Author + information
- Received October 25, 2010
- Revision received December 21, 2010
- Accepted January 5, 2011
- Published online May 1, 2011.
- Edward L. Hannan, PhD⁎,⁎ (, )
- Ye Zhong, MD⁎,
- Stephen J. Lahey, MD†,
- Alfred T. Culliford, MD‡,
- Jeffrey P. Gold, MD§,
- Craig R. Smith, MD∥,
- Robert S.D. Higgins, MD¶,
- Desmond Jordan, MD∥ and
- Andrew Wechsler, MD#
- ↵⁎Reprint requests and correspondence:
Dr. Edward L. Hannan, School of Public Health, State University of New York, University at Albany, One University Place, Rensselaer, New York, 12144-3456
Objectives The aim of this study was to identify reasons for and predictors of readmission.
Background Short-term readmissions have been identified as an important cause of escalating health care costs, and coronary artery bypass graft (CABG) surgery is 1 of the most expensive procedures.
Methods We retrospectively analyzed 30-day readmissions for 33,936 New York State patients who underwent CABG surgery between January 1, 2005, and November 30, 2007. The main reasons for readmission (principal diagnoses) and the significant independent predictors of readmission were identified. The hospital-level relationship between risk-adjusted mortality rate and risk-adjusted readmission rate was explored to determine the value of readmission rate as a complementary measure of quality.
Results The most common reasons for readmission were post-operative infection (16.9%), heart failure (12.8%), and “other complications of surgical and medical care” (9.8%). Increasing age, female sex, African-American race, higher body mass index, numerous comorbidities, 2 post-operative complications (renal failure and unplanned cardiac reoperation), Medicare or Medicaid status, discharges to a skilled nursing facility, saphenous vein grafts, and longer lengths of stay were all associated with higher rates of readmission. The correlation between the risk-adjusted 30-day readmission rate of hospitals and risk-adjusted in-hospital/30-day mortality rate was 0.32 (p = 0.047). The range across hospitals in the readmission rate was from 8.3% to 21.1%.
Conclusions The 30-day readmission rate for CABG surgery remains high, despite decreases in short-term mortality. Patients with any of the numerous risk factors for readmission should be closely monitored. Hospital readmission rates are not highly correlated with mortality rates and might serve as an independent quality measure.
Hospital readmissions within a short period after the initial admission contribute substantially to the overall cost of health care in the United States. In a study of Medicare fee-for-service beneficiaries discharged from the hospital in 2003 and 2004, Jencks et al. (1) found that 1 in every 5 patients was readmitted within 30 days. The estimated cost of unplanned hospital readmissions in 2004 was $17.4 billion of a total of the $102.6 billion in total hospital payments made by Medicare in that year.
As part of the new health care reform bill, the Centers for Medicare and Medicaid Services will begin tracking hospital readmission rates and initiating financial incentives to reduce preventable readmissions. Consequently, it is important to be able to estimate the extent of the problem and to examine in more detail the nature of and reasons for hospital readmissions.
Coronary artery bypass graft (CABG) surgery is of interest in this regard, because it has been demonstrated to have reasonably high short-term readmission rates (2,3) and because it is 1 of the most expensive procedures, with mean charges of nearly $100,000 for the index admission (4). One of these studies was conducted in New York with data from 1999 (2).
We hypothesize that the rates of readmission within 30 days of discharge have dropped considerably in New York, given the substantial decreases in short-term mortality that have occurred since the last study. We also hypothesize that the reasons for readmission and the significant predictors of readmission are essentially the same. An additional purpose of the study is to compare hospital level risk-adjusted readmission rates with their risk-adjusted mortality rates to determine whether readmissions provide an independent component of quality not measured by mortality.
Study group and endpoints
From January 1, 2005, to November 30, 2007, there were 33,936 isolated CABG procedures (procedures with no other major concomitant cardiac surgery such as valve surgery) performed in New York hospitals. We excluded patients who had previous cardiac surgery within 30 days (n = 1), who were not New York State residents (n = 1,333), who died during the index hospital stay (n = 527), and who were transferred to another acute care facility (n = 449). After then removing 673 patients whose medical records could not be found in administrative data, there were 30,953 CABG procedures remaining that were the subjects of the study. Residents of other states were excluded, because we were unable to capture readmissions to hospitals in other states.
