Eur J Cardiothorac Surg 2000;17:747-751
© 2000 Elsevier Science NL
Clinical versus actual outcome in cardiac surgery: a post-mortem study
Andrew T. Goodwina,
Martin Goddardb,
Gordon J. Taylora,
Andrew J. Ritchiea
a Department of Cardiac Surgery, Papworth Hospital, Papworth Everard, Cambridge CB3 8RE, UK
b Department of Pathology, Papworth Hospital, Papworth Everard, Cambridge CB3 8RE, UK
Corresponding author. Tel.: +44-1480-830-541; fax: +44-1480-364-338
e-mail: goodwinat{at}hotmail.com
 |
Abstract
|
|---|
Background: Clinical attribution of the cause of death can be misleading, with the only true outcome measure being post-mortem analysis. Despite this there is very little published data on post-mortems following cardiac surgery. Methods: Prospective consecutive post-mortem data were collected on 167 patients (84.4% of all in-hospital cardiac surgical deaths) in a single institution. Clinical diagnoses were compared with post-mortem findings. Results: The mean age at death was 69.8 with 67.6% male. The proportion undergoing coronary artery bypass graft (CABG) alone was 52.1%, valve surgery 18.6%, valve+CABG 19.2% and other procedures 10.1%. The mean time to death was 7.9 days (range 0–87). The causes of death were cardiac 67.7%, gastrointestinal 9.6%, respiratory 8.4%, haemorrhage/technical failure 4.8%, stroke (cerebrovascular accident) 3.6%, multiorgan failure 3.0%, sepsis 1.8%, malignancy 0.6% and trauma 0.6%. Post-mortem revealed an unsuspected cause of death in 19 (11.4%). These were gastrointestinal (infarction nine, perforation two), cardiac three, adult respiratory distress syndrome two, technical two and pulmonary embolus one. In addition, an unsuspected lung cancer was found in 1 patient who died of cardiac causes. When cardiac deaths were compared with non-cardiac causes the Parsonnet score was higher 20.0 (±1.4) vs. 15.3 (±1.6), P=0.07; and a greater proportion tended to have poor ejection fractions (34 vs. 15%), P=0.12. There was no significant difference between the groups in terms of age, sex, operation, hypertension, diabetes, creatinine and body mass. Conclusions: Post-mortem can determine unsuspected diagnoses in a significant proportion of cases. Pre-operative risk factors do not correlate with eventual cause of death. Post-mortem still has an important role to play in cardiac surgery.
Key Words: Post-mortem • Cardiac surgery • Outcome • Audit • Risk stratification
|
1. Introduction
|
|---|
Despite modern advances in surgical techniques and myocardial protection, death following cardiac surgery continues to be a significant problem. In order for improvements in outcome to be made it is important that lessons are learned from past mistakes and experiences. It is important therefore that as much information as possible is obtained in patients dying following surgery. Clinical attribution of the cause of death can be misleading, with the only true outcome measure being post-mortem analysis. Despite this there are very few published reports of post-mortem findings following cardiac surgical procedures [1–3]. In the series by Zehr et al. only 24% of all deaths underwent post-mortem. Post-mortem rates are low due to various factors. There may be misconceptions that the clinical cause of death is accurate [4]. In addition, there may be fear that litigation will ensue if unexpected findings are revealed [5]. As surgeons are increasingly operating on older and sicker patients, with poorer ventricular function and a multitude of concomitant medical conditions, it remains important to define what exactly are the causes of failure. Risk stratification systems, although based on statistical populations, are increasingly being utilized in practice to define the ‘risk’ of a procedure for an individual patient [6–8].
We present a study in which the post-mortem rate is high. We aimed to compare presumed versus actual cause of death. In addition, we wished to see whether any pre-operative risk factors correlated with the eventual cause of death.
|
2. Methods
|
|---|
Prospective post-mortem data were collected on all in-hospital cardiac surgical deaths between March 1996 and January 1999 in a single institution. Transplant and aortic surgery deaths were excluded. There were 198 deaths from 5103 procedures during this period, resulting in an overall in-hospital mortality of 3.9%. Post-mortems were performed on 167 patients (84.4% of all in-hospital cardiac surgical deaths). By law in the United Kingdom any patient where there is uncertainty as to the cause of death must be referred to the coroner. In addition, all patients who die in the immediate post-operative period at our institution are referred to the coroner. In all other cases the families are approached for their consent for a hospital post-mortem to be performed. In this series 160 post-mortems were performed for the coroner (95%) with the remainder performed after consent by the relatives. The high post-mortem rate in this study compares with a mean hospital post-mortem rate in the United Kingdom of 15–20% [5,9].
