Signal detection algorithms to identify drug effects and drug interactions
An algorithm according to an embodiment of the present invention provides for latent signal detection of adverse events. Embodiments infer the presence of adverse drug events from large observational databases housed by the FDA, WHO, and other governmental organizations. The disclosed algorithms do not require the adverse event to be reported explicitly. Instead, the algorithms infer the presence of adverse events through more common secondary effects. In an embodiment, machine learning techniques are used for this purpose.
1. A computer-implemented method for detection of latent signals in adverse event information, comprising:
receiving a set of drug and event information that includes a first set of adverse event information and further includes prescription and morbidity information;
identifying a second set of events associated with the first set of adverse events;
computing covariances in drug co-prescription from the set of drug and event information;
computing covariances in co-morbities from the set of drug and event information;
approximating adverse event biases based on the covariances in drug co-prescription and comorbidities;
applying a statistical analysis to the second set of events to determine a subset of the second set of events that occurs above a predetermined level with the first set of adverse events, wherein the statistical analysis is corrected based on the approximated adverse event biases;
receiving a training dataset that includes drug and event information;
training a predictive model using the subset of the second set of events and the training dataset, wherein the predictive model is trained to detect a detected set of adverse events; and
applying the predictive model to a test dataset to determine the detected set of adverse events.
2. The computer-implemented method of claim 1 , wherein the set of drug and event information is received from the Food and Drug Administration.
3. The computer-implemented method of claim 1 , wherein drug labels are used to identify the second set of events.
4. The computer-implemented method of claim 1 , wherein training the predictive model is performed using machine learning techniques.
5. The computer-implemented method of claim 1 , wherein the training dataset and test dataset are independent.
6. The computer-implemented method of claim 1 , further comprising drugs or drug-drug pairs by a likelihood to be associated with the first adverse event.
7. The computer-implemented method of claim 6 , further comprising removing drugs or drug-drug pairs with predetermined mechanisms.
8. A non-transitory computer-readable medium including instructions that, when executed by a processing unit, cause the processing unit to detect latent signals in adverse event information, by performing the steps of:
receiving a set of drug and event information that includes a first set of adverse event information and further includes prescription and morbidity information;
identifying a second set of events associated with the first set of adverse events;
computing covariances in drug co-prescription from the set of drug and event information;
computing covariances in co-morbities from the set of drug and event information;
approximating adverse event biases based on the covariances in drug co-prescription and comorbidities;
applying a statistical analysis to the second set of events to determine a subset of the second set of events that occurs above a predetermined level with the first set of adverse events, wherein the statistical analysis is corrected based on the approximated adverse event biases;
receiving a training dataset that includes drug and event information;
training a predictive model using the subset of the second set of events and the training dataset, wherein the predictive model is trained to detect a detected set of adverse events; and
applying the predictive model to a test dataset to determine hidden the detected set of adverse events.
9. The non-transitory computer-readable medium of claim 8 , wherein the set of drug and event information is received from the Food and Drug Administration.
10. The non-transitory computer-readable medium of claim 8 , wherein drug labels are used to identify the second set of events.
11. The non-transitory computer-readable medium of claim 8 , wherein training the predictive model is performed using machine learning techniques.
12. The non-transitory computer-readable medium of claim 8 , wherein the training dataset and test dataset are independent.
13. The non-transitory computer-readable medium of claim 8 , further comprising drugs or drug-drug pairs by a likelihood to be associated with the first adverse event.
14. The non-transitory computer-readable medium of claim 13 , further comprising removing drugs or drug-drug pairs with predetermined mechanisms.
15. A computing device comprising:
a data bus;
a memory unit coupled to the data bus;
a processing unit coupled to the data bus and configured to
receive a set of drug and event information that includes a first set of adverse event information and further includes prescription and morbidity information;
identify a second set of events associated with the first set of adverse events;
compute covariances in drug co-prescription from the set of drug and event information;
compute covariances in co-morbities from the set of drug and event information;
approximate adverse event biases based on the covariances in drug co-prescription and comorbidities;
apply a statistical analysis to the second set of events to determine a subset of the second set of events that occurs above a predetermined level with the first set of adverse events, wherein the statistical analysis is corrected based on the approximated adverse event biases;
receive a training dataset that includes drug and event information;
train a predictive model using the subset of the second set of events and the training dataset, wherein the predictive model is trained to detect a detected set of adverse events; and
apply the predictive model to a test dataset to determine the detected set of adverse events.