Reducing latency in medication reconciliation after hospital discharge
Techniques for discovering or predicting that a patient is eligible for a post-hospital discharge medication reconciliation process are disclosed. The prediction is based on automatically applying a machine learning model to one or more incoming prescription orders, where the model has been generated from statistically analyzing a plurality of historical prescriptions to determine prescription attributes and/or values thereof that are more strongly correlated to hospital discharges than are other prescription attributes and/or values. When a prescription order is discovered to be associated with a hospital discharge, the named patient is deemed eligible for medication reconciliation, and an alert is generated so that pharmacy personnel may initiate a post-hospital discharge medication reconciliation process for the patient. As such, patient health outcomes may be improved after being discharged from the hospital, as post-hospital discharge medication reconciliations are able to be completed within 30 days of the patient's discharge using the disclosed techniques.
1 . A computer-implemented method of reducing data latency in initiating medication reconciliation processes for patients after the patients are discharged from hospitals, the computer-implemented method comprising:
training, by one or more processors, a plurality of neural networks using historical data of prescription orders issued in conjunction with hospital discharges of patients and prescription orders that were not issued in conjunction with hospital discharges of patients to discover one or more prescription attributes and/or one or more prescription attribute values that are more strongly correlated to hospital discharges than are other prescription attributes and/or prescription attribute values;
receiving, by the one or more processors, a prescription order that is to be filled for a particular patient;
predicting, by the one or more processors based upon receiving the prescription order, that the prescription order is associated with a hospital discharge of the particular patient, including:
obtaining patient data including information indicative of a hospital discharge of the particular patient,
extracting one or more attributes and/or attribute values indicative of the hospital discharge of the particular patient from one or more of the prescription order or the patient data,
based upon the extracted one or more attributes and/or attribute values, obtaining a trained neural network from the plurality of trained neural networks, each of the plurality of trained neural networks trained using a respective set of attributes and/or attribute values, the trained neural network trained using the respective set of attributes and/or attribute values most correlated to the extracted one or more attributes and/or attribute values,
providing the extracted one or more attributes and/or attribute values as input to the trained neural network,
obtaining, from a corresponding output of the trained neural network, an indication of a likelihood that the received prescription order is correlated to the hospital discharge of the particular patient, and
determining that the likelihood exceeds a threshold; and
responsive to determining that the likelihood exceeds a threshold, generating, by the one or more processors at a user device of one or more of healthcare providers of the particular patient, an alert indicating that the particular patient is eligible for medication reconciliation based on the prediction.
2 . The computer-implemented method of claim 1 , wherein:
the prescription order includes an indication of a medication; and
predicting that the prescription order is associated with the hospital discharge of the particular patient includes including an indication of the medication in the input provided to the trained neural network.
3 . The computer-implemented method of claim 1 , wherein:
the prescription order includes an identification of a prescriber of the prescription order; and
predicting that the prescription order is associated with the hospital discharge of the particular patient includes including an indication of the prescriber in the input provided to the trained neural network.
4 . The computer-implemented method of claim 1 , further comprising:
storing, by the one or more processors, the trained neural network on one or more memories.
5 . The computer-implemented method of claim 4 , wherein:
the trained neural network is generated based on performing one or more statistical analyses on the historical data of prescription orders in conjunction with one or more sets of historical data of a plurality of patients, at least some of which had been discharged from hospitals;
the computer-implemented method further comprises obtaining historical data of the particular patient; and
predicting that the prescription order is associated with the hospital discharge of the particular patient includes including the one or more attributes and/or attribute values of the prescription order and one or more attributes of the historical data of the particular patient in the input provided to the trained neural network.
6 . The computer-implemented method of claim 5 , wherein the one or more sets of historical data of the plurality of patients include data indicative of at least one of: patient demographics, patient prescription fill history, patient pre-existing health conditions, or patient allergies.
7 . The computer-implemented method of claim 1 , wherein:
the one or more attributes and/or attribute values of the prescription order that are provided as the input to the trained neural network are included in one or more prescription attributes and/or one or more prescription attribute values that are more strongly correlated to hospital discharges; and
the one or more attributes and/or attribute values of the prescription order that are included in the input provided to the trained neural network are indicative of one or more of: a medication, a regimen of the medication, a group of medications, a prescriber, or a location of the prescriber.
8 . The computer-implemented method of claim 1 , wherein generating the alert indicating that the particular patient is eligible for medication reconciliation comprises generating the alert (i) at a user interface of a computing system of a pharmacy enterprise, and/or (ii) at a user interface of a personal computing device of the particular patient or of an agent of the particular patient.
