative binomial (ZINB) models to study the association between dual use and HF-related emergency department visits, hospitalization, and hospital readmission. 33,34 The ZINB model is a mixture of a negative binomial model for the count outcome (including some zeros) combined with a logit model to determine the probability for excess zeros. Thus, the negative binomial model handles the problem of overdispersion, and the zero-inflated model handles the excess zeros. The parameters in the ZINB model have conditional or latent class interpretations, which correspond to a susceptible subpopulation at risk for the condition (in our case ED visit, hospitalization, or readmission) with counts generated from a negative binomial distribution and a nonsusceptible subpopulation that provides the extra or excess zeros. 35 We used a stepwise variable selection method based on statistical information criterion. 36 All analysis is done in SAS PROC COUNTREG 9.4 (SAS Institute, Inc, Cary, NC). The authors are solely responsible for the design and conduct of this study, all study analyses, and the drafting and editing of the manuscript, and its final contents.
Results The study cohort consisted of 13,977 veterans with HF with ED visits or hospitalizations during the study time interval (Table I). Compared to VA-only users and dual users, individuals receiving all of their ED and hospital care outside the VA tended to be older (73.5 vs 67.0 years [only VA users] and 68.2 years [dual users]), more likely to be NHW (82.2% vs 60.2% [only VA users] and 62.0% [dual users]) and married (69.9% vs 57.6% [only VA users] and 55.0% [dual users]). Individuals receiving all their ED and hospital care outside the VA were also less likely to have high levels of service connected disability than the other 2 groups (14.6% vs 22.8% [only VA users] and 21.7% [dual users]). Comorbidity burden varied across the 3 exposure categories with the burden of disease depending upon the comorbidity in question (Table I). Unadjusted rates of ED visits, hospitalizations, and hospital
Axon et al 3
readmissions are summarized in Table II. Considering visits or admissions from any primary cause, dual users had the highest rate of ED visits, hospitalizations, and 30-day hospital readmissions, whereas only non-VA users had the lowest rates. Hospitalization rates were 53.7 per 10,000 patient-days in dual users and 26.0 per 10,000 patient-days in only non-VA users. When limited to visits or admissions with HF as the primary diagnosis, ED visits were again highest in dual users (7.0 per 10,000 patient-days) and lowest in only non-VA users (3.6 per 10,000 patient-days) as were HF hospitalizations (7.5 vs 2.8 per 10,000 patient-days). Thirty-day hospital readmission rates after a HF hospitalization follow the same pattern regardless of whether the primary diagnosis on readmission was HF. Table III shows the adjusted rate ratio estimates of the association between count outcomes and dual use after adjusting for age, race, gender, year of visit, and comorbidities that were found to be significant. Among those who were at risk for having an ED visit, the rate of ED visits from any cause was 1.18 (1.14-1.22) times higher in dual users than in VA-only users, whereas the rate for non-VA users was 0.62 (0.60-0.64) times lower than VA-only users. The adjusted hospitalization rate from any cause was 1.93 (1.85-2.01) times higher in dual users than VA-only users and similar in non–VA-only users and VA-only users. The adjusted 30-day hospital readmissions rate from any cause was 1.82 (1.74-1.90) times higher in dual users than VA-only users. As depicted in Supplementary Table I, anemia, depression, fluid and electrolyte disorders, and psychoses were comorbidities selected for inclusion in models focused on health care utilization for all diagnoses. Focusing on primary diagnosis of HF, the rate of ED visits for HF were 1.15 (1.04-1.27) times higher and the rate of hospitalizations for HF were 1.40 (1.26-1.56) times higher in dual users than VA-only users. Moreover, after a hospitalization for HF, 30-day hospital readmission rates for any cause and HF specifically as the primary diagnosis are 1.46 (1.30-1.65) times higher and 1.46 (1.31-1.63) times higher, respectively, in dual users when compared to VA only users. As depicted in Supplementary Table II, fluid and electrolyte disorders and lung conditions were most frequently included in models focused on health care utilization where HF was the primary diagnosis at index encounter.
Discussion We examined a cohort of veterans hospitalized for HF and found that dual users had significantly higher rates of acute health care utilization as compared to VA-only users and non–VA-only users. This was true for ED visits, hospitalizations, and hospital readmissions for any diagnosis and for HF-specific health care visits. To our knowledge, this is the first published report of the impact of dual use on HF-related utilization and outcomes.
