Automatic calculation of a modified APACHE II score using a patient data management system (PDMS)

Abstract

Objective: To investigate a fully automated and modified APACHE II score calculation exclusively based on routine data supplied by patient data management system, the ICUData, and to assess the predictive performance of this score using analysis of discrimination and calibration at an operative ICU. Method: SQL scripts (calculation programs) were developed to calculate the scores of 524 patients who stayed at the ICU between April 1st, 1999 and March 31st, 2000. The calculation programs considered unavailable data as ‘not pathological’. The main outcome measure was survival status at ICU discharge. The discriminative power on mortality of this modified APACHE II score was checked with a receiver operating characteristic (ROC) curve. Calibration was tested using the Hosmer–Lemeshow goodness-of-fit test. Results: The 459 survivors had an average APACHE score of 17.8±5.3. The score of the 65 deceased patients averaged 22.7±4.6. The area under the ROC curve of 0.790 was significantly >0.5 ( P<0.01) and had a 95% confidence interval (CI) of 0.712–0.825. The goodness-of-fit test showed a good calibration ( H=4.89, P=0.70, dof 7, C=6.96, P=0.541, dof 8). Conclusion: A prediction model based on completely automatically calculated ‘modified APACHE II scores’ can be constructed using data collected with PDMS. However, due to differences in the patient collective and methods used, the results need validation and can only be partially compared to results from other studies.