Prognosis related to organ dysfunction in intensive care unit

Objective

To develop a model for assessing severity of organ dysfunction (OD) among patients on the first 24 h of Intensive Care Unit (ICU) stay, using a score to determine the probability of ICU mortality.

Design

Prospective cohort study.

Setting

General medical and surgical ICU in a tertiary teaching hospital in City of São Paulo, Brazil.

Patients

Three hundred and seventy-eight consecutive, unselected patients over the period from March to October of 1996: developmental sample. Three hundred patients over the period from February to June of 1997: validation sample.

Outcome measure

Patients vital status at ICU discharge. None intervention was considered.

Statistical analysis

APACHE II score was calculated for all patients. A Lowess Regression model, using the variables that demonstrated P ≤ 0.10 in the univariate analysis was made to identify the level of severity of each variable. The variables were then entered into a multiple logistic regression analysis resulting in a probability of ICU mortality equation. The Goodness-of-fit test was used to evaluate model calibration; discrimination was evaluated using area under the receiver operating characteristic curve (ROC), in the developmental and validation samples.

Main results

OD was considered in five systems: neurologic, pulmonary, renal, cardiovascular and hematologic, plus the presence of chronic disease. The points were assigned from 1 to 4 according to the levels of severity (Table). The results showed good calibration (P = 0.96; C = 2.33; dF = 8 and P = 0.90; C = 3.01; dF =10) respectively in the developmental and validation samples, and good discrimination (ROC curve of 0.81 and 0.82, respectively).

Conclusion

Cardiovascular dysfunction was the most severe organ dysfunction, followed by pulmonary, renal and neurologic dysfunction. Hematologic dysfunction and the presence of chronic disease were less severe. This model can be used as end point in epidemiologic studies of organ dysfunction in our ICU when the points are summed according to the horizontal lines (severity within an organ system), or as a predictor of death when the points are summed vertically, once the β is the same for all variables.

Table 1