The impact of missing components of the Acute Physiology Score on the standardized mortality ratio calculated by the APACHE III prognostic model

Introduction

In the APACHE III prognostic model, a weight of zero is given to missing and normal Acute Physiology Score (APS) values. The standardized mortality ratio (SMR) is used to evaluate the performance of an ICU. The objective of this study was to determine the impact of missing APS values on SMR.

Methods

The study involved a retrospective analysis of the prospectively collected APACHE III database. The study population included adults treated in four intensive care units of a tertiary medical center with the APACHE III database. We obtained the APACHE III predicted mortality rate and hospital outcome. We noted whether the first day values of the variables used to calculate the APS were missing. We created eight categories based on the number of missing values. We summarized descriptive data as mean (standard deviation [SD]), median (interquartile range [IQR]) and percentage. We created a logistic regression model to determine the association of the number of missing values with hospital mortality by entering the number of missing APS variables and the APACHE III predicted mortality rate as predictor variables. P < 0.05 was considered significant.

Results

We had 49,682 admissions during the study period. Excluding patients who did not authorize their medical records to be reviewed for research and those with four or less available laboratory values, 42,607 patients were included in the study. Patients' mean (SD) age was 62.7 (17.3) years; 40,417 (94.9%) were white and 24,294 (57.0%) were male. The overall predicted and observed hospital mortality rates were 9.7% (4151/42607) and 12.0%, respectively, with SMR (95% confidence interval [CI]) of 0.813 (0.789–0.838). The median (IQR) predicted hospital mortality rate was 4.36 (1.65–13.20)%. Complete data were available in 934 (2.2%) admissions. The predicted mean and observed mortality rates as well as the SMR with the 95% CI for each of the missing categories are presented in Table 1. Four or more missing APS values were independently associated with higher observed hospital mortality.

Table 1 Table 1

Conclusions

The majority of patients admitted to the ICU have one or more missing laboratory values needed to calculate the APS scores of the APACHE III prognostic system. When using such prognostic systems to evaluate the performance of ICUs, the number of missing values should be taken into consideration.