Mortality probability model II (MPM_0–72) in 1667 patients with acute cardiovascular disorders

Objectives

Aim of this prospective study was to evaluate the mortality probability model (MPM II) in terms of outcome prediction (hospital mortality [HM]) and calibration in patients with predominantly cardiovascular disorders admitted to a medical intensive care unit (MICU).

Methods

1677 patients (pts) (age 64 ± 13 years, 68.9% male, SAPS II 27 ± 15.9, 418 pts acute myocardial infarction, 337 pts unstable angina, 217 pts rhythm disturbances, 141 pts heart failure, 103 pts cardiac arrest, 416 pts other admission diagnosis) were included between April 1999 and April 2001. SAPS II and MPM_0–72 were determined according to the published guidelines. Discrimination power of SAPS II and MPM_0–72 for survivors (S) and non-survivors (NS) (HM) was assessed by the area under the Receiver Operating Characteristic (AUROC) curve, calibration of the models with the Goodness of Fit H-Test (GOF-H) and standardized mortality ratio (SMR).

Results

Two hundred and sixty-one (15.6%) pts died. ICU mortality was 10.9%. AUROC for SAPS II was 0.83 (0.79–0.81), for MPM₀ 0.80 (0.76–0.84), for MPM₂₄ 0.83 (0.79–0.87), for MPM₄₈ 0.82 (0.78–0.86) and for MPM₇₂ 0.87 (0.82–0.89). MPM_0–72 II significantly overestimated mortality (Table 1). Moreover stratifying pts in subgroups according to age, admission process and diagnostic categories revealed a poor calibration with overestimation of mortality too.

Table 1 χ² value GOF-H and SMR (95% confidence interval)

Conclusion

Discrimination of MPM_0–72 II was reliable. However, all models showed a most significant lack of calibration overestimating mortality mainly in the low risk strata and other pt related criteria. These results evidence the importance of a difference in the uniformity of fit and case mix of the present study population compared to the original population in which the model was developed.