Readmissions were classified as types that were likely to be complications of CABG surgery by applying the International Classification of Diseases-Ninth Revision-Clinical Modification (ICD-9-CM) codes identified for this purpose in earlier studies in New York and Pennsylvania (2,3).
The primary database used in the study is the New York State Cardiac Surgery Reporting System (CSRS), which contains detailed clinical information about every patient undergoing CABG surgery in New York State since January 1, 1989. The CSRS contains numerous demographic variables; patient clinical risk factors and complications; dates of admission, surgery, and discharge, and discharge status. Participation in CSRS is voluntary, but hospitals must obtain a Certificate of Need to perform CABG surgery in the state, and CSRS is used to generate reports that evaluate hospital and surgeon quality/risk-adjusted mortality.
Concerted efforts are made to maintain its completeness and accuracy by matching to the New York administrative database (Statewide Planning and Research Cooperative System [SPARCS]) to ensure that all patients undergoing CABG surgery are contained in CSRS and that their in-hospital mortality is accurate. The accuracy of patient risk factors in CSRS is assured by concerted medical recoding auditing by the Department of Health utilization review agent, and 30-day mortality outside of the hospital is obtained by matching CSRS to the New York vital statistics data.
The CSRS was linked with SPARCS, because SPARCS contains all acute care admissions for non-federal hospitals in the state. Information contained in SPARCS includes patient demographic data (age, sex, race); the principal and up to 14 secondary diagnoses; the primary and up to 14 secondary procedures; admission, discharge, and procedure dates; and discharge disposition. The CSRS and SPARCS records were matched with unique hospital and patient identifiers and date of birth.
Reasons for readmissions within 30 days after discharge after the index procedure that could be related to the CABG surgery were identified with definitions from earlier studies (2,3). These reasons were identified with ICD-9-CM codes, and the number and percentage of each reason for readmission were calculated.
A wide variety of potential risk factors related to readmissions within 30 days, including patient demographic data, other patient preoperative risk factors, post-operative complications, and other variables were selected. Demographic data included age, female sex, and African-American race. Other preoperative risk factors included body mass index (BMI), comorbidities, and medical history before the index CABG surgery, including previous open heart procedures. Post-operative complications included shock, Q-wave myocardial infarction, bleeding requiring reoperation, sepsis, gastrointestinal bleeding, renal failure, respiratory failure, and unplanned cardiac reoperation or interventional procedure. Other variables included hospital risk-adjusted mortality in 2005 to 2007, hospital CABG volume in 2005 to 2007, surgeon CABG volume in 2005 to 2007, type of primary insurance (Medicare, Medicaid, private insurance, self-pay, or “others”), discharge destination (home, skilled nursing home, inpatient physical medicine/rehabilitation, and “other”), length of stay during the admission for CABG surgery, the type of graft used (internal mammary artery [IMA], saphenous vein, or both), and type of surgery (off-pump, on-pump).
The independent relationship between readmissions within 30 days after discharge and all aforementioned risk factors was examined in a stepwise logistic regression model (p < 0.05) to determine which of them were independently associated with readmission within 30 days. Then all significant independent predictors were entered into a logistic regression model with generalized estimating equations to account for clustering of patients within hospitals (5).
