Intended for healthcare professionals


Mortality in adult cardiac surgery

BMJ 2005; 330 doi: (Published 03 March 2005) Cite this as: BMJ 2005;330:489
  1. Tom Treasure, professor of cardiothoracic surgery (tom.treasure{at}
  1. Guy's and St Thomas' Hospital, London SE1 9RT

    Named surgeons' outcomes have arrived

    Twenty years ago in the BMJ the Society of Cardiothoracic Surgeons of Great Britain and Ireland published data from its cardiac surgical register.1 Figures for 1977-82 showed wide variation in provision of coronary operations—from 14 per million of the population in south west England to 200 per million in the London area. The north west and Mersey were well down at 50 per million (see figure on Reporting was by hospital and the identities were concealed before collation. In this issue the north west surgeons, in response to a clamour that has been around for some time, provide survival figures for 25 named surgeons.2 Others have posted surgeon specific mortality on the web.3 These data are believed to be “robust, rigorous, and risk adjusted,” as required by Alan Milburn, former secretary of state for health, in parliament.4

    Does disclosure of results serve patients' best interests? To put the question another way, what purpose is served by concealment? For an analysis, read Atul Gawande (who wrote Complications5) in the New Yorker.6 He describes the experience of the family of a child with cystic fibrosis and explores their choices. Should they travel to the unit at the top of the league table or work with knowledge gained from comparison of results to make the local programme as good as it can be? As with coronary surgery, only those at the better end of performance will publish voluntarily. Looked at cynically, this may be no better than puffing their wares. But we also know that for coronary surgery in the United Kingdom, these are the same units who have published data analyses over the years.711 I argue that insights gained from knowing and working with detailed outcome data have helped their results to be among the best. More in the patients' interests might be knowledge of hospitals' overall outcomes for patients with ischaemic heart disease whether they are managed medically, have an angioplasty, or have surgery. There are ways of shifting the risk—with a tendency for the highest risk cases to arrive on the surgeon's operating table.

    Is it self evident that hospital death after coronary surgery is an outcome dependent on the surgeon, or is this a wrong assumption? Risk factors that depend on the patient (age, previous infarction, diabetes, hypertension, renal failure) must be adjusted for,7 10 11 but that done the surgeon is a key determinant of survival. The major perioperative factors are myocardial function and bleeding. More specifically, they are the effectiveness of the grafts, protection of the myocardium from ischaemia while the grafts are fashioned, and making the grafts blood tight at completion. These are dependent on the surgeon.

    In contrast, consider a disease like cystic fibrosis. Variation in survival between centres, although large,6 cannot be related to a single intervention but to the efforts of a multidisciplinary team over a lifetime. Knowledge of these differences allows teams with poorer performance to find out how others get better results.6 That was exactly the motivation behind the cardiac surgical register.1 Naming and shaming was avoided.

    Attributing outcome to individual surgeons within a service may not be in the interest of patients. Here are two detrimental effects. The very patients who have the most to gain in terms of added life years and quality face the highest perioperative risk of death. If a surgeon has had a death in the previous 10 patients the pressure is on to avoid any deaths in the next 40 to achieve 98% survival—the average for published data. Risk adjustment and use of confidence intervals should prevent incorrect inferences being drawn but a surgeon might simply become selective. Or, if a patient rescued at the point of death has suffered irretrievable brain damage for reasons beyond the surgeon's control, the surgeon may continue to strive unrealistically for survival in the interests of protecting figures. Both of these are happening now.

    Published outcomes for coronary operations have arrived. What next? Surgery for lung cancer surgery is the other major component of cardiothoracic surgeons' workload. If we take pneumonectomy, for example, here is an operation with a hospital mortality of 10-15%.12 Contrast that with mortality rates for coronary artery bypass graft.2 Unlike a coronary operation, where the procedure improves the heart and equips patients better to survive, lung resection inevitably takes away function. Most candidates are ageing lifelong smokers, and a major determinant is the patient. The important outcome is cure from lung cancer not just short term survival,13 and in any case it would be wrong to make judgments for a patient that are aimed at protecting the operator's figures.14

    In sum, survival is determined by factors in perioperative care that probably outweigh those attributable to variation in technical skill of the specialised surgeons involved. It sits somewhere in the middle ground between a single, skill dependent event such as a coronary operation and the lifelong management of cystic fibrosis by a team. In an attempt to analyse these factors, the Association of Cardiothoracic Anaesthetists is collecting data during 2005 to give insight into variation in case selection, work up, intraoperative management, and postoperative care of pneumonectomy. The outcome is clearly dependent on team effort, and yet mindful of what is happening to their cardiac surgical colleagues, a deep resistance exists among a minority of surgeons to the collection of data on pneumonectomy.


    • Embedded Image A figure showing the average annual number of operations for ischaemic heart disease in the United Kingdom from 1977 to 1982 is on

    • Papers p 506

    • Competing interests None declared


    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.
    6. 6.
    7. 7.
    8. 8.
    9. 9.
    10. 10.
    11. 11.
    12. 12.
    13. 13.
    14. 14.
    View Abstract