The APACHE III Prognostic System
1992; Elsevier BV; Volume: 102; Issue: 6 Linguagem: Inglês
10.1016/s0012-3692(16)40910-4
ISSN1931-3543
Autores Tópico(s)Disaster Response and Management
ResumoTo the Editor: We agree. We used a figure from Table 6 of the article by Lemeshow et al, which was for a restricted set of patients. The correct overall classification from the study was 84.9 percent, achieved with the patients from a single hospital as compared with a correct classification of 88.1 percent with APACHE III on a 40-hospital data base. We emphasize, however, that because of the bias inherent in trying to compare correct classification rates across data files that have varying baseline outcome rates, we prefer receiver operating characteristic (ROC) areas. For the specific issue of predicting hospital death rates, ROC areas are even more useful, since the conventional threshold of a 0.50 risk of death used in the above correct classification calculations is arbitrary. The APACHE III system achieved a 0.90 ROC area. We also would like to take this opportunity to correct one error in our APACHE III article. The sex ratio was reversed. The correct distribution of sex across ICU admissions is 55.2 percent male and 44.8 percent female. Sex has no relationship with outcome and is not used in any APACHE outcome predictions. The APACHE III Prognostic SystemCHESTVol. 102Issue 6PreviewTo the Editor: Full-Text PDF
Referência(s)