9 . The computer-implemented method of claim 1 , further comprising transmitting, based on the prediction that prescription order is associated with the hospital discharge of the particular patient, a request for information associated with the particular patient to at least one of: a health plan provider of the particular patient, or a medication therapy management provider.
10 . The computer-implemented method of claim 9 , further comprising receiving a response to the request for information associated with the particular patient, and completing a medication reconciliation process for the particular patient based on the response.
11 . A system for reducing data latency in initiating medication reconciliation processes for patients after the patients are discharged from hospitals, the system associated with a pharmacy enterprise, and the system comprising:
one or more processors;
a communication interface;
one or more memories; and
computer-executable instructions that are stored on the one or more memories and that, when executed by the one or more processors, cause the system to:
train a plurality of neural networks by using historical data of prescription orders issued in conjunction with hospital discharges of patients and prescription orders that were not issued in conjunction with hospital discharges of patients to discover one or more prescription attributes and/or one or more prescription attribute values that are more strongly correlated to hospital discharges than are other prescription attributes and/or prescription attribute values;
receive a prescription order that is to be filled for a particular patient;
predict based upon receiving the prescription order, that the prescription order is associated with a hospital discharge of the particular patient by:
obtaining patient data including information indicative of a hospital discharge of the particular patient,
extracting one or more attributes and/or attribute values indicative of the hospital discharge of the particular patient from one or more of the prescription order or the patient data,
based upon the extracted one or more attributes and/or attribute values, obtaining a trained neural network from the plurality of trained neural networks, each of the plurality of trained neural networks trained using a respective set of attributes and/or attribute values, the trained neural network trained using the respective set of attributes and/or attribute values most correlated to the extracted one or more attributes and/or attribute values,
providing the extracted one or more attributes and/or one or more attribute values of the prescription order as input to a trained neural network,
obtaining, from a corresponding output of the trained neural network, an indication of a likelihood that the received prescription order is correlated to the hospital discharge of the particular patient, and
determining that the likelihood exceeds a threshold; and
responsive to determining that the likelihood exceeds a threshold, generate, at a user device of one or more of healthcare providers of the particular patient based on the prediction, an alert indicating that the particular patient is eligible for medication reconciliation.
12 . The system of claim 11 , wherein:
the prescription order includes an identification of a medication; and
the input to the trained neural network includes an indication of the medication.
13 . The system of claim 11 , wherein:
the prescription order includes an identification of a prescriber of the prescription order; and
the input to the trained neural network includes an indication of at least one of the prescriber or a location of the prescriber.
14 . The system of claim 11 , wherein the computer-executable instructions are further executable to cause the system to:
store the trained neural network in the one or more memories.
15 . The system of claim 14 , wherein:
the trained neural network is generated based on performing one or more statistical analyses on one or more sets of historical prescription data in conjunction with one or more sets of historical data of a plurality of patients, some of which had been discharged from hospitals;
the computer-executable instructions are further executable to cause the system to obtain historical data of the particular patient; and
the input to the trained neural network further includes one or more attributes of the historical data of the particular patient.
16 . The system of claim 15 , wherein the one or more sets of historical data of the plurality of patients include data indicative of at least one of: patient demographics, patient prescription fill history, patient pre-existing health conditions, or patient allergies.
17 . The system of claim 11 , wherein:
the one or more attributes and/or the one or more attribute values of the prescription order provided as the input to the trained neural network are included in one or more prescription attributes and/or prescription attribute values that are more strongly correlated to hospital discharges; and
the one or more attributes and/or attribute values of the prescription order that are included in the input to the trained neural network are indicative of one or more of: a medication, a regimen of the medication, a group of medications, a prescriber, or a location of the prescriber.
18 . The system of claim 11 , further comprising a user interface, and wherein the alert is generated at (i) the user interface of the system, and/or (ii) at a user interface of a personal computing device of the particular patient or of an agent of the particular patient.
19 . The system of claim 11 , wherein the computer-executable instructions are further executable to cause the system to transmit, via the communication interface to a medication therapy management provider, a request for information associated with the particular patient based on the prediction that the particular patient is eligible for medication reconciliation.
20 . The system of claim 19 , wherein the computer-executable instructions are further executable to cause the system to complete a medication reconciliation process for the particular patient based on a response to the request for information associated with the particular patient.