Statistical details :
The ability of ordinal variables (E/A, S/D, E/Vp) to predict PCWP was
tested for previously determined cutoffs as per extant guidelines. Based
on diastolic function guidelines4, sensitivity and
specificity were calculated for prefixed cutoff values
i.e.,(E/A<1, E/A1-2, E/A>2, S/D<1 &
E/Vp>2.5). For continuous data, optimal cut off values were
determined by constructing ROC curves i.e., for echocardiographic
variables ( SF, E/Ea, DT).
As per new revision of diastolic function guidelines
201616, evaluation of diastolic function should
proceed on different lines in patients with normal versus depressed left
ventricular systolic function because of high prevalance of diastolic
dysfunction in the latter. Accordingly, when the pretest probability of
diastolic dysfunction is much higher in patients with systolic
dysfunction, any attempt to distinguish presence or absence of diastolic
dysfunction should stress on specificity.
For patients with normal systolic function, cut off values using ROC
curve analysis were calculated by focusing on sensitivity. So, for
variables (SF, E/Ea, DT), cut off values were calculated based on both
maximizing sensitivity and specificity. In all the cases, validity
parameters such as sensitivity, specificity, predictive value of
positives, predictive value of negatives & accuracy were computed.