Age was represented in the model as a continuous variable with a spline “knot” at age 70 years, because that function was more strongly related to readmissions than various age ranges. African-American race and “other race” were used as candidate indicator variables, with white race serving as the reference category. Each of the comorbidities was treated as a candidate indicator variable, as were previous open heart surgery and previous percutaneous coronary intervention. The BMI was subdivided into 7 categories on the basis of ranges commonly used to denote levels of obesity. Hospital risk-adjusted mortality in 2005 to 2007 was divided into tertiles with the lowest tertile as the reference group. Hospital annual CABG volume was divided into 4 groups (<200, 200 to 299, 300 to 399, 400 and more), and fewer than 200 was set as the reference group. Surgeon CABG volume was divided into 4 groups (<100, 100 to 199, 200 to 299, 300 and more), and fewer than 100 was set as the reference group. Primary payer was represented by 4 categories, with private insurance used as the reference category, and Medicare, Medicaid, self-pay/other used as candidate indicator variables. Discharged to home after surgery was used as the reference for discharge destination, and discharged to skilled nursing facility, and discharged to inpatient physical medicine/rehab, and discharged to “other” place were used as candidate indicator variables. Length of stay in admission for CABG surgery was subdivided into 5 categories, with 4 days or fewer used as the reference category. Any IMA grafting used in CABG surgery was defined as a binary variable, and no IMA grafting was used as the reference category. Type of CABG surgery was set as a binary variable to indicate whether the surgery was off- or on-pump, with off-pump as the reference category. Throughout this process, categories were combined with the reference category when they were not significantly different as predictors of readmission.
The process described in the preceding text was then repeated with “readmission” restricted to readmission within 30 days because of infection (which was found to be the most frequent reason for readmission within 30 days after CABG surgery), with the candidate independent variables from the bivariate analyses remaining the same.
The relationship between the observed and risk-adjusted readmission rates of hospitals was examined by calculating the hospital-level correlation between the 2 rates. This was done to determine how important it is to adjust for differences in hospital case mix when assessing hospital performance on the basis of readmission rates.
To determine whether hospital readmission rates capture another component of quality that might be separate and distinct from hospital mortality rates, we computed the risk-adjusted in-hospital/30-day mortality and 30-day readmission rates of hospitals, and calculated the correlation between the 2 rates. Hospital risk-adjusted mortality rates were the ones used in the New York State Department of Health report that covers the period from 2005 to 2007 (6). Risk-adjusted readmission rates were obtained by developing a logistic regression model similar to the one described in the preceding text for identifying significant predictors of readmission. However, the model for obtaining hospital risk-adjusted rates differed in that complications of CABG and some other factors (including hospital mortality, hospital and surgeon CABG volume, primary payer, discharge place, and length of stay after CABG surgery) were not allowed as candidates for the model, because the model was used to assess quality of care. The resulting model was used to obtain hospital risk-adjusted complication rates in the same manner that the New York State Department of Health reports obtain the risk-adjusted mortality rates of hospitals (6).
The discrimination of each of the previously described models was measured with the C-statistic (7), and calibration was measured with the Hosmer-Lemeshow statistic (8). All analyses were performed with SAS software (version 8.2, SAS Institute, Cary, North Carolina).
The 30-day all-cause readmission rate was 16.5%, with a range across hospitals of 8.3% to 21.1%. The 30-day all-cause mortality rate after discharge among patients who were not readmitted was 0.76%. Among the 30-day readmissions, 87.3% were for reasons related to CABG surgery, so the 30-day readmission rate related to complications of CABG surgery was 14.4%. Patients readmitted within 30 days experienced a 2.79% in-hospital mortality rate.
The most common reason for readmission within 30 days for complications was post-operative infection (16.9%), followed by heart failure (12.8%), “other complications of surgical and medical care” (9.8%), cardiac dysrhythmia (6.3%), and angina/chest pain (4.7%) (Table 1). Note that the ICD-9-CM code for “other complications of surgical and medical care” refers to 1 or more of the nervous system, circulatory system, respiratory system, digestive system, and urinary system. Table 1 also notes that the median time for readmission ranged from 5 days (acute respiratory failure) to 13 days (cellulitus), with an overall median of 9 days.
The significant independent risk factors for readmission within 30 days after discharge after CABG operations are presented in Table 2. As noted, increasing age, female sex, African-American race, higher BMI, numerous comorbidities, 2 post-operative complications (renal failure and unplanned cardiac reoperation), primary payer of Medicare or Medicaid, discharges to a skilled nursing facility, the use of only saphenous vein grafts (no IMA grafting), and longer lengths of stay were all independently associated with higher rates of readmission within 30 days.
