Successful Linking of The Society of Thoracic Surgeons Adult Cardiac Surgery Database to Centers for Medicare and Medicaid Services Medicare Data

Transcription

1 Successful Linking of The Society of Thoracic Surgeons Adult Cardiac Surgery Database to Centers for Medicare and Medicaid Services Medicare Data Jeffrey Phillip Jacobs, MD, Fred H. Edwards, MD, David M. Shahian, MD, Constance K. Haan, MD, MS, John D. Puskas, MD, David L. S. Morales, MD, James S. Gammie, MD, Juan A. Sanchez, MD, J. Matthew Brennan, MD, Sean M. O Brien, PhD, Rachel S. Dokholyan, MPH, Bradley G. Hammill, MS, Lesley H. Curtis, PhD, Eric D. Peterson, MD, MPH, Vinay Badhwar, MD, Kristopher M. George, MD, John E. Mayer, Jr, MD, W. Randolph Chitwood, Jr, MD, Gordon F. Murray, MD, and Frederick L. Grover, MD The Congenital Heart Institute of Florida (CHIF), All Children s Hospital and Children s Hospital of Tampa, University of South Florida College of Medicine, Cardiac Surgical Associates of Florida (CSAoF), Saint Petersburg and Tampa, Florida; Shands Jacksonville, University of Florida College of Medicine-Jacksonville, Jacksonville, Florida; Massachusetts General Hospital, Harvard University Medical School, Boston, Massachusetts; Emory University Hospital Midtown, Emory University, Atlanta, Georgia; Texas Children s Hospital, Baylor College of Medicine, Houston, Texas; University of Maryland Medical Center, Baltimore, Maryland; Saint Mary s Hospital, Waterbury, Connecticut; Duke Clinical Research Institute, Duke University Medical Center, Durham, North Carolina; Cardiac Surgical Associates of Florida, Orlando, Florida; Children s Hospital Boston, Harvard University Medical School, Boston, Massachusetts; Brody School of Medicine, East Carolina University, Greenville, North Carolina; West Virginia University, Morgantown, West Virginia; and University of Colorado Denver, School of Medicine, Aurora, Colorado Background. The Centers for Medicare and Medicaid Services (CMS) Medicare database complements The Society of Thoracic Surgeons (STS) database by providing information about long-term outcomes and cost. This study demonstrates the feasibility of linking STS data to CMS data and examines the penetration, completeness, and representativeness of the STS database. Methods. Using variables common to both STS and CMS databases, STS operations were linked to CMS data for all CMS coronary artery bypass graft surgery (CABG) hospitalizations discharged between 2000 and 2007, inclusive. For each CMS CABG hospitalization, it was determined whether a matching STS record existed. Results. Center-level penetration (number of CMS sites with at least one matched STS participant divided by the total number of CMS CABG sites) increased from 45% to 78%. In 2007, 854 of 1,101 CMS CABG sites (78%) were linked to an STS site. Patient-level penetration (number of CMS CABG hospitalizations done at STS sites divided by the total number of CMS CABG hospitalizations) increased from 51% to 84%. In 2007, 94,409 of 111,967 CMS CABG hospitalizations (84%) were at an STS site. Completeness of case inclusion at STS sites (number of CMS CABG cases at STS sites linked to STS records divided by the total number of CMS CABG cases at STS sites) increased from 88% to 97%. In 2007, 88,857 of 91,363 CMS CABG hospitalizations at STS sites (97%) were linked to an STS record. Conclusions. The successful linking of STS and CMS databases demonstrates high and increasing penetration and completeness of the STS database. Linking STS and CMS data will facilitate studying long-term outcomes of cardiothoracic surgery. (Ann Thorac Surg 2010;90:1150 7) 2010 by The Society of Thoracic Surgeons The Adult Cardiac Surgery Database of The Society of Thoracic Surgeons (STS) is the largest clinical cardiac surgical database in the world [1]. Begun in 1989, this Accepted for publication May 10, Presented at the Forty-sixth Annual Meeting of The Society of Thoracic Surgeons, Fort Lauderdale, FL, Jan 25 27, Address correspondence to Dr Jacobs, Cardiac Surgical Associates of Florida, 625 Sixth Ave S, Ste 475, St Petersburg, FL 33701; jeffjacobs@msn.com. For related article, see page 1065 database is the premier clinical data registry for cardiac surgery. As of December 31, 2009, the STS Adult Cardiac Surgery Database has 1,004 participants, representing 1,020 hospitals in the United States. The STS Adult Cardiac Surgery Database currently contains records from more than 3.9 million cardiac surgical operations by The Society of Thoracic Surgeons /$36.00 Published by Elsevier Inc doi: /j.athoracsur

