If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Death is an important outcome of procedural interventions. The death rate, or mortality rate, is subject to variability by definition. The Society of Thoracic Surgeons Adult Cardiac Surgery Database definition of “operative” mortality originally included all in-hospital deaths and deaths occurring within 30 days of the procedure. In recent versions of the Society of Thoracic Surgeons Adult Cardiac Surgery Database, “in-hospital” has been modified to include “patients transferred to other acute care facilities,” and “deaths within 30 days unless clearly unrelated to the procedure” has been changed to “deaths within 30 days regardless of cause.” This study addresses the impact of these redefinitions on outcome reporting.
The California Office of Statewide Health Planning and Development hospitalized patient discharge database was queried for the year 2009, the most recent year that data files could be linked to the vital statistics death files to include all-cause mortality. Isolated coronary artery bypass grafting, isolated valve, coronary artery bypass grafting valve, and percutaneous coronary intervention procedures were identified by International Classification of Diseases, Ninth Edition, Clinical Modification procedure codes. Percutaneous coronary intervention procedures were further divided into acute coronary syndrome (percutaneous coronary intervention acute coronary syndrome) and all other percutaneous coronary intervention (percutaneous coronary intervention no acute coronary syndrome). Deaths were counted by 5 methods depending on the time and place of occurrence: (1) in-hospital or during the index hospitalization; (2) in-hospital + connected hospitalization, defined as a transfer to another acute care facility on the same day or within 24 hours of discharge; (3) in-hospital + 30 day, death during index hospitalization or within 30 days after the procedure; (4) in-hospital + connected + 30 day readmission, death during index hospitalization, transfer to acute care facility, or deaths during readmission within 30 days; and (5) in-hospital + connected + 30 day. To study the impact of these operative mortality definitions, we examined 5 different methods to track mortality and performed 2 separate analyses. The first analysis did not exclude any patients, and the second analysis excluded any patient who could not be accurately tracked after hospital discharge.
In the first analysis with no patients excluded, a total of 17% (117/697) of surgical deaths and 31% (409/1324) of percutaneous coronary intervention deaths were counted after the original hospitalization. The highest percentage of posthospital deaths occurred after elective percutaneous coronary intervention: 45% (135/301). In surgical patients, the highest percentage of posthospital deaths occurred in coronary artery bypass grafting procedures: 20% (57/284). In the second analysis, with untrackable patients excluded, hospital deaths included 12% (161/1324) for percutaneous coronary intervention compared with 4% (30/697) for surgical procedures.
A significant percentage of procedural deaths occur after transfer or discharge from the index hospital. This is especially evident in the percutaneous coronary intervention group. These findings illustrate the importance of the definition of “operative” mortality and the need to ensure accuracy in the reporting of data to voluntary clinical registries, such as the Society of Thoracic Surgeons Adult Cardiac Surgery Database and National Cardiovascular Data Registry.
Evaluating the care delivered by hospitals has been the goal of quality reporting programs, with accurate reporting being imperative for analysis and comparisons of outcomes. This is particularly emphasized when outcomes for procedural interventions are analyzed and compared. Although postoperative death is an important measurable outcome, operative morality does not have a universal definition.
Several definitions of operative morality have been used in various quality programs, with some defining postoperative mortality as a death that occurs during the same admission, and others defining postoperative death within a standardized time interval after surgery,
with 30-day follow-up most common. However, it is difficult for some hospitals to track 30-day mortality, because patients may be referred from remote locations and some centers may not have the resources to track all mortality after discharge.
The Society of Thoracic Surgeons National Database (STS-NDB) separately collects both in-hospital and 30-day mortality status, and has revised its definition of operative mortality to ensure more accurate and consistent measurement of outcomes. The original definition of operative mortality included (1) all deaths occurring during the hospitalization in which the operation was performed, even if after 30 days; and (2) those deaths occurring after discharge from the hospital but within 30 days of the procedure unless the cause of death is clearly unrelated to the operation. This definition was further updated in 2011 to address the issue of patients transferred to other acute care facilities. The most recent iteration of operative mortality appeared in 2014, which defined operative mortality as (1) all deaths, regardless of cause, occurring during the hospitalization in which the operation was performed, even if after 30 days (including patients transferred to other acute care facilities); and (2) all deaths, regardless of cause, occurring after discharge from the hospital but before the end of the 30th postoperative day.