American Heart Journal Month Year
4 Axon et al
Table I. Subject characteristics by dual use status Characteristic Age Continuous (mean, y) ≥65 y (%) Gender (%) Male Female Race/ethnicity (%) NHW Non-Hispanic black Hispanic Other Missing Marital status (%) Married Divorced Widowed Never married Unknown Service connected disability (%) N50% Rurality (%) Urban Large town Small town Isolated rural Missing Comorbidities (%) Anemia Cancer Cardiovascular disease Cerebrovascular disease Depression Diabetes Fluid electrolyte disorder Hypertension Hypothyroidism Liver disease Lung conditions Obesity Peripheral vascular disease Psychoses Substance abuse Other⁎
Only VA users, n = 2242
Only non-VA users, n = 8825
Dual users, n = 2910
Total, N = 13,977
P
67.0 52.5
73.5 78.4
68.12 56.0
71.3 69.6
b.0001 b.0001
97.5 2.5
98.2 1.8
97.9 2.1
98.0 2.0
.0801
60.2 37.2 0.4 1.2 1.0
82.2 17.0 0.2 0.7 0.0
62.0 36.6 0.3 1.2 0.0
74.4 24.3 0.3 0.8 0.2
b.0001
57.6 20.0 10.2 10.8 1.4
69.9 9.2 12.7 4.8 3.4
55.0 19.2 12.2 11.3 2.2
64.8 13.0 12.2 7.1 2.8
b.0001
22.8
14.6
21.7
17.4
b.0001
88.5 5.7 1.2 1.1 3.6
88.2 7.0 1.4 0.9 2.5
88.8 5.6 1.2 1.4 3.0
88.4 6.5 1.3 1.0 2.8
.0027
27.7 26.2 74.0 16.6 42.9 61.7 61.4 97.5 13.4 12.6 19.0 41.2 32.0 12.9 25.2 84.7
30.0 28.2 79.4 25.4 35.2 57.8 67.3 97.7 20.9 11.4 21.8 27.9 44.2 10.1 13.1 85.4
40.6 27.9 81.6 33.6 50.4 65.1 81.1 98.4 17.9 17.6 27.1 36.2 44.8 19.6 28.0 91.5
31.8 27.8 79.0 25.7 39.6 60.0 69.2 97.8 19.1 12.9 22.4 31.8 42.4 12.5 18.2 86.6
b.0001 .1646 b.0001 b.0001 b.0001 b.0001 b.0001 .0658 b.0001 b.0001 b.0001 b.0001 b.0001 b.0001 b.0001 b.0001
⁎ Includes AIDS, chronic pulmonary disease, coagulopathy, peptic ulcer disease, renal failure, and rheumatoid arthritis.
Notably, our data sources included information on younger veterans and on episodes of care paid for by carriers other than VA and Medicare. These results are important because they highlight a novel system-level risk factor for HF hospitalization and readmission. These findings also add to the building body of evidence suggesting that dual health care system use may be less efficient and less safe. Previous work has demonstrated that many veterans receive care from both VA and non-VA sources. For example, Weeks et al 16 analyzed patterns of hospital utilization for veterans residing in New York State and found that veterans of all ages relied more heavily on VA for mental health and substance use disorders and less
heavily for circulatory, respiratory, or digestive disorders. In the outpatient realm, rural veterans have been observed to rely on Medicare more for primary care and on VA services for specialty and mental health care. 37 We observed proportions of non-VA hospital utilization that were similar to these prior reports. For instance, Jia et al 18 examined 12-month hospital readmission rates for veterans after stroke, and they found adjusted odds ratios of 1.5 (95% CI 1.1-1.9) for readmission among VA/ Medicare dual users and 2.3 (95% CI 1.2-4.5) among VA/ Medicaid dual users as compared to VA-only users. 18 There are several potential reasons why dual users may face higher hospitalization and readmission rates. First, dual use may be systematically associated with greater
American Heart Journal Volume 0, Number 0
Axon et al 5
Table II. Unadjusted rates of emergency department visits, hospitalization, and hospital readmission per 10,000 patient-days Only VA users, Only non-VA n = 2242 users, n = 8825
Dual users, n = 2910
Table III. Adjusted rate ratios of emergency department visits, hospitalizations, and hospital readmissions⁎ Only VA users, n = 2242
Mean (SD)⁎ All diagnoses ED visits 72.6 Hospitalizations 31.5 (total) 30.6 30-d hospital readmissions (all cause) HF as primary diagnosis ED visits 6.2 HF hospitalizations 6.5 (total) 30-d hospital 7.4 readmissions (all cause) 6.4 30-d hospital readmissions (principle diagnosis HF)
Only non-VA users, n = 8825
Dual users, n = 2910
Rate ratio (95% CI)
(79.0)† (56.7)
45.0 (67.5) 26.0 (34.5)
94.3 (135.4) 53.7 (64.8)
(54.6)
23.2 (32.4)
50.2 (59.5)
(22.8) (27.1)
3.6 (12.6) 2.8 (12.4)
7.0 (17.7) 7.5 (26.8)
(36.5)
2.6 (11.4)
8.8 (31.8)
(27.0)
2.2 (10.2)
7.0 (24.4)
All diagnoses ED visits Hospitalizations (total) 30-d hospital readmissions (all cause) HF as primary diagnosis ED visits HF hospitalizations (total) 30-d hospital readmissions (all cause) 30-d hospital readmissions (principle diagnosis HF)
