Portal de Periódicos da CAPES

Validity of the Society of Thoracic Surgeons National Adult Cardiac Surgery Database

2004; Elsevier BV; Volume: 77; Issue: 4 Linguagem: Inglês

10.1016/j.athoracsur.2003.07.030

1552-6259

Karl F. Welke, T. Bruce Ferguson, Laura P. Coombs, Rachel S. Dokholyan, Cindy J Murray, M Schrader, Eric D. Peterson,

Cardiac Valve Diseases and Treatments

In 1989, The Society of Thoracic Surgeons (STS) created a national voluntary cardiac surgery database as a means of supporting national quality improvement efforts [1Ferguson Jr, T.B. Dziuban S.W. Edwards F.H. et al.The STS National Database current changes and challenges for the new millennium.Ann Thorac Surg. 2000; 69: 680-691Abstract Full Text Full Text PDF PubMed Scopus (220) Google Scholar]. The database has grown to become the largest clinical database of its kind, and includes 638 hospitals and clinical information from 2,164,079 surgical procedures. This provider-supported database allows participants to benchmark their risk-adjusted results against regional and national standards. In addition, it provides data for research projects that can improve the process of patient care and outcomes. It has also been used to support quality improvement efforts. Each of these uses, however, is dependent on an accurate and complete clinical database. The purpose of this article is to provide the STS membership with a description of the ongoing efforts to assure validity of the National Adult Cardiac Surgery Database (NCD). In particular, we will highlight examples of data quality assurance mechanisms occurring at three levels: internally, regionally, and nationally.Internal data quality improvement effortsThe site data managers play a crucial role in ensuring the quality of the data being submitted to the NCD. To support these key individuals, the STS has initiated annual site data manager meetings. These well-attended meetings generate in-depth discussions and suggestions about how to best collect the data, the application of clinical data definitions, and other topics related to the overall quality of the database.In addition, several steps have been added to the data collection process of the NCD with the intent of improving data quality. First, the STS has worked extensively with database vendors to develop point of entry data quality controls. For instance, if a patient is identified as having died during the original hospitalization and the field “date of death” is left blank, the software requires that the expected date of death be entered. These checks assure internal data consistencies within a given procedural record. Second, at the point of file submission to the Database Warehouse and Analysis Center at the Duke Clinical Research Institute, sites receive a detailed automated data quality report that warns site data managers about potential problems with various characteristics of the data file including records with inconsistent data, missing data values, and outlier or extreme data values. This proactive system allows sites the option of resubmitting their data as many times as is necessary within the harvest window to correct initial data issues. Experience has shown that a majority of STS sites take advantage of this option and submit their data multiple times per harvest. Official final sign-off from the data managers at each site on data to be included in analyses for that harvest is an additional safeguard to ensure that the appropriate data are entering the NCD. At the Duke Clinical Research Institute, the site data go through a third round of data quality control checks to assure completeness and consistency before being integrated into the NCD. In addition, after each harvest a detailed letter summarizing key data quality concerns for a particular site is sent from STS headquarters to the primary surgeon at each siteAs a result of these efforts, the quality of the data in the NCD has increased remarkably [2Ferguson Jr, T.B. Hammill B.G. DeLong E.R. et al.A decade of change risk profiles and outcomes for isolated CABG procedures, 1990–1999.Ann Thorac Surg. 2002; 73: 480-490Abstract Full Text Full Text PDF PubMed Scopus (554) Google Scholar]. For example, in 2002 the average percent of missing values on the 28 clinical variables used for mortality risk modeling was 1.6%. Of these, only two variables were missing more than 5% of the time: New York Heart Association functional class (6.9%) and ejection fraction (8.9%)Regional validation efforts in IowaThe STS data quality is also assisted by many regional collaborative quality improvement efforts. One such example is the collaboration between the Iowa Society of Thoracic Surgeons (IASTS), and the Iowa Foundation for Medical Care (IFMC). The IASTS is a voluntary, not-for-profit organization whose purpose is to bring together the thoracic surgeons of Iowa to engage in activities of mutual benefit including continuous quality improvement activities aimed at improving outcomes for patients who have had cardiac surgery. Currently, all sites performing cardiothoracic surgery in Iowa