|
|
| |||||||||||||||||||||||||||||||||||||||||||||
Background Some have the opinion that patients cared for in Veterans Health Administration (VHA) hospitals receive care of poorer quality than those cared for in non-VHA institutions. To assess the quality of care in VHA hospitals, we compared the outcome of acute myocardial infarction among patients in VHA and non-VHA institutions while controlling for potential confounders, including coexisting conditions and severity of illness.
Methods We studied 2486 veterans discharged from 81 VHA hospitals and 29,249 Medicare patients discharged from 1530 non-VHA hospitals, restricting our samples to men at least 65 years of age who were discharged with confirmed acute myocardial infarction. We compared coexisting conditions, severity of illness, and 30-day and 1-year mortality in the two samples.
Results VHA patients were significantly more likely than Medicare patients to have a recorded history of hypertension (64.3 percent vs. 57.3 percent), chronic obstructive pulmonary disease or asthma (30.9 percent vs. 23.5 percent), diabetes (34.8 percent vs. 29.0 percent), stroke (20.4 percent vs. 14.2 percent), or dementia (7.2 percent vs. 4.8 percent) (P<0.001 for all comparisons). According to both multivariate logistic regression and an analysis using 2265 matched pairs of VHA and Medicare patients, there were no significant differences in 30-day or 1-year mortality. The matched-pairs analysis found that the difference in mortality at 30 days (the mortality rate among Medicare patients minus the mortality rate among VHA patients), averaged over the 5-year age groups, was –0.8 percent (95 percent confidence interval, –2.8 to 1.3), and the difference in mortality at 1 year was –1.3 percent (95 percent confidence interval, –3.9 to 1.3).
Conclusions VHA patients had more coexisting conditions than Medicare patients. Nevertheless, we found no significant difference in mortality between VHA and Medicare patients, a result that suggests a similar quality of care for acute myocardial infarction.
In contrast to health care provided by the VHA, Medicare coverage for the majority of its 39 million beneficiaries consists of indemnity insurance combined with fee-for-service payments to physicians.2 A patient 65 years of age or older becomes eligible by having worked for at least 10 years in Medicare-covered employment, and a younger patient becomes eligible because of disability (including end-stage renal disease).
Some have the opinion that patients cared for in VHA institutions receive care of poorer quality than those cared for in non-VHA institutions.3 However, valid comparisons between VHA and non-VHA care are difficult to carry out, since VHA patients may have more coexisting conditions and greater severity of illness than patients in non-VHA institutions,4,5 confounding comparisons of outcome as a measure of quality of care. Comparisons between VHA and non-VHA care may also be confounded by differences in the patients' age, sex, and socioeconomic status6 and by unmeasured differences in patients' preferences for site of care. Differences among hospitals in the availability of services and in teaching status may also affect the validity of comparisons.
Given these problems with comparisons, there are few published data to address the relative quality of VHA and non-VHA health care. We are aware of only two studies, conducted at a single VHA hospital, that used clinical data to compare the outcome of care in VHA and non-VHA institutions.5,7 Our goal was to compare the coexisting conditions, severity of illness, and outcome of acute myocardial infarction in VHA and non-VHA hospitals using nationally representative clinical data. We assessed outcome (as defined by Donabedian8) to measure quality. Our first hypothesis was that VHA patients had a greater burden of illness at the time of admission with acute myocardial infarction. Our second hypothesis was that after adjustment for patients' characteristics, mortality rates would be higher among VHA patients than among Medicare patients, reflecting poorer quality of care.
Methods
We created two cohorts retrospectively: one consisted of all fee-for-service Medicare beneficiaries discharged with a diagnosis of acute myocardial infarction from acute care hospitals located in seven states, and the other consisted of a national sample of veterans discharged with a diagnosis of acute myocardial infarction from nonpsychiatric VHA facilities.