2 Ann Thorac Surg JACOBS ET AL 2010;90: STS-CMS LINK 1151 The STS Adult Cardiac Surgery Database captures detailed clinical data on adults undergoing cardiac surgical procedures performed by participants throughout the United States. The wealth of clinical information in the STS database has been the basis of well over 100 peer-reviewed publications and multiple presentations at national and international meetings. The information collected over a 20-year period has been used to improve patient outcomes, primarily through feedback provided by center-specific reports that allow participants to compare local results against national risk-adjusted benchmarks. Using risk-adjusted data from the STS Adult Cardiac Surgery Database, STS has created multiple statistical risk-adjustment models for numerous outcomes [2 4]. In July, 2009, The Annals of Thoracic Surgery published a comprehensive three-part series of articles describing the development and validation of 27 new STS cardiac surgery risk models (one for each of nine outcomes and three major categories of procedures isolated coronary artery bypass graft surgery [CABG], isolated valve surgery, and CABG plus valve surgery). These articles, which include the intercepts and coefficients for all models, established a new standard of transparency for health care risk models. In 2005, STS worked with the National Quality Forum to obtain national endorsement of 21 cardiac surgery performance measures. Much of the evidence base for these measures was obtained from the STS database. These measures are the first set of national standardized performance measures to assess the performance and outcomes in cardiac surgery [5]. Evaluation of long-term patient outcomes is a key goal of STS. Despite the many strengths of the STS database, an ongoing drawback has been the lack of any follow-up information beyond hospital discharge and 30 days after a procedure. Furthermore, the accuracy of information collected postdischarge and up to 30 days postprocedure varies among institutions, and addressing this variability is a major goal of STS. The Medicare database of the Centers for Medicare and Medicaid Services (CMS Medicare Database) complements the STS database by providing information about long-term outcomes and cost. The purpose of this study is to demonstrate the feasibility of linking the STS database to CMS data and to examine the penetration, completeness, and representativeness of the STS Adult Cardiac Surgery Database. Material and Methods Institutional Review Board Approval This study was approved by the Duke University Health System Institutional Review Board. Because the data used in analysis represent a limited data set (no direct patient identifiers) that was originally collected for nonresearch purposes, and the investigators do not know the identity of individual patients, the analysis of these data was declared by the Duke University Health System Institutional Review Board to be research not involving human subjects [6]. Medicare Database of Centers for Medicare and Medicaid Services Medicare is health insurance provided by the federal government of the United States for the following groups of patients: (1) people aged 65 years or older, (2) people younger than 65 with certain disabilities, and (3) people of any age with end-stage renal disease, defined as permanent kidney failure requiring dialysis or a kidney transplant. The CMS administrative claims data source for this study is the 100% Medicare inpatient claims file, which contains information on hospitalizations of patients enrolled in fee-for-service Medicare. It includes dates of service and diagnostic codes from the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). The database contains anonymous patient identifiers that enable follow-up of beneficiaries over time, but does not enable identification of any beneficiary through their Medicare health insurance number. In addition, the 100% Medicare denominator file, which links to the inpatient file, contains information on beneficiary eligibility, demographic characteristics, and date of death. The STS database was linked to CMS claims files using a probabilistic matching algorithm that overcomes the need for a universal patient identifier [1, 6, 7]. Records in the two databases were considered to be the same patient if they matched on a set of indirect identifiers including hospital, date of birth, sex, admission date, and discharge date. The matching algorithm required an exact match on a few of these variables or partial matching on a larger number of variables. In a pilot investigation using claims data for patients with heart failure, almost all (99.9%) CMS records from the same hospital were uniquely identified using date of birth, along with any combination of admission date and discharge date, regardless of patient sex. The Duke Clinical Research Institute used variables common to both the STS Adult Cardiac Surgery Database and the CMS Medicare Database to link STS operations to CMS inpatient claims data for all hospitalizations during which a patient underwent CABG surgery, with or without concomitant procedures. These hospitalizations are termed CMS CABG hospitalizations. This study was restricted to patients discharged from 2000 through 2007, inclusive, who were 65 years of age or older on the date of discharge. The vast majority of patients included in this analysis were enrolled in traditional fee-for-service Medicare. Although Medicare tends not to receive claims for other types of patients, this analysis includes a few very rare cases of patients who were not traditional fee-for-service Medicare because their managed care plans sent claims to Medicare. For each CMS CABG hospitalization, it was determined whether a matching STS record existed. The CMS CABG hospitalizations were identified by ICD-9 procedure codes to For each CABG hospitalization, it was determined whether there was a matching record

