Predictive modeling of respiratory disease risk and events
An application server predicts respiratory disease risk, rescue medication usage, exacerbation, and healthcare utilization using trained predictive models. The application server includes model modules and submodel modules, which communicate with a database server, data sources, and client devices. The submodel modules train submodels by determining submodel coefficients based on training data from the database server. The submodel modules further determine statistical analysis data and estimates for medication usage events, healthcare utilization, and other related events. The model modules combine submodels to predict respiratory disease risk, exacerbation, rescue medication usage, healthcare utilization, and other related information. Model outputs are provided to users, including patients, providers, healthcare companies, electronic health record systems, real estate companies and other interested parties.
1. A method comprising:
accessing a respiratory disease risk model comprising a logistic regression function trained to predict a respiratory disease risk for a user based on an expected incidence of medication usage events for the user, the respiratory disease risk model trained using a dataset of users, wherein each user of the training dataset is associated with an expected incidence of medication usage events and is labeled with a respiratory disease risk;
for a given user, inputting an expected incidence of medication usage events for the given user to the respiratory disease risk model;
generating, by the respiratory disease risk model, a respiratory disease risk for the given user; and
sending a respiratory disease risk notification to a computing device associated with the given user including the respiratory disease risk and the expected incidence of medication usage events for the given user.
2. The method of claim 1 , further comprising:
determining a recommended action based on the respiratory disease risk for the given user; and
updating the respiratory disease risk notification to include the recommended action; and
sending the updated respiratory disease risk notification to the computing device associated with the given user.
3. The method of claim 1 , wherein the respiratory disease risk model is further trained to predict a respiratory disease risk for a population of users sharing common characteristics, the respiratory disease risk model trained based on a population dataset of users with the common characteristics.
4. The method of claim 1 , further comprising:
accessing a land use regression sub-model trained to predict air quality data for a location based on land use characteristics of the location;
determining a current location of the given user;
inputting one or more land use characteristics to the land user regression sub-model to predict air quality data for the current location of the given user; and
generating the respiratory disease risk for the given user by inputting the expected incidence of medication usage events for the given user and the predicted air quality data for the current location to the respiratory disease risk model.
5. The method of claim 1 , further comprising:
training the respiratory disease risk model to predict a geographic regional risk for a geographic region by assigning significance values to a plurality of surrounding parameters, wherein each significance value describes the effect of a surrounding parameter of the plurality on the geographic regional risk;
accessing measurements of one or more surrounding parameters from a third-party database at a given time;
inputting the accessed measurements to the respiratory disease risk model to predict a geographic regional risk for the given time; and
sending a respiratory disease risk notification to a plurality of computing devices in the geographic region including the geographic regional risk.
6. The method of claim 5 , further comprising:
accessing measurements of one or more surrounding parameters for a plurality of geographic regions;
generating, by the respiratory disease risk model, a geographic regional risk for each geographic region of the plurality based on the accessed measurements; and
generating a respiratory disease risk map by aggregating the geographic regional risk scores for each geographic region.
7. The method of claim 1 , wherein the respiratory disease risk model is trained to generate an immediate expected respiratory disease risk or a time-lagged expected respiratory disease risk, wherein the time-lagged expected respiratory disease risk describes an impact of a time-lagged exposure on the expected incidence of medication usage events for the given user.
8. A non-transitory computer-readable storage medium storing instructions encoded thereon that, when execute by a processor, cause the processor to:
access a respiratory disease risk model comprising a logistic regression function trained to predict a respiratory disease risk for a user based on an expected incidence of medication usage events for the user, the respiratory disease risk model trained using a dataset of users, wherein each user of the training dataset is associated with an expected incidence of medication usage events and is labeled with a respiratory disease risk;
for a given user, input an expected incidence of medication usage events for the given user to the respiratory disease risk model;
generate, by the respiratory disease risk model, a respiratory disease risk for the given user; and
send a respiratory disease risk notification to a computing device associated with the given user including the respiratory disease risk and the expected incidence of medication usage events for the given user.
9. The non-transitory computer-readable storage medium of claim 8 , further comprising instructions that cause the processor to:
determine a recommended action based on the respiratory disease risk for the given user; and
update the respiratory disease risk notification to include the recommended action; and
send the updated respiratory disease risk notification to the computing device associated with the given user.
10. The non-transitory computer-readable storage medium of claim 8 , wherein the respiratory disease risk model is further trained to predict a respiratory disease risk for a population of users sharing common characteristics, the respiratory disease risk model trained based on a population dataset of users with the common characteristics.
11. The non-transitory computer-readable storage medium of claim 8 , further comprising instructions that cause the processor to:
access a land use regression sub-model trained to predict air quality data for a location based on land use characteristics of the location;
determine a current location of the given user;
input one or more land use characteristics to the land user regression sub-model to predict air quality data for the current location of the given user; and
generate the respiratory disease risk for the given user by inputting the expected incidence of medication usage events for the given user and the predicted air quality data for the current location to the respiratory disease risk model.
12. The non-transitory computer-readable storage medium of claim 8 , further comprising:
train the respiratory disease risk model to predict a geographic regional risk for a geographic region by assigning significance values to a plurality of surrounding parameters, wherein each significance value describes the effect of a surrounding parameter of the plurality on the geographic regional risk;
access measurements of