Methods and apparatus to estimate costs of prescriptions
Example methods, apparatus, and articles of manufacture to estimate costs of prescriptions are disclosed herein. An example computer-implemented method, executed by a processor, to estimate a cost of a prescription for a medication includes receiving, with a processor, a prescription for a medication for a patient, receiving, with the processor from an entity, a request for an estimated cost for the prescription, processing, with a machine learning model, the prescription to determine the estimated cost of the medication for the patient, and providing, with the processor, the estimated cost to the entity.
1 . A computer-implemented method, executed by one or more processors, to estimate a cost of a prescription for a medication, the method comprising:
receiving a prescription for a medication for a patient;
receiving a request for an estimated cost for the prescription for the patient from an entity;
forming an input vector including prescription information for the received prescription, and payor information for the patient;
processing the input vector for the received prescription, with a trained machine learning model, to determine the estimated cost of the medication for the patient, wherein the machine learning model is trained until it converges with data representing a plurality of sold prescriptions by updating the machine learning model based upon computed differences between (i) predicted costs determined by the machine learning model based on prescription information for the plurality of sold prescriptions and payor information, and (ii) actual sold costs for the plurality of sold prescriptions;
providing the estimated cost to the entity; and
deploying the updated machine learning model.
2 . The method of claim 1 , wherein the entity is a person.
3 . The method of claim 2 , wherein the prescription is received from the patient.
4 . The method of claim 1 , wherein the entity is a prescriber for the prescription, or a representative of the prescriber.
5 . The method of claim 4 , further comprising providing an alternative medication and estimated cost for the alternative medication to the entity.
6 . The method of claim 4 , wherein the prescription and the request are received, and the estimated cost is provided while the patient is in a vicinity of the prescriber, or the representative of the prescriber.
7 . The method of claim 6 , further comprising when, in response to the providing of the estimated cost, an indication to fulfill the prescription, or an alternative medication for the patient is received, cause the prescription or the alternative medication to be filled for delivery to the patient.
8 . The method of claim 1 , wherein the input vector further includes one or more of demographic information, socioeconomic information, or geographic information.
9 . The method of claim 1 , wherein the machine learning model is updated by applying gradient boosting.
10 . The method of claim 1 , wherein the machine learning model is validated with second data representing a second plurality of sold prescriptions, the second data including for each sold prescription: (i) prescription information, (ii) payor information, and (iii) an actual sold cost.
11 . The method of claim 1 , further comprising providing to the entity information representing a likelihood that the medication will be picked up by the patient.
12 . The method of claim 1 , wherein the machine learning model is converged when a metric computed over one or more of the differences satisfies a criteria.
13 . A non-transitory computer-readable storage medium storing computer-readable instructions that, when executed by one or more processors, cause a machine to:
receive a prescription for a medication for a patient;
receive, from an entity, a request for an estimated cost for the prescription for the patient;
form an input vector including prescription information for the received prescription, and payor information for the patient;
process the input vector with a trained machine learning model to determine the estimated cost of the medication for the patient, wherein the machine learning model is trained until it converges with data representing a plurality of sold prescriptions by updating the machine learning model based upon computed differences between (i) predicted costs determined by the machine learning model based on prescription information for the plurality of sold prescriptions and payor information, and (ii) actual sold costs for the plurality of sold prescriptions;
provide the estimated cost to the entity; and
deploy the updated machine learning model.
14 . The non-transitory computer-readable storage medium of claim 13 , wherein the entity is a person.
15 . The non-transitory computer-readable storage medium of claim 13 , wherein the entity is a prescriber for the prescription, or a representative of the prescriber, and wherein the instructions, when executed by the one or more processors, cause the machine to:
provide an alternative medication and estimated cost for the alternative medication to the entity;
provide information representing a likelihood that the medication will be picked up by the patient; and
cause the prescription to be filled when an indication to fulfill the prescription is received in response to at least one of the estimated cost, the alternative medication, of the likelihood.
16 . The non-transitory computer-readable storage medium of claim 13 , wherein the input vector further includes one or more of demographic information, socioeconomic information, or geographic information.
17 . The non-transitory computer-readable storage medium of claim 13 , wherein the machine learning model is updated using gradient boosting.
18 . The non-transitory computer-readable storage medium of claim 13 , wherein the machine learning model is validated with second data representing a second plurality of sold prescriptions, the second data including for each sold prescription: (i) prescription information, (ii) payor information, and (iii) an actual sold cost.
19 . The non-transitory computer-readable storage medium of claim 13 , wherein the machine learning model is converged when a metric computed over one or more of the differences satisfies a criteria.
20 . A computer-implemented method, executed by one or more processors, to train a machine learning model for estimating costs of prescriptions, the method comprising:
obtaining data representing a plurality of sold prescriptions, the data including for each sold prescription: (i) prescription information, (ii) payor information, and (iii) an actual sold cost;
for each sold prescription of the plurality of sold prescriptions until the machine learning model converges:
forming an input vector for the sold prescription that includes the prescription information and the payor information for the sold prescription,
processing the input vector for the sold prescription with the machine learning model to generate an estimated cost for the sold prescription,
determining a difference between the estimated cost for the sold prescription and the actual sold cost for the sold prescription, and
updating the machine learning model based on the difference; and
when the machine learning model converges, deploying, on one or more processors, the updated trained machine learning model for use in estimating costs of prescriptions for medications.
21 . The method of claim 20 , further comprising:
obtaining second data representing a second plurality of sold prescriptions, the second data including for each sold prescription of the second plurality of sold prescriptions: (i) prescription information, (ii) payor information, and (iii) an actual sold cost;
for each sold prescription of the second plurality of sold prescriptions:
forming an input vector for the sold prescription that includes the prescription information and the payor information for the sold prescription,
processing the input vector with the trained machine learning model to generate an estimated cost for the sold prescription, and
determining a difference between the estimated cost for the sold prescription and the actual sold cost for the sold prescription;
computing a statistical metric of the differences for the second plurality of sold prescriptions; and
determining whether the training of the machine learning model has converged based on the statistical metric.
22 . The method of claim 20 , wherein updating the machine learning model includes applying gradient boosting.
23 . The method of claim 20 ,
wherein the prescription information includes one or more of a medical diagnosis, drug usage information, a drug name, a drug strength, a drug amount, a drug type, a drug class, a drug subclass, or a drug form, and
wherein the payor information includes one or more of a payor identifier, a plan group identifier, a subscriber identifier, fees information, or benefits information.
24 . The method of claim 20 , further comprising:
receiving a prescription for a medication for a patient;
receiving a request for an estimated cost for the prescription for the patient;
forming an input vector including prescription information for the received prescription, and payor information for the patient;
processing the input vector for the received prescription, with the trained machine learning model, to determine the estimated cost of the medication for the patient;
obtaining an actual cost of the medication for the patient; and
updating the machine learning model based upon a difference between the estimated cost of the medication for the patient and the actual cost of the medication for the patient.