3 1152 JACOBS ET AL Ann Thorac Surg STS-CMS LINK 2010;90: in the STS Adult Cardiac Surgery Database. The term CMS site is used as shorthand for sites with a unique CMS provider number, which is the identification number of the institutional provider certified by Medicare to provide services to the beneficiary. (A single hospital can have more than one CMS provider number.) The CMS sites were classified as performing CABG and termed CMS CABG sites if they had more than 10 CABG hospitalizations in a calendar year. This rule was intended to reduce the chance of misclassifying a non- CABG hospital as a CABG hospital owing to ICD-9 coding errors. A CMS site is considered to have STS participation at the time of discharge if there was at least one matched STS record at the site that month. A CMS CABG claim was considered to be a legitimate CABG admission (ie, not a coding error) if there were at least 10 other CABG claims submitted from the same CMS site (billing center [MCARE_ID]) during the same calendar year. Thus, the analysis only includes claims from CMS sites with more than 10 CABG claims in a given calendar year. Once the individual patients are linked, longitudinal records can be created containing follow-up information, including subsequent death and subsequent hospitalizations. From this follow-up information, it is possible to identify and capture readmissions, and to obtain data including primary and secondary diagnoses, any repeat surgical procedures, mortality, major morbidity events, and charges. Penetration, Completeness, and Representativeness Analyses performed included calculation of the percentage of CMS CABG sites that had at least one matching STS participant, the percentage of CMS CABG hospitalizations that were linked to the STS Adult Cardiac Surgery Database, and variations of each of these. Two types of penetration are defined. Center-level penetration was defined as the number of CMS sites with at least one matched STS participant divided by the total number of CMS CABG sites. A CMS site was defined as a CMS CABG site if it had more than 10 CABG hospitalizations in a calendar year. Patient-level penetration was defined as the number of CMS CABG hospitalizations done at STS sites divided by the total number of CMS CABG hospitalizations. Completeness of case inclusion at STS sites was defined as the number of CMS CABG cases at STS sites linked to STS records divided by the total number of CMS CABG cases at STS sites. Representativeness of STS data was studied by comparing linked and unlinked CMS CABG operations at STS sites. Differences were assessed using the Wilcoxon rank sum test for continuous variables (age) and 2 tests for categorical variables (sex, race, admission type, inhospital mortality). Thirty-day mortality rates were estimated for linked and unlinked CABG hospitalizations using the Kaplan-Meier method and compared by a Z-test using the Greenwood standard error estimator. Results Center-Level Penetration A total of 1,371 CMS sites had at least one CABG claim during 2000 to Of these, 926 (68%) had CABG claims in all 8 years, and 445 had CABG claims in fewer than 8 years. Among the 10,968 possible combinations of CMS site and calendar year (ie, 8 1,371 10,968 site-years), there were 8,887 site-years with at least one CABG claim. Of these, 8,422 (94.8%) met the more-than-10-case inclusion threshold and were included in the analysis of penetration and completeness for that calendar year. Conversely, 465 site-years fell below the more-than-10- case inclusion criterion and were excluded. Among 1,172,167 CABG claims submitted in , all but 1,974 of them (1,974 of 1,172, %) were done at site-years meeting the more-than-10-case volume threshold and were included. Among the 1,173 CMS centers that submitted CABG claims in 2007, 1,101 of them (93.9%) met the more-than- 10-case inclusion threshold and were included in the analysis of penetration and completeness in Conversely, 72 centers fell below the more-than-10-case inclusion criterion in 2007 and were excluded. Among 112,325 CABG claims in 2007, all but 358 of them (358 of 112, %) were done at sites meeting the morethan-10-case volume threshold and were included. In 2007, the final analysis population included 111,967 CABG admissions from 1,101 CMS sites. Center-level penetration (number of CMS sites with at least one matched STS participant divided by the total number of CMS CABG sites) increased from 45% to 78% from 2000 to In 2007, 854 of 1,101 CMS CABG sites (78%) were linked to an STS site. Table 1 documents center-level penetration during the years 2000 to 2007, inclusive. Figure 1 documents center-level penetration of the STS Adult Cardiac Surgery Database stratified by state, for the year Table 1. Center-Level Results: Number of CABG Sites in CMS Linked to an STS Participant Number of CMS CABG Sites Center-Level Year All STS Sites Penetration N 1 N 2 N 2 /N % , % , % , % , % , % , % , % N 1 number of CMS sites performing CABG in a calendar year; N 2 number of CMS CABG sites that participated in STS for at least 1 month of the year. CABG coronary artery bypass graft surgery; CMS Centers for Medicare and Medicaid Services; STS The Society of Thoracic Surgeons.