submit data to the Society of Thoracic Surgeons NCD. The IFMC is a private, nonprofit organization composed of health care professionals and is the Centers for Medicare and Medicaid Services Quality Improvement Organization for Iowa. In March 2000, the IASTS and the IFMC entered into an agreement to work together in support of the Society's quality improvement goals and as part of an Agency for Healthcare Research and Quality sponsored continuous quality improvement grant to the STS.As a part of this effort, the IFMC performed a series of independent external data quality audits in the spring of 2001. Initially, five randomly selected isolated records of patients who had undergone coronary artery bypass grafting (CABG) were audited from each of the 12 Iowa sites. These audits focused on CABG operative risk of mortality data elements. After a masked review, a comparison was made between data collected by the IFMC staff and data submitted by the hospital staff. The overall agreement rate for the audited data variables was 95.8%, ranging from 80% to 100% for individual items.After initial review by the IASTS, a second audit process was repeated by the IFMC in the spring of 2002. This audit also assessed coding for postoperative atrial fibrillation owing to the wide range of occurrence noted in the state. Ten randomly selected records were reviewed at all sites. The overall agreement rate for the audited data variables was again high at 96.2%, with postoperative atrial fibrillation at 96.4%. A third round of audits are planned for 2003 for data elements that fell below a 95% agreement rate in 2001 or 2002.National comparison of the NCD to Medicare part A data filesAs a final level of data quality assurance, the STS has joined with the University of Iowa to compare the NCD to Medicare Provider and Analysis Review (MEDPAR) Part A public use data files. This effort will allow validation of the yearly CABG volumes of cases submitted to NCD with those claims submitted to Medicare for patients aged 65 years or older. Additionally, this will allow comparison of unadjusted mortality rates between the two data sets. In the first round of analyses, all Medicare patients 65 years of age and older who underwent CABG in US hospitals from 1994 to 1999 were identified in the MEDPAR Part A public use data files, obtained from the Center for Medicare and Medicaid Services. Only patients covered by fee-for-service arrangements are included in MEDPAR files (excluding up to 10% of Medicare patients enrolled in health maintenance organizations during this time period). Isolated CABG cases were identified by the appropriate diagnosis-related group codes (106, 107).A comparable sample of patients was created from the NCD by identifying all patients 65 years of age or older who underwent isolated CABG from 1994 to 1999.Hospitals common to the MEDPAR and NCD data sets were identified by matching on hospital name and city (n = 519 hospitals).Findings from this comparison are summarized in Table 1. The volume of patients was consistently slightly higher in the NCD. In-hospital mortality, while following the same trend toward lower mortality over time, was consistently higher in the NCD. The lesser number of patients in the MEDPAR data may be explained in part by the omission of patients enrolled in health maintenance organizations from Medicare files. In addition, the identification of CABG patients by diagnosis-related group code in the MEDPAR data may have resulted in the exclusion of patients who underwent CABG, but were assigned a higher diagnosis-related group code owing to another procedure. As this latter group is likely to represent higher risk patients, this may in part explain the higher mortality rate in the NCD.Table 1Comparison of Hospital Coronary Artery Bypass Graft Volume and In-Hospital Mortality in the STS and Medicare DatasetsYear199419951996199719981999Number of hospitals249312363393412359Medicare volume43,34260,30472,84576,17073,58761,675STS volume46,24763,59479,59785,04385,86874,906Medicare deaths1,6242,1732,6162,6312,4662,101STS deaths2,0502,5973,2173,3283,2292,899Medicare mortality (%)3.73.63.63.53.43.4STS mortality (%)4.44.14.03.93.83.9STS = The Society of Thoracic Surgeons. Open table in a new tab These preliminary findings of higher volumes and mortality rates in the NCD suggest that the vast majority of STS participants are providing complete case records and complete mortality results. These results are consistent with a similar comparison of STS mortality rates for valve surgery with those from the New York State Cardiac Surgery Reporting System database, and the trends seen over the last decade in both the STS and the Department of Veterans Affairs national cardiac surgery databases [3Grover F.L. Edwards C. Similarity between the STS and New York State databases for valvular heart disease.Ann Thorac Surg. 2000; 70: 1143-1144Abstract Full Text Full Text PDF PubMed Scopus (16) Google Scholar, 4Grover F.L. Shroyer A.L. Hammermeister K. et al.A decade's experience with quality improvement in cardiac surgery using the