Intended for healthcare professionals

Paper

Mortality data in adult cardiac surgery for named surgeons: retrospective examination of prospectively collected data on coronary artery surgery and aortic valve replacement

BMJ 2005; 330 doi: https://doi.org/10.1136/bmj.330.7490.506 (Published 03 March 2005) Cite this as: BMJ 2005;330:506
  1. Ben Bridgewater the adult cardiac surgeons of north west England, consultant surgeon (ben.bridgewater{at}smuht.nwest.nhs.uk)1
  1. 1 South Manchester University Hospital, Manchester M23 9LT
  1. Correspondence to: B Bridgewater
  • Accepted 14 February 2005

Abstract

Objectives To present named surgeon mortality for isolated first time coronary artery surgery and aortic valve surgery.

Design Retrospective analysis of prospectively collected data.

Setting All NHS hospitals undertaking adult cardiac surgeryin north west England.

Participants 25 consultant surgeons carrying out coronary arterysurgery and aortic valve replacement between April 2001 andMarch 2004.

Main outcome measures Mortality for both operations accordingto surgeon. EuroSCORE to stratify patients into low and highrisk.

Results 10 163 patients underwent surgery under 25 surgeons.The average number of patients per surgeon was 363 for coronaryartery surgery and 44 for aortic valve replacement. Seventeenper cent of the patients undergoing coronary artery surgeryand half of those undergoing aortic valve surgery were consideredhigh risk. The average mortality was 1.8% (range 0-3.8%) forcoronary surgery and 1.9% (0-12.5%) for aortic valve surgery.Mortality for all surgeons fell below 99% control limits ofthe national mean for both operations.

Conclusions The presented mortality figures for the two cardiacoperations fell within accepted limits for all surgeons. Thedivision of outcomes according to low and high risk patientsis imperfect but may help to inform the public about the complexitiesof this type of analysis and prevent surgeons avoiding highrisk patents who may benefit from an operation.

Introduction

Recent years have seen a move towards increased openness andtransparency in healthcare delivery. This has been acceleratedby a series of events, including the Bristol Royal Infirmaryinquiry into paediatric cardiac surgery deaths.1 One recommendationof the inquiry was that patients must be able to see informationabout the relative performance of individual consultants operatingwithin hospitals. The Society of Cardiothoracic Surgeons ofGreat Britain and Ireland therefore published a study in 2004of activity and performance of all consultants undertaking adultcardiac surgery in the United Kingdom.2 The history leadingto this analysis and the underlying methods have been comprehensivelydescribed.3 The study was conducted on a single operation: firsttime isolated coronary artery surgery. Because of a lack ofcomprehensive data on which to perform a complete analysis thatwould allow adjustments to be made for differing case mix, thebenchmarking was done on “crude” non-adjusted mortality data.The exact mortality for individual surgeons was not given, butinstead surgeons were listed with a comment about whether theymet the Society of Cardiothoracic Surgeons standards, whichwere defined as being acceptable if the surgeon fell within99.99% confidence intervals of the national average.

Janet Smith has commented that the General Medical Council couldbe criticised for putting the interests of doctors before theinterests of patients.4 When it comes to publishing mortalitydata for individual surgeons there is potentially a conflictbetween the interests of these two groups and the confidenceintervals used recently by the Society of Cardiothoracic Surgeonshave been criticised as leaning too far towards protecting thereputations of surgeons.5 To be useful to the public and fairto surgeons it is important that any analysis of surgeon specificdata should give easily understood information that will bemeaningful, be based on robust data, and compare “like withlike.” It is also essential that any such publication shouldnot ultimately disadvantage patients by engendering a culturewhere surgeons are anxious to obtain good overall mortalityresults by turning down higher risk patients,2 6 who often havethe most to gain from successful surgery.

Recently the Freedom of Information Act has become law in Englandand Wales. This gives individuals the right to obtain data frompublic organisations. Under the act it is inevitable that individualsurgeon data will come into the public domain. Some individualhospitals are responding by putting results on the internet.We feel that it is important that any such analysis should becomprehensive and well informed. We have collected a full dataseton all patients undergoing surgery in north west England since1997 and can stratify for patient risk.