4 Ann Thorac Surg JACOBS ET AL 2010;90: STS-CMS LINK 1153 Fig 1. Center-level penetration of The Society of Thoracic Surgeons Adult Cardiac Surgery Database stratified by state, for the year Patient-Level Penetration Patient-level penetration (number of CMS CABG hospitalizations done at STS sites divided by the total number of CMS CABG hospitalizations) increased from 51% to 84% from 2000 to In 2007, 94,409 of 111,967 CMS CABG hospitalizations (84%) were at an STS site. Table 2 documents patient-level penetration during the years 2000 to 2007, inclusive. Figure 2 documents patient-level penetration of the STS Adult Cardiac Surgery Database stratified by state, for the year Completeness Completeness of case inclusion at STS sites (number of CMS CABG cases at STS sites linked to STS records divided the by total number of CMS CABG cases at STS sites) increased from 88% to 97% from 2000 to In 2007, 88,857 of 91,363 CMS CABG hospitalizations at STS sites (97%) were linked to an STS record. Table 2 documents the completeness of case inclusion at STS sites during the years 2000 to 2007, inclusive. Representativeness Table 3 compares linked and unlinked CABG operations in 2007 at STS sites. The data in Table 3 are an analysis of CMS records at sites with STS participation in Table 3 presents a comparison of CABG claims that were matched versus those that were not matched Table 2. Patient-Level Results: Number of CMS CABG Hospitalizations Matched to an STS Record Number of CMS CABG Hospitalizations CMS CABG Hospitalizations at Active STS Sites Year All At STS Sites Patient-Level Penetration All Linked to STS Completeness N 3 N 4 N 4 /N 3 N 5 N 6 N 6 /N ,633 86,358 51% 79,198 69,408 88% ,243 86,887 53% 82,267 73,353 89% ,240 97,260 59% 91,648 80,017 87% ,685 95,671 62% 91,777 87,554 95% ,936 97,397 66% 92,155 88,770 96% ,169 97,840 72% 94,643 91,649 97% ,320 97,870 79% 94,789 92,016 97% ,967 94,409 84% 91,363 88,857 97% N 3 number of CMS CABG hospitalizations; N 4 number of CMS CABG hospitalizations at sites with STS participation during same year; N 5 number of CMS CABG hospitalizations at sites with STS participation during same month; N 6 number of CMS CABG hospitalizations linked to STS record. CABG coronary artery bypass graft surgery; CMS Centers for Medicare and Medicaid Services; STS The Society of Thoracic Surgeons.

