Augmenting healthcare stewardship using machine learning

1 . A computer-implemented method comprising:

receiving a set of clinical care dataset, historical care dataset, and regional care pathway dataset associated with a plurality of patients in a healthcare stewardship program;

training a machine learning model using the set of the clinical care dataset, the historical care dataset, and the regional care pathway dataset, the set of the clinical care dataset, the historical care dataset, and the regional care pathway dataset including known care events of interest, patient criticality levels, and patient care treatments to train the machine learning model, the machine learning model including a neural network that learns a corresponding weight to assign to each one of clinical care data, historical care data, and regional care pathway data of different patients for generating a prediction of care events of interest, patient criticality levels, and patient care treatments as outputs;

receiving an input of clinical care data, historical care data, and regional care pathway data associated with a patient from the plurality of patients;

determining, using the machine learning model and the input of clinical care data, historical care data, and regional care pathway data associated with the patient, a first output of a patient criticality level, a second output of a care event of interest associated with the patient criticality level, and a third output of a recommendation of a treatment associated with the patient criticality level and the care event of interest;

receiving, via a secure portal provided by a machine learning module, feedback correcting one or more of the first output of the patient criticality level, the second output of the care event of interest, and the third output of the recommendation of the treatment;

updating, by the machine learning module, identification hints for one or more of the known care events of interest, patient criticality levels, and patient care treatments in the set of clinical care dataset, the historical care dataset, and the regional care pathway dataset based on the feedback;

retraining, by the machine learning module, the machine learning model using the updated identification hints for one or more of the known care events of interest, patient criticality levels, and patient care treatments in the set of clinical care dataset, the historical care dataset, and the regional care pathway dataset;

determining, using the retrained machine learning model and the input of clinical care data, historical care data, and regional care pathway data associated with the patient, a fourth output of a patient criticality level, a fifth output of a care event of interest associated with the patient criticality level, and a sixth output of a recommendation of a treatment associated with the patient criticality level and the care event of interest, the retrained machine learning model assigning a corresponding weight to each one of the input of the clinical care data, the historical care data, and the regional care pathway data associated with the patient in determining the third output of the patient criticality level, the fourth output of the care event of interest, and the sixth output of the recommendation of the treatment, the recommendation of the treatment including timely transitioning of medication from a broad-spectrum to a narrow-spectrum, from a first dosage to a second dosage, and from a first route of administering the medication to a second route of administering the medication;

automatically surfacing the recommendation of the treatment for review by a healthcare personnel within a workflow of the healthcare stewardship program based on the fourth output, the fifth output, and the sixth output of the retrained machine learning model; and

administering the treatment including timely transitioning of the medication from the broad-spectrum to the narrow-spectrum, from the first dosage to the second dosage, and from the first route of administering the medication to the patient to the second route of administering the medication to the patient responsive to the review by the healthcare personnel.

2 . The computer-implemented method of claim 1 , further comprising:

filtering the clinical care data, the historical care data, and the regional care pathway data associated with the patient;

generating a clinical dashboard based on the filtering, the first output of the patient criticality level, the second output of the care event of interest, and the third output of the recommendation of the treatment; and

presenting the clinical dashboard to the healthcare personnel within the workflow of the healthcare stewardship program.

3 . The computer-implemented method of claim 2 , further comprising:

generating an alert notification of the care event of interest; and

automatically surfacing, via the clinical dashboard, the alert notification of the care event of interest for review by the healthcare personnel.

4 . The computer-implemented method of claim 2 , further comprising:

generating a listing of the plurality of patients in the clinical dashboard;

sorting the patient in the listing of the plurality of patients based on the patient criticality level; and

associating a graphical indicator with the patient in the listing of the plurality of patients, the graphical indicator indicating a status associated with a review of the patient by the healthcare personnel within the workflow of the healthcare stewardship program.

5 . The computer-implemented method of claim 2 , further comprising:

receiving, via the clinical dashboard, a feedback from the healthcare personnel on the first output of the patient criticality level, the second output of the care event of interest, and the third output of the recommendation of the treatment;

updating the plurality of the clinical care dataset, the historical care dataset, and the regional care pathway dataset based on the feedback; and

retraining the machine learning model using the updated plurality of the clinical care dataset, the historical care dataset, and the regional care pathway dataset.

6 . The computer-implemented method of claim 5 , wherein the feedback includes at least one from a group of acceptance, rejection, and correction.

7 . The computer-implemented method of claim 2 , wherein the clinical dashboard includes healthcare personnel notes, patient vitals trend, patient medication timeline, patient laboratory results, and an activity log associated with the treatment of the patient within the workflow of the healthcare stewardship program.

8 . The computer-implemented method of claim 1 , wherein the patient criticality level is one from a group of low, medium, and high.

9 . The computer-implemented method of claim 1 , wherein the treatment includes a therapeutic procedure, a surgical procedure, a non-surgical procedure, a laboratory test, a medical test, an imaging test, a medication prescription, and a follow-up care.

10 . The computer-implemented method of claim 1 , wherein the machine learning model includes a convolutional neural network assigning a weight to each of the clinical care data, the historical care data, and the regional care pathway data associated with the patient.

