Intelligent generation of personalized CQL artifacts
Described is a system for receiving a disease specific treatment algorithm corresponding to a disease of a patient from a first Electronic Health Record (EHR) system, processing data corresponding to the disease specific treatment algorithm using a Large Language Model (LLM), receiving one or more CQL models from the LLM based on the processing of the data corresponding to the disease specific treatment algorithm, receiving an update to a patient record of the patient from the first EHR system, based on the update to the patient record and the one or more CQL models, triggering a Clinical Decision Support (CDS) hook to generate an alert, and causing transmission of the alert for a medical practitioner associated with the first EHR system.
1. A system comprising:
at least one processor; and
at least one memory component storing instructions that, when executed by the at least one processor, cause the at least one processor to perform operations comprising:
receiving a disease specific treatment algorithm corresponding to a disease of a patient from a first Electronic Health Record (EHR) system, the disease specific treatment algorithm including a decision tree that includes guidelines for treating a specific disease;
processing data corresponding to the disease specific treatment algorithm by inputting the data into a Large Language Model (LLM), the LLM being trained to process disease specific treatment algorithms to generate Clinical Quality Language (CQL) models compatible for Fast Healthcare Interoperability Resources (FHIR) and configured to trigger Clinical Decision Support (CDS) hooks;
receiving one or more first COL models from the LLM based on the processing of the data corresponding to the disease specific treatment algorithm, the one or more first COL models include at least a first CDS hook;
receiving an update to a patient record of the patient from the first EHR system;
generating an updated disease specific treatment algorithm based on the update;
inputting the updated disease specific treatment algorithms to the LLM to receive one or more second CQL models from the LLM;
executing the one or more second CQL models triggering at least a second CDS hook to generate an alert; and
causing transmission of the alert for a medical practitioner associated with the first EHR system.
2. The system of claim 1 , wherein the disease specific treatment algorithm comprises a decision tree that applies information of a patient's EHR and outputs one or more treatment procedures for the patient.
3. The system of claim 1 , wherein the one or more second CQL models include a decision tree that applies information of a patient's electronic health record and outputs one or more treatment procedures for the patient in a standardized form compatible with FHIR systems.
4. The system of claim 1 , wherein the LLM is trained to retrieve a patient's EHR and apply the patient's EHR to the one or more second COL models to generate one or more treatment procedures for the patient.
5. The system of claim 4 , wherein the update of the patient records is applied to the one or more treatment procedures, wherein triggering at least the second CDS hook comprises an indication of a failure for the one or more treatment procedures.
6. The system of claim 4 , wherein retrieving the patient's EHR comprises retrieving the patient's EHR from the first EHR system.
7. The system of claim 4 , wherein the first EHR system selects the disease specific treatment algorithm for the patient, wherein retrieving the patient's EHR comprises retrieving the patient's EHR from a second EHR system different than the first EHR system.
8. The system of claim 4 , wherein the patient's EHR comprises a digital version of a paper chart of an assessment by a medical practitioner of the patient.
9. The system of claim 4 , wherein the patient's EHR indicates a plurality of diseases, wherein the operations further comprise processing data associated with a plurality of disease specific treatment algorithms for individual diseases of the plurality of diseases using the LLM to generate the one or more treatments.
10. The system of claim 1 , wherein the operations further comprise;
retrieving a patient's EHR; and
applying the patient's EHR to the one or more second CQL models to generate one or more treatment procedures for the patient, wherein the update of the patient records is applied to the one or more treatment procedures.
11. The system of claim 10 , wherein triggering at least the second CDS hook comprises an indication of a failure for the one or more treatment procedures.
12. The system of claim 1 , wherein triggering at least the second CDS hook is further based on an indication of a failure in the update to the patient record to prescribe a correct dosage of a medication.
13. The system of claim 1 , wherein triggering at least the second CDS hook is further based on an indication of a failure in the update to the patient record to prescribe a medication at a particular time.
14. The system of claim 1 , wherein triggering at least the second CDS hook is further based on an indication of a newly prescribed medication in the update to the patient record.
15. The system of claim 1 , wherein triggering at least the second CDS hook is further based on an indication of a new patient record opened in the update to the patient record.
16. The system of claim 1 , wherein the alert comprises a user selectable interface element, wherein in response to a user selection of the user selectable interface element, the system initiates communication with a third party server associated with a pharmacy to create, update, modify, or cancel a medication request.
17. The system of claim 1 , wherein the operations further comprise:
training the LLM by:
identifying training disease specific treatment algorithms and corresponding expected training CQL models for the training disease specific treatment algorithms;
applying the training disease specific treatment algorithms to the LLM to receive output COL models;
compare the output CQL models with the expected training CQL models to determine a loss parameter for the LLM; and
update a characteristic of the LLM based on the loss parameter.
18. The system of claim 1 , wherein the operations further comprise:
determining an update to the COL model; and
retraining the LLM using training CQL models that are based on the updated COL model.
19. A method comprising:
receiving a disease specific treatment algorithm corresponding to a disease of a patient from a first Electronic Health Record (EHR) system, the disease specific treatment algorithm including a decision tree that includes guidelines for treating a specific disease;
processing data corresponding to the disease specific treatment algorithm by inputting the data into a Large Language Model (LLM), the LLM being trained to process disease specific treatment algorithms to generate Clinical Quality Language (CQL) models compatible for Fast Healthcare Interoperability Resources (FHIR) and configured to trigger Clinical Decision Support (CDS) hooks;
receiving one or more first COL models from the LLM based on the processing of the data corresponding to the disease specific treatment algorithm, the one or more first COL models include at least a first CDS hook;
receiving an update to a patient record of the patient from the first EHR system;
generating an updated disease specific treatment algorithm based on the update;
inputting the updated disease specific treatment algorithm to the LLM to receive one or more second CQL models from the LLM;
executing the one or more second CQL models triggering at least a second CDS hook to generate an alert; and
causing transmission of the alert for a medical practitioner associated with the first EHR system.
20. A non-transitory computer-readable storage medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform operations comprising:
receiving a disease specific treatment algorithm corresponding to a disease of a patient from a first Electronic Health Record (EHR) system, the disease specific treatment algorithm including a decision tree that includes guidelines for treating a specific disease;
processing data corresponding to the disease specific treatment algorithm by inputting the data into a Large Language Model (LLM), the LLM being trained to process disease specific treatment algorithms to generate Clinical Quality Language (CQL) models compatible for Fast Healthcare Interoperability Resources (FHIR) and configured to trigger Clinical Decision Support (CDS) hooks;
receiving one or more first CQL models from the LLM based on the processing of the data corresponding to the disease specific treatment algorithm, the one or more first CQL models include at least a first CDS hook;
receiving an update to a patient record of the patient from the first EHR system;
generating an updated disease specific treatment algorithm based on the update;
inputting the updated disease specific treatment algorithm s to the LLM to receive one or more second CQL models from the LLM;
executing the one or more second CQL models triggering at least a second CDS hook to generate an alert; and
causing transmission of the alert for a medical practitioner associated with the first EHR system.