Although revascularization procedures are increasingly done in the elderly, information on their outcomes has come almost exclusively from single-institution reviews and a few multicenter registries [6-15]. Although these case series provide important clinical information, they have limitations. First, cases contained in these reviews often date back more than a decade and may not be representative of current cardiovascular practice. Sample size is often limited in institutional reviews, which partially explains the considerable interstudy variation in procedure-related mortality results. Second, published case series represent the efficacy of revascularization procedures at major academic and high-volume centers and may overestimate the effectiveness of these procedures in general community practice [16, 17]. Finally, secular trends in procedure-related outcomes from single-institution reviews are difficult to follow because they may be biased by variations in an institution's referral population or by publication biases [18-20].

We examined use and outcome of revascularization procedures in the Medicare population. Because Medicare provides insurance coverage for most elderly persons in the United States, our results represent national benchmarks for angioplasty and bypass surgery mortality. We also had sufficient sample size to examine procedure-related outcomes according to age, race, and sex. Finally, Medicare's longitudinal database enabled us to examine secular changes in mortality after revascularization procedures for 1987 to 1990.

Methods

Top Methods Results Discussion Author & Article Info References

Data Sources

We obtained data from the Medicare Provider Analysis and Review file provided by the Health Care Financing Administration on Medicare Part A enrollees who had revascularization procedures from 1987 to 1990. This file contains demographic and limited clinical information on all inpatient hospitalizations that are billed to Medicare. These data include the enrollees' Health Insurance Claims number, age, sex, race, discharge status (including death), and as many as five discharge diagnoses and three procedures as identified by International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) codes [21]. We obtained Medicare eligibility and health maintenance organization enrollment from Medicare's finder files by matching insurance claims numbers. Mortality data through the end of 1991 came from Medicare records and date-of-death files. Using individual patient identifiers, we linked subsequent hospital admissions and procedures to form a longitudinal record of a patient's hospitalized care for 1987 to 1990.

Patients

Our study sample consisted of all patients aged 65 years or older listed in the Medicare files who had angioplasty or bypass surgery between 1987 and 1990. Patients in the angioplasty cohort were selected by the first angioplasty procedure coded in the file (ICD-9-CM codes 36.01, 36.02, or 36.05). Patients in the bypass surgery cohort were selected by the first bypass surgery procedure coded in the file (ICD-9-CM codes 36.10 through 36.19). To avoid counting patients twice, we included patients who had multiple Medicare-covered revascularization procedures during the study period only once, defined by the initial revascularization procedure done during the study period. Patients coded as having had both angioplasty and bypass surgery on the same hospital day were considered to have had angioplasty with a subsequent complication that required bypass surgery; these patients were thus assigned to the angioplasty cohort.

We excluded patients from the total Medicare revascularization cohort for any of the following reasons, with some patients meeting more than one of the criteria: 1) nonelderly Medicare eligibility: We excluded patients who received Medicare benefits for reasons other than being older than 65 years, including being part of the Railroad Retirement Board (1.4%), being eligible for Medicare but younger than 65 years (4.8%), and having end-stage renal disease [0.2%] as a reason for Medicare entitlement; 2) incomplete data: Because Medicare does not have continuous, reliable data on patients enrolled in health maintenance organizations, we excluded patients who were enrolled in such organizations during the study (4.1%), as well as patients who were admitted to non-U.S. hospitals or federal hospitals [0.5%]; 3) valve replacement surgery: We wished to focus on outcomes after coronary revascularization in the treatment of ischemic heart disease. Patients having concomitant cardiac valve surgery (ICD-9-CM codes 35.0 to 35.3) and bypass surgery may have primarily presented with valve dysfunction and had bypass grafts placed at the time of surgery to obviate the need for a second open-chest procedure. Patients having valve replacement and bypass surgery are also at greater risk for procedure-related mortality than are patients having only bypass surgery [22, 23]. To establish a more uniform cohort, we excluded patients who had both valve and bypass surgery procedures (5.0%).