11 . The computer-implemented method of claim 1 , wherein the healthcare stewardship program is an antibiotics stewardship program.

12 . A system comprising:

one or more processors; and

a memory, the memory storing instructions, which when executed cause the one or more processors to:

receive a set of clinical care dataset, historical care dataset, and regional care pathway dataset associated with a plurality of patients in a healthcare stewardship program;

train a machine learning model using the set of the clinical care dataset, the historical care dataset, and the regional care pathway dataset, the set of the clinical care dataset, the historical care dataset, and the regional care pathway dataset including known care events of interest, patient criticality levels, and patient care treatments to train the machine learning model, the machine learning model including a neural network that learns a corresponding weight to assign to each one of clinical care data, historical care data, and regional care pathway data of different patients for generating a prediction of care events of interest, patient criticality levels, and patient care treatments as outputs;

receive an input of clinical care data, historical care data, and regional care pathway data associated with a patient from the plurality of patients;

determine, using the machine learning model and the input of clinical care data, historical care data, and regional care pathway data associated with the patient, a first output of a patient criticality level, a second output of a care event of interest associated with the patient criticality level, and a third output of a recommendation of a treatment associated with the patient criticality level and the care event of interest;

receiving, via a secure portal provided by a machine learning module, feedback correcting one or more of the first output of the patient criticality level, the second output of the care event of interest, and the third output of the recommendation of the treatment;

updating, by the machine learning module, identification hints for one or more of the known care events of interest, patient criticality levels, and patient care treatments in the set of clinical care dataset, the historical care dataset, and the regional care pathway dataset based on the feedback;

retraining, by the machine learning module, the machine learning model using the updated identification hints for one or more of the known care events of interest, patient criticality levels, and patient care treatments in the set of clinical care dataset, the historical care dataset, and the regional care pathway dataset;

determining, using the retrained machine learning model and the input of clinical care data, historical care data, and regional care pathway data associated with the patient, a fourth output of a patient criticality level, a fifth output of a care event of interest associated with the patient criticality level, and a sixth output of a recommendation of a treatment associated with the patient criticality level and the care event of interest, the retrained machine learning model assigning a corresponding weight to each one of the input of the clinical care data, the historical care data, and the regional care pathway data associated with the patient in determining the third output of the patient criticality level, the fourth output of the care event of interest, and the sixth output of the recommendation of the treatment, the recommendation of the treatment including timely transitioning of medication from a broad-spectrum to a narrow-spectrum, from a first dosage to a second dosage, and from a first route of administering the medication to a second route of administering the medication;

automatically surface the recommendation of the treatment for review by a healthcare personnel within a workflow of the healthcare stewardship program based on the fourth output, the fifth output, and the sixth output of the retrained machine learning model; and

administer the treatment including timely transitioning of the medication from the broad-spectrum to the narrow-spectrum, from the first dosage to the second dosage, and from the first route of administering the medication to the patient to the second route of administering the medication to the patient responsive to the review by the healthcare personnel.

13 . The system of claim 12 , wherein the instructions further cause the one or more processors to:

filter the clinical care data, the historical care data, and the regional care pathway data associated with the patient;

generate a clinical dashboard based on the filtering, the first output of the patient criticality level, the second output of the care event of interest, and the third output of the recommendation of the treatment; and

present the clinical dashboard to the healthcare personnel within the workflow of the healthcare stewardship program.

14 . The system of claim 13 , wherein the instructions further cause the one or more processors to:

generate an alert notification of the care event of interest; and

automatically surface, via the clinical dashboard, the alert notification of the care event of interest for review by the healthcare personnel.

15 . The system of claim 13 , wherein the instructions further cause the one or more processors to:

generate a listing of the plurality of patients in the clinical dashboard;

sort the patient in the listing of the plurality of patients based on the patient criticality level; and

associate a graphical indicator with the patient in the listing of the plurality of patients, the graphical indicator indicating a status associated with a review of the patient by the healthcare personnel within the workflow of the healthcare stewardship program.

16 . The system of claim 13 , wherein the instructions further cause the one or more processors to:

receive, via the clinical dashboard, a feedback from the healthcare personnel on the first output of the patient criticality level, the second output of the care event of interest, and the third output of the recommendation of the treatment;

update the plurality of the clinical care dataset, the historical care dataset, and the regional care pathway dataset based on the feedback; and

retrain the machine learning model using the updated plurality of the clinical care dataset, the historical care dataset, and the regional care pathway dataset.

17 . The system of claim 16 , wherein the feedback includes at least one from a group of acceptance, rejection, and correction.

18 . The system of claim 13 , wherein the clinical dashboard includes healthcare personnel notes, patient vitals trend, patient medication timeline, patient laboratory results, and an activity log associated with the treatment of the patient within the workflow of the healthcare stewardship program.

19 . The system of claim 12 , wherein the patient criticality level is one from a group of low, medium, and high.

20 . The system of claim 12 , wherein the treatment includes a therapeutic procedure, a surgical procedure, a non-surgical procedure, a laboratory test, a medical test, an imaging test, a medication prescription, and a follow-up care.