1.00 1.00
0.62 (0.60-0.64) 1.18 (1.14-1.22) 0.98 (0.95-1.02) 1.93 (1.85-2.01)
1.00
0.87 (0.83-0.90) 1.82 (1.74-1.90)
1.00 1.00
0.60 (0.55-0.66) 1.15 (1.04-1.27) 0.61 (0.55-0.68) 1.40 (1.26-1.56)
1.00
0.51 (0.45-0.57) 1.46 (1.30-1.65)
1.00
0.51 (0.46-0.57) 1.46 (1.31-1.63)
⁎ Count outcomes for ED visits, hospitalizations, and hospital readmissions demonstrated heavy zero inflation and skewness. Thus, SDs are wide. † Event rates per 10,000 patient-days.
⁎ Calculated using ZINB regression. All models adjusted for age, race, gender, year of visit, dual use category, and comorbidities that were found to be significant using a stepwise selection procedure. Zero model also adjusted for year of visit.
variation in care quality leading to overtreatment or undertreatment of HF patients. Higher HF care quality, as measured by Centers for Medicare and Medicaid Services core measures compliance at the hospital level, has been associated with modestly but significantly lower 30-day hospital admission rates for HF patients. 38 Second, dual use may be a marker for poorer actual or perceived access to HF care either in the primary care setting or in a cardiology clinic. Before a hospitalization, poorer care continuity and less timely access to care for patients with acute HF symptoms could lead to less effective management and/or delays in treatment seeking. Heart failure is considered by the Agency for Healthcare Quality and Research to be an ambulatory care sensitive condition, and several prior studies have demonstrated that lower access to timely primary care is associated with higher hospital admission and readmission rates. 27,39,40 It may also be true that patients seeking care outside their usual system of care (ie, VA) may be more likely to be admitted to the hospital than those for whom medical records and follow-up options are more readily available. Although direct evidence of this phenomenon is lacking, hospital admission rates clearly vary widely by area and are a key determinant of hospital readmission rates. 9,10,41 This study should be evaluated in light of certain limitations. First, this is an observational study using administrative data sets, which is subject to the inherent limitations of this analytic approach. Although we attempted to control for variations in known likely confounders, there appears to be significant baseline
differences across the patient groups. Specifically, non– VA-only subjects tended to be older, more likely to be NHW and married, and less likely to have high levels of service connected disability. Demographically, they seem less similar to the other 2 groups. In addition, a higher proportion of dual users had several comorbidities including depression, diabetes, fluid/electrolyte disorders, lung conditions, and liver disease. Although we accounted for these differences in our models, our analysis may be subject to some residual confounding due to unmeasured comorbidities or overall illness burden. Although our research team had access to information on mortality for censoring purposes, we did not analyze this as a specific outcome in our analyses. Finally, we were not able to include information on the severity of HF in this analysis, such as New York Heart Associate class or ejection fraction. If this were available, it might allow for adjustment based on HF severity. Despite these limitations, our results add to a body of evidence suggesting that dual use is associated with higher rates of health care utilization and worse health care outcomes. Our analyses are strengthened by the use of robust statistical methods to account for heavy zero inflation and overdispersion. The VA and other integrated health delivery systems have reason for concern about fractured cross-system care, and care transitions measures such as hospital readmission have received increasing scrutiny in the era of the Accountable Care Act. 42 From a health systems perspective, even if outcomes for dual
American Heart Journal Month Year
6 Axon et al
users are influenced by severity of illness, dual use remains a marker for a patient population that may require more intensive clinical management and care coordination. Additional research is needed to better understand the factors leading to dual use for HF and other conditions in the veteran population and to measure the impacts of dual use in the general population of HF patients.