We excluded 14.9% of the 685 624 initially identified patients who had revascularization, producing a final cohort of 225 915 patients who had angioplasty and 357 885 patients who had bypass surgery.

Data Analysis

We determined the frequencies of demographic and clinical characteristics separately for the angioplasty and bypass surgery cohorts. Age was calculated at the time of revascularization and was used directly or subdivided into four categories: ages 65 to 69 years, 70 to 74 years, 75 to 79 years, and 80 years and older. Patients were considered to have been admitted for an acute myocardial infarction if the patient's primary diagnosis was ICD-9-CM code 410.xx or if the patient had a secondary diagnosis of myocardial infarction and a primary diagnosis of a complication of infarction (such as papillary muscle rupture). In addition, for an acute myocardial infarction admission, the patient's total length of stay must have been 3 days or longer, after linkage of interhospital transfers, if the patient was discharged alive [24].

We used two methods to adjust for the effects of comorbidity on patient outcomes. First, we examined the frequency of individual comorbid diseases (for example, congestive heart failure or pulmonary disease) using previously developed ICD-9-CM mapping strategies [25, 26]. For analysis, we included each disease condition that was coded in more than 1.0% of our patients. Second, we calculated a previously developed composite measure of comorbidity, a modified Charlson index, using the ICD-9-CM diagnoses coded at hospital discharge after revascularization [25-27]. Because we intended to examine separately the effect of acute myocardial infarction on patient outcomes, we removed this condition from our calculated Charlson index scores.

We calculated procedure use rates by dividing the number of initial procedures done during a given year by the number of Medicare-eligible patients for that year. The denominator for this rate was derived from Medicare's Enrollment Database after we excluded patients for the previously described reasons. Patients in the angioplasty cohort were classified as having had either single-vessel angioplasty (ICD-9-CM codes 36.01 or 36.02) or multivessel angioplasty (ICD-9-CM code 36.05). Using ICD-9-CM codes 36.15 to 36.16, we also identified the use of left or right internal mammary artery grafts during bypass surgery.

The survival rate after revascularization procedure was calculated from the procedure date forward. The procedure date was missing for 329 (<0.1%) of the records; in these cases, the admission date was substituted. Because we had complete survival follow-up without censoring through the end of 1991, unadjusted mortality rates for as long as 1 year after revascularization procedures were calculated as simple ratios.

To examine the significance of temporal trends in patient characteristics, we applied linear regression models (for continuous variables) and categorical models (for categorical variables). In these models, the year that the procedure was done was the independent variable, and the patient characteristic was the dependent variable.

Using logistic regression models, we calculated adjusted odds ratios estimating the risk for 30-day and 1-year mortality after revascularization procedures. We selected variables entered into the models on the basis of univariate association with mortality after procedures (P < 0.01) or their clinical relevance, or both. We adjusted for the effects of comorbidity on procedure-related mortality using two techniques. First, we entered individual comorbid diseases that were coded in more than 1% of the revascularization cases as separate independent variables in our models. We also examined the Charlson index as a single composite variable reflecting comorbidity. Because these methods of adjustment did not significantly differ in the changes in procedure-related mortality over time, we selected the composite variable, the Charlson index, for simplicity. The final regression models simultaneously adjusted for patient age, race, sex, Charlson index score, whether the procedure occurred during an admission for acute myocardial infarction, and the year the procedure was done. From these models, we calculated adjusted odds ratios for procedure-related mortality in a given year and 95% confidence intervals around these mortality estimates.

Results

Top Methods Results Discussion Author & Article Info References

Procedure Use

The rate of angioplasty procedures done in the Medicare-eligible population increased by 55% from 15.7 per 10 000 cases in 1987 to 24.3 per 10 000 cases in 1990. The percentage of these procedures coded as multivessel angioplasty also increased from 2.2% to 12.3%. The rate of bypass surgery procedures done in the Medicare-eligible population increased by 18% from 28.7 per 10 000 cases in 1987 to 34.0 per 10 000 cases in 1990. The use of internal mammary graft or grafts during bypass surgery procedures also increased from 27.9% to 40.4% during the study period.