Methods

The north west quality improvement programme in cardiac interventionsis a regional consortium involving all four NHS centres thatperform cardiac surgery and percutaneous coronary interventionsin adults in north west England.7 We have collected prospectivedata on all patients undergoing cardiac surgery since 1 April1997, including preoperative and operative variables, to enablea predicted mortality to be calculated. Data were collectedin each institution and returned to a central source for analysis.Validation of data was conducted in each centre. Mortality wasdefined as any postoperative death in hospital during the admissionfor surgery.

For this analysis we looked at all patients undergoing surgerybetween 1 April 2001 and 31 March 2004, including patients undergoingprivate surgery within the NHS hospitals but excluding waitinglist initiative and private practice cases undertaken outsidethe NHS hospitals. For this analysis we have reported outcomesfor isolated first time coronary artery surgery and isolatedfirst time aortic valve surgery. To adjust for predicted riskwe have used a well accepted risk prediction algorithm, theEuroSCORE.8 If a factor necessary to calculate the score wasmissing in a record, we assumed that factor to be absent (thisoccurred in less than 2% of cases). To present the data we havesubdivided the cases into low risk (score ≤ 5) and high risk(score ≥ 6) groups.9 10 To compare mortality with national datawe used funnel plots with exact (Clopper-Pearson) 99% controllimits.11 All analysis was performed with SAS for Windows version 8.2.

Results

Tables 1 and 2 present the results for the two operations. Tomake the tables easier to read we have not shown confidence intervals.

Table 1

Outcome for isolated first time coronary artery bypass graft (CABG). Mortality for all surgeons falls below the 99% upper control limit of national mortality for crude data, and thus indicates satisfactory performance

View this table:
Table 2

Outcome for isolated first time aortic valve replacement (AVR) with survival rounded to nearest integer. Mortality for all surgeons falls below the 99% upper control limit of national mortality for crude data, and thus indicates satisfactory performance

View this table:

A total of 10 163 patients had surgery under the care of 25consultant surgeons. Each surgeon operated on an average of363 patients (range 41-567) for coronary artery bypass surgeryand 44 (3-71) for aortic valve replacement. These differencesreflect several issues, including date of the consultant's appointment,the nature of the individual's surgical practice (pure cardiacor combined cardiothoracic), subspecialist skill, referral patterns,and other commitments such as management and research. Seventeenper cent (range 7-24% between surgeons) of the patients undergoingcoronary artery surgery and half (33-69%) of those undergoingaortic valve surgery were considered high risk. Again thesevariations reflect many variables, including referral patterns,perceived skill, and the surgeon's threshold for accepting highrisk cases. For all isolated surgery the mortality was 1.8%(range 0-3.2%) for coronary artery surgery and 1.9% (0-12%)for aortic valve replacement. For low risk patients the overallmortality was 1% (0-2.4%) for coronary artery surgery was and0.9% (0% to 6%) for aortic valve replacement.

Figures 1 and 2 show the number of cases and mortality for thetwo operations. We compared mortality outcomes with publishednational data for the United Kingdom. In 2002-3 the numberswere 2.0% for coronary artery surgery and 3.2% for aortic valvereplacement.2 The figures also show 99% control limits as afunnel plot.12 For all surgeons, mortality fell within the limits,suggesting satisfactory performance.

Fig 1
Fig 2

Discussion

Principal findings

We have presented mortality data for named surgeons for isolatedfirst time coronary artery surgery and isolated first time aorticvalve surgery for all surgeons operating in north west Englandin 2001-4. We subdivided the patients into low and high riskgroups. Members of the public can now see the outcomes of theseindividual surgeons and can be reassured that all are performingto satisfactory standards. This type of analysis will becomeincreasingly common under the Freedom of Information Act.