According to Table 3, the significant predictors of readmission within 30 days for post-operative infections were female sex, obesity, peripheral vascular disease, chronic obstructive pulmonary disease (COPD), diabetes, 3-vessel disease, higher hospital mortality, lower surgeon CABG volume, and a length of stay >4 days. Patients with a BMI >40 had the highest risk for readmission (adjusted odds ratio: 3.11, 95% confidence interval: 2.41 to 4.00).
The correlation between the risk-adjusted 30-day readmission rate and the observed 30-day readmission rate was 0.94 (p < 0.001) (Fig. 1). The correlation between hospital risk-adjusted 30-day readmission rate and risk-adjusted in-hospital/30-day mortality rate was 0.32 (p = 0.047) (Fig. 2).
Hospital readmissions within a short period after initial discharge occur frequently and add substantial costs to the healthcare system of our country. Decreasing short-term readmissions were identified as 1 of the targets in the President's health care reform bill for reducing unnecessary costs, and the intention is to eliminate or reduce payment for 30-day readmissions after treatment for various medical conditions and procedures. Initially, readmissions for acute myocardial infarction, congestive heart failure, and pneumonia will have reduced payments, but this list is likely to expand.
The main purposes of this study were to document the extent of 30-day readmissions after CABG surgery, identify the reasons (principal diagnosis on readmission) for those readmissions, and determine the independent predictors of readmissions for patients undergoing CABG operations in New York State.
Among the findings was that the 30-day all-cause readmission rate was 16.5% and the 30-day all-cause mortality rate after discharge from the index admission for patients who were not readmitted was 0.76%. The most common reason for readmission within 30 days for complications was post-operative infection (16.9%), followed by heart failure (12.8%) and “other complications of surgical and medical care” (1 or more of nervous system, circulatory system, respiratory system, digestive system, and urinary system). A total of 6.3% of the readmissions were for cardiac dysrhythmia, and 4.7% were for angina/chest pain.
Patients who were at highest risk for readmission included elderly persons, women, African Americans, patients with 1 or more of several comorbidities (cerebrovascular disease, peripheral vascular disease, shock, congestive heart failure, COPD, extensive aortic atherosclerosis, diabetes, immune system deficiency), high BMI, 1 or more of 2 complications of surgery (renal failure and unplanned cardiac reoperation). Also, patients with Medicare or Medicaid who were discharged to a skilled nursing facility, who only received saphenous vein grafts, and who experienced longer lengths of stay all tended to have higher rates of readmission within 30 days, all other factors being equal.
Earlier studies have also found that elderly persons (2,3,9,10), women (2,3,9–12), and African Americans (2,3) were more likely to experience readmissions; in addition to these factors, our earlier study also found that congestive heart failure, COPD, diabetes, and longer lengths of stay were independently associated with higher readmission rates (2). Another study found that longer lengths of stay in the index admission were associated with higher readmission rates (13). High BMI was not reported as a predictor in earlier studies but was a significant predictor in our study. This might be a sign of the escalating rate of obesity in the United States, and the high cost of readmissions is yet another reason why the obesity explosion must be controlled.
Although these results would seem to imply that readmissions were related to patient-related factors rather than system-related factors such as hospital or surgeon volume (which were tested in the statistical model), we hypothesize that the important system-related factors were things like coordination with outpatient care that we did not have access to in our database.
We also found that patients at the highest risk of readmission for post-operative infections included women, obese patients, and patients with unplanned reoperations and longer lengths of stay. In particular, obesity and reoperations are intuitively risks for infections, and these patients might benefit from additional monitoring before and after discharge after the index procedure.
The 4 most common reasons for readmission in our earlier study and this study were identical: post-operative infections (28.3% and 16.9%), heart failure (15.7% and 12.8%), “other complications of surgical and medical care” (11.4% and 9.8%), and dysrhythmias (7.7% and 6.3%). Our earlier study, with 1999 New York data, found that 15.3% of CABG surgery patients were readmitted within 30 days of discharge, and 12.9% of all live discharges after the index CABG surgery were readmitted for reasons identified as related to the index admission (compared with 16.5% and 14.3%, respectively, in the current study).
