Generating medication orders from a clinical encounter

1. A computer-implemented method, comprising:

receiving a plurality of text transcriptions of medical encounters involving a first medical professional;

training a machine learning (ML) model for the first medical professional based on (i) speech patterns of the first medical professional and (ii) prescription patterns of the first medical professional, based on the plurality of text transcriptions;

applying one or more natural language processing (NLP) algorithms to a new text transcription of a medical encounter dialogue to determine a plurality of features of the dialogue;

upon determining, based on the new transcription, that the first medical professional participated in the dialogue, retrieving the ML model trained for the first medical professional;

processing, by the ML model executing on a processor, the new text transcription and the plurality of features of the dialogue to identify a plurality of candidate medications from a knowledge base, comprising:

generating a score for each of a plurality of statements made by the first medical professional during the medical encounter; and

identifying at least one of the plurality of candidate medications based on (i) a first statement of the plurality of statements, (ii) a medication history of a patient involved in the medical encounter, and (iii) a pattern of prescriptions for a geographical location of the medical encounter;

outputting an indication of the plurality of candidate medications for display;

receiving feedback from the first medical professional, wherein the feedback specifies to exclude a particular medication from the plurality of candidate medications; and

updating the ML model based on the feedback.

2. The computer-implemented method of claim 1 , wherein the plurality of features comprise: (i) a sentiment, (ii) a tone, (iii) a concept, and (iv) a grammatical feature of each of a plurality of statements in the dialogue, wherein the ML model specifies a plurality of attributes of a speech of the medical professional.

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

determining that the first statement comprises a first concept;

determining, by the ML model, that the first concept is related to a first type of medication;

determining, by the ML model, the plurality of candidate medications from the knowledge base based on the first type of medication.

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

receiving selection of a first candidate medication of the plurality of candidate medications;

generating a prescription for the first candidate medication, wherein the prescription comprises an indication of: (i) a patient associated with the prescription, (ii) a name of the first candidate medication, (iii) a dose of the first candidate medication, (iv) a route of the first candidate medication, and (v) a set of instructions for taking the first candidate medication; and

transmitting the prescription for fulfillment.

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

capturing an audio recording of the dialogue;

generating, by the one or more NLP algorithms, the new text transcription of the dialogue; and

generating, by the one or more NLP algorithms, a plurality of annotations of the new text transcription, wherein the plurality of annotations are processed by the ML model to generate the plurality of candidate medications.

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

determining a first set of candidate medications based on a patient history data, a medical condition of a patient, and a prescription history of the medical professional, wherein the first set of candidate medications are processed by the ML model to generate the plurality of candidate medications.

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

determining, by the one or more NLP algorithms, that a second statement in the dialogue is not associated with medical concepts; and

discarding the second statement, thereby refraining from processing the second statement by the ML model.

8. A computer program product, comprising:

a non-transitory computer-readable storage medium having computer readable program code embodied therewith, the computer readable program code executable by a processor to perform an operation comprising:

receiving a plurality of text transcriptions of medical encounters involving a first medical professional;

training a machine learning (ML) model for the first medical professional based on (i) speech patterns of the first medical professional and (ii) prescription patterns of the first medical professional, based on the plurality of text transcriptions;

applying one or more natural language processing (NLP) algorithms to a new text transcription of a medical encounter dialogue to determine a plurality of features of the dialogue;

upon determining, based on the new transcription, that the first medical professional participated in the dialogue, retrieving the ML model trained for the first medical professional;

processing, by the ML model, the new text transcription and the plurality of features of the dialogue to identify a plurality of candidate medications from a knowledge base, comprising:

generating a score for each of a plurality of statements made by the first medical professional during the medical encounter; and

identifying at least one of the plurality of candidate medications based on (i) a first statement of the plurality of statements, (ii) a medication history of a patient involved in the medical encounter, and (iii) a pattern of prescriptions for a geographical location of the medical encounter;

outputting an indication of the plurality of candidate medications for display;

receiving feedback from the first medical professional, wherein the feedback specifies to exclude a particular medication from the plurality of candidate medications; and

updating the ML model based on the feedback.

9. The computer program product of claim 8 , wherein the plurality of features comprise: (i) a sentiment, (ii) a tone, (iii) a concept, and (iv) a grammatical feature of each of a plurality of statements in the dialogue, wherein the ML model specifies a plurality of attributes of a speech of the medical professional.