The post-mortem request forms were analyzed to obtain the presumed clinical cause of death. The post-mortem and clinical cause of death were compared. Multiple pre-operative variables for each patient were obtained from a hospital database recorded at the time of surgery. The variables collected included age, sex, operation, previous cardiac surgery, left ventricular function, diabetes, hypertension, pre-existing renal failure, body mass index, urgency of operation, pre-operative intra-aortic balloon use and various catastrophic states. This allowed risk stratification according to the system devised by Parsonnet for each patient [6,7]. In addition data were collected on pre-operative creatinine and atrial fibrillation.
At post-mortem evidence was sought to confirm the presence of the disease being operated on (e.g. coronary atheroma, evidence of ventricular hypertrophy/distension in valve patients). In addition, evidence was sought of surgical misadventure. Following this the cause of death is established and whether or not it is related to the heart. In the absence of any other demonstrable abnormality it is assumed that the cause of death is due to the underlying cardiac problem which led to the operation being performed. Histological examination was performed in all patients. Each patient underwent standard sampling of the coronaries and myocardium as well as any other relevant organs.
Data were 100% complete for both post-mortem and risk data. Results are expressed as mean±SEM. Data were compared using a chi-square test or Mann–Whitney test as appropriate.
|
3. Results
|
|---|
The mean age at death was 69.8±0.8, with 67.6% male. The proportion undergoing coronary artery bypass graft (CABG) alone was 52.1%, valve surgery 18.6%, valve+CABG 19.2% and other procedures 10.1% (Table 1). The mean time to death was 7.9 days (range 0–87 days). Post-mortem confirmed that the pre-operative diagnosis for which the operation was being performed was correct in all cases (100%). The causes of death are shown in Table 2. Post-mortem revealed an unsuspected cause of death in 19 (11.4%). The unsuspected causes of death are summarized in Table 3. In 11 of the 16 patients dying of gastrointestinal causes the findings were unsuspected. The operations performed on patients dying of gastrointestinal causes were CABG seven, valve two, valve+CABG four, CABG+abdominal aortic aneurysm repair one, ischaemic ventricular septal rupture one and pericardiectomy one. The causes of death in patients undergoing the different operation types are shown in Table 4. There was no significant difference in the causes of death between the different operation types.
There were eight deaths due to surgical misadventure. In patients dying from haemorrhage the sites were aortotomy (three) and coronary anastomosis (one). The deaths due to technical error were vein graft thrombosis (one), aortic dissection due to clamping (one), retro-peritoneal haemorrhage due to an intra-aortic balloon pump (one) and a liver capsule tear during insertion of a chest drain (one).
When cardiac deaths were compared with non-cardiac causes, the Parsonnet score was higher: 20.0±1.4 vs. 15.2±1.6, P=0.07; and greater proportion tended to have poor ejection fractions (34 vs. 15%), P=0.12. There was no significant difference between the cardiac and non-cardiac groups in terms of age, sex, operation, hypertension, diabetes, body mass index, creatinine and pre-operative atrial fibrillation (Table 5).
|
4. Discussion
|
|---|
In this study we have shown that post-mortem can reveal an unsuspected cause of death in 11.4% of the patients studied. The commonest unsuspected diagnosis was large or small bowel ischaemia/infarction. It is possible that diagnosis during life would have altered the management strategies and the possible eventual outcome in these patients. This should lead to earlier referral for laparoscopy or laparotomy in this group of patients.
It has been suggested that modern diagnostic techniques leave the cause of death in much less doubt [4,5]. Post-mortem analysis, however, remains the ‘gold standard’ and final arbiter. The major advantage of post-mortem over other tests is that it allows the sampling of tissues for histological examination. This can be particularly useful in determining myocardial infarction, small pulmonary emboli and adult respiratory distress syndrome, as well as many other histological diagnoses. There are other potential benefits in the collection of such information, such as the utilization of histological tissue for research and study. Also, in order for progress to be made, the information available to clinicians from post-mortems is invaluable [5,9,10]. Such information underpins the activity of clinical audit where the basic presumption is of accurate information. If the outcome data in any study are inaccurate then any information derived from them remains flawed.