Tables 1 and 2 show secular changes in the characteristics of patients having revascularization procedures. Medicare patients who had a revascularization procedure in 1990 were significantly older and less likely to be white in 1990 than were patients in 1987. The percentage of women who had angioplasty increased marginally, whereas the proportions of men and women who had bypass surgery did not significantly change. Finally, patients who had angioplasty and bypass surgery in 1990 had significantly higher Charlson comorbidity scores and were more likely to have their procedure after admission for an acute myocardial infarction than patients who had these procedures in 1987.

Changes in the frequency of individual coded comorbid conditions in the angioplasty and bypass surgery cohorts are listed in the Appendix Table. Between 1987 and 1990, the percentage of patients who had angioplasty and bypass surgery for revascularization and who were coded as having congestive heart failure, pulmonary disease, or diabetes with complications increased significantly, whereas the percentage of patients with peripheral vascular disease decreased slightly.

Procedure-related Outcomes

Overall unadjusted 30-day and 1-year mortality rates after angioplasty were 3.3% and 8.0%, respectively. Between 1987 and 1990, unadjusted 30-day mortality rates after angioplasty decreased by 25% (95% CI, 22% to 28%), and 1-year mortality decreased by 10% (CI, 8% to 11%) (Figure 1). Overall unadjusted 30-day and 1-year mortality rates after bypass surgery were 5.8% and 11.0%, respectively. During the 4 years of study, unadjusted 30-day mortality rates after bypass surgery also decreased by 12% (CI, 10% to 14%), and 1-year mortality decreased by 8% (CI, 7% to 10%) (Figure 2).

View larger version (25K): [in this window] [in a new window]   Figure 1. Unadjusted 30-day and 1-year mortality after angioplasty by year of procedure (1987 to 1990).

View larger version (24K): [in this window] [in a new window]   Figure 2. Unadjusted 30-day and 1-year mortality after bypass surgery by year of procedure (1987 to 1990).

Tables 3 and 4 show the effect of age, race, sex, comorbidity, and acute myocardial infarction on mortality after revascularization procedures. Unadjusted 30-day and 1-year mortality rates after revascularization procedures increased markedly with increasing patient age. Female and black patients had higher procedure-related mortality than did male and white patients, respectively. Patients who were admitted for an acute myocardial infarction, who had more coded discharge diagnoses, and who had higher Charlson comorbidity scores had higher procedure-related mortality.

Figures 3 and 4 show changes over time in adjusted mortality after angioplasty and bypass surgery. We calculated adjusted odds ratios for mortality to examine shifts in procedure-related outcome at 30 days and at 1 year after controlling for age, race, sex, Charlson index score, and whether the procedure occurred during an admission for myocardial infarction. Patients who had angioplasty in 1990 were 0.63 (CI, 0.59 to 0.68) times as likely to die within 30 days and 0.78 (CI, 0.75 to 0.82) times as likely to die within 1 year as were patients who had angioplasty in 1987. Similarly, patients who had bypass surgery in 1990 were 0.82 (CI, 0.79 to 0.86) times as likely to die within 30 days and 0.81 (CI, 0.79 to 0.84) times as likely to die within 1 year as were patients who had bypass surgery in 1987.

View larger version (14K): [in this window] [in a new window]   Figure 3. Adjusted 30-day and 1-year odds ratio mortality after angioplasty for 1990 and 1987 after controlling for patient age, race, sex, Charlson index score, and acute myocardial infarction on admission. The 95% confidence intervals for the odds ratio are shown as bars on each side of the point estimate. If the odds ratio and confidence interval are less than 1.0, the adjusted likelihood of mortality after angioplasty in that year is significantly lower than the mortality for procedures done in 1987.