Veterans Affairs and Society of Thoracic Surgeons national databases.Ann Surg. 2001; 234: 464-472Crossref PubMed Scopus (196) Google Scholar]. Efforts to validate the NCD with the Medicare data continue. A MEDPAR dataset defined by International Classification of Diseases, ninth revision, codes rather than diagnosis-related group codes is being created to help identify procedures that were missed in the previous match and the above analysis will be repeated. Comparison of mortality at the institutional level will follow, allowing the STS to identify “outlier” institutions with respect to mortality rates.CommentThe validity of the NCD is crucial if it is to continue to serve as a benchmark for participants and a resource for outcomes research and quality improvement efforts. The STS has made and continues to make efforts to assure the validity of the NCD. Prospectively defined and collected clinical data are more detailed and accurate than administrative data, but both can suffer from the question of validity. The STS validation initiatives have been directed at maximizing the benefits of a clinical data set: collection of data by clinical personnel who should have a better understanding of the data than administrative personnel, more accurate recording of data of interest in cardiac surgery because of a database specifically designed for that purpose, and standardization of definitions. While efforts to confirm the validity of the NCD are in progress, work done to date suggests that the NCD is accurate.The goal of the STS is not just to produce a high quality clinical data set, but also to improve the process and outcomes of cardiac surgical care. Unlike administrative data, clinical process and outcomes data derived from and immediately relevant to surgeons' patients can be distributed in a timely manner to support quality improvement efforts. Indeed, the STS has just completed the first successful national randomized trial of continuous quality improvement in medicine, built on the backbone of the STS NCD [5Ferguson Jr, T.B. Peterson E.D. Coomb L.P. et al.Use of continuous quality improvement to increase use of process measures in patients undergoing coronary artery bypass graft surgery.JAMA. 2003; 290: 49-56Crossref PubMed Scopus (162) Google Scholar]. In addition, clinical data sets can be modified to include and exclude data elements to support specific areas of interest. The NCD has made possible numerous research projects and quality improvement initiatives. Continual efforts to assure validity of the STS NCD make such work possible. In 1989, The Society of Thoracic Surgeons (STS) created a national voluntary cardiac surgery database as a means of supporting national quality improvement efforts [1Ferguson Jr, T.B. Dziuban S.W. Edwards F.H. et al.The STS National Database current changes and challenges for the new millennium.Ann Thorac Surg. 2000; 69: 680-691Abstract Full Text Full Text PDF PubMed Scopus (220) Google Scholar]. The database has grown to become the largest clinical database of its kind, and includes 638 hospitals and clinical information from 2,164,079 surgical procedures. This provider-supported database allows participants to benchmark their risk-adjusted results against regional and national standards. In addition, it provides data for research projects that can improve the process of patient care and outcomes. It has also been used to support quality improvement efforts. Each of these uses, however, is dependent on an accurate and complete clinical database. The purpose of this article is to provide the STS membership with a description of the ongoing efforts to assure validity of the National Adult Cardiac Surgery Database (NCD). In particular, we will highlight examples of data quality assurance mechanisms occurring at three levels: internally, regionally, and nationally. Internal data quality improvement effortsThe site data managers play a crucial role in ensuring the quality of the data being submitted to the NCD. To support these key individuals, the STS has initiated annual site data manager meetings. These well-attended meetings generate in-depth discussions and suggestions about how to best collect the data, the application of clinical data definitions, and other topics related to the overall quality of the database.In addition, several steps have been added to the data collection process of the NCD with the intent of improving data quality. First, the STS has worked extensively with database vendors to develop point of entry data quality controls. For instance, if a patient is identified as having died during the original hospitalization and the field “date of death” is left blank, the software requires that the expected date of death be entered. These checks assure internal data consistencies within a given procedural record. Second, at the point of file submission to the Database Warehouse and Analysis Center at the Duke Clinical Research Institute, sites receive a detailed automated data quality report that warns site data managers about potential problems with various characteristics