Our findings differ from previously published series in number of ways (Table 6). Our post-mortem rate is high (84%) compared with the series by Zehr et al. (24%) but similar to that of Lee et al. (88%) [2,3]. The most striking difference is in the proportion of deaths due to cardiac causes, 68 vs. 27%, in the American paper. This may be because the post-mortem rate is low and different diagnostic criteria may exist in the different patient populations. In addition, the proportion of deaths of unknown cause in the United States series is 25% whereas in this series there were none. In patients within our series where there was little to find at post-mortem, such as following a sudden or arrhythmic death, the cause has been attributed to the underlying cardiac condition for which the operation was performed. The other major difference between this series and others is the high proportion of deaths due to gastrointestinal causes (10 vs. 3 and 1%). This may also be due to different populations of patients undergoing different operations. For example, there were 53% of patients undergoing CABG in our series compared with 32 and 34% in the other published series [2,3]. In addition, the series by Zehr et al. includes transplant and aortic surgery. Other possible differences between the institutions may include differences in intra-operative technique (such as bypass perfusion pressures and systemic hypothermia) and post-operative management (such as inotrope usage and drug therapy). There were no significant differences in the pre-operative risk factors and operations performed in the patients dying of gastrointestinal causes (Tables 4 and 5).
In this study we have also shown that the correlation of risk factors appears to have little relationship to the actual cause of death. We analyzed whether any of the risk factors collected for calculation of the risk of death correlated with the eventual outcome. The patients dying from cardiac causes had higher Parsonnet scores pre-operatively and tended to have worse ejection fractions when compared with the non-cardiac deaths. There was no difference between all the other factors studied. The numbers of patients in the non-cardiac deaths category were too small to further subdivide into the different causes of death. It is perhaps not surprising that the cardiac deaths are more likely in patients with poorer pre-operative cardiac function. It is not clear why the patients with better pre-operative cardiac function should be more likely to die of the non-cardiac causes, and in particular gastrointestinal infarction. The risk of intestinal infarction is increased in patients with generalized atherosclerosis, left heart chamber thrombus, and in patients with low perfusion pressures peri-operatively. It could be argued that all of these might be commoner in patients with poorer pre-operative cardiac function.
In conclusion, post-mortem can identify undiagnosed causes of death in a significant proportion of patients following cardiac surgery. In addition, the risk factors utilized to determine risk stratification correlate poorly with the actual cause of death. Post-mortem information remains the gold standard outcome measure which underpins the improvements in clinical standards we all seek.
|
Footnotes
|
|---|
Presented at the 13th Annual Meeting of the European Association for Cardio-thoracic Surgery, Glasgow, Scotland, UK, September 5–8, 1999.
|
Appendix A Conference discussion
|
|---|
Dr M. Irarrazaval (Santiago, Chile): The first question is a reversal. How often did this post-mortem study not show any answer when you clinically didn't know for sure what was the reason of the death?
Mr Goodwin: How often did the post-mortem not show an answer? Is that what are you asking?
Dr Irarrazaval: Yes, because sometimes, particularly in the acute situation, the patient dies and you are not sure what was the reason. How often did the path study not show a clear answer?
Mr Goodwin: In some of the cardiac cases it can be very difficult to give a definite cause of death; however, in the absence of other pathologies, then a cardiac cause was given by the pathologist, having searched thoroughly for other causes.
Dr Irarrazaval: A lot of these patients have several causes, for instance, multiple organ failure or sepsis very frequently started with something like cardiac or respiratory failure. How did you handle that to assign the responsibility?
Mr Goodwin: If at post-mortem it was found to be due to multi-organ failure, then the death was assigned to that. If there was nothing else found and there were cardiac causes underlying it then it was assigned as a cardiac death.
Dr B. Buxton (Victoria, Australia): As you have indicated in your post-mortem study, we are always going to have trouble predicting patient-specific outcome as opposed to group predictions. The problem is that we have these isolated events in the patient-specific category, for example calcified aortas, liver transplantations and gastrointestinal complications which don't show up in group data but are very highly specific to patient outcome. You have brought this issue to our attention very clearly. It is essentially quite difficult to improve patient-specific prediction. We use estimates of outcome derived from group data and build on top of that a second tier of predictors which are then applied to an individual patient.
Mr Goodwin: I think one of the reasons for performing this study is perhaps because individual patients are being refused surgery on the basis of risk scoring systems which assign them as being high risk, whereas you cannot actually tell for any individual patient. Obviously there is clinical suspicion, and with a large population you can predict outcome. It doesn't seem particularly fair for one individual patient sitting in a clinic to refuse them an operation just on the basis of an overall risk scoring system.
Dr Buxton: It is very difficult to define the importance of specific patient predictors. For example, what is the risk of a calcified aorta? It is high, but we don't have enough patients with calcified aortas to to know exactly what the probability of death is, nor can we stratify those small groups. You have illustrated some of the major problems in patient-specific prediction.
Dr B. Messmer (Aachen, Germany): To go back to the first question about cardiac death or death of unknown reasons, most people are a little bit reluctant to do a post-mortem because often you don't see anything, and then you assume it's cardiac. Of course the patient dies finally because the heart stops. But in the very early phase the pathologist almost never can give you an exact cardiac cause. It takes a certain time, and that is one of the crucial points. Regarding your unexpected deaths due to intestinal complications, did you draw any conclusions from your results? We are pretty liberal by doing explorative laparotomy especially when the lactate level goes up and the patient has abdominal symptoms. In such cases mesenteric infarction or scattered bleeding is often found. If you don't do anything such a patient will die. Therefore again I ask my question: Did you draw any conclusions from your results?
Mr Goodwin: Yes. The surprising finding of this study was the high level of gastrointestinal complications, and we have subsequently gone back and re-reviewed the patients’ ITU charts, looking for whether there was anything which we could perhaps have detected prior to death. The commonest finding was, as you suggested, the lactic acidosis or metabolic acidosis with no other explanation. And certainly we have a much more liberal policy with either laparotomy, or even a diagnostic laparoscopy, looking for gastrointestinal problems in patients with unexplained metabolic acidosis.
Dr J. Pirk (Prague, Czech Republic): Well, I would suggest you should go further. You should examine these patients before the surgery, maybe for Helicobacter, or whatever. After surgery it is too late.
Mr Goodwin: Thank you.
|
References
|
|---|
-
Schoen F.J., Titus J.L., Lawrie G.M. Autopsy-determined causes of death after cardiac valve replacement. J Am Med Assoc 1983;249:899-902.[Abstract]
-
Zehr K.J., Liddicoat J.R., Salazar J.D., Gillinov A.M., Hruban R.H., Hutchins G.M., Cameron D.E. The autopsy: still important in cardiac surgery. Ann Thorac Surg 1997;64:380-383.[Abstract/Free Full Text]
-
Lee A.H., Borek B.T., Gallagher P.J., Saunders R., Lamb R.K., Livesey S.A., Tsang V.T., Monro J.L. Prospective study of the value of necropsy examination in early death after cardiac surgery. Heart 1997;78:34-38.[Abstract/Free Full Text]
-
Lee P.N. Comparison of autopsy, clinical and death certificate diagnosis with particular reference to lung cancer. A review of the published data. APMIS Suppl 1994;45:1-42.[Medline]
-
Underwood J.C., Cotton D.W., Stephenson T.J. Audit and necropsy. Lancet 1989;i:442.
-
Parsonnet V., Dean D., Bernstein A.D. A method of uniform stratification of risk for evaluating the results of surgery in acquired adult heart disease. Circulation 1989;79(Suppl I):I3-I12.
-
Nashef S.A., Carey F., Silcock M.M., Oommen P.K., Levy R.D., Jones M.T. Risk stratification for open heart surgery: trial of the Parsonnet system in a British hospital. Br Med J 1992;305:1066-1067.
-
Orr R.K., Maini B.S., Sottile F.D., Dumas E.M., O'Mara P. A comparison of four severity-adjusted models after coronary artery bypass graft surgery. Arch Surg 1995;130:301-306.[Abstract]
-
Mosquera D.A., Goldman M.D. Surgical audit without autopsy: tales of the unexpected. Ann R Coll Surg Engl 1993;75:115-117.[Medline]
-
Barendregt W.B., de Boer H.H., Kubat K. Autopsy analysis in surgical patients: a basis for clinical audit. Br J Surg 1992;79:1297-1299.[Medline]
Received September 6, 1999;
received in revised form March 17, 2000;
accepted March 21, 2000.
Top
Abstract
1. Introduction
