Efh. Vanbommel et al., USE OF APACHE-II CLASSIFICATION TO EVALUATE OUTCOME AND RESPONSE TO THERAPY IN ACUTE-RENAL-FAILURE PATIENTS IN A SURGICAL INTENSIVE-CARE UNIT, Renal failure, 17(6), 1995, pp. 731-742
This study objective was to determine the applicability of the acute p
hysiology and chronic health evaluation (APACHE) II score in surgical
patients with acute renal failure (ARF) requiring dialytic support, an
d to assess its utility in evaluating data from this specific disease
group. This was a retrospective, partly prospective follow-up study of
patients who developed ARF during their course of stay on the surgica
l intensive care unit (ICU) of a Dutch university hospital from Januar
y 1, 1986 to January 31, 1994. A total of 111 patients were identified
, of whom 104 patients were considered eligible for this study. Data f
or the individual APACHE II scores were calculated from the most deran
ged values during the initial 24 h of ICU admission (APACHE II1) and o
n the day dialytic support was instituted (APACHE II2). the ratio betw
een the two APACHE II scores was also calculated for each patient (AP2
/AP1 ratio). Receiver operating characteristic curves (ROC) were const
ructed. Other variables evaluated included age, sex, serum creatinine,
diagnostic category, time from ICU admission to start of dialytic sup
port, and the type of dialytic support. Of these 104 patients (median
age 64; range 23-85 years), 51 (50%) survived to leave the ICU, of who
m 47 (46%) survived to leave hospital. The APACHE II2 score (27.0 +/-
4.4 vs. 22.4 +/- 3.5; p < 0.001) and AP(2)/AP(1) ratio (1.12 +/- 0.09
vs. 0.97 +/- 0.06; p < 0.001) were significantly higher for nonsurvivo
rs as compared to survivors. The ROC curve was most discriminative for
the AP(2)/AP(1) ratio (area under the curve 0.92) and to a lesser ext
ent for the APACHE II2 score (area under the curve 0.78). Estimated ri
sk of death with the APACHE II equation did not improve predictive pow
er Multivariate analysis of various variables revealed the AP(2)/AP(1)
ratio as the single most important factor predicting death (odds rati
o 13.8, p < 0.001). Adjusting for the AP(2)/AP(1) ratio, no impact on
outcome was observed for age, diagnostic category time from ICU admiss
ion to start of dialytic support, and the type of dialytic support. Ab
ove a value of 1.0 of the AP(2)/AP(1) ratio, logistic regression revea
led a sharp increase in death probability with increasing AP(2)/AP(1)
ratio. APACHE II, when used at the time of initiation of dialytic supp
ort, proved to be a valid way in our surgical ICU to stratify ARF pati
ents by the severity of their illness. Moreover; use of the AP(2)/AP(1
) ratio further improved the usefulness of this severity index and may
help to identify patients who have little chance of survival. Predict
ing death with the APACHE II equation did not improve predictive power
.