To study the impact of these changes to the STS-NDB operative mortality definitions, we examined 5 different methods to track mortality and performed 2 separate analyses. The first analysis did not exclude any patients, and the second analysis excluded any patient who could not be accurately tracked after hospital discharge. The aim of this study was to demonstrate the effect of differences in mortality definition methodology and the impact of the definition on morality rates.
Materials and Methods
The California Office of Statewide Health Planning and Development (OSHPD) hospitalized Patient Discharge Database was queried for the year 2009, the most recent year that data files could be linked to the vital statistics death files to include all-cause mortality. The OSHPD collects inpatient discharge data from all licensed inpatient hospitals in California. Each record in the dataset corresponds to an individual inpatient hospital discharge.
The California Cardiac Surgery and Intervention Project, a project of the California Society of Thoracic Surgeons, collects data from the OSHPD Patient Discharge Database on all cardiac surgery and percutaneous coronary intervention (PCI) procedures performed in the state. The data files have patient-level data and include information on patient demographics, diagnosis and procedure codes, payment source, admission source, and all-cause mortality. Because California requires licensed hospitals to submit data on all discharged patients biannually, a 100% collection rate is ensured for patients undergoing cardiac surgery or PCI. The OSHPD includes confidential information about health care outcomes to qualified California Society of Thoracic Surgeons investigators who are faculty members of the University of California. The California Cardiac Surgery and Intervention Project reports volume and outcomes on all cardiac surgery and interventions on its Web site to provide quality improvement data for health care givers and other stakeholders.
We queried the database from January 1, 2009, to December 31, 2009, for all patients undergoing a primary procedure of isolated coronary artery bypass grafting (CABG), isolated valve, CABG valve, and PCI procedures. These patients were identified by International Classification of Diseases, 9th Revision, Clinical Modification procedure codes. PCI procedures were further divided into acute coronary syndrome (ACS) (PCI with ACS) and all other PCI (PCI without ACS) on the basis of the presence of an admission diagnosis code for acute myocardial infarction code (410.xx).
The OSHPD inpatient discharge record also includes patient disposition. Patients transferred to another facility can be tracked through connected hospitalizations on the basis of a record linkage number that uniquely identifies the patient. We included in our mortality outcome deaths that occurred after the index admission or a death that occurred after a subsequent connected acute care admission. Patients who were discharged to home, a skilled nursing facility, long-term care, home health service, or nursing home were considered survivors. Two admissions were considered connected if the disposition at the transferring facility was acute care or the admission source at the facility the patient was transferred to was acute care and the type of care provided was acute inpatient care. In addition, the discharge date at the transferring facility had to be within 1 day of the admission date at the facility a patient was transferred to.
Deaths were counted by 5 methods depending on the time and place of occurrence (Figure 1):
Death during surgery admission: This definition includes any death that occurs during the hospital stay where the procedure was performed, regardless of time.
Death during surgery admission and connected admission: This includes the previous definition and any deaths occurring after connected transfer to another acute care facility, regardless of time elapsed after the procedure.
Death during surgery admission and within 30 days of surgery: death during the surgery admission or within 30 days after the procedure regardless of location. This definition is equivalent to older STS-NDB definitions of operative mortality.
Death during surgery admission and connected admission and at readmission within 30 days of surgery: death during the procedure admission, during connected hospitalization, plus any deaths occurring within 30 days during readmission to another acute care facility.
Death during surgery admission and connected admission and within 30 days of surgery: death during procedure admission or connected hospitalization, or within 30 days of procedure from any cause or at any location. This definition is equivalent to the current STS-NDB definition of operative mortality.
Approximately 8% of patients did not have unique identifiers and could not be followed through connected hospitalizations. For this reason, 2 analyses were performed. The first included all patients without exclusion. The second analysis excluded any patient who could not be followed longitudinally after an in-hospital procedure, regardless of the reason. Table 1 shows the results with all patients included regardless of the ability to determine transfer status and Figure 2 illustrates these results further in chart format.
Table 1Number of patients excluded
Deaths N (%)
Deaths N (%)
Deaths N (%)
Deaths N (%)
Deaths N (%)
PCI with ACS
PCI without ACS
A total of 17% (117/697) of surgical deaths and 31% (409/1324) of PCI deaths were counted after the original hospitalization. The highest percentage of post-hospital deaths occurred after elective PCI: 45% (135/301). In surgical patients, the highest percentage of post-hospital deaths occurred in CABG procedures: 20% (57/284). CABG, Coronary artery bypass grafting; PCI, percutaneous coronary intervention; ACS, acute coronary syndrome.
