Data analysis
Based on the data collected, we performed sensitivity and specificity
analyses to compare the accuracy of the custom alert to the DDI warning.
Sensitivity and specificity are properties of diagnostic tests and are
not predictive of disease in individual patients. Sensitivity is the
ability for a test to correctly determine if a patient has the disease,
or its ability to avoid false negatives. Specificity is the ability for
a test to correctly determine if a patient does not have the disease, or
its ability to avoid false positives.10 Sensitivity of
the custom alert and the DDI warning was calculated by dividing true
positives or the sum of patients with a custom alert or a DDI warning,
by the sum of patients with QTc prolongation. Specificity was calculated
by dividing true negatives or the sum of patients without a custom alert
or a DDI warning, by the sum of patients with no QTc prolongation. The
results will help us determine if the risk score is sensitive enough to
detect the presence of QTc prolongation risk, and/or specific enough to
avoid false positive results.
When evaluating the feasibility or the success of a screening test, we
also considered the positive and negative predictive values. A positive
predictive value (PPV) is the probability that subjects with a positive
screening test truly have the disease, whereas a negative predictive
value (NPV) is the probability that subjects with a negative screening
test truly do not have the disease.11 PPV and NPV were
calculated for both the custom alert and the DDI warning.
Following data collection, sensitivity and specificity calculations, we
performed a multiple logistic regression to analyze the components of
the risk score and develop the most optimal prediction model. We used
the logworth to determine the contribution of the components to the data
model. Components with a p-value of < 0.05 were considered
significant, thus contributing to the data model. We used the
missing-indicator method rather than the imputation method to deal with
missing values.