5 1154 JACOBS ET AL Ann Thorac Surg STS-CMS LINK 2010;90: Fig 2. Patient-level penetration of The Society of Thoracic Surgeons Adult Cardiac Surgery Database stratified by state, for the year to records in the STS database during periods of STS participation. Comment The STS Adult Cardiac Surgery Database has been successfully linked to CMS Medicare data. Initial examination of the combined dataset provides reassuring information about the penetration and completeness of the STS Adult Cardiac Surgery Database. This combined dataset will be a rich source of information about longterm mortality, rehospitalization rates, and health care economics. Administrative datasets such as the CMS Medicare Database can complement the clinical data contained in the STS database. This pilot study was designed to prove Table 3. Representativeness of STS Data in 2007 a Not matched to STS (n 2,506) Matched to STS (n 88,857) p Value Age, years, mean (SD) 74.3 (6.1) 74.1 (6.0) Sex, male 1,612 (64.3%) 59,583 (67.1%) Race White 2,185 (87.2%) 81,745 (92.0%) Black 208 (8.3%) 4,091 (4.6%) Other/unknown 113 (4.5%) 3,021 (3.4%) Admission type Emergent 718 (28.7%) 20,766 (23.4%) Urgent 613 (24.5%) 23,435 (26.4%) Elective 1,167 (46.6%) 44,482 (50.1%) Other/unknown 8 (0.3%) 174 (0.2%) Mortality b In-hospital 144 (5.7%) 3,316 (3.7%) day (Kaplan-Meier) 8.8% (7.7%, 9.9%) 5.4% (5.3%, 5.6%) a Only includes admissions during months of STS participation. Sample size (n 91,363) matches the number N 5 in Table 2. coronary artery bypass graft surgery (CABG) cases: isolated CABG and CABG combined with other operations. SD standard deviation; STS The Society of Thoracic Surgeons. b Mortality is for all

6 Ann Thorac Surg JACOBS ET AL 2010;90: STS-CMS LINK 1155 the feasibility of linking the STS database to CMS Medicare data and to examine the penetration, completeness, and representativeness of the STS Adult Cardiac Surgery Database. From 2000 to 2007, the penetration and completeness of the STS Adult Cardiac Surgery Database has increased. Center-level penetration (defined as the number of CMS sites with at least one matched STS participant divided by the total number of CMS CABG sites) increased from 45% to 78%. Patient-level penetration (defined as the number of CMS CABG hospitalizations done at STS sites divided by the total number of CMS CABG hospitalizations) increased from 51% to 84%. Completeness of case inclusion at participating sites (defined as the number of CMS CABG cases at STS sites linked to STS records divided by the total number of CMS CABG cases at STS sites) increased from 88% to 97%. The STS Adult Cardiac Surgery Database currently has 1,004 participants who are either practice groups of cardiothoracic surgeons or individual cardiothoracic surgeons. These 1,004 participants represent 1,020 hospitals in the United States. Based on previous internal studies, STS believes that current STS participating hospitals represent over 90% of hospitals performing adult cardiac surgery in the United States. Thus, the STS Adult Cardiac Surgery Database has very high penetration in the United States. This penetration has increased every year since 2000 and should continue to increase. The completeness of case inclusion at STS sites is also quite high at 97%. This completeness also has shown a steady trend toward improvement over time. Because all hospitals do not submit claims to CMS (ie, Kaiser and military hospitals), it is possible that the data reported in this manuscript about center-level penetration either slightly overestimate or underestimate the true center-level penetration of the STS Adult Cardiac Surgery Database. Furthermore, because data from Medicare do not include all CABG surgery performed in the United States (75% of STS cases matched to CMS claims), it is possible that the data reported in this manuscript about patient-level penetration and completeness actually either slightly overestimate or underestimate the true penetration and completeness of the STS Adult Cardiac Surgery Database. Finally, because of matching errors, we may be overestimating or underestimating completeness. Higher mortality was observed for unmatched cases at STS sites; the reasons for this finding are unclear and bear further investigation. Several potential explanations might contribute: (1) Selective underreporting of patients with poor outcomes may occur. (2) Higher mortality may occur among patients cared for by surgeons who do not participate in the STS database at a given hospital in comparison with the mortality of patients cared for by surgeons who do participate in the STS database at the same hospital. (3) The STS database may fail to capture patients who are not part of the routine pathway secondary to lack of systematic protocols to capture these nonpathway patients in the database. (These nonpathway patients are patients who are outside the normal pathway of care, including those who die in the operating room and those who have very prolonged hospitalization. Such nonpathway patients are more likely to be accidently omitted.) (4) Errors about mortality could exist in the CMS Medicare data as well. (The eventual verification of life status with the Social Security Death Master File or the National Death Index might address this possibility). (5) Finally, the case-mix of matched and unmatched cases is different. (Unmatched