APACHE III was developed in 1991 to expand and improve the prognostic estimates provided by APACHE II (Knaus. etal 1991). The APACHE III database was collected between 1988 and 1990 and included data on patient and institutional characteristics within a nationally representative sample of 17 440 intensive care admissions. The study involved 42 ICUs at 40 United States hospitals, including 26 randomly selected to represent institutions with more than 200 beds according to geographic region, size, and teaching status. This process resulted in about 50 per cent of all United States hospitals and 85 per cent of all intensive care beds being eligible to participate.
APACHE III was developed to achieve the following goals:
1. to re-evaluate the selection and weighting of physiological variables using objective statistical modeling;
2. to update and expand the size and representativeness of the reference database;
3. to examine the relationship between outcome and patient selection for and timing of intensive care admission;
4. to distinguish clearly the use of predictive estimates for patient groups from mortality estimates for individual patients.
To develop APACHE III, the database was randomly split into two. Using the estimation dataset and multivariable logistic regression analysis, weights were developed to capture the prognostic impact of each physiological variable's deviation from normal. During this process, APACHE II was used to control for disease and chronic health status. Next, the significance of 34 items reflecting chronic health status, comorbidities, and age were evaluated, and weights were estimated using the entire data file. Validity was then tested by comparing predicted hospital survival with that observed in the validation dataset.
APACHE III is similar to APACHE II in principle but is more complex. The APACHE III score consists of points for 17 physiological abnormalities, points for age, and points for chronic health evaluation. Scoring for the 17 physiological variables (the acute physiology score) reflects the extent of abnormality of five vital signs, 11 laboratory tests, and neurological status. Figure! illustrates how GCS variables have been incorporated into APACHE III scoring. In contrast with the GCS, APACHE III scoring for neurological abnormalities is based on the worst rather than the best response during each ICU day. APACHE III scoring for each GCS variable is weighted to reflect the explanatory power of interactions between worst ocular, motor, and verbal responses for hospital survival for the validation half of 17 440 ICU
admissions to 42 ICUs at 40 United States hospitals (Knaus.efa/ 1991). Based on reliability testing, the GCS variables were also reformatted by simplifying the evaluation of eye opening and eliminating the distinctions between incomprehensible words and inappropriate sounds, between flexion withdrawal and decorticate rigidity, and between decerebrate rigidity and no response. As shown in Fig..^ and Fig 3, points for the five vital signs and 11 laboratory tests range from zero to 23.
Fig. 3 APACHE III scoring for acid-base disturbances. (Reproduced with permission from Knauseia/ (1991))
The APACHE III score also includes a score for age and chronic health evaluation (Table.!). Although 34 chronic health items were tested, only seven comorbid conditions had a significant impact on short-term mortality. Each of these seven comorbidities predisposes to infection, the cause of a large number of hospital deaths. We believe that prior cardiac, respiratory, and renal disease were not significant because their prognostic impact is captured by improved physiological measurement. The sum of points for the components of APACHE III yields a numerical score that theoretically ranges from zero to 299; physiological scoring contributes the majority (score 0-252) compared with age (score 0-24) and chronic health evaluation (score 0-23).
□UUV IihRH MhJHUI [CMW»IVH1 CiWtHlllMQ
AlUi J 3
Table 1 APACHE III scoring for age and chronic health evaluation
The APACHE III system has three major uses. First, the APACHE III score can be used to measure severity of disease and to risk-stratify patients within a single diagnostic category or independently defined patient group. This is because an increased score is associated with an increased risk of hospital death. Second, APACHE III scores can be used to compare patient outcomes, but can only be used for intensive care admissions meeting diagnostic and selection criteria similar to that used in the APACHE III study. Third, APACHE III can be used to predict patient outcomes. However, to do this, the score must be related to diagnosis, selection criteria, and a reference database.
Was this article helpful?