The effect of intensive care on in-hospital survival

Intensive Care Units (ICU) are one of the most powerful and expensive technologies within inpatient care. However, its effect on survival is still an issue under discussion. The objective of this paper is to assess the effect of General ICU on in-hospital survival. We assessed the effect of ICU on survival using Linear and Probit regressions. Since admission to IC is not random and depends on unobserved (to the researcher) heterogeneity, we reassessed the IC effect by Instrumental Variables (IV) and Bivariate Probit techniques, using crowding in the IC unit as an instrument. The results show that a simple Probit of the IC effect on survival is 7-10 percentage-points (pts). The IV estimate of the IC effect on survival is 21-34 pts, and the Bivariate Probit estimate is 17-21 pts. We conclude that although admitted patients are at lower risk of death, as determined by their observable (to the researcher) characteristics, controlling for observable differences, those with a higher unobserved risk of mortality are more likely to be admitted. The implications for an optimal admission policy are discussed.