Medicare Sample
The Medicare sample was obtained through the Cooperative Cardiovascular Project, which was undertaken by HCFA to improve the quality of care for Medicare patients with acute myocardial infarction.9 As part of the Cooperative Cardiovascular Project, HCFA studied all patients discharged with a principal diagnosis of acute myocardial infarction (code 410 of the International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM],10 excluding a fifth digit of 2, which would indicate acute myocardial infarction in the preceding eight weeks) from all nonfederal acute care hospitals in each state during a specified eight-month period between January 1, 1994, and June 30, 1995. Our cohort is a subgroup of the larger HCFA study, including all patients in the Cooperative Cardiovascular Project who were discharged from hospitals in California, Florida, Massachusetts, New York, Ohio, Pennsylvania, and Texas. These states were selected because they were known to differ in the frequency of use of cardiac procedures, are large and geographically diverse, and served as the basis of a study examining the appropriateness of care after acute myocardial infarction.11,12 Eliminating women, we initially identified 41,754 male Medicare beneficiaries who were 65 years of age or older.
We excluded 12,505 patients from the study. The numbers of excluded patients and the reasons for their exclusion were as follows (some patients met more than one of the exclusion criteria): 4498 patients did not meet the clinical criteria for acute myocardial infarction9 (creatine kinase MB fraction above 0.05; lactate dehydrogenase level exceeding 1.5 times the upper limit of normal, with isoenzyme 1 level higher than isoenzyme 2 level; or the presence of two of the following three conditions: chest pain, a doubling of the creatine kinase level, or detection of a new acute myocardial infarction on an electrocardiogram); 6576 patients were discharged from the hospital to which they were transferred without a diagnosis of ICD-9-CM code 410; 423 patients were discharged alive after a length of stay of less than three days; data were incomplete for 1969 patients who resided outside the United States, whose medical records were unavailable, or who were transferred more than once; 3196 patients were enrolled in a health maintenance organization at the time of the index event, and we could not ascertain the details of their subsequent care; 641 patients were admitted or discharged outside the study period; 187 patients were transferred to hospitals out of the seven study states; 10 patients were hospitalized for more than 180 days; and an incorrect date of death was recorded for 10 patients (e.g., a date of death before admission). After these exclusions, there remained 29,249 male Medicare patients who were at least 65 years of age and who had been discharged from 1530 non-VHA hospitals.
VHA Sample
Since there are many fewer VHA hospitals than non-VHA hospitals in the seven sampled states, we used a national VHA sample. We used the Patient Treatment File, a national centralized data base that records all use of VHA facilities, to identify all male patients with a primary diagnosis of acute myocardial infarction (ICD-9-CM code 410) who were discharged between January 1, 1994, and September 30, 1995. Only patients discharged from nonpsychiatric VHA facilities with a length of stay of at least three days (if discharged alive) and without a fifth digit ICD-9-CM code of 2 were included.13 We initially identified 13,310 VHA patients.
We sampled patients stratified according to the capability of the hospital cardiac service.14,15,16 Each of the 139 VHA hospitals was classified as one of four types. Noncatheterization hospitals do not perform on-site catheterization, percutaneous transluminal coronary angioplasty, or coronary-artery bypass grafting. The hospitals in the lowest quartile of admissions for acute myocardial infarction were defined as having a low volume, and all other hospitals were defined as having a high volume. Cardiac-catheterization-only hospitals perform on-site catheterization but not revascularization. Cardiac-surgery hospitals perform all the cardiac procedures named above. Within each type of hospital, we randomly included up to 100 patients, generating a stratified national random sample of 5503 patients from 81 VHA hospitals. We reviewed the records of 5193 (94.4 percent) of these patients.
We excluded 2707 patients for one or more reasons: 433 who did not meet the clinical criteria for acute myocardial infarction,9 2029 who were less than 65 years of age, 181 who were discharged to an acute care non-VHA facility, and 64 for whom information was incomplete (missing discharge date or date of birth). After these exclusions, there remained 2486 male veterans who were at least 65 years of age and who had been discharged from 81 VHA hospitals.
Data Sources
We used the Cooperative Cardiovascular Project17 structured review instrument to obtain medical-records data for Medicare patients. The records of transfer patients were linked with the initial admitting hospital. The data were abstracted at two abstraction centers under contract with HCFA.9,18 The overall agreement between abstracters averaged 95 percent.9 Mortality was determined from the Health Insurance Master File. The HCFA Provider of Service File and the American Hospital Association data bases were linked to our sample to obtain structural characteristics of the hospitals, such as availability of cardiac services, teaching affiliation, and number of beds.
