Direct comparison of mortality rates has limited value because most deaths are due to the disease process. Predicting the risk of death accurately remains a challenge.
Methods
A cross-sectional study compared the expected mortality rate as calculated with an administrative model to a physiological model, Acute Physiology and Chronic Health Evaluation IV. The combined cohort and stratified samples (< 0.1, 0.1-0.5, or > 0.5 predicted mortality) were considered. A total of 47,982 patients were scored from 1 July 2013 to 30 June 2014, and 46,061 records were included in the analysis.
Results
A moderate correlation was shown for the combined cohort (Pearson correlation index, 0.618; 95% confidence interval [CI], 0.380-0.779; R2 = 0.38). A very good correlation for the less than 10% stratum (Pearson correlation index, 0.884; R2 = 0.78; 95% CI, 0.79-0.937) and a moderate correlation for 0.1 to 0.5 predicted mortality rates (Pearson correlation index, 0.782; R2 = 0.61; 95% CI, 0.623-0.879). There was no significant positive correlation for the greater than 50% predicted mortality stratum (Pearson correlation index, 0.087; R2 = 0.007; 95% CI, − 0.23 to 0.387).
Conclusion
At less than 0.1, the models are interchangeable, but in spite of a moderate correlation, greater than 0.1 hospital standardized mortality ratio cannot be used to predict mortality.
