Evaluation of APACHE II and the ICU cancer mortality model in an Indian cancer hospital ICU

Introduction

The performance of a prognostic score must be validated before it may be used in an ICU, especially where there is a specific patient case mix.

Objective

To validate a cancer-specific ICU scoring system – the ICU Cancer Mortality Model (ICMM) [1] – and to compare its performance with that of a general ICU scoring system – APACHE II.

Methods

Six hundred and two consecutive admissions (369 males, 233 females, mean age 50.3 ± 14.1 years) from July 2004 to September 2005 were prospectively studied in a 21-bed ICU-HDU. Two hundred and sixty-three patients were admitted following surgery (Group 1), 209 with solid tumours were admitted from the wards (Group 2) and 130 had haematological malignancies, lymphoma or myeloma (Group 3). In case of multiple ICU admissions during a single hospital admission, data from the last ICU admission were recorded. Discrimination was determined by computing the area under the receiver-operating characteristic curve (AUC, represented as area ± SE). Calibration was calculated using the Hosmer-Lemeshow goodness-of-fit test.

Results

The average APACHE II score was 15.3 ± 9.8, and ICU length of stay (LOS) 5.2 ± 6.6 days. Predicted mortality in hospital was 24.8% by APACHE II and 50.3% by ICMM. Two hundred and eleven patients died in hospital (observed mortality 35.0%), yielding a standardized mortality ratio (SMR) of 1.4 and 0.7, respectively. Both APACHE II and ICMM discriminated well (AUC 0.90 ± 0.013 vs 0.87 ± 0.016, respectively). Both scoring systems calibrated poorly (APACHE II, ^H 59.0, P < 0.001, df = 10 and ICMM, ^H 97.6, P < 0.001, df = 10). The mean APACHE II score in Group 1 (8.6 ± 5.8) was significantly lower (P < 0.001) than in Groups 2 (17.9 ± 8.5) and 3 (24.9 ± 7.9). Mortality was 11%, 39% and 77% in Groups 1, 2 and 3, respectively. The SMR for Group 1 was 1.19 for APACHE II and 0.32 for ICMM; for Group 2, 1.36 and 0.72, respectively; and for Group 3, 1.54 and 1.02. APACHE II showed better discrimination than ICMM in Group 1 (AUC 0.86 ± 0.04 vs 0.79 ± 0.04) and calibrated well for Group 1 (^H = 14.9, df = 10, P = 0.14).

Conclusion

Both APACHE II and the cancer-specific ICMM discriminated well but were poorly calibrated, and could not be validated for use in our patients. APACHE II worked well for a subgroup consisting of surgical patients with a lower severity of illness, while the ICMM tended to better for patients with hematological malignancy with a high severity of illness and mortality.