Systems and methods for generating predictive data models using large data sets to provide personalized action recommendations

1. A non-transitory computer readable medium including instructions that are executable by one or more processors to cause a system to perform a method for generating a personalized action recommendation, the method comprising:

acquiring a request for a service from a first user, wherein the request is associated with a second user and the second user's condition;

receiving user information from a plurality of data sources in a plurality of formats;

converting the received user information into a bundle by normalizing the received user information in the plurality of formats to a consistent format and mapping the normalized user information to a plurality of aspects of the bundle;

storing the converted user information;

identifying one or more features of the second user based on the stored user information;

assigning the second user to two or more segments of users based on the identified one or more features, wherein assigning the second user to the segment two or more segments comprises determining confidence values of the two or more segments;

generating a set of two or more recommended actions for the second user based on the two or more segments;

determining an expected value of each of the two or more recommended actions based on a probability of the second user performing the two or more recommended actions;

determining a rank of the two or more recommended actions based on the expected value of each of the two or more recommended actions; and

outputting a recommended action to the first user with a highest expected value for the second user in response to the request for the service.

2. The non-transitory computer readable medium of claim 1 , wherein the expected value of each of the two or more recommended actions is determined based on an expected value of resolving the second user's condition, a probability that the two or more recommended actions resolves the second user's condition, a probability that the second user's condition is resolved without the two or more recommended actions, and a cost of taking the two or more recommended actions.

3. The non-transitory computer readable medium of claim 2 , wherein the expected value of resolving the second user's condition is determined by subtracting an expected cost of care if the second user's condition is resolved from an expected cost of care if the second user's condition is not resolved.

4. The non-transitory computer readable medium of claim 1 , wherein the expected value of each of the two or more recommended actions is determined using a machine learning model.

5. The non-transitory computer readable medium of claim 4 , wherein the machine learning model comprises at least one of cohort matching or a regression model.

6. The non-transitory computer readable medium of claim 4 , wherein the machine learning model is trained using the stored user information.

7. The non-transitory computer readable medium of claim 1 , wherein the one or more features comprise a state of the second user, and wherein the state of the second user comprises at least one of a gap in care, a medical condition, medications, attributes, behaviors, genetics data, or social determinants of health associated with the second user.

8. The non-transitory computer readable medium of claim 1 , wherein the one or more features comprise a goal associated with the second user, and wherein the goal associated with the second user is extrapolated based on stored medical history of the second user.

9. The non-transitory computer readable medium of claim 1 , wherein:

the second user is a patient;

the request for the service is from the first user comprising at least one of a marketer, a care coordinator, or a clinician.

10. The non-transitory computer readable medium of claim 1 , wherein the two or more segments of users comprise a set of users who share at least one of the one or more features or the second user's condition.

11. A method performed by a system for generating a personalized action recommendation, the method comprising:

acquiring a request for a service from a first user, wherein the request is associated with a second user and the second user's condition;

receiving user information from a plurality of data sources in a plurality of formats;

converting the received user information into a bundle by normalizing the received user information in the plurality of formats to a consistent format and mapping the normalized user information to a plurality of aspects of the bundle;

storing the converted user information;

identifying one or more features of the second user based on the stored user information;

assigning the second user to two or more segments of users based on the identified one or more features, wherein assigning the second user to the two or more segments comprises determining confidence values of the two or more segments value of the segment;

generating a set of two or more recommended actions for the second user based on the two or more segments;

determining an expected value of each of the two or more recommended actions based on a probability of the second user performing the two or more recommended actions;

determining a rank of the two or more recommended actions based on the expected value of each of the two or more recommended actions; and

outputting a recommended action to the first user with a highest expected value for the second user in response to the request for the service.

12. The method of claim 11 , wherein the expected value of each of the two or more recommended actions is determined based on an expected value of resolving the second user's condition, a probability that the two or more recommended actions resolves the second user's condition, a probability that the second user's condition is resolved without the two or more recommended actions, and a cost of taking the two or more recommended actions.

13. The method of claim 12 , wherein the expected value of resolving the second user's condition is determined by subtracting an expected cost of care if the second user's condition is resolved from an expected cost of care if the second user's condition is not resolved.

14. The method of claim 11 , wherein the expected value of each of the two or more recommended actions is determined using a machine learning model, and wherein the machine learning model comprises at least one of cohort matching or a regression model.

15. The method of claim 14 , wherein the machine learning model is trained using the stored user information.

16. The method of claim 11 , wherein the one or more features comprise a state of the second user, and wherein the state of the second user comprises at least one of a gap in care, a medical condition, medications, attributes, behaviors, genetics data, or social determinants of health associated with the second user.

17. The method of claim 11 , wherein the one or more features comprise a goal associated with the second user, and wherein the goal associated with the second user is extrapolated based on stored medical history of the second user.

18. The method of claim 11 , wherein:

the second user is a patient;

the request for the service is from the first user comprising at least one of a marketer, a patient, a care coordinator, or a clinician.

19. The method of claim 11 , wherein the two or more segments of users comprises a set of users who share at least one of the one or more features or the second user's condition.

20. A personalized action recommendation system comprising:

one or more memory devices storing processor-executable instructions; and

one or more processors configured to execute the instructions to cause the personalized action recommendation system to perform:

acquiring a request for a service from a first user, wherein the request is associated with a second user and the second user's condition;

receiving user information from a plurality of data sources in a plurality of formats;

converting the received user information into a bundle by normalizing the received user information in the plurality of formats to a consistent format and mapping the normalized user information to a plurality of aspects of the bundle;

storing the converted user information;

identifying one or more features of the second user based on the stored user information;

assigning the second user to two or more segments of users based on the identified one or more features, wherein assigning the second user to the two or more segments comprises determining confidence values of the two or more segments;

generating a set of two or more recommended actions for the second user based on the two or more segments;

determining an expected value of each of the two or more recommended actions based on a probability of the second user performing the two or more recommended actions;

determining a rank of the two or more recommended actions based on the expected value of each of the two or more recommended actions; and

outputting a recommended action to the first user with a highest expected value for the second user in response to the request for the service.