View larger version (15K): [in this window] [in a new window]   Figure 4. Adjusted 30-day and 1-year odds ratio mortality after bypass surgery for 1990 and 1987 after controlling for patient age, race, sex, Charlson index score, and acute myocardial infarction on admission. The 95% confidence intervals for the odds ratio are shown as bars on each side of the point estimate. If the odds ratio and confidence interval are less than 1.0, the adjusted likelihood of mortality after bypass surgery in that year is significantly lower than the mortality for procedures done in 1987.

Patients who received a mammary artery graft during bypass surgery had markedly lower mortality than did patients who received saphenous vein grafts both at 30 days (3.0% and 7.3%, respectively [P < 0.001]) and at 1 year (6.4% and 13.3%, respectively [P < 0.001]). This inverse association between mammary grafts and procedure-related mortality persisted after adjustment for age, race, sex, admission for myocardial infarction, year of procedure, and Charlson index score. After adjustment for these characteristics, 30-day and 1-year odds ratios for mortality in patients who received mammary grafts were 0.46 (CI, 0.44 to 0.47) and 0.51 (CI, 0.50 to 0.53), respectively, compared with patients who received saphenous vein grafts.

Discussion

Top Methods Results Discussion Author & Article Info References

Revascularization procedures are increasingly used in the elderly. We found that between 1987 and 1990, the rates of angioplasty and bypass surgery done in the elderly increased by 55% and 18%, respectively. The risk profile for patients having these interventions also appears to have increased during this period. Despite apparently higher risk, unadjusted 30-day and 1-year mortality after revascularization actually decreased between 1987 and 1990. After adjustment for these changes in patient characteristics, 30-day mortality after angioplasty and bypass surgery decreased by 37% and 18%, respectively, and 1-year mortality decreased by 22% and 19%, respectively.

Explanation for Improved Outcomes

There are several potential explanations for these secular improvements in revascularization outcome. First, revascularization procedures may have been applied to healthier patients with less-advanced disease or comorbidity. Medicare administrative files lack several important clinical measures of coronary disease severity, including coronary anatomy. Compared with clinical databases, administrative data also may result in undercoding of comorbid illness [26-28]. However, available measures of the case mix of patients (including patient age; the percentage of patients admitted with an acute myocardial infarction before revascularization; and the frequency of patients coded as having congestive heart failure, pulmonary disease, and diabetes with complications) increased significantly between 1987 and 1990. Because each of these characteristics was an independent predictor of procedure-related mortality Table 3, patients who had angioplasty and bypass surgery in 1990 appear to have increased baseline risk profiles compared with patients who had those procedures in 1987.

Temporal changes in coding practices, however, could have partially explained the increased comorbidity noted in our cohort [29, 30]. To examine the potential effect of coding "creep" on our adjusted mortality estimates, we conservatively assumed that all increases in Charlson comorbidity index scores were caused by inflationary coding practices rather than by actual changes in the patients' comorbidity. To explore this possibility, we rescaled ("deflated") the Charlson index scores from 1988 to 1990 back to the 1987 levels and recalculated our risk-adjusted mortality estimates. Even with this conservative assumption regarding comorbidity, however, we found equivalent decreases in estimated procedure-related and 1-year mortality over the 4-year period. Finally, we examined the case mix of patients in previously published institutional reviews and found similar increases in disease severity and comorbidity among patients who are referred for revascularization procedures [31-35]. Although it remains possible that patients who had revascularization in general community practices were healthier in 1990 than in 1987, available evidence argues against this conclusion.

Second, improved procedure-related outcomes may have resulted from increasing experience in performing revascularization procedures in the elderly. Institutions doing more bypass surgery have been shown to have lower procedure-related mortality [36-38]. This relation is particularly pronounced in the high-risk patient populations in which many of the elderly are classified [39]. Most recently, a similar inverse relation between angioplasty volume and procedure-related complications was shown in a California administrative database [17] and in Medicare files [40]. The reasons for this relation between experience and outcome include gains in performance skills, alterations in patient referral patterns, improvement in patient selection processes, and superior medical management after procedures [36, 41, 42].