Medical-records data for the VHA patients were abstracted by trained nurses using the Cooperative Cardiovascular Project interactive software.19 Overall agreement between abstracters with respect to key variables was 96 percent. Mortality was determined from the inpatient discharge status in the Patient Treatment File as well as from the Veterans Affairs Beneficiary Identification and Record Location Subsystem (BIRLS).20,21 We obtained the characteristics of the VHA hospitals from the American Hospital Association data base, the Department of Veterans Affairs Cardiac Services Directory, and the 1995 version of the Federal Practitioner. For both samples, a hospital was considered to have a university teaching affiliation if it had at least one intern or resident in an accredited allopathic or osteopathic residency training program according to the American Hospital Association data base.
Statistical Analysis
Differences in Coexisting Conditions and Severity of Acute Myocardial Infarction
We calculated the frequency of coexisting conditions and measures of the severity of acute myocardial infarction in both samples.22 Chi-square tests were used to examine differences between the two groups in discrete variables, and t-tests were used to examine differences in continuous variables. For continuous variables, means ±SD were calculated. When appropriate, we also calculated the frequency with which a variable was missing or a test was not performed. We adjusted for multiple comparisons using a Bonferroni adjustment, so that the overall level of significance was 0.05 within each class of comparisons.
Differences in Mortality
In addition to calculating crude 30-day and 1-year rates of mortality from all causes, we estimated differences in mortality by two adjustment methods. First, we calculated the risk-adjusted odds of mortality for Medicare patients relative to VHA patients using logistic regression, the standard method of controlling simultaneously for observed covariates.23 We included in the model sociodemographic features, coexisting conditions, severity of acute myocardial infarction, hospital characteristics (availability of cardiac services and presence or absence of university teaching affiliation), and other admission characteristics known to affect mortality, as well as a binary variable indicating to which cohort (Medicare or VHA) the patient belonged.22
Second, because we might not have accounted for all confounders appropriately, we created a matched sample using a propensity-score approach to compare the survival of VHA and Medicare patients.24 This technique balances patients from each cohort on the basis of observed characteristics to replicate a randomized, controlled trial. To create the propensity scores, we created a logistic-regression model in which the response was the log of the odds of belonging to the VHA cohort. Characteristics of the patient (sociodemographic features, coexisting conditions, and severity of acute myocardial infarction) and the hospital (availability of cardiac services and presence or absence of university affiliation) were included in the model. Once the model was fitted, we stratified the sample according to the cardiac services available in the hospital (noncatheterization services, catheterization only, or cardiac surgery) and five-year age group. Within each of these strata, we matched each VHA patient to the Medicare patient with the closest estimated propensity score. We included in our analyses only the matches that were within 0.60 of the pooled standard error of the estimated propensity score.25 If no match was found, the VHA patient was excluded from the analyses. To measure how well we balanced the two cohorts in terms of observed covariates, we calculated how far apart the two cohorts were in terms of each observed covariate. To achieve this, we measured the average difference in each covariate, expressed as a percentage of the pooled standard deviation of the covariate, before and after matching. For example, if the fraction of VHA patients who were black minus the fraction of Medicare patients who were black, as measured by the percentage of the pooled standard deviation, was 20 percent, the means of the two samples were considered to differ by 2/10 of a standard deviation. Differences larger than 10 percent indicate that the two samples were far apart in terms of the distribution of the covariate and therefore might not be appropriately balanced.26
Using the matched pairs, we then estimated differences in 30-day and 1-year mortality. Paired differences (and their corresponding standard errors) were calculated within each five-year age group and then within each hospital type. We estimated the overall average difference by combining the paired differences among the five-year age groups using precision weights27,28 and constructed a 95 percent confidence interval for the overall average difference. We repeated this across the hospital strata, calculating the precision-weighted average difference in mortality across the three types of hospitals.
