System and method for determining triage categories

1. A method for determining patient triage categories in pre-hospital settings, the method comprising:

creating a training data set from patient data of a subset of patients of a plurality of patients, the patient data comprising attributes describing the plurality of patients in a pre-hospital setting, the attributes including triage categories assigned to the plurality of patients in the pre-hospital setting, the pre-hospital setting representing a pre-hospital environment in a particular community or geographical region before arrival of the plurality of patients at one or more health care facilities in the particular community or geographical region;

training a machine learning (ML) model using the training data set to generate triage categories for the subset of patients of the plurality of patients;

validating the ML model using a test data set from the patient data;

receiving patient attribute data from a plurality of devices in the pre-hospital setting, the patient attribute data describing a patient in the pre-hospital setting, the receiving performed by a triage system having a processor, a non-transitory computer-readable medium, an interface, and the ML model thus trained;

applying, by the triage system, the ML model to the patient attribute data received from the plurality of devices to generate a triage category for the patient in the pre-hospital setting;

presenting, by the triage system through the interface of the triage system, the triage category generated by the ML model for the patient in the pre-hospital setting, wherein the presented triage category is used for transporting or delivering the patient into care of a trauma center in the particular community or geographical region, the trauma center having a level designation corresponding to the triage category generated for the patient; and

tuning the ML model according to a desired over triage rate or a desired under triage rate so that triage categories generated by the ML model are more likely to yield the desired over triage rate or under triage rate.

2. The method according to claim 1 , wherein the ML model comprises an ensemble classifier configured for utilizing multiple classification techniques to produce a statistically significant predictor of a patient triage category.

3. The method according to claim 2 , wherein the ensemble classifier comprises a random forest classifier, a rotation forest classifier, or a boosting classifier.

4. The method according to claim 2 , wherein, responsive to application of the ML model to the patient attribute data received from the plurality of devices, the ensemble classifier is operable to generate a triage classification score and wherein, from the triage classification score, the ML model is operable to generate the triage category for the patient in the pre-hospital setting based on the desired over triage rate, the desired under triage rate, a triage guideline, or a combination thereof.

5. The method according to claim 1 , wherein the attributes further include vital signs of the plurality of patients determined at multiple points in the pre-hospital setting, patient demographics of the plurality of patients in the pre-hospital setting, any mechanism of injury pertaining to the plurality of patients in the pre-hospital setting, interventions given to the plurality of patients in the pre-hospital setting, any pre-hospital fluid given to the plurality of patients in the pre-hospital setting, medications given to the plurality of patients in the pre-hospital setting, patient management characteristics pertaining to the plurality of patients in the pre-hospital setting, any site of injury pertaining to the plurality of patients in the pre-hospital setting, any disposition pertaining to the plurality of patients in the pre-hospital setting, accuracy of any initial triage level assigned in the pre-hospital setting, any triage level assigned based on patient outcome or treatment, any bleeding status pertaining to the plurality of patients in the pre-hospital setting, any pulse character pertaining to the plurality of patients in the pre-hospital setting, or a combination thereof.

6. The method according to claim 5 , wherein the patient attribute data received by the triage system from the plurality of devices in the pre-hospital setting includes a number of attributes that is the same as or fewer than the attributes describing the plurality of patients in the pre-hospital setting.

7. The method according to claim 1 , wherein the pre-hospital setting represents a pre-hospital environment in which first responders or emergency medical service personnel are assessing the patient or transporting the patient to a trauma center or hospital.

8. A triage system for determining patient triage categories in pre-hospital settings, the triage system comprising:

a processor;

a non-transitory computer-readable medium;

an interface;

a machine learning (ML) model; and

stored instructions translatable by the processor for:

creating a training data set from patient data of a subset of patients of a plurality of patients, the patient data comprising attributes describing the plurality of patients in a pre-hospital setting, the attributes including triage categories assigned to the plurality of patients in the pre-hospital setting, the pre-hospital setting representing a pre-hospital environment in a particular community or geographical region before arrival of the plurality of patients at one or more health care facilities in the particular community or geographical region;

training a machine learning (ML) model using the training data set to generate triage categories for the subset of patients of the plurality of patients;

validating the ML model using a test data set from the patient data;

receiving patient attribute data from a plurality of devices in the pre-hospital setting, the patient attribute data describing a patient in the pre-hospital setting;

applying the ML model to the patient attribute data received from the plurality of devices to generate a triage category for the patient in the pre-hospital setting;

presenting, through the interface of the triage system, the triage category generated by the ML model for the patient in the pre-hospital setting, wherein the presented triage category is used for transporting or delivering the patient into care of a trauma center in the particular community or geographical region, the trauma center having a level designation corresponding to the triage category generated for the patient; and

tuning the ML model according to a desired over triage rate or a desired under triage rate so that triage categories generated by the ML model are more likely to yield the desired over triage rate or under triage rate.

9. The triage system of claim 8 , wherein the ML model comprises an ensemble classifier configured for utilizing multiple classification techniques to produce a statistically significant predictor of a patient triage category.

