Background Postprocedure length of stay (LOS) remains an important determin
ant of medical costs after coronary stenting. Variables that predict LOS in
this setting have not been well characterized.
Methods We evaluated 359 consecutive patients who underwent coronary stenti
ng with antiplatelet therapy. Sequential multiple linear regression (MLR) m
odels were constructed with use of 4 types of variables to predict log-tran
sformed LOS: preprocedure, intraprocedure, and postprocedure factors and ad
verse outcomes.
Results Preprocedure factors alone explained more than one third of the var
iability in postprocedure LOS (adjusted R-2 = 0.37). The addition of proced
ural variables added little to the model (adjusted R-2 = 0.39). Entering no
noutcome postprocedure variables significantly enhanced the predictive capa
city of the model, explaining more than half the variability in postprocedu
re LOS (adjusted R-2 = 0.54). In the final model, addition of outcome varia
bles increased its predictive capacity only slightly (adjusted R-2 = 0.61).
In this model, significant preprocedure factors included: myocardial infar
ction (MI) within 24 hours, MI within 1 to 30 days, women with peripheral v
ascular disease, intravenous heparin, and chronic atrial fibrillation. High
-risk intervention was the only significant intraprocedure variable. Signif
icant postprocedure factors included periprocedure ischemia; cerebrovascula
r accident or transient ischemic attack; treatment with intravenous heparin
or nitroglycerin or intra-aortic balloon pump; and need for blood transfus
ion. Significant adverse outcomes included contrast, nephropathy, gastroint
estinal bleeding, arrhythmia, vascular complication, and repeat angiography
.
Conclusion This prediction model identifies a number of potentially reversi
ble factors responsible for prolonging LOS. and may enable the development
of more accurate risk-adjusted methods with which to improve or compare car
e.