cases are more likely to be black race, emergent admission type, and associated with mortality.) The power of the STS database will increase substantially by transforming it into a tool for longitudinal follow-up [1]. Linking the STS database to administrative claims databases through probabilistic matching represents one of several potential strategies that will allow longitudinal follow-up with the STS database: (1) Through probabilistic matching, the STS database can be linked to administrative claims databases and become a valuable source of information about long-term mortality, rates of rehospitalization, long-term morbidity, and cost. (2) Through deterministic matching using shared unique identifiers, the STS database can be linked to national death registries like the Social Security Death Master File and the National Death Index, to verify life-status over time [1, 8]. (3) Through either probabilistic matching or deterministic matching, the STS database can link to multiple other clinical registries, such as the National Cardiovascular Data Registry of the American College of Cardiology, to provide enhanced clinical follow-up. (4) The STS database can develop clinical longitudinal follow-up modules of its own to provide detailed clinical follow-up. On January 1, 2008, the STS database began collecting data with unique patient identifiers. Unique patient identifiers are included in the data that are routinely harvested by the STS Adult Cardiac Surgery Database for all operations as of January 1, These unique patient identifiers are being collected and stored in compliance with The Health Insurance Portability and Accountability Act of The rationale for this initiative is to facilitate the linkage of the clinical information contained in the STS database to other clinical and administrative databases. Links to other clinical databases will include those clinical databases maintained by the American College of Cardiology and additional clinical databases maintained by other medical and surgical specialties. Links to administrative databases will include the CMS Medicare Database and databases maintained by other governmental and nongovernmental agencies, as well as the Social Security Death Master File. Studies based on linked observational data may be more robust and reflective of the real world than studies based on prospective randomized clinical trials (RCT), the putative gold standard. Patients enrolled in RCT may not represent a true cross-section of the population. Furthermore, some diseases may be so rare that they will never be studied with RCT. Studies based on such linked observational data may (1) be more reflective of the clinical realities of the world, (2) be more finan-

7 1156 JACOBS ET AL Ann Thorac Surg STS-CMS LINK 2010;90: cially feasible to perform, (3) have the potential to involve many more patients, and (4) have the ability to answer questions about comparative effectiveness that would be impossible to answer with RCT. As the techniques are developed for linking clinical databases with administrative claims data on resource utilization, the ability to determine the long-term comparative effectiveness of various treatment options will be established. In addition, long-term data on resource utilization and comparative effectiveness will be necessary as technological and scientific advances place increasing pressure on societal healthcare resources. These linkages will allow the examination of a wide variety of data elements across the entire spectrum of care and will serve as a model for other clinical database networks in the future. As the demand and expectation of public reporting of outcomes increases, the ability to verify data in the STS database with other databases and national death registries, as well as the ability for the STS database to function as a tool for longitudinal follow-up, will become even more important [9]. In conclusion, the successful linking of the STS Adult Cardiac Surgery Database and the CMS Medicare Database demonstrates high and increasing penetration and completeness of the STS database. Higher mortality was observed for unmatched cases at STS sites; the reasons for this are unclear and bear further investigation. Linking STS and CMS data will facilitate studying long-term outcomes of cardiothoracic surgery. References 1. Jacobs JP, Haan CK, Edwards FH, et al. The rationale for incorporation of HIPAA compliant unique patient, surgeon, and hospital identifier fields in the STS database. Ann Thorac Surg 2008;86: Shahian DM, O Brien SM, Filardo G, et al, for the Society of Thoracic Surgeons Quality Measurement Task Force. The Society of Thoracic Surgeons 2008 cardiac surgery risk models: part 1 coronary artery bypass grafting surgery. Ann Thorac Surg 2009;88(Suppl): O Brien SM, Shahian DM, Filardo G, et al, for the Society of Thoracic Surgeons Quality Measurement Task Force. The Society of Thoracic Surgeons 2008 cardiac surgery risk models: part 2 isolated valve surgery. Ann Thorac Surg 2009; 88(Suppl): Shahian DM, O Brien SM, Filardo G, et al, for the Society of Thoracic Surgeons Quality Measurement Task Force. The Society of Thoracic Surgeons 2008 cardiac surgery risk models: part 3 valve plus coronary artery bypass grafting surgery. Ann Thorac Surg 2009;88(Suppl): Available at The National Quality Forum website: Accessed August 1, Dokholyan RS, Muhlbaier LH, Falletta J, et al. Regulatory and ethical considerations for linking clinical and administrative databases. Am Heart J 2009;157: Hammill BG, Hernandez AF, Peterson ED, Fonarow GC, Schulman KA, Curtis LH. Linking inpatient clinical registry data to Medicare claims data using indirect identifiers. Am Heart J 2009;157: Morales DLS, McClellan AJ, Jacobs JP. Empowering a database with national long-term data about mortality: the use of national death registries. Cardiol Young 2008;18(Suppl 2): Jacobs JP, Cerfolio RJ, Sade RM. The ethics of transparency: publication of cardiothoracic surgical outcomes in the lay press. Ann Thorac Surg 2009;87: DISCUSSION DR JOHN S. IKONOMIDIS (Charleston, SC): While members of the audience are gathering their thoughts, Jeff, I really think we need to spend a little bit more time talking about the unmatched cohort of patients, because, clearly, addition of those patients to the database would change results. You have nicely described potential reporting biases and also potential systemic errors inherent with the process, and I was wondering if you could conjecture to us about what measures can be put into place or are being put into place to encourage more accurate reporting and also address the inherent systemic problems with the database? DR JACOBS: I think the question you raise is of extreme importance. First, we should remember that these unmatched cases represent 3% or less of all operations being done at these hospitals. So clearly, unmatched cases are the minority of cases at hospitals participating in the STS Database. Nevertheless, the higher mortality in unmatched cases is very concerning. The STS currently has a very active audit process in place to try to encourage complete and accurate submission of data; and, in the last year, STS has made the decision (1) to increase the number of sites that will be audited each year, and (2) to explore additional methodologies of audit to facilitate auditing even a greater number of sites. Clearly, audit is one important mechanism to help sites to (1) identify where they are missing cases, (2) identify why those cases are not getting enrolled, and (3) develop mechanisms to solve the problem and capture the minority of cases that currently are not being captured. DR JACOB BERGSLAND (Oslo, Norway): I was wondering if you have tried to do a similar study using the New York State database? That would be quite interesting, because New York State, of course, is a compulsory database, not voluntary. So it would be interesting to see how those data match up as well. DR JACOBS: I agree. I myself have not done any research with the New York State database, but I would agree with you that doing research with this linkage with a database that has mandatory participation would lead to very interesting findings. This type of research could be done both with the New York State database and also with the United Kingdom Central Cardiac Audit Database, which is a national database with compulsory participation. Thank you. DR JEFFREY B. RICH (Norfolk, VA): Jeff, that is great work. In the state of Virginia, we have done our database match, STS and the CMS data, for about 6 years now and find it a very effective clinical financial tool for looking at quality improvement and cost reduction in cardiac surgery. Have you started thinking along those lines about the use of this data? The second question relates to going back to the unmatched patients: can you query on a hospital-by-hospital basis, finding those hospitals that are participants in the STS database, and tell if there is a mismatch between the patients that CMS is reporting out of that particular hospital and those not appearing in the STS database? That is, in effect, an immediate national audit.

8 Ann Thorac Surg JACOBS ET AL 2010;90: STS-CMS LINK 1157 DR JACOBS: Thank you, Jeff. Your contributions to our profession in this domain are important and commendable. Let s talk about your second question initially. I think looking at the matched versus unmatched cases on a hospital-by-hospital basis is a very important question. It would be interesting to find out if these unmatched cases are evenly distributed across all hospitals as opposed to being concentrated in a small subset of hospitals. We have not yet done this analysis. However, Jeff, the analysis that you propose would provide some very important information. Your first question asks about actually using these data to study clinical questions about long-term outcomes and cost. This study was designed to help lay the foundation for us to have the ability to do this type of research. In fact, we now have a funded research initiative, under the leadership of Fred Edwards, that will use this linkage to examine long-term outcomes of treatment of coronary artery disease. A second funded research initiative is looking at long-term outcomes of surgery for aortic valve disease. We have recently submitted a grant to look at long-term outcomes of treatment of atrial fibrillation, and another grant will be submitted soon to examine long-term outcomes of surgery for lung cancer. So these data presented today are just the tip of the iceberg of the power that we will have from this linkage, which will allow us to look at long-term outcomes, comparative effectiveness, and health care economics.