Prediction and Prevention of Complications in General Surgery

Abstract

This thesis examines two concepts that impact on surgical complications – prediction and prevention. The starting point was a randomised controlled trial (RCT) which compared the frequency and the cost of infection for two prophylactic antibiotics. A parallel study also examined the surgeons’ ability to predict wound infection and major complications. The results generated a number of important clinical questions, and it is the studies that were performed to investigate these questions which determined the direction and development of the thesis.

Although prediction is fundamental to clinical practice, there was minimal objective data on the ability of the surgeon to predict complications. How good is the surgeon at predicting major complications? Can prediction be improved? How important is it when the surgeon changed the prediction of complications at the end of the procedure? This thesis addresses these questions. In contrast the ability of prophylactic antibiotics to prevent infection has been extensively studied and the theory behind their use is well understood. We wanted to assess the ‘best prophylactic antibiotic’ to use in abdominal surgery, both in terms of frequency of infection and cost effectiveness. We were also interested in the ability of the surgeon to predict wound infection and on the impact of prophylactic antibiotics on chest infection and urinary infection.

The RCT comparing the use of ceftriaxone and cefotaxime showed that ceftriaxone had some advantages, including better prophylaxis for patients undergoing appendicectomy, and better prophylaxis against chest and/or urinary infection. This also demonstrated that the cost of infection was significantly different between the two antibiotics, and that this could be used to demonstrate significant differences in the severity of infection. For some endpoints there was a significant difference in the cost of infection even when the frequency of infection was the same. Differences in the frequency of chest infection and urinary infection were the impetus for performing a meta-analysis of all RCT assessing the prophylactic use of ceftriaxone. This showed that ceftriaxone was significantly better at preventing infection than other appropriately selected antibiotics. This difference was noted mainly in high risk cases. For the wound this was most noticeable after clean-contaminated and contaminated surgery. For chest infections this was after upper abdominal surgery and for urinary infections after pelvic surgery.

The study looking at the prediction of wound infection, using a 100mm visual analogue scale (VAS), was performed immediately before and after surgery. This showed that the surgeon was poor at predicting who would, or would not, go on to develop a wound infection. However an increase in the prediction of infection postoperatively identified a high risk group of patients who developed significantly more infections. The finding that microbiological and surgical factors were only weakly predictive lead to a study which aimed to assess the importance of patient factors, as indicated by the American Society of Anesthesiologists (ASA) classification of physical status, on the development of wound infection. This demonstrated that in the context of optimal prophylactic antibiotic use that the impact of wound contamination was minimized and that host factors, as represented by ASA score, become more important in preventing wound infection. The prediction of major complications by the surgeon was then assessed in a three part prospective study. In part one prediction was made using a 100mm VAS immediately before and after surgery. In part two a preoperative multifactorial VAS, using six additional subscales was introduced. In part three predictions were completed following the presentation of detailed outcome feedback. Surgical prediction was initially ‘as good as’ the Physiology and Operative Severity Score for the enUmeration of Mortality and Morbidity (POSSUM) score. After the presentation of clinically relevant feedback there was an improvement in the prediction of major complications (from very good to excellent). This is the first time that feedback has been shown to improve the prediction of major complication.

The prediction of major complications study also supported the following observations. Increasing the prediction of major complications postoperatively was important, and confirmed the importance of intraoperative events in determining the outcome of surgery. Surgical prediction was able to improve both the discrimination and goodness of fit of a multifactorial model for predicting complications. One reason for this is the ability of the surgeon to identify infrequent, but clinically important, risk factors. The prediction of complications by the surgeon using a VAS was accurate, versatile (when compared to a number of risk models), broadly applicable (able to be used by a number of surgeons and for a wide range of procedures) and adaptable. A major limitation is the need for formal reliability testing.

When a VAS is intended to be used as a linear scale it is helpful to clearly label (or ‘anchor’) the VAS. Further studies need to look at the reliability and potential applications of surgical prediction of major complications using a VAS score.