of the data file including records with inconsistent data, missing data values, and outlier or extreme data values. This proactive system allows sites the option of resubmitting their data as many times as is necessary within the harvest window to correct initial data issues. Experience has shown that a majority of STS sites take advantage of this option and submit their data multiple times per harvest. Official final sign-off from the data managers at each site on data to be included in analyses for that harvest is an additional safeguard to ensure that the appropriate data are entering the NCD. At the Duke Clinical Research Institute, the site data go through a third round of data quality control checks to assure completeness and consistency before being integrated into the NCD. In addition, after each harvest a detailed letter summarizing key data quality concerns for a particular site is sent from STS headquarters to the primary surgeon at each siteAs a result of these efforts, the quality of the data in the NCD has increased remarkably [2Ferguson Jr, T.B. Hammill B.G. DeLong E.R. et al.A decade of change risk profiles and outcomes for isolated CABG procedures, 1990–1999.Ann Thorac Surg. 2002; 73: 480-490Abstract Full Text Full Text PDF PubMed Scopus (554) Google Scholar]. For example, in 2002 the average percent of missing values on the 28 clinical variables used for mortality risk modeling was 1.6%. Of these, only two variables were missing more than 5% of the time: New York Heart Association functional class (6.9%) and ejection fraction (8.9%) The site data managers play a crucial role in ensuring the quality of the data being submitted to the NCD. To support these key individuals, the STS has initiated annual site data manager meetings. These well-attended meetings generate in-depth discussions and suggestions about how to best collect the data, the application of clinical data definitions, and other topics related to the overall quality of the database. In addition, several steps have been added to the data collection process of the NCD with the intent of improving data quality. First, the STS has worked extensively with database vendors to develop point of entry data quality controls. For instance, if a patient is identified as having died during the original hospitalization and the field “date of death” is left blank, the software requires that the expected date of death be entered. These checks assure internal data consistencies within a given procedural record. Second, at the point of file submission to the Database Warehouse and Analysis Center at the Duke Clinical Research Institute, sites receive a detailed automated data quality report that warns site data managers about potential problems with various characteristics of the data file including records with inconsistent data, missing data values, and outlier or extreme data values. This proactive system allows sites the option of resubmitting their data as many times as is necessary within the harvest window to correct initial data issues. Experience has shown that a majority of STS sites take advantage of this option and submit their data multiple times per harvest. Official final sign-off from the data managers at each site on data to be included in analyses for that harvest is an additional safeguard to ensure that the appropriate data are entering the NCD. At the Duke Clinical Research Institute, the site data go through a third round of data quality control checks to assure completeness and consistency before being integrated into the NCD. In addition, after each harvest a detailed letter summarizing key data quality concerns for a particular site is sent from STS headquarters to the primary surgeon at each site As a result of these efforts, the quality of the data in the NCD has increased remarkably [2Ferguson Jr, T.B. Hammill B.G. DeLong E.R. et al.A decade of change risk profiles and outcomes for isolated CABG procedures, 1990–1999.Ann Thorac Surg. 2002; 73: 480-490Abstract Full Text Full Text PDF PubMed Scopus (554) Google Scholar]. For example, in 2002 the average percent of missing values on the 28 clinical variables used for mortality risk modeling was 1.6%. Of these, only two variables were missing more than 5% of the time: New York Heart Association functional class (6.9%) and ejection fraction (8.9%) Regional validation efforts in IowaThe STS data quality is also assisted by many regional collaborative quality improvement efforts. One such example is the collaboration between the Iowa Society of Thoracic Surgeons (IASTS), and the Iowa Foundation for Medical Care (IFMC). The IASTS is a voluntary, not-for-profit organization whose purpose is to bring together the thoracic surgeons of Iowa to engage in activities of mutual benefit including continuous quality improvement activities aimed at improving outcomes for patients who have had cardiac surgery. Currently, all sites performing cardiothoracic surgery in Iowa submit data to the Society of Thoracic Surgeons NCD. The IFMC is a private, nonprofit organization composed of health care