10. The triage system of claim 9 , wherein the ensemble classifier comprises a random forest classifier, a rotation forest classifier, or a boosting classifier.

11. The triage system of claim 9 , wherein, responsive to application of the ML model to the patient attribute data received from the plurality of devices, the ensemble classifier is operable to generate a triage classification score and wherein, from the triage classification score, the ML model is operable to generate the triage category for the patient in the pre-hospital setting based on the desired over triage rate, the desired under triage rate, a triage guideline, or a combination thereof.

12. The triage system of claim 8 , wherein the attributes further include vital signs of the plurality of patients determined at multiple points in the pre-hospital setting, patient demographics of the plurality of patients in the pre-hospital setting, any mechanism of injury pertaining to the plurality of patients in the pre-hospital setting, interventions given to the plurality of patients in the pre-hospital setting, any pre-hospital fluid given to the plurality of patients in the pre-hospital setting, medications given to the plurality of patients in the pre-hospital setting, patient management characteristics pertaining to the plurality of patients in the pre-hospital setting, any site of injury pertaining to the plurality of patients in the pre-hospital setting, any disposition pertaining to the plurality of patients in the pre-hospital setting, accuracy of any initial triage level assigned in the pre-hospital setting, any triage level assigned based on patient outcome or treatment, any bleeding status pertaining to the plurality of patients in the pre-hospital setting, any pulse character pertaining to the plurality of patients in the pre-hospital setting, or a combination thereof.

13. The triage system of claim 12 , wherein the patient attribute data received by the triage system from the plurality of devices in the pre-hospital setting includes a number of attributes that is the same as or fewer than the attributes describing the plurality of patients in the pre-hospital setting.

14. The triage system of claim 8 , wherein the pre-hospital setting represents a pre-hospital environment in which first responders or emergency medical service personnel are assessing the patient or transporting the patient to a trauma center or hospital.

15. A non-transitory computer program product for determining patient triage categories in pre-hospital settings, the computer program product comprising a non-transitory computer-readable medium storing instructions translatable by the processor for:

creating a training data set from patient data of a subset of patients of a plurality of patients, the patient data comprising attributes describing the plurality of patients in a pre-hospital setting, the attributes including triage categories assigned to the plurality of patients in the pre-hospital setting, the pre-hospital setting representing a pre-hospital environment in a particular community or geographical region before arrival of the plurality of patients at one or more health care facilities in the particular community or geographical region;

training a machine learning (ML) model using the training data set to generate triage categories for the subset of patients of the plurality of patients;

validating the ML model using a test data set from the patient data;

receiving patient attribute data from a plurality of devices in the pre-hospital setting, the patient attribute data describing a patient in the pre-hospital setting;

applying the ML model to the patient attribute data received from the plurality of devices to generate a triage category for the patient in the pre-hospital setting;

presenting, through an interface of the triage system, the triage category generated by the ML model for the patient in the pre-hospital setting, wherein the presented triage category is used for transporting or delivering the patient into care of a trauma center in the particular community or geographical region, the trauma center having a level designation corresponding to the triage category generated for the patient; and

tuning the ML model according to a desired over triage rate or a desired under triage rate so that triage categories generated by the ML model are more likely to yield the desired over triage rate or under triage rate.

16. The non-transitory computer program product of claim 15 , wherein the ML model comprises an ensemble classifier configured for utilizing multiple classification techniques to produce a statistically significant predictor of a patient triage category.

17. The non-transitory computer program product of claim 16 , wherein the ensemble classifier comprises a random forest classifier, a rotation forest classifier, or a boosting classifier.

18. The non-transitory computer program product of claim 16 , wherein, responsive to application of the ML model to the patient attribute data received from the plurality of devices, the ensemble classifier is operable to generate a triage classification score and wherein, from the triage classification score, the ML model is operable to generate the triage category for the patient in the pre-hospital setting based on the desired over triage rate, the desired under triage rate, a triage guideline, or a combination thereof.

19. The non-transitory computer program product of claim 15 , wherein the attributes further include vital signs of the plurality of patients determined at multiple points in the pre-hospital setting, patient demographics of the plurality of patients in the pre-hospital setting, any mechanism of injury pertaining to the plurality of patients in the pre-hospital setting, interventions given to the plurality of patients in the pre-hospital setting, any pre-hospital fluid given to the plurality of patients in the pre-hospital setting, medications given to the plurality of patients in the pre-hospital setting, patient management characteristics pertaining to the plurality of patients in the pre-hospital setting, any site of injury pertaining to the plurality of patients in the pre-hospital setting, any disposition pertaining to the plurality of patients in the pre-hospital setting, accuracy of any initial triage level assigned in the pre-hospital setting, any triage level assigned based on patient outcome or treatment, any bleeding status pertaining to the plurality of patients in the pre-hospital setting, any pulse character pertaining to the plurality of patients in the pre-hospital setting, or a combination thereof.

20. The non-transitory computer program product of claim 19 , wherein the patient attribute data received by the triage system from the plurality of devices in the pre-hospital setting includes a number of attributes that is the same as or fewer than the attributes describing the plurality of patients in the pre-hospital setting.