In the second analysis (Table 2), patients who could not be tracked were excluded. For method 1, the definition includes only deaths that occurred in the hospital where the procedure occurred, so all procedures and outcomes were included. In method 2, patients who were missing across admissions patient identifiers (usually an encrypted Social Security number) were excluded, which resulted in a reduction in the total number of patients (denominator). To retain an accurate percentage of operative mortality, deaths (numerator) that occurred in “untrackable” patients were also excluded. In methods 3, 4, and 5, out-of-state patients, patients who left against medical advice, and patients with missing across admissions patient identifiers were excluded. This also caused a reduction in both the denominator and the numerator in these groups. Figure 3 further illustrates these findings.
Despite the reduction in patient numbers, the mortality percentages were essentially the same. Table 2 could be considered statistically correct because the denominator includes the population at risk for the event being “counted” in the numerator (in this case the population at risk is the number of procedure patients, and the numerator is the event, which is death). Exclusions resulted in a variable number of procedures, depending on the definition. Not only is mortality (numerator) affected depending on the definition used, but also the denominator is dependent on the definition and method used. Despite these effects, mortality percentages were basically unchanged.
In both tables, method 1 (death during surgery admission) resulted in the lowest mortality number for all procedures, and method 5 (death during surgery admission or connected admission or within 30 days of surgery) gave the highest mortality percentage for all procedures. The PCI groups were found to have a larger percentage of deaths occurring after hospital discharge and within 30 days of the procedure compared with surgical groups. Out of hospital deaths are captured by methods 3 and 5, and the difference between method 4, which captures all in hospital deaths, and method 5 is an estimate of out of hospital mortality. Out of hospital deaths were 12% (161/1324) for PCI and 4% (30/697) for surgical procedures. Method 4 also captures the largest percentage of “operative” deaths for surgical patients compared with method 5, which is the current STS-NDB definition: 96% (667/697).
We evaluated the impact of redefinition of operative mortality by counting deaths by 5 methods, each of which was dependent on the time and place of occurrence. The data collected show that a large percentage of procedural deaths occur after transfer or discharge from the index hospital, particularly in the PCI groups. We found that a larger percentage of deaths occurred after transfer or discharge from the index hospital in the PCI groups, which was in contrast to the mortality counts in the surgical groups. PCI mortality was more dependent on the method used to define mortality compared with surgical patients, with a larger percentage of deaths occurring after hospital discharge and within 30 days of the procedure. This finding has implications in terms of how procedural quality assessment should be defined.
Method 5 includes mortality after the index admission, mortality after connected acute care admissions, and deaths that occurred within 30 days of the procedure regardless of location. This definition results in a higher mortality level compared with definition 1 in patients undergoing PCI. Approximately half the mortality after PCI without ACS occurred after hospital discharge and within 30 days of the procedure. On the basis of the data in this study, definition 5 is the most encompassing and is the current definition used by the STS and should be applied to other clinical registries, such as the American College of Cardiology National Cardiovascular Data Registry (ACC-NCDR). According to the NCDR Web site, outcome measures include PCI in-hospital mortality and 30-day readmission; however, overall 30-day mortality is not included.
In comparing the PCI groups and the surgery groups, a patient death is more likely to be picked up through an inpatient readmission within 30 days for the surgical group, whereas information after home discharge is important for a complete picture of mortality after PCI. Of note, patients discharged to skilled nursing facilities did not count as a connected admission, because skilled nursing facilities are not required to report these data. The issue of whether death in a skilled nursing facility should be counted is a subject of some debate. By using definitions 3, 4, and 5, patients who die within 30 days of a procedure would be counted as an operative death, but those patients who die after 30 days at a skilled nursing facility would not be counted. This is a potential source of “gaming” the operative mortality measure. Furthermore, because of issues with subject hospitals tracking readmission to other facilities, and the occurrence of out of hospital death, substantially more diligence and resources are needed to ensure accurate follow-up on the part of reporting hospitals. It is likely that this is a variable that affects accuracy in self-reporting databases.
