The Waterlow score for risk assessment in surgical patients

Introduction

Perioperative scoring systems aim to predict outcome following surgery and are used in preoperative counselling to guide management and to facilitate internal or external audit. The Waterlow score is used prospectively in many UK hospitals to stratify the risk of decubitus ulcer development. The primary aim of this study was to assess the potential value of this existing scoring system in the prediction of mortality and morbidity in a general surgical and vascular cohort.

Methods

A total of 101 consecutive moderate to high risk emergency and elective surgical patients were identified through a single institution database. The preoperative Waterlow score and outcome data pertaining to that admission were collected. The discriminatory power of the Waterlow score was compared against that of the American Society of Anesthesiologists (ASA) grade and the Portsmouth Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity (P-POSSUM).

Results

The inpatient mortality rate was 17% and the 30-day morbidity rate was 29%. A statistically significant association was demonstrated between the preoperative Waterlow score and inpatient mortality (p<0.0001) and 30-day morbidity (p=0.0002). Using a threshold Waterlow score of 20 to dichotomise risk, accuracies of 0.84 and 0.76 for prediction of mortality and morbidity were demonstrated. In comparison with P-POSSUM, the preoperative Waterlow score performed well on receiver operating characteristic analysis. With respect to mortality, the area under the curve was 0.81 (0.80–0.85) and for morbidity it was 0.72 (0.69–0.76). The ASA grade achieved a similar level of discrimination.

Conclusions

The Waterlow score is collected routinely by nursing staff in many hospitals and might therefore be an attractive means of predicting postoperative morbidity and mortality. It might also function to stratify perioperative risk for comparison of surgical outcome data. A prospective study comparing these risk prediction scores is required to support these findings.