Third, technologic improvements in revascularization procedures may account for decreasing procedure-related mortality. Multiple technical advances in bypass surgery gained widespread use during the late 1980s, including refinements in cardioplegia techniques, improvements in cardiac anesthesia, and increased use of mammary artery grafts [43]. Although examining the effect of many of these advances was beyond the scope of our study, we were able to examine the use of mammary artery grafts. In the mid-1980s, investigators reported that patients who received mammary artery grafts had improved long-term patency and survival outcomes compared with patients who received saphenous vein grafts [44, 45]. We found that the use of these grafts during bypass surgery in the elderly increased by 45% between 1987 and 1990, reflecting the diffusion of these study findings into general community practice. We also found that patients who received mammary artery grafts had significantly lower 30-day and 1-year mortality after bypass surgery than did patients who received vein grafts. These results are consistent with previous smaller institutional reviews that showed improved early and long-term outcomes in elderly patients who received mammary grafts [46-48]. Thus, secular changes in the use of mammary artery grafts may partially explain improved bypass surgery outcomes.

Three caveats must be noted in interpreting these results. First, the accuracy of procedure-related coding of mammary artery grafts within Medicare administrative data has not been verified. Although random miscoding of these grafts would tend to diminish any relation between their use and procedure-related outcome, systematic misclassification cannot be rejected. Second, the correlation between mammary artery graft use and improved survival may also be confounded by differences in patient characteristics between those who received and those who did not receive these grafts. For instance, patients who have more severe coronary disease or those who had emergent procedures may be less likely to receive a mammary artery graft. Although administrative data cannot rule out this possibility, we found that their use remained a powerful predictor of 30-day and 1-year survival after adjustment for differences in patient demographics, myocardial infarction status, and available measures of comorbidity. These results are also consistent with the findings of Gardner and colleagues [47], who reported that mammary artery graft use strongly predicted both short- and long-term survival in the elderly, even after adjustment for differences in age, sex, ejection fraction, unstable angina, diabetes mellitus, year the procedure was done, and use of intra-aortic balloon pump. Third, the relation between mammary artery graft use and patient outcome may be confounded by differences in operator characteristics. Surgeons who preferentially use these grafts may also use other technical advances or provide better overall quality of care.

Improvement in angioplasty outcomes could also have resulted from technologic advances. During the study period, many new angioplasty devices and techniques were introduced and increasingly applied, including improvements in antithrombotic and anticoagulation therapy, steerable guidewires, perfusion catheters, arthrectomy catheters, and other technologic improvements [49]. Unfortunately, because these procedures or therapies are not specifically coded within Medicare administrative data, we could not investigate their relation to angioplasty-related mortality. Ongoing randomized clinical trials and device registries will better define the effect of these technologies on patient outcome [50-52].

Use of Claims Data

Global claims databases such as the Medicare files are now accessible for providing national population-based outcomes of various diseases and clinical interventions. Three main strengths of this form of data include its size, inclusiveness, and the continuous nature in which it is collected. With their size, which allow analysis of more than 150 000 revascularization cases per year, Medicare data can provide exceedingly stable and precise estimates of procedure-related mortality. Medicare data also cover a sufficiently large patient population to permit analysis of revascularization results in various minority groups, such as blacks and the very elderly, who were under-represented in previous clinical series.

For example, few studies have examined racial differences in revascularization outcomes. An early case series, based on 56 patients, found higher operative mortality after bypass surgery in blacks than in whites [53]. Other institutional reviews of similar size, however, reported no racial difference in surgical mortality results [54, 55]. We followed the revascularization outcomes in more than 16 000 elderly black patients. After adjusting for age, sex, year the procedure was done, Charlson index score, and myocardial infarction on admission, we found higher 30-day mortality in black patients than in white patients (adjusted odds ratio after angioplasty, 1.12 [CI, 0.98 to 1.27]; odds ratio after bypass surgery, 1.18 [CI, 1.09 to 1.28]). These postprocedure-related mortality differences between blacks and whites increased by 1 year for both angioplasty (odds ratio, 1.29 [CI, 1.19 to 1.40]) and bypass surgery (odds ratio, 1.33 [CI, 1.25 to 1.41]). The causes for these differences in revascularization outcomes between blacks and whites are unclear and deserve further investigation within clinical databases.