It is interesting that, given all the quality improvement initiatives in New York and the country in the past several years, the 30-day readmission rate has actually increased. For example, the in-hospital mortality rate in New York decreased from 2.24% in 1999 to 1.54% in 2007 (2). However, a countervailing reason why readmission rates have not decreased might be that the mean length of stay after CABG surgery decreased from 8.2 days in 1999 to 7.7 days in 2007. Also, there is some evidence in the current study of higher prevalence of significant risk factors that were common to both studies (dialysis rose from 1.2% to 1.9%, COPD from 16.2% to 19.9%, and diabetes from 31.6% to 35.7%).
A notable finding of the study is that the correlation between the risk-adjusted in-hospital/30-day mortality rates of hospitals and risk-adjusted readmission rates was 0.32 (p = 0.047). This is considerably higher than the correlation in our previous study (0.09, p = 0.64), although it should be noted that correlations between separately adjusted variables can lead to bias. It should be noted that in this study we had access to a better measure of mortality (in-hospital/30-day), whereas in the earlier study 30-day mortality was not available. Although the correlation between mortality and readmission was significantly different from 0, it is not extremely high. Thus, readmission seems to represent a different measure of quality than short-term mortality. Hospitals might perform well on 1 of the measures and poorly on the other measure.
We also found that there was a very high hospital-level correlation between risk-adjusted and observed readmission rates (R = 0.95, p < 0.001). This would seem to imply that crude hospital readmission rates are good proxies for risk-adjusted readmission rates in assessing hospital performance. However, it would seem to be prudent to use risk-adjusted rates whenever possible. First, some hospitals might have a mix of patients who are at much higher risk for readmission. Second, the C-statistic used to measure discrimination in the statistical model for risk-adjusting the readmission rates was a modest 0.65. This is not uncommon for models that predict complications or readmissions, but it means that unavailable patient-level predictors of readmission could have improved the discrimination and/or that unmeasured hospital quality of care/process measures would have substantially improved the ability to predict readmissions. Further research is required to identify these patient and process indicators.
This study has demonstrated that short-term readmissions after CABG surgery remain a problem, and that there are several risk factors associated with this adverse outcome that enhance the chances that patients will require readmissions within 30 days. Hospital readmissions are not highly correlated with hospital short-term mortality rates and consequently might serve as another independent quality-of-care measure. We speculate that readmissions have not decreased over time despite decreasing mortality rates, because efforts to coordinate inpatient and outpatient care have been insufficient. Since the announcement by the Centers for Medicare and Medicaid Services that reimbursement for readmissions would be curtailed or eliminated, numerous organizations have sponsored symposia for coordinating inpatient and outpatient care, and it is hypothesized that hospitals that heed these suggestions will experience considerably reduced readmission rates.
The authors would like to thank the New York State Cardiac Advisory Committee (CAC) for their encouragement and support of this study and Kimberly S. Cozzens, Rosemary Lombardo, and the cardiac surgery departments of the participating hospitals for their tireless efforts to ensure the timeliness, completeness, and accuracy of the registry data.
Dr. Wechsler is a consultant to Bioventrix and Bayer; and is a medical director for Estech. All other authors have reported that they have no relationships to disclose.
- Abbreviations and Acronyms
- body mass index
- coronary artery bypass graft
- chronic obstructive pulmonary disease
- Cardiac Surgery Reporting System
- International Classification of Diseases-Ninth Revision-Clinical Modification
- internal mammary artery
- Statewide Planning and Research Cooperative System
- Received October 25, 2010.
- Revision received December 21, 2010.
- Accepted January 5, 2011.
- American College of Cardiology Foundation
- Jencks S.F.,
- Williams M.V.,
- Coleman E.A.
- Pennsylvania Health Care Cost Containment Council
- American Heart Association
- Zeger S.L.,
- Liang K.Y.,
- Albert P.S.
- New York State Department of Health
- Hosmer D.W.,
- Lemeshow S.