Strengths and weaknesses

We carried out this study on a dataset produced from all patientsundergoing cardiac surgery in north west England over a threeyear period. These data have been validated locally and havethe confidence of clinicians, but they have not been subjectedto external validation. It is essential to stratify outcomesby risk to be fair to surgeons and prevent risk averse behaviour.Outcomes are influenced by “the socioeconomic status of thelocal population; severity of cardiac illness; prevalence ofco-morbidities; threshold of referral from both the generalpractitioner and the cardiologist; threshold of acceptance bythe surgeon; standards of anaesthesia, surgery and intensivecare; adequacy of the facilities and staffing levels; attitudeto training; interpersonal relationships between staff; andthe geographical layout of the unit.”3 As the surgeon is onlyone of these factors, albeit an important one, we have somereservations about publishing data for named surgeons. It does,however, seem to be what the public wants and was recommendedby the Bristol Royal Infirmary inquiry.1 As surgeons are importantin institutional change13 this type of analysis is useful asit could stimulate improvements in systems of care as well asreassure the public, as long as it does not lead to denial ofsurgery to higher risk patients.6 2

We included in our analysis all patients undergoing surgeryin NHS hospitals. During the study period, several low riskNHS patients were operated on by the same surgeons listed inthis analysis in private hospitals under waiting list initiativeschemes and have not been included in the results. This wasseen to differing extents in the four hospitals and potentiallyskews the results. Similarly, there were differing rates ofprivate patients (who are usually low risk) undergoing surgerywithin the NHS hospitals, which is a further potentially confoundingfactor.

The volume of patients undergoing coronary artery surgery waslarge, with an average of 363 per surgeon over the three years,allowing for meaningful comparisons to be made. The number ofaortic valve replacements per surgeon was much smaller. We includedthese data to be open and transparent about our results butbecause of the small numbers involved, this study has only limitedpower to detect any potentially outlying performance for aorticvalve surgery.

Comparison with other studies

There is a precedent for publishing data on named surgeons,with programmes in place in Pennsylvania, New Jersey, and NewYork. In New York data are validated so comprehensively thatthey are some three years out of date by the time they are published.The New York system produces a risk adjusted mortality for eachsurgeon, rather than the risk stratified approach we have taken,and has been criticised for encouraging risk averse behaviour.2 6While there is some disagreement as to what extent publishingnamed surgeon data actually leads to denial of surgery to highrisk patients14 our belief from experience and discussion withcardiology colleagues is that it is a serious clinical concernthat is potentially damaging to the public. Our analysis shouldhelp to moderate the consequence of publishing outcomes forindividual surgeons.

We have used the additive EuroSCORE to divide patients intolow and high risk groups.9 10 Most patients undergoing coronaryartery surgery are low risk but of all patients undergoing thisoperation half of the deaths are in the high risk patients.9 We have also shown that, while the EuroSCORE is a useful overalltool for risk prediction, it is not reliable at adjusting forrisk in high risk patients.9 We therefore simply used the EuroSCOREto subdivide the patients into the two groups and presentedthe results without further adjustment. We think that this shouldmake our analysis clear, but it is important to note that patientswithin the high risk cohort will be a heterogeneous group, rangingfrom the very elderly patient with little cardiac dysfunctionand no co-morbidities (who most surgeons would consider likelyto have a pretty good outcome with surgery) to others who haveextensive cardiac dysfunction and severe multisystem abnormalitieswho will have a high mortality with or without surgery. Alsoincluded will be patients who develop severe complications duringcardiac catheterisation who require cardiopulmonary resuscitationuntil emergency surgery is performed. These high risk patientsare often those who have the most to gain from successful surgeryand a small number of operative deaths in this group could adverselyaffect a surgeon's results but in no way be a result of pooroperative care. Because of the varied nature of high risk patientsand the limitations of existing risk adjustment models, we wouldurge caution is the comparison of outcomes between surgeonsin this group.

Similarly, patients considered to be low risk are not completelyhomogeneous. Surgeons may regard as high risk some patientscategorised as low risk by EuroSCORE—for example, thosewith very severe diffuse calcified coronary artery disease inthe absence of other risk factors, or previously fit patientsundergoing surgery with ongoing chest pain after an acute coronarysyndrome. Again a small number of fatalities in these patients,which may be due to no surgical fault, might easily distortthe results.

The cut off for risk adjustment was based on previous analysesof isolated coronary artery surgery.9 10 We used a similar cutoff for our analysis of aortic valve surgery (EuroSCORE ≤ 5).The mortality in the low risk groups was similar for both operations,but half of the patients undergoing aortic valve surgery werehigh risk compared with 17% of patients undergoing coronaryartery surgery. This may simply reflect that aortic valve surgeryoverall carries a greater risk than coronary artery surgerybut may also reflect limitations in the risk adjustment methods,which will require further investigation.

