Explanation of the table of results of coronary surgery

The Society of Cardiothoracic Surgeons is the first surgical or medical group in the UK to collect and publicly release their national results. The table of results for coronary surgery is based on information provided by surgeons to the UK Cardiac Surgical Register, which is run by the Society of Cardiothoracic Surgeons. The data is very basic. It does not take into account any factors which might influence the results of surgery such as how sick the patients might be. This is a serious shortcoming of this type of reporting. The Society has made considerable strides towards collecting the sort of comprehensive data required to cater for differences between the types of patient seen in different units through the National Adult Cardiac Surgical Database, but not all units yet have the facilities to collect the required data. This means we cannot yet make really meaningful comparisons between all units in the UK.

The data is reported for several categories of operations:

• All open heart operations 

• Isolated, first time coronary operations which represent bulk of work of most heart surgeons. 

• Isolated Aortic valve replacement

• Aortic valve replacement accompanied by coronary bypass surgery. This is a more complex procedure in sicker patients

Information on each of these is laid out as described below for each of these categories.

What are “Confidence Limits”? Confidence limits area statistical tool which attempts to allow for variability arising from inevitable and unpredictable differences between patients and surgeons, operations and outcomes. Any analysis based only on a limited series of cases will always be subject to potential error. If a surgeon operates on two patients and they both survive he has a mortality of 0%, but if one dies he has an statistical operative mortality of 50%, but no one would believe that either percentage represented the real risk for patients being operated on by that surgeon until he had done more cases. He might do another 48 with no deaths which would change his mortality to 2%. So the greater the number of cases with known outcomes the more certain one can be of the true risk of an operation by that surgeon. Statisticians cope with this by creating mathematical “confidence limits” around any outcome based on the number of observations or cases performed. In the example above where one of two patients dies we cannot be absolutely sure that that surgeon’s mortality is 50% but we can be 95% sure that the actual risk lies between 1.2% and 98%. By the time the surgeon has done 50 cases with only one death we can be 95% sure that his real operative mortality lies between 0.5% and 10.6%. So the greater the number of cases the more accurate the assessment becomes. In the current table we have used 99% confidence limits.

In practical terms we look at data over defined time periods: one year, two years or three years in order to be able to analyse a meaningful number of cases. But even so these short series represent only a snapshot in time and may not necessarily be representative of overall performance; even if a surgeon or unit had no deaths within the timeframe, we would not necessarily believe that this happy situation would continue forever. Similarly, a short run of bad outcomes should be regarded with similar caution.

Limitations of the reported results The way the tables have been constructed is the best we can do from the data available at the moment, but many other conditions such as underlying heart function, lung function, smoking history, diabetes, obesity, high blood pressure, kidney function and other vascular conditions all have an impact on the risk of a heart operation. These factors must all be taken into account when calculating surgical risk, particularly if meaningful comparisons between units or surgeons are to be made. For example, if a patient has had a heart attack which has significantly reduced his heart function to the point where he is very breathless he is ten times more likely to die during an operation than the average patient, yet he may have the most to gain. Such patients may be seen in some units in greater numbers than in others because some units serve older and sicker populations, or because those units have special expertise; unless such factors are taken into account unfair comparisons may be made. This type of important and clinically relevant information is simply not available in the current NHS information systems. In fact, of the four most important risk factors for coronary bypass surgery (advanced age, emergency status, poor heart function and whether this is a repeat operation), only the first two are accounted for in the published tables.

This needs to be rectified, so we are working closely with the NHS information people to ensure that the right kind of information is collected in all heart surgery units.

We favour publishing measures of how well a unit is working. We don’t think that simply ranking units according to how many patients die is necessarily the best way, because the sickest patients most in need of surgery are those who are most likely to die as a result of the operation and also the most likely to benefit. So we need to work out a better set of measures that don’t penalise units who may be trying very hard for very sick patients in their area. If we cannot sort this issue out quickly then those units that feel misrepresented by improperly weighted data may find an easy solution by simply avoiding high risk cases. Nobody want this so we need to explore different ways of finding out how well a unit is working.