Interobserver variability of the measurement of the mortality probability models (MPM II) in the assessment of severity of illness.

Abstract

To assess the interobserver variability of the measurement of the MPM II system. Random sample of an inception cohort. Fifteen adult medical and surgical intensive care units (ICUs) in Spain. A 5% random sample (n = 119) of 2332 patients consecutively admitted in the ICU, aged 18 years or older. Prospective data collection during the stay of the patient in the ICU. Repeated data collection, after the patient's hospital discharge. Demographic information, length-of-stay and vital status at both ICU and hospital discharge, as well as all variables necessary for computing the MPM II system were measured. Interobserver variability for categorical variables was measured computing the kappa index. For interval variables the interobserver variability was assessed by the intraclass correlation coefficient, the paired t-test, and linear regression. In the MPM II0 index, the variables coma, heart rate, systolic blood pressure, chronic renal insufficiency, metastatic neoplasm, acute renal failure, cardiac dysrhythmia and cardiopulmonary resuscitation, showed moderate or low levels of agreement. In the MPM II24 index, the variables coma or deep stupor, PaO2 and prothrombin time showed low or moderate interobserver agreement. Agreement for the MPM II probability of death was high, although the external observer obtained a lower estimation than the hospital observer. The MPM II severity system showed a high level of stability when used in groups of patients. Nevertheless, the observed variability in some variables means that it should be used with caution for individual patients.