Hospital deaths and adverse events in Brazil

Table 3 Discriminatory capacity and adjustment of models for predicting death

Model	χ² of Model*	Hosmer-Lemeshow χ² test	C Statistic (95% CI)
1. Baseline model**:	137.236 (df 9, sig. 0.000)	13.898 (sig. 0.053)	0.85 (0.82-0.89)
2. Baseline model + Length of stay***	148.015 (df 10, sig. 0.000)	13.638 (sig. 0.092)	0.85 (0.82-0.89)
3. Final model: Model 2 + Occurrence of adverse event ****	187.316 (df 12, sig. 0.000)	11.271 (sig. 0.187)	0.88 (0.85-0.91)

* Initial likelihood χ² with intercept: 484.387 (Wald).
** Baseline model includes: gender, age, Charlson index applied to primary and secondary diagnoses, and type of admission.
*** Surgical or invasive procedure and clinical specialty variables were also tested, but were excluded due to lack of statistical significance.
**** Occurrence of adverse event = no, yes non-preventable, yes preventable.

ISSN: 1472-6963