14.
15. 16.
Appendix. Supplementary data Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.ahj.2015.09.023.
17.
18.
References 1. Go AS, Mozaffarian D, Roger VL, et al. Heart disease and stroke statistics—2013 update: a report from the American Heart Association. Circulation 2013;127:e6-245. 2. Lloyd-Jones D, Adams R, M. C. Heart Disease and Stroke Statistics 2009 Update. A report from the American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Circulation 2008;108: 191261. 3. Blecker S, Ladapo JA, Doran KM, et al. Emergency department visits for heart failure and subsequent hospitalization or observation unit admission. Am Heart J 2014;168:901-8. [e901]. 4. Heidenreich PA, Trogdon JG, Khavjou OA, et al. Forecasting the future of cardiovascular disease in the United States: a policy statement from the American Heart Association. Circulation 2011;123:933-44. 5. Cheng RK, Cox M, Neely ML, et al. Outcomes in patients with heart failure with preserved, borderline, and reduced ejection fraction in the Medicare population. Am Heart J 2014;168:721-30. 6. Chaudhry SI, McAvay G, Chen S, et al. Risk factors for hospital admission among older persons with newly diagnosed heart failure: findings from the Cardiovascular Health Study. J Am Coll Cardiol 2013;61:635-42. 7. Lupon J, Vila J, Bayes-Genis A. Risk prediction tools in patients with heart failure. JACC Heart fail 2015;3:267. 8. Ross JS, Mulvey GK, Stauffer B, et al. Statistical models and patient predictors of readmission for heart failure: a systematic review. Arch Intern Med 2008;168:1371-86. 9. Goodman DC, Fisher ES, Chang C-H. After hospitalization: a Dartmouth Atlas Report on post-acute care for Medicare beneficiaries. In: Bronner KK, ed. The Dartmouth Atlas of Healthcare; 2011. 10. Goodman DC, Fisher ES, Chiang-Hua C. The revolving door: a report on U.S. hospital readmissons. Robert Wood Johnson Foundation. 2013. 11. Adams Jr KF, Fonarow GC, Emerman CL, et al. Characteristics and outcomes of patients hospitalized for heart failure in the United States: rationale, design, and preliminary observations from the first 100,000 cases in the Acute Decompensated Heart Failure National Registry (ADHERE). Am Heart J 2005;149:209-16. 12. Heidenreich PA, Hernandez AF, Yancy CW, et al. Get With The Guidelines program participation, process of care, and outcome for Medicare patients hospitalized with heart failure. Circ Cardiovasc Qual Outcomes 2012;5:37-43. 13. Heidenreich PA, Lewis WR, LaBresh KA, et al. Hospital performance recognition with the Get With The Guidelines Program and mortality
19.
20.
21.
22. 23.
24. 25.
26.
27.
28. 29.
30.
31.
32.