Results
Characteristics of the Patients and the Hospitals
Table 1 shows the sociodemographic features of the patients and the characteristics of the hospitals. VHA patients were younger, were less likely to be white, and had a longer inpatient stay than Medicare patients. Almost half the VHA patients were initially admitted to a noncatheterization hospital, whereas half the Medicare patients were initially admitted to a cardiac-surgery hospital. More VHA patients than Medicare patients were initially admitted to a university-affiliated hospital.
|
Table 2 shows the clinical characteristics of the patients. VHA patients were significantly more likely than Medicare patients to have a history of various coexisting conditions, such as hypertension, chronic obstructive pulmonary disease or asthma, diabetes, stroke, and dementia.
|
|
The unadjusted 30-day mortality rates for VHA and Medicare patients were 17.3 percent and 18.1 percent, respectively (P=0.30) (Table 4). At one year, the unadjusted mortality rates were also nearly indistinguishable in the two cohorts: 31.5 percent for VHA patients and 31.8 percent for Medicare patients (P=0.77).
|
Table 4 shows the odds ratios and 95 percent confidence intervals for each covariate used in the logistic model. The estimated odds of 30-day mortality for Medicare patients as compared with VHA patients were 0.94 (95 percent confidence interval, 0.82 to 1.07; receiver-operating-characteristic curve, 0.800). The corresponding estimate at one year was 0.94 (95 percent confidence interval, 0.84 to 1.05; receiver-operating-characteristic curve, 0.799).
Matched Analysis
Using a model with 29 covariates to predict VHA use, we were able to obtain an accuracy of 88 percent (receiver-operating-characteristic curve, 0.88) and to match 2265 (91.1 percent) of the VHA patients to Medicare patients. Before matching, 16 of the 29 covariates had a standardized difference larger than 10 percent, whereas after matching, all standardized differences were less than 5 percent.
Using the matched sample, we observed no statistically significant difference in 30-day or 1-year mortality between Medicare and VHA patients. The difference in mortality at 30 days (mortality among Medicare patients minus mortality among VHA patients) averaged over the 5-year age groups was –0.8 percent (95 percent confidence interval, –2.8 to 1.3), and the difference in mortality at 1 year was –1.3 percent (95 percent confidence interval, –3.9 to 1.3). After averaging among types of hospital, similar results were observed: –0.8 percent for 30-day mortality (95 percent confidence interval, –2.9 to 1.3) and –0.9 percent for 1-year mortality (95 percent confidence interval, –3.6 to 1.7).
Our findings with regard to mortality were not consistent with our prior hypotheses. To try to explain our findings, and because survival is improved by the use of a number of medications after acute myocardial infarction,29 we calculated the crude rates of use of thrombolytic agents at the time of arrival at the hospital and of use of beta-blockers, angiotensin-converting–enzyme inhibitors, aspirin, or combinations of these drugs at the time of discharge. A similar percentage of VHA and Medicare patients underwent thrombolysis at the time of arrival (15.8 percent vs. 16.9 percent, P=0.16). Of the patients who survived to discharge, more VHA patients than Medicare patients were receiving beta-blockers (49.7 percent vs. 41.6 percent, P<0.001), angiotensin-converting–enzyme inhibitors (44.6 percent vs. 32.5 percent, P<0.001), or aspirin (77.2 percent vs. 68.6 percent, P<0.001) at discharge.
Discussion
We conducted a national study using clinical data to compare the coexisting conditions, severity of acute myocardial infarction, and outcome in patients cared for in VHA hospitals and patients cared for under fee-for-service Medicare financing. To minimize potential confounding, we collected comparable data from a uniform clinical cohort, restricted our samples to men 65 years or age or more, and matched the cohorts according to patient and hospital characteristics to carry out risk-adjusted comparisons of mortality.
We found that among elderly men with acute myocardial infarction, those treated at VHA hospitals are more likely to have coexisting conditions, such as chronic obstructive pulmonary disease or asthma, diabetes, hypertension, stroke, and dementia, than those cared for under Medicare financing. Given these base-line differences, the ideal way to answer our question of whether the quality of care is poorer in VHA hospitals would be to randomly assign patients with acute myocardial infarction to either fee-for-service Medicare or VHA care and then to examine their long-term outcome. Given that we could not perform this experiment, we used a propensity-score approach to replicate randomization within our observational data set. Using this technique to match patients on the basis of their propensity to use VHA services, we compared survival among those who were treated within the VHA system and those who were not, within each type of hospital and five-year age group. After this matching technique had been applied, there were no significant differences between the two groups of patients in 30-day or 1-year mortality.