Because of their inclusiveness, Medicare claims analyses show outcomes across various geographic regions, institutional types, and provider characteristics. These results, therefore, are representative of the national effectiveness of these procedures. Previous studies have shown that differences between the efficacy of a procedure within clinical trials at high volume at academic institutions and the effectiveness of the procedure in general community practice can often be striking [56, 57]. Our overall procedure-related and 1-year mortality results for angioplasty and bypass surgery (see Tables 3 and 4) are, in general, higher than those from previous clinical series [6-15]. Much of this apparent difference, however, reflects both the advanced age of patients in our cohort and the inclusion of patients who had revascularization procedures after admission for an acute myocardial infarction.

Finally, because claims data are collected in a continuous and standardized fashion, temporal changes in procedure-related outcomes can be correlated with changes in medical practice or health policy initiatives. For example, between 1989 and 1991, the New York State Department of Health reported that unadjusted and adjusted mortality rates for bypass surgery decreased for their state. It has been postulated that this improvement in outcome resulted from the State's hospital and physician profiling efforts. By supplying these profiles to providers, as well as to the public, the authors suggested that overall quality of care was improved [58]. We found that bypass surgery mortality decreased not only in New York but throughout the country over a similar time period. Thus, the release of "public cardiovascular scorecards" may only partially explain New York's improvement in outcomes. With the prospect of major health care reform initiatives, claims data such as these provide baseline and continuous follow-up measurements that may be used to monitor both the potential positive and negative effects of these health policy changes.

Certain precautions, however, should be considered when interpreting the results from claims data. As previously noted, claims data lack many important clinical details that affect revascularization outcome. For the individual patient, mortality predictions are likely to be determined more by their severity of disease and comorbidity than by their age, race, or sex. Comparisons of these national outcomes to individual institutions or physicians must likewise be done with caution. Hospitals or physicians, whose patient referral base differs markedly from Medicare's patient profile, can be expected to have different mortality rates. Variations in outcomes at the individual provider level should also be expected in part because of limited sample size resulting in random fluctuations in mortality rates [59-62].

Our results also do not specifically address the complex question of the appropriateness of these revascularization procedures in the elderly. We found that mortality increased markedly with increasing patient age, reaching a 1-year mortality of nearly 20% in patients older than 80 years. On the other hand, most elderly patients, even the very elderly, who have bypass surgery and angioplasty survive. Because previous investigators have noted that higher-risk patients often derive the most absolute survival benefit from revascularization therapy [42, 63], one must look both at long-term potential benefits from revascularization and at the up-front procedure-related risk. To properly discuss the appropriateness of angioplasty or bypass surgery, all risks and benefits of various treatment options must be known. Beyond potentially increasing survival, revascularization procedures can benefit the elderly by relieving angina and improving quality of life. By having interventions, however, the elderly are subjected to risks for cerebrovascular accidents, wound infections, and other procedure-related morbid conditions. Because of the observational nature of our study, the limited measures of disease severity, and the limited outcomes assessed, we cannot compare the results of one interventional strategy with that of another, nor can we compare these interventional outcomes with the natural history of the disease.

In conclusion, our study highlights the dynamic nature of patient outcome measurements. In our 4-year study, national mortality results after revascularization procedures significantly improved. It appears that much of this improvement is caused by the result of technical advances and increasing clinical experience. These improvements in national outcomes are important to patients and clinicians because they provide estimates of the risks for a potential revascularization procedure. These results are also of interest to health policy experts who are currently evaluating the appropriateness of these costly interventional technologies in the elderly. Given that outcomes measurements can rapidly change, guidelines for care must be considered in an evolutionary context. Appropriate mechanisms for the continual assessment of patient outcomes and revision of clinical guidelines should be established.