professionals and is the Centers for Medicare and Medicaid Services Quality Improvement Organization for Iowa. In March 2000, the IASTS and the IFMC entered into an agreement to work together in support of the Society's quality improvement goals and as part of an Agency for Healthcare Research and Quality sponsored continuous quality improvement grant to the STS.As a part of this effort, the IFMC performed a series of independent external data quality audits in the spring of 2001. Initially, five randomly selected isolated records of patients who had undergone coronary artery bypass grafting (CABG) were audited from each of the 12 Iowa sites. These audits focused on CABG operative risk of mortality data elements. After a masked review, a comparison was made between data collected by the IFMC staff and data submitted by the hospital staff. The overall agreement rate for the audited data variables was 95.8%, ranging from 80% to 100% for individual items.After initial review by the IASTS, a second audit process was repeated by the IFMC in the spring of 2002. This audit also assessed coding for postoperative atrial fibrillation owing to the wide range of occurrence noted in the state. Ten randomly selected records were reviewed at all sites. The overall agreement rate for the audited data variables was again high at 96.2%, with postoperative atrial fibrillation at 96.4%. A third round of audits are planned for 2003 for data elements that fell below a 95% agreement rate in 2001 or 2002. The STS data quality is also assisted by many regional collaborative quality improvement efforts. One such example is the collaboration between the Iowa Society of Thoracic Surgeons (IASTS), and the Iowa Foundation for Medical Care (IFMC). The IASTS is a voluntary, not-for-profit organization whose purpose is to bring together the thoracic surgeons of Iowa to engage in activities of mutual benefit including continuous quality improvement activities aimed at improving outcomes for patients who have had cardiac surgery. Currently, all sites performing cardiothoracic surgery in Iowa submit data to the Society of Thoracic Surgeons NCD. The IFMC is a private, nonprofit organization composed of health care professionals and is the Centers for Medicare and Medicaid Services Quality Improvement Organization for Iowa. In March 2000, the IASTS and the IFMC entered into an agreement to work together in support of the Society's quality improvement goals and as part of an Agency for Healthcare Research and Quality sponsored continuous quality improvement grant to the STS. As a part of this effort, the IFMC performed a series of independent external data quality audits in the spring of 2001. Initially, five randomly selected isolated records of patients who had undergone coronary artery bypass grafting (CABG) were audited from each of the 12 Iowa sites. These audits focused on CABG operative risk of mortality data elements. After a masked review, a comparison was made between data collected by the IFMC staff and data submitted by the hospital staff. The overall agreement rate for the audited data variables was 95.8%, ranging from 80% to 100% for individual items. After initial review by the IASTS, a second audit process was repeated by the IFMC in the spring of 2002. This audit also assessed coding for postoperative atrial fibrillation owing to the wide range of occurrence noted in the state. Ten randomly selected records were reviewed at all sites. The overall agreement rate for the audited data variables was again high at 96.2%, with postoperative atrial fibrillation at 96.4%. A third round of audits are planned for 2003 for data elements that fell below a 95% agreement rate in 2001 or 2002. National comparison of the NCD to Medicare part A data filesAs a final level of data quality assurance, the STS has joined with the University of Iowa to compare the NCD to Medicare Provider and Analysis Review (MEDPAR) Part A public use data files. This effort will allow validation of the yearly CABG volumes of cases submitted to NCD with those claims submitted to Medicare for patients aged 65 years or older. Additionally, this will allow comparison of unadjusted mortality rates between the two data sets. In the first round of analyses, all Medicare patients 65 years of age and older who underwent CABG in US hospitals from 1994 to 1999 were identified in the MEDPAR Part A public use data files, obtained from the Center for Medicare and Medicaid Services. Only patients covered by fee-for-service arrangements are included in MEDPAR files (excluding up to 10% of Medicare patients enrolled in health maintenance organizations during this time period). Isolated CABG cases were identified by the appropriate diagnosis-related group codes (106, 107).A comparable sample of patients was created from the NCD by identifying all patients 65 years of age or older who underwent isolated CABG from 1994 to 1999.Hospitals common to the MEDPAR and NCD data sets were identified by matching on hospital name and city (n = 519 hospitals).Findings from this comparison are summarized