Systems and methods for automatically predicting incorrect drug dispensed events in a pharmacy
A system comprising one or more processors and one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform functions comprising: receiving a drug profile for a drug identified as part of a high-risk look-alike-sound-alike (LASA) drug pair prior to filling a prescription for the drug; generating an anomaly score based on direction components of the drug profile; and transmitting an alert to a fulfilment screen when the anomaly score exceeds a predetermined threshold. Other embodiments are disclosed.
1 . A system comprising:
one or more processors; and
one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform functions comprising:
receiving a drug profile for a drug identified as part of a high-risk look-alike-sound-alike (LASA) drug pair prior to filling a prescription for the drug;
generating an anomaly score based on direction components of the drug profile; and
transmitting an alert to a fulfilment screen when the anomaly score exceeds a predetermined threshold.
2 . The system of claim 1 , wherein the high-risk LASA drug pair is identified based on at least one of:
an orthographic similarity;
a phonetic similarity; or
a frequency of occurrence of incorrect dispensing in historical events data.
3 . The system of claim 1 , wherein high-risk LASA drug pair is identified based on:
extracting keywords from the direction components of the drug profile of the drug; and
comparing the keywords to an acceptable profile for the drug.
4 . The system of claim 1 , wherein the computing instructions when executed on the one or more processors, further cause the one or more processors to perform a function comprising:
creating a top-k LASA list of high-risk LASA pairs based on weighted LASA scores for drug pairs.
5 . The system of claim 4 , wherein the computing instructions when executed on the one or more processors, further cause the one or more processors to perform a function comprising:
updating the top-k LASA list at periodic intervals using a feedback loop by:
deleting a first drug pair of the drug pairs from the top-k LASA list based on a deletion criteria comprising a percentage of overridden alerts for the first drug pair exceeds a predetermined threshold during a time period of the periodic intervals; and
adding a second drug pair of the drug pairs to the top-k LASA list based at least in part on a LASA score.
6 . The system of claim 1 , wherein generating the anomaly score comprises:
generating, using an isolation forest machine-learning model, the anomaly score based on the direction components of the drug profile.
7 . The system of claim 1 , wherein the anomaly score is based on (i) a univariate score and (ii) a multivariate score.
8 . The system of claim 7 , wherein the univariate score is calculated based on:
comparing a component of the direction components of the drug profile to a preconfigured limit for the component.
9 . The system of claim 7 , wherein the multivariate score is calculated using a conditional probability.
10 . The system of claim 1 , wherein transmitting the alert comprises:
triggering the alert for a patient prescribed a first drug based on at least a drug purchase history of the patient and a patient profile for the patient, wherein triggering the alert occurs when one of:
the first drug is part of the high-risk LASA drug pair and a counterpart drug of the high-risk LASA drug pair is found in the drug purchase history of the patient; or
a non-electronic prescription is received for the first drug with and the patient has no previous drug purchase history.
11 . A method being implemented via execution of computing instructions configured to run on one or more processors and stored at one or more non-transitory computer-readable media, the method comprising:
receiving a drug profile for a drug identified as part of a high-risk look-alike-sound-alike (LASA) drug pair prior to filling a prescription for the drug;
generating an anomaly score based on direction components of the drug profile; and
transmitting an alert to a fulfilment screen when the anomaly score exceeds a predetermined threshold.
12 . The method of claim 11 , wherein the high-risk LASA drug pair is identified based on at least one of:
an orthographic similarity;
a phonetic similarity; or
a frequency of occurrence of incorrect dispensing in historical events data.
13 . The method of claim 11 , wherein high-risk LASA drug pair is identified based on:
extracting keywords from the direction components of the drug profile of the drug; and
comparing the keywords to an acceptable profile for the drug.
14 . The method of claim 11 , further comprising:
creating a top-k LASA list of high-risk LASA pairs based on weighted LASA scores for drug pairs.
15 . The method of claim 14 , further comprising:
updating the top-k LASA list at periodic intervals using a feedback loop by:
deleting a first drug pair of the drug pairs from the top-k LASA list based on a deletion criteria comprising a percentage of overridden alerts for the first drug pair exceeds a predetermined threshold during a time period of the periodic intervals; and
adding a second drug pair of the drug pairs to the top-k LASA list based at least in part on a LASA score.
16 . The method of claim 11 , wherein generating the anomaly score comprises:
generating, using an isolation forest machine-learning model, the anomaly score based on the direction components of the drug profile.
17 . The method of claim 11 , wherein the anomaly score is based on (i) a univariate score and (ii) a multivariate score.
18 . The method of claim 17 , wherein the univariate score is calculated based on:
comparing a component of the direction components of the drug profile to a preconfigured limit for the component.
19 . The method of claim 17 , wherein the multivariate score is calculated using a conditional probability.
20 . The method of claim 11 , wherein transmitting the alert comprises:
triggering the alert for a patient prescribed a first drug based on at least a drug purchase history of the patient and a patient profile for the patient, wherein triggering the alert occurs when one of:
the first drug is part of the high-risk LASA drug pair and a counterpart drug of the high-risk LASA drug pair is found in the drug purchase history of the patient; or
a non-electronic prescription is received for the first drug with and the patient has no previous drug purchase history.