Our data show that death should have a clear definition in any registry database, because the performance can significantly vary depending on the definition of death used. Previous studies have shown that altering the parameters of the definition of mortality results in a redistribution of outcomes. In a study by Johnson and colleagues,
outcomes at 43 Veterans Affairs hospitals were measured by redefining operative mortality in patients who underwent CABG surgery. The definition of operative death was changed from the traditional 30-day mortality definition (any death occurring within 30 days after surgery, regardless of cause, in or out of the hospital) to 30-day mortality plus all procedure-related mortality (any death that occurred 30 days to 6 months after surgery that was deemed to be a direct result of a complication of the surgery). With this change in definition, the ranking of 86% of the hospitals in the study changed, with 21% changing their quartile rank and 11% changing outlier status. Their study was able to show that by altering the definition of mortality, the hospital profiles of postoperative mortality from CABG surgery were affected. Likewise, by readjusting the definition of mortality by using different parameters for time and place of occurrence, we were able to show deaths that may have not been captured under previous definitions of mortality.
Monitoring surgical outcomes is necessary to promote quality improvement.
It is vital that definitions be standardized to provide achievable benchmarks and identify hospitals with potential quality problems. It is equally important that data be accurate and have standardization of definitions, because data collection and public reporting without clear parameters can have important implications on comparative health care outcomes. Furthermore, it is imperative that data are accurate and be collected and reported by sources that are verified and audited. There has been a push for more outcome data to be made publically available, but this should be done with caution. Reporting from public consumer groups should be approached conservatively, because data reported from these sources can vary depending on the definition, as shown by our data. This has been shown in previous studies, with Joynt and colleagues
showing that states with mandatory public reporting for PCI outcomes have been found to have substantially lower rates of PCI among patients with acute myocardial infarction compared with states without public reporting. This finding was also demonstrated in a study by Aggarwal and colleagues,
who recommended the inclusion of 30-day mortality in the definition, because approximately half of the PCI-related deaths occurred within 30 days postdischarge. Several studies have shown that public release of outcome data does not change the behavior of consumers,
Health care providers and clinical institutions should be the main audience of these data, because results from outcomes can be used for targeting areas of improvement for both individuals and institutions. This approach has been shown to lead to improvement in outcomes without public release of data.
A potential limitation of this study is that the data from this project come from an administrative dataset that is dependent on administrative and billing coding. Administrative databases are designed for their own individual purposes (ie, administrative claims) and not for clinician inference.
Furthermore, models using administrative data to predict postoperative mortality have been shown to be improved with the addition of a small number of clinical variables, making them useful when used for quality improvement efforts.
The administrative data are limited by the need to use International Classification of Diseases, 9th Revision, Clinical Modification codes for diagnoses and procedures, and do not support clinical risk stratification. The data also are dependent on the California Vital Statistics death file, and the currently available death file only allows us to complete assessment of death status within 30 days of surgery for 2009 and earlier years, which leads to a gap of 5 years when used as a source of outcome information after cardiac procedures. The death files are only made available after a period of 5 years, so 2009 is the most recent data that are accessible. Furthermore, the inability to follow patients who did not have patient identifiers leads to changes in both the numerator and the denominator of our mortality calculation depending on the definition. Patients who were transferred to skilled nursing facilities were also not counted in mortality calculations unless they died within 30 days of the procedure, as previously discussed. Any patient who died in a skilled nursing facility beyond 30 days was not included in the numerator or denominator in this report because they were considered to be discharged. However, it is important and unique to be able to collect accurate data on all revascularization procedures in the state of California, including both CABG and PCI, because reporting is mandatory. Outcome information is secure in that data points are actual hospital admissions or validated deaths. These results may not be applicable across the United States, because currently only New York, New Jersey, Pennsylvania, Massachusetts, and California require outcome reporting after cardiac intervention.
This report is the first to recognize the substantial and potentially unreported mortality after hospital discharge in California patients who have undergone PCI. It is evident that mortality rates for the same procedure can be variable and depend on the definition chosen. The findings of this study illustrate the importance of the definition of operative mortality and the need to ensure accuracy in the reporting of data to voluntary clinical registries such as the STS-NDB and ACC-NCDR.
Conflict of Interest Statement
J.K. reports consulting fees for Edwards Lifesciences. All other authors have nothing to disclose with regard to commercial support.
What is operative mortality? Defining death in a surgical registry database: a report from the STS Congenital Database Task Force and the Joint EACTS-STS Congenital Database Committee.