in Table 1. The volume of patients was consistently slightly higher in the NCD. In-hospital mortality, while following the same trend toward lower mortality over time, was consistently higher in the NCD. The lesser number of patients in the MEDPAR data may be explained in part by the omission of patients enrolled in health maintenance organizations from Medicare files. In addition, the identification of CABG patients by diagnosis-related group code in the MEDPAR data may have resulted in the exclusion of patients who underwent CABG, but were assigned a higher diagnosis-related group code owing to another procedure. As this latter group is likely to represent higher risk patients, this may in part explain the higher mortality rate in the NCD.Table 1Comparison of Hospital Coronary Artery Bypass Graft Volume and In-Hospital Mortality in the STS and Medicare DatasetsYear199419951996199719981999Number of hospitals249312363393412359Medicare volume43,34260,30472,84576,17073,58761,675STS volume46,24763,59479,59785,04385,86874,906Medicare deaths1,6242,1732,6162,6312,4662,101STS deaths2,0502,5973,2173,3283,2292,899Medicare mortality (%)3.73.63.63.53.43.4STS mortality (%)4.44.14.03.93.83.9STS = The Society of Thoracic Surgeons. Open table in a new tab These preliminary findings of higher volumes and mortality rates in the NCD suggest that the vast majority of STS participants are providing complete case records and complete mortality results. These results are consistent with a similar comparison of STS mortality rates for valve surgery with those from the New York State Cardiac Surgery Reporting System database, and the trends seen over the last decade in both the STS and the Department of Veterans Affairs national cardiac surgery databases [3Grover F.L. Edwards C. Similarity between the STS and New York State databases for valvular heart disease.Ann Thorac Surg. 2000; 70: 1143-1144Abstract Full Text Full Text PDF PubMed Scopus (16) Google Scholar, 4Grover F.L. Shroyer A.L. Hammermeister K. et al.A decade's experience with quality improvement in cardiac surgery using the Veterans Affairs and Society of Thoracic Surgeons national databases.Ann Surg. 2001; 234: 464-472Crossref PubMed Scopus (196) Google Scholar]. Efforts to validate the NCD with the Medicare data continue. A MEDPAR dataset defined by International Classification of Diseases, ninth revision, codes rather than diagnosis-related group codes is being created to help identify procedures that were missed in the previous match and the above analysis will be repeated. Comparison of mortality at the institutional level will follow, allowing the STS to identify “outlier” institutions with respect to mortality rates. As a final level of data quality assurance, the STS has joined with the University of Iowa to compare the NCD to Medicare Provider and Analysis Review (MEDPAR) Part A public use data files. This effort will allow validation of the yearly CABG volumes of cases submitted to NCD with those claims submitted to Medicare for patients aged 65 years or older. Additionally, this will allow comparison of unadjusted mortality rates between the two data sets. In the first round of analyses, all Medicare patients 65 years of age and older who underwent CABG in US hospitals from 1994 to 1999 were identified in the MEDPAR Part A public use data files, obtained from the Center for Medicare and Medicaid Services. Only patients covered by fee-for-service arrangements are included in MEDPAR files (excluding up to 10% of Medicare patients enrolled in health maintenance organizations during this time period). Isolated CABG cases were identified by the appropriate diagnosis-related group codes (106, 107). A comparable sample of patients was created from the NCD by identifying all patients 65 years of age or older who underwent isolated CABG from 1994 to 1999. Hospitals common to the MEDPAR and NCD data sets were identified by matching on hospital name and city (n = 519 hospitals). Findings from this comparison are summarized in Table 1. The volume of patients was consistently slightly higher in the NCD. In-hospital mortality, while following the same trend toward lower mortality over time, was consistently higher in the NCD. The lesser number of patients in the MEDPAR data may be explained in part by the omission of patients enrolled in health maintenance organizations from Medicare files. In addition, the identification of CABG patients by diagnosis-related group code in the MEDPAR data may have resulted in the exclusion of patients who underwent CABG, but were assigned a higher diagnosis-related group code owing to another procedure. As this latter group is likely to represent higher risk patients, this may in part explain the higher mortality rate in the NCD. STS = The Society of Thoracic Surgeons. These preliminary findings of higher volumes and mortality rates in the NCD suggest that the vast majority of STS participants are providing complete case records and complete mortality results. These results are consistent with a similar comparison of STS mortality rates for valve