Our findings with regard to mortality were not consistent with our prior hypotheses. The use of thrombolytic agents was similar in both groups, but more VHA patients than Medicare patients were receiving beta-blockers, angiotensin-converting–enzyme inhibitors, or aspirin at discharge. We postulate that the similar outcomes with regard to mortality are at least partly accounted for by the higher rates of use of medications that are known to decrease mortality in the VHA cohort. Of course, none of these unadjusted rates of use account for important differences among patients (in, for example, age or coexisting conditions), hospitals, or indications for or contraindications to treatment. Thus, process of care, as defined by Donabedian,8 in VHA hospitals may be better (or worse), but our data do not allow us to judge that dimension of the quality of health care.
To assess whether our exclusions from the VHA cohort might have biased our mortality findings in favor of the VHA, we determined the one-year mortality rate for the patients who were discharged to an acute care hospital from a VHA facility and were excluded from our sample. Among these 181 patients, one-year mortality was 30 percent, suggesting that the transfer of sicker patients out of VHA hospitals did not explain our findings. Since only 8.9 percent of our patients were transferred from another VHA hospital, it is unlikely that our mortality findings could be accounted for by transfer into VHA hospitals of patients who received care elsewhere. Because there is little variation in the population-based rates of hospitalization for acute myocardial infarction in most areas30 and little disagreement about the appropriateness of hospitalization for this condition, it is unlikely that differences in the threshold for admission for VHA and Medicare patients confounded our findings. Furthermore, the findings in Table 3 do not support the view that VHA hospitals selectively admit patients with less severe disease.
Our findings with regard to mortality are consistent with those of a previous single-institution study that used clinical data to compare outcomes of VHA patients and those treated by the private sector,7 although the authors of that study noted that their power to demonstrate even a 50 percent difference in mortality was only 55 percent. Other comparative studies have used administrative data and either have reported unadjusted mortality rates31 or have not been able to account for a number of possible confounders of their results.32,33
From data-base studies of patients who use both VHA and Medicare services, we know that most users of VHA services are initially hospitalized for acute myocardial infarction under Medicare financing.34 This is because ambulances that are called to assist patients with cardiac symptoms may be required to take them to the nearest emergency department, which may not be a VHA facility. Thus, the findings of this study cannot be generalized to all veterans with acute myocardial infarction, but only to veterans who are cared for in VHA hospitals for this condition.
There are many methodologic challenges to be overcome in carrying out comparisons between different systems of health care, but given the size, scope, and budget of the VHA health care system, the lack of such comparative data is surprising. We have demonstrated that patients cared for in VHA hospitals have a greater burden of illness than patients cared for under Medicare financing. With extensive risk adjustment, we have found no differences in mortality between VHA patients and fee-for-service Medicare patients, suggesting that VHA patients receive care of similar quality for acute myocardial infarction.
Supported by grants from the Department of Veterans Affairs Health Services Research and Development Service (IIR 94-054), the Department of Veterans Affairs Cooperative Studies Program (EPC 97-069), and the Agency for Healthcare Quality and Research (R01-HS08071). Dr. Petersen has received a Research Career Development Award (RCD 95-306) from the Veterans Affairs Health Services Research and Development Service.
We are indebted to Margaret Volya, M.Sc., Harvard Medical School, and Caterina Brown, Boston Veterans Affairs Health Care System, for expert SAS programming and data management; and to George E. Thibault, M.D., for advice and support.
Source Information
From the Houston Center for Quality of Care and Utilization Studies, Houston Veterans Affairs Medical Center, and the Section for Health Services Research, Department of Medicine, Baylor College of Medicine, Houston (L.A.P.); the Department of Health Care Policy, Harvard Medical School, Boston (S.-L.T.N., B.J.M.); the Institute for Health Policy, Partners Health Care System, and the Department of Medicine, Massachusetts General Hospital, Boston (J.D.); the Department of Biostatistics, Harvard School of Public Health, Boston (S.-L.T.N.); and the Department of Radiology, Brigham and Women's Hospital, Boston (B.J.M.). The views expressed in this article are solely those of the authors and do not necessarily represent those of the Department of Veterans Affairs.
Address reprint requests to Dr. Petersen at Health Services Research and Development (152) (T110), Houston Veterans Affairs Medical Center, 2002 Holcombe Blvd., Houston, TX 77030.
References
| |||||||||||||||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
