From 954 attempts to resuscitate patients from out-of-hospital cardiac
arrest two datasets were derived, namely 861 cases of cardiac arrest
and 906 cases of either cardiac or primary respiratory arrest. For eac
h dataset, multivariate analysis was performed by fitting a number of
explanatory variables with respect to the outcomes of admission to hos
pital and discharge home in logistic regression models. There were num
erous interactions between these variables. Being conscious at the tim
e of the arrival of the ambulance crew and subsequently having cardiac
arrest strongly predicted survival, as did both the presence of a wit
ness to the arrest and the initiation of cardiopulmonary resuscitation
(CPR) by a bystander; this latter effect was a marker for early CPR.
The strongest predictor of a poor outcome was delay to CPR or delay to
advanced cardiac life support. (C) 1997 Elsevier Science IreIand Ltd.