surgery with those from the New York State Cardiac Surgery Reporting System database, and the trends seen over the last decade in both the STS and the Department of Veterans Affairs national cardiac surgery databases [3Grover F.L. Edwards C. Similarity between the STS and New York State databases for valvular heart disease.Ann Thorac Surg. 2000; 70: 1143-1144Abstract Full Text Full Text PDF PubMed Scopus (16) Google Scholar, 4Grover F.L. Shroyer A.L. Hammermeister K. et al.A decade's experience with quality improvement in cardiac surgery using the Veterans Affairs and Society of Thoracic Surgeons national databases.Ann Surg. 2001; 234: 464-472Crossref PubMed Scopus (196) Google Scholar]. Efforts to validate the NCD with the Medicare data continue. A MEDPAR dataset defined by International Classification of Diseases, ninth revision, codes rather than diagnosis-related group codes is being created to help identify procedures that were missed in the previous match and the above analysis will be repeated. Comparison of mortality at the institutional level will follow, allowing the STS to identify “outlier” institutions with respect to mortality rates. CommentThe validity of the NCD is crucial if it is to continue to serve as a benchmark for participants and a resource for outcomes research and quality improvement efforts. The STS has made and continues to make efforts to assure the validity of the NCD. Prospectively defined and collected clinical data are more detailed and accurate than administrative data, but both can suffer from the question of validity. The STS validation initiatives have been directed at maximizing the benefits of a clinical data set: collection of data by clinical personnel who should have a better understanding of the data than administrative personnel, more accurate recording of data of interest in cardiac surgery because of a database specifically designed for that purpose, and standardization of definitions. While efforts to confirm the validity of the NCD are in progress, work done to date suggests that the NCD is accurate.The goal of the STS is not just to produce a high quality clinical data set, but also to improve the process and outcomes of cardiac surgical care. Unlike administrative data, clinical process and outcomes data derived from and immediately relevant to surgeons' patients can be distributed in a timely manner to support quality improvement efforts. Indeed, the STS has just completed the first successful national randomized trial of continuous quality improvement in medicine, built on the backbone of the STS NCD [5Ferguson Jr, T.B. Peterson E.D. Coomb L.P. et al.Use of continuous quality improvement to increase use of process measures in patients undergoing coronary artery bypass graft surgery.JAMA. 2003; 290: 49-56Crossref PubMed Scopus (162) Google Scholar]. In addition, clinical data sets can be modified to include and exclude data elements to support specific areas of interest. The NCD has made possible numerous research projects and quality improvement initiatives. Continual efforts to assure validity of the STS NCD make such work possible. The validity of the NCD is crucial if it is to continue to serve as a benchmark for participants and a resource for outcomes research and quality improvement efforts. The STS has made and continues to make efforts to assure the validity of the NCD. Prospectively defined and collected clinical data are more detailed and accurate than administrative data, but both can suffer from the question of validity. The STS validation initiatives have been directed at maximizing the benefits of a clinical data set: collection of data by clinical personnel who should have a better understanding of the data than administrative personnel, more accurate recording of data of interest in cardiac surgery because of a database specifically designed for that purpose, and standardization of definitions. While efforts to confirm the validity of the NCD are in progress, work done to date suggests that the NCD is accurate. The goal of the STS is not just to produce a high quality clinical data set, but also to improve the process and outcomes of cardiac surgical care. Unlike administrative data, clinical process and outcomes data derived from and immediately relevant to surgeons' patients can be distributed in a timely manner to support quality improvement efforts. Indeed, the STS has just completed the first successful national randomized trial of continuous quality improvement in medicine, built on the backbone of the STS NCD [5Ferguson Jr, T.B. Peterson E.D. Coomb L.P. et al.Use of continuous quality improvement to increase use of process measures in patients undergoing coronary artery bypass graft surgery.JAMA. 2003; 290: 49-56Crossref PubMed Scopus (162) Google Scholar]. In addition, clinical data sets can be modified to include and exclude data elements to support specific areas of interest. The NCD has made possible numerous research projects and quality improvement initiatives. Continual efforts to assure validity of the STS NCD make such work possible.