33. 34.
for acute myocardial infarction and heart failure. Am Heart J 2009;158:546-53. Hernandez AF, Fonarow GC, Liang L, et al. The need for multiple measures of hospital quality: results from the Get with the Guidelines– Heart Failure Registry of the American Heart Association. Circulation 2011;124:712-9. Nasir K, Lin Z, Bueno H, et al. Is same-hospital readmission rate a good surrogate for all-hospital readmission rate? Med Care 2010;48:477-81. Weeks WB, West AN, Wallace AE, et al. Comparing the characteristics, utilization, efficiency, and outcomes of VA and non-VA inpatient care provided to VA enrollees: a case study in New York. Med Care 2008;46:863-71. Helmer D, Sambamoorthi U, Shen Y, 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. Jia H, Zheng Y, Reker DM, et al. Multiple system utilization and mortality for veterans with stroke. Stroke 2007;38:355-60. Tarlov E, Lee TA, Weichle TW, et al. Reduced overall and event-free survival among colon cancer patients using dual system care. Cancer Epidemiol Biomarkers Prev 2012;21:2231-41. Wolinsky FD, An H, Liu L, et al. 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. Wolinsky FD, Miller TR, An H, et al. Dual use of Medicare and the Veterans Health Administration: are there adverse health outcomes? BMC Health Serv Res 2006;6:131. Ross JS, Keyhani S, Keenan PS, et al. Dual use of Veterans Affairs services and use of recommended ambulatory care. Med Care 2008;46:309-16. Carey K, Montez-Rath ME, Rosen AK, et al. Use of VA and Medicare services by dually eligible veterans with psychiatric problems. Health Serv Res 2008;43:1164-83. Liu CF, Bolkan C, Chan D, et al. Dual use of VA and non-VA services among primary care patients with depression. J Gen Intern Med 2009;24:305-11. Petersen LA, Byrne MM, Daw CN, et al. Relationship between clinical conditions and use of Veterans Affairs health care among Medicare-enrolled veterans. Health Serv Res 2010;45:762-91. Zhu CW, Penrod JD, Ross JS, et al. 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. 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. South Carolina Revenue and Fiscal Affairs Office. website accessed at, http://rfa.sc.gov/healthcare 2015. [on July 16, 2015.]. Egede LE, Gebregziabher M, Hunt KJ, et al. Regional, geographic, and racial/ethnic variation in glycemic control in a national sample of veterans with diabetes. Diabetes Care 2011;34:938-43. Egede LE, Gebregziabher M, Hunt KJ, et al. Regional, geographic, and ethnic differences in medication adherence among adults with type 2 diabetes (February). Ann Pharmacother 2011;45:169-78. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;43:1130-9. Payne E, Hardin JW, Egede LE, et al. Approaches for dealing with various sources of overdispersion in modeling count data: scale adjustment versus modeling. Stat Methods Med Res May 31, 2015, http://dx.doi.org/10. 1177/0962280215588569. [Epub ahead of print]. Lambert D. Zero-inflated Poisson regression, with an application to defects in manufacturing. Technometrics 1992;34:1-14. Greene WH. Accounting for excess zeros and sample selection in Poisson and negative binomial regression models. New York University Stern School of Business Working Papers. 1994.
American Heart Journal Volume 0, Number 0
35. Atkins DC, Gallop RJ. Re-thinking how family researchers model infrequent outcomes: a tutorial on count regression and zero-inflated models. J Fam Psychol 2007;21:726-35. 36. Cameron AC, Trivedi PK. Regression analysis of count data. Cambridge, UK: Cambridge University Press. 1998. 37. Weeks WB, Bott DM, Lamkin RP, et al. Veterans Health Administration and Medicare outpatient health care utilization by older rural and urban New England veterans. J Rural Health 2005;21:167-71. 38. Heidenreich PA, Zhao X, Hernandez AF, et al. Patient and hospital characteristics associated with traditional measures of inpatient quality of care for patients with heart failure. Am Heart J 2012;163: 239-45. [e233].
Axon et al 7
39. Ambulatory care sensitive conditions: age-standardized acute care hospitalization rate for conditions where appropriate ambulatory care prevents or reduces the need for admission to the hospital, per 100,000 population younger than age 75 years. Available at: http://www.qualitymeasures.ahrq.gov/content.aspx?id=35186 2014. Accessed 3/21/2014. 40. Axon RN. Frequency and costs of hospital transfers for ambulatory care sensitive conditions. Am J Manag Care 2015;21:11-9. 41. Pines JM, Mutter RL, Zocchi MS. Variation in emergency department admission rates across the United States. Med Care Res Rev 2013;70: 218-31. 42. Axon RN, Williams MV. Hospital readmission as an accountability measure. J Am Med Assoc 2011;305:504-5.