Use of Acute Physiology and Chronic Health Evaluation II in nontraumatic neurological intensive care patients: calibration and discrimination analysis

Context

The Acute Physiology, Age and Chronic Health Evaluation (APACHE) II model has been extensively used since its publication in 1985, and there is a recommendation from the Ministry of Health for its use in Brazilian intensive care units. The severity scoring systems were developed to describe populations of intensive care patients from the perspective of gravity of disease. To meet this objective it is necessary to analyze the capacity of the scoring systems to describe the population in which they are to be used.

Objective

To analyze the calibration and discrimination properties of the APACHE II severity score system in nontraumatic neurological intensive care unit (NICU) patients.

Design, settings and population

A prospective study of all nontraumatic neurological patients admitted to two NICUs at two tertiary care level hospitals located in São Paulo city metropolitan area, between March 2002 and February 2003. The patients were followed until death or hospital discharge.

Statistical analysis

The area under the receiver operator characteristic curve was used to analyze discrimination and calibration through a Hosmer–Lemeshow goodness-of-fit test.

Results

There were 499 nontraumatic neurological patients admitted in the study period, with a mortality rate of 7.6% and a total standardized mortality ratio of 1.20. The neurosurgical patients were responsible for 76.9% of admissions (8% of emergency surgery); 52.3% of the neurological patients and 39.5% of neurosurgical patients were classified (using the APACHE II classification system) as other neurological/neurosurgical disease as the main diagnosis cause of admission. The discrimination was found to be excellent and the inclusion of the admission cause diagnosis did not appear to increase the discrimination further (area under the curve of 0.932 for APACHE II score and of 0.925 for APACHE II mortality). However, the goodness of fit was not adequate (18.73; P = 0.02; 8 degrees of freedom).

Conclusion

The APACHE II method showed an excellent discrimination and an inadequate calibration in this nontraumatic NICU population. A total 42.6% of patients were classified as 'other neurological/neurosurgical disease' as the main cause of admission, which could give room for improvement of the method.