10. The computer program product of claim 9 , the operation further comprising:

determining that the first statement comprises a first concept;

determining, by the ML model, that the first concept is related to a first type of medication;

determining, by the ML model, the plurality of candidate medications from the knowledge base based on the first type of medication.

11. The computer program product of claim 8 , the operation further comprising:

receiving selection of a first candidate medication of the plurality of candidate medications;

generating a prescription for the first candidate medication, wherein the prescription comprises an indication of: (i) a patient associated with the prescription, (ii) a name of the first candidate medication, (iii) a dose of the first candidate medication, (iv) a route of the first candidate medication, and (v) a set of instructions for taking the first candidate medication; and

transmitting the prescription for fulfillment.

12. The computer program product of claim 8 , the operation further comprising:

capturing an audio recording of the dialogue;

generating, by the one or more NLP algorithms, the new text transcription of the dialogue; and

generating, by the one or more NLP algorithms, a plurality of annotations of the new text transcription, wherein the plurality of annotations are processed by the ML model to generate the plurality of candidate medications.

13. The computer program product of claim 8 , the operation further comprising:

determining a first set of candidate medications based on a patient history data, a medical condition of a patient, and a prescription history of the medical professional, wherein the first set of candidate medications are processed by the ML model to generate the plurality of candidate medications.

14. The computer program product of claim 8 , the operation further comprising:

determining, by the one or more NLP algorithms, that a second statement in the dialgo is not associated with medical concepts; and

discarding the second statement, thereby refraining from processing the second statement by the ML model.

15. A system, comprising:

a processor; and

a memory storing one or more instructions which, when executed by the processor, performs an operation comprising:

receiving a plurality of text transcriptions of medical encounters involving a first medical professional;

training a machine learning (ML) model for the first medical professional based on (i) speech patterns of the first medical professional and (ii) prescription patterns of the first medical professional, based on the plurality of text transcriptions;

applying one or more natural language processing (NLP) algorithms to a new text transcription of a medical encounter dialogue to determine a plurality of features of the dialogue;

upon determining, based on the new transcription, that the first medical professional participated in the dialogue, retrieving the ML model trained for the first medical professional;

processing, by the ML model, the new text transcription and the plurality of features of the dialogue to identify a plurality of candidate medications from a knowledge base, comprising:

generating a score for each of a plurality of statements made by the first medical professional during the medical encounter; and

identifying at least one of the plurality of candidate medications based on (i) a first statement of the plurality of statements, (ii) a medication history of a patient involved in the medical encounter, and (iii) a pattern of prescriptions for a geographical location of the medical encounter;

outputting an indication of the plurality of candidate medications for display;

receiving feedback from the first medical professional, wherein the feedback specifies to exclude a particular medication from the plurality of candidate medications; and

updating the ML model based on the feedback.

16. The system of claim 15 , wherein the plurality of features comprise: (i) a sentiment, (ii) a tone, (iii) a concept, and (iv) a grammatical feature of each of a plurality of statements in the dialogue, wherein the ML model specifies a plurality of attributes of a speech of the medical professional.

17. The system of claim 16 , the operation further comprising:

determining that the first statement comprises a first concept;

determining, by the ML model, that the first concept is related to a first type of medication;

determining, by the ML model, the plurality of candidate medications from the knowledge base based on the first type of medication.

18. The system of claim 15 , the operation further comprising:

receiving selection of a first candidate medication of the plurality of candidate medications;

generating a prescription for the first candidate medication, wherein the prescription comprises an indication of: (i) a patient associated with the prescription, (ii) a name of the first candidate medication, (iii) a dose of the first candidate medication, (iv) a route of the first candidate medication, and (v) a set of instructions for taking the first candidate medication; and

transmitting the prescription for fulfillment.

19. The system of claim 15 , the operation further comprising:

capturing an audio recording of the dialogue;

generating, by the one or more NLP algorithms, the new text transcription of the dialogue; and

generating, by the one or more NLP algorithms, a plurality of annotations of the new text transcription, wherein the plurality of annotations are processed by the ML model to generate the plurality of candidate medications.

20. The system of claim 15 , the operation further comprising:

determining a first set of candidate medications based on a patient history data, a medical condition of a patient, and a prescription history of the medical professional, wherein the first set of candidate medications are processed by the ML model to generate the plurality of candidate medications;

determining, by the one or more NLP algorithms, that a second statement in the dialog is not associated with medical concepts; and

discarding the second statement, thereby refraining from processing the second statement by the ML model.