We have presented the data in funnel plots12 comparing mortalityfor each surgeon with the national mean for that operation.2 We benchmarked ourselves against national data, and, in theabsence of risk stratified data to use for comparison, we hadto use crude mortality. All surgeons' mortality in this studyfell below the 99% prediction interval of the national mean.This level seemed appropriate given the number of surgeons inthe study and the relatively small number of operations eachsurgeon had performed to try to balance the risks of missingunacceptable performance and the chance of wrongly labellingthe performance of a good surgeon as unacceptable.

Implications

This study will inform the public about outcome data for individualsurgeons. The overall analysis shows that results in north westEngland are good and no surgeon shows unacceptable performancefor either operation. The results compare favourably with nationaland international data.2 Patients in north west England willnow be able to scrutinise an individual surgeon's comparativeoutcomes, and this should provide reassurance about overallquality and give the ability to exercise choice of surgeon.As data for all surgeons fall within predefined limits on afunnel plot, however, we would argue that there would be littlebasis for using the data presented here to select one, ratherthan another, surgeon in this list. This type of analysis shouldbe updated regularly to give ongoing reassurance to patientsand should be fed into local systems of surgeon appraisal andnational programmes for professional revalidation.

Unanswered questions and future research

There has been considerable attention placed on results in cardiacsurgery but so far these have not been explicit on a named surgeonbasis in the United Kingdom. The implications for patients anddoctors of the type of analysis presented here are not fullyunderstood and will warrant further investigation. The Societyof Cardiothoracic Surgeons will soon have a full dataset onall patients undergoing surgery in Great Britain and Ireland,and a complete analysis with results for all surgeons may proveuseful. Medical and surgical specialties with less well evolveddata collection systems and techniques of risk adjustment arestill subject to disclosure of outcome data for named cliniciansunder the Freedom of Information Act, and the implications ofthis on patients and the medical profession are not yet clear.

What is already known on this topic

Analysis of crude postoperativemortality associated with isolated first time coronary arterysurgery suggests that all surgeons currently operating in theUnited Kingom are achieving predefined standards

Actual mortalitydata for individual surgeons has not been published on a nationalor regional basis

Publication of mortality data that have notbeen adjusted for risk has been criticised as being unfair tosurgeons and for encouraging surgeons to turn down high riskcases

What this study adds

Mortality associated with isolatedfirst time coronary artery surgery and isolated first time aorticvalve replacement for low and high risk patients in one areain England has been shown for named surgeons

The overall resultsare good and all surgeons perform to satisfactory standards

Use of this type of analysis should help to prevent risk aversebehaviour by surgeons and promote a culture of openness andtransparency in healthcare delivery

Footnotes

  • GraphicThe named consultant surgeons were operating at the end of March2004 in north west England. All analyses were undertaken atthe Cardiothoracic Centre Liverpool. The surgical steering groupof the programme comprises John Au, Ben Bridgewater, Brian Fabri,Antony Grayson, Geir Grotte, Mark Jackson, Mark Jones, DannyKeenan, and Russell Millner. We thank the audit officers workingin each centre for their hard work in collecting and validatingthe data. All the named surgeons approved this paper.

  • Contributors This study was conducted on behalf of all cardiacsurgeons in north west England. BB was responsible for studydesign and prepared the manuscript with Antony Grayson. AG andMark Jackson performed the analysis. All surgeons reviewed themanuscript and contributed to its final draft. BB is guarantor.

  • Funding Funding for the north west quality improvement programmein cardiac interventions collaboration has been received fromall primary care trusts in north west England. All involvedwith this study were independent from the funding.

  • Competing interests John Au, Ben Bridgewater, Brian Fabri,Antony Grayson, Geir Grotte, Mark Jackson, Mark Jones, DannyKeenan, and Russell Millner are members of the steering groupof the NWQIP. BB is a member of the tripartite group (Departmentof Health, Healthcare Commission, and Society of CardiothoracicSurgeons of GB and Ireland), responsible for steering nationalcardiac surgery audit in England and Wales.

  • Ethical approval The north west quality improvement programmewas approved by the regional ethical committee.

References

  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