| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ann Thorac Surg 2006;81:1557-1560
© 2006 The Society of Thoracic Surgeons
Division of Cardiovascular Surgery, University of Toronto and The Hospital for Sick Children, Toronto, Ontario, Canada
* Address correspondence to Dr Williams, Division of Cardiovascular Surgery, The Hospital for Sick Children, 555 University Ave, Room 1525, Toronto, ON, Canada M5G 1X8 (Email: bill.williams{at}sickkids.ca).
Jacobs and colleagues [1] presented a reasoned approach to defining death in the context of a registry database for congenital heart surgery. Their purpose is to standardize data entry to allow comparisons of operative mortality risk among institutions (and surgeons). It is interesting that we continue to focus on death after congenital heart surgery rather than focusing on the remarkable improvement in survival rate.
The inference of their paper, unproven although intuitively satisfying, is that lower operative mortality rate indicates better institutional performance. However one could imagine a scenario in which an institution adopts a protocol that accepts a higher operative risk procedure for the perceived benefit of a better long-term outcome. Such is the case in babies with a borderline left ventricle in which a biventricular repair may be attempted rather than accepting the late problems of single ventricle palliation. The change from atrial to arterial repair of infants with transposition is an historical example of accepting a higher early risk for the potential of long-term patient benefit of the more anatomic repair. Therefore some variance in institutional outcome must be tolerated for those willing to explore the limits of surgical intervention.
Comparisons of institutional risk in the surgical management of children with congenital heart disease is complicated by the diverse nature and rarity of congenital heart disease, differing case-mix among institutions, and the current low operative risk. Analysis of results after congenital heart surgery is affected by data quality, case-mix differences, and chance alone. Data quality is dependent on dedicated personnel who must achieve and maintain data integrity, and the use of standard nomenclature and definitions. Jacobs and colleagues [1] address the later point in defining death after surgery.
The essence of Jacob and colleagues' [1] definition of operative mortality is:
"... any death, regardless of cause, occurring1. within 30 days after surgery (either) in or out of the hospital and
2. any death after 30 days (if) during the same hospitalization subsequent to the operation."
Why 30 days? Why death "in or out" of hospital before 30 days?
The mortality rate after operation (or any other event) is 100%, and only the timing of death is variable. The risk of death after operation changes with time. Typically the risk is high at the time of operation but falls rapidly to a lower risk that is constant but not zero, and much later the risk rises in a third phase. These 3 phases of changing instantaneous risk of death are identified by the statistical technique of hazard function analysis [2]. Risk factors that influence outcome, either positively or negatively, in one phase may or may not affect outcome in any of the other phases. The early hazard phase of risk is unaffected by either the first 30-day period after operation or by hospital discharge.
To illustrate the early hazard phase of postoperative death, I have used our institutional registry database, including all children operated on between July 1, 1982 and October 30, 2005 (Fig 1). The early phase of risk extends well beyond either the first 30 days after the operation or hospital discharge. The timing of in-hospital mortality is shown in Table 1. The arbitrary limits of defining "operative mortality" within 30 days or prior to hospital discharge include only part of the early hazard of death after operation. Nevertheless, some finite and consistent endpoint is required to define operative mortality, and the basic definition of Jacobs and colleagues [1] is reasonable within these limitations.
|
|
The theoretical consideration of where children are discharged to (Appendix Rules 7–9) is of limited practicality and unnecessarily complicates a registry database; an example of the ideal compromising the pragmatic. As The Society of Thoracic Surgeons (STS)–European Association of Cardiothoracic Surgery (EACTS) database experience has already shown with failure to assign mortality to a specific operation, "... incomplete data entry prevented reliable statistical analysis" [1].
Appendix Rule number 12 appropriately eliminates thoracic and interventional cardiology procedures; presumably their outcomes are reported within their specialty databases [1]. However, I do take issue with eliminating extracorporeal membrane oxygenation and the "Other" category of cardiac operations. The litmus test to determine data entry in a registry database should be:
Selective data entry will create an incomplete dataset and analysis. It is the total surgical experience of the institution in managing children with cardiovascular disease that is of interest and importance in a registry database. The registry should focus on dispassionate reporting of results rather than justifying variance.
Therefore I prefer a simplification of the previous definition of Jacobs and colleagues [1] as:
"... any death, regardless of cause, occurring prior to hospital discharge."
When a child has more than one operation during the same hospital admission (Appendix Rule 2), Jacobs and colleagues [1] assign mortality to the initial operation. Their proposal is useful, pragmatic, and ideally suited to a registry database. Some will question the assignment of mortality to the initial operation, rather than the "most important" operation of that admission. However, the registry database should consider the success or failure of each surgical admission for surgery rather than that of a specific operation. The assignment of mortality to a specific operation is a different topic. The database must be able to decipher the outcome for each of a child's operations to analyze operation-specific mortality. However, the results of specific operations is a separate and secondary task to the assessment of institutional performance.
For children who have more than one admission for cardiac surgery (Appendix Rule 3), Jacobs and colleagues [1] proposes that the registry consider these children as new patients. From a registry database viewpoint this is reasonable. In clinical practice, children readmitted for subsequent surgery have a risk profile that may be either similar or very different than for their previous admission(s). It is reasonable to analyze the outcome of these children as independent events for assessment of institutional performance. The database should also allow analysis of children with more than one hospital admission for surgery, but similar to the calculation of operation-specific mortality risk, this is of secondary importance to the registry's primary function of surveying institutional performance. Repeat admissions for the same child do become an important confounding variable if one were to analyze the 1-year survival, because the child's death could be incorrectly attributed to each admission.
In the final section of their paper, Jacobs and colleagues [1] discuss the important concept of data verification. The experience in Britain is noteworthy. In Britain, every child's record in the Central Cardiac Surgery Audit Database is linked to the government's Office of National Statistics to provide an outcome (alive or dead) 1 year after the operation [3]. Their access to central tracking of mortality identified errors in institutional data and provided a more complete assessment of institutional performance without the limits imposed by the arbitrary 30-day or in-hospital limit. Central tracking of 5,494 pediatric cardiac procedures in Britain identified 469 deaths within 1 year, including 194 within the first 30 days. The institutions had identified only 78% of the deaths within 30 days. Furthermore, institutional records for 30-day mortality accounted for only 32% of all deaths within 1 year. These data further support the concept that the early hazard of death extends well beyond the initial 30-day period.
A link to government mortality data in North America would be an invaluable independent cross-reference to the registry database outcomes. It would avoid the additional expense of post-discharge data collection. The 1-year outcomes would circumvent much of the controversial issues raised by Jacobs and colleagues [1] with respect to the timing of death and the nature of hospital discharge (Appendix Rules 7–9). One-year mortality includes most of the early hazard phase of risk after operation rather than the limited and variable proportion of deaths that occur within the arbitrary 30-day or in-hospital limits. Is access to government mortality registries a possibility in the United States?
In 2001, the United States National Committee on Vital and Health statistics presented a vision for a National Health Information Infrastructure (NHII) [4]. The mandate of the NHII is to:
"... build a 21st Century support system — a comprehensive, knowledge-based system capable of providing information to all who need it to make sound decisions about health."
Included in the NHII vision for healthcare providers is the promotion of quality patient care "... by providing access to more complete and accurate patient data ..." and "... to create and monitor programs and services including health education programs and clinical research." The committee report addresses the important issues of privacy and confidentiality and states that the NHII strategy of enhancing data exchange "... is not a threat to our privacy but a vital set of resources for preventing and addressing personal and collective health threats." In their view, the Health Insurance Portability and Accountability Act "... provides a platform for exchange of financial, clinical, and administrative information in healthcare transaction." The Health Insurance Portability and Accountability Act standards "... should make it possible to have a network architecture that is all but invisible to the end users." The NHII proposes that the use of health statistics be used to "... design, implement, monitor, and evaluate specific health programs and policies."
The document led to the formation of the National Center for Health Statistics. Although each state is responsible for collecting mortality data, the National Center for Health Statistics collates these data for the nation. Perhaps the National Center for Health Statistics would develop a secure link to The STS in the future as has been accomplished in Britain. We must capitalize on expanding information technology and resist the threat of unreasonable restrictions on the use of data coordination threatened by external forces [5]. Independent central tracking of mortality would obviate the complexities inherent in Jacobs and colleagues' [1] Appendix Rules.
The joint EACTS-STS International Congenital Heart Surgery Nomenclature and Database Committee have done a remarkable job in achieving agreement on diagnostic and procedural coding. The committee now proposes criteria for defining postoperative death. The complexities imposed by some of Jacobs and colleagues [1] Appendix Rules may discourage participation by many centers. However the essence of Jacobs and colleagues' [1] definition, especially if modified to simply include all in-hospital mortality, is excellent. Until independent central tracking of mortality is available, the proposal by Jacobs and colleagues [1] for defining early mortality after cardiovascular surgery is a best approximation for monitoring institutional performance.
 |
Acknowledgments |
|---|
 Top Acknowledgments References |
|---|
|
References |
|---|
|
TopAcknowledgmentsReferences |
|---|
Related Article
This article has been cited by other articles:
|
|
 |
|
  J. P. Jacobs, M. L. Jacobs, C. Mavroudis, B. Maruszewski, C. I. Tchervenkov, F. G. Lacour-Gayet, D. R. Clarke, T. Yeh Jr, H. L. Walters III, H. Kurosawa, et al. What is Operative Morbidity? Defining Complications in a Surgical Registry Database Ann. Thorac. Surg., October 1, 2007; 84(4): 1416 - 1421. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
