Table 2.

Disorder of water balance over the course of the ICU stay and ICU mortality

Variate	Hazard ratio	95% confidence interval	P value	Bayes information criteriona
Crude analysis
Hypernatremia	3.34	1.55 to 6.88	0.002	313.29
DDAVP use	8.23	3.24 to 19.52	<0.001	304.75
Adjusted for baseline risk of death
Hypernatremia	3.00	1.34 to 6.51	0.003	291.17
DDAVP use	5.48	2.13 to 13.21	<0.001	286.07
Adjusted for baseline risk of death and for each other
Hypernatremia	2.04	0.81 to 4.84	0.092	288.09
DDAVP use	3.88	1.40 to 10.33	0.005
Hypernatremia adjusted for baseline risk and stratified according to DDAVP use
Hypernatremia with DDAVP use	0.58	0.07 to 3.67	0.57	256.53
Hypernatremia without DDAVP use	4.20	1.62 to 10.17	0.004

Hazard ratio, 95% confidence intervals, and P values associated with hypernatremia and desmopressin acetate (DDAVP) use in the intensive care unit (ICU), estimated by six different Cox proportional-hazards regression models. Hazard ratios associated with hypernatremia and DDAVP use were first estimated in separate models, without adjusting for confounding factors (Crude analysis), and after adjusting for baseline risk (Adjusted for baseline risk of death). They were then estimated including hypernatremia and DDAVP use in the same model in order to isolate the effect of each variate independently of the other (Adjusted for baseline risk of death and for each other). Finally, the hazard ratio associated with hypernatremia was estimated stratifying the Cox regression model for DDAVP use (Hypernatremia adjusted for baseline risk and stratified according to DDAVP use). Baseline risk is represented by the score from the International Mission for Prognosis and Analysis CT prognostic model [11]. ^aA measure of both model fit and parsimony; the better the model, the smaller the associated BIC value.