Simplified approach to ICU severity scoring with MPM and EUROSCORE

Introduction

Severity scoring is a powerful tool for quality control in the ICU. We introduced a new severity scoring system to our ICU. The aim of this report is to describe the database and present the results of the first 12 months.

Methods

We analysed needs for database and severity scoring. We chose on-admission scoring with EUROSCORE for cardiac surgical patients, and Mortality Prediction Model II time zero for other patients. Data were collected via highly simplified forms onto a spreadsheet. Demographic entry was by ward clerk, on-admission severity scoring was by resident medical staff, and risk of death calculation and data cleaning was by senior attending medical staff. Individual patient risk of death was presented as a logit. Combined risk of death for the whole cohort was calculated from the arithmetic mean of individual logits. We estimated the time to completion of each step in data acquisition to calculate a total time spent per patient.

Results

There were 1,355 admissions, mean (SD) age 69.2 (15.9) years, 57.1% male; median (range) length of stay 23.1 (1.7 to 1,882.5) hours. Fifty-four per cent of patients were ventilated for a median (range) 7.5 (0.8 to 1,877) hours. Predicted mortality for cardiothoracic and noncardiothoracic patients combined was 8.64% and the observed mortality was 2.8%. EUROSCORE-predicted mortality for the 535 cardiothoracic patients was 5.72% and the observed mortality was 0.75% (four patients). MPM-predicted mortality for the 820 noncardiothoracic patients was 11.23% and the observed mortality was 4% (34 patients). Estimated time to complete the severity scoring form was 30 seconds, to enter a new patient on the database was 90 seconds and to calculate risk of death and check data integrity was 90 seconds. Total was 3.5 minutes per patient.

Conclusions

Data collection/analysis is essential for quality management in the ICU. Proprietary systems are expensive. Traditional scoring systems (APACHE II) are poorly calibrated to some case mixes. To overcome these problems, we devised a simple, inexpensive and highly valuable ICU database. The key features were well calibrated, on-admission severity scoring, highly simplified forms, a basic spreadsheet and collaborative staff involvement. Senior medical staff performed the final data checking. The project provided abundant high-quality data with a total input of 3.5 minutes per patient. We are trialling the database in a large ICU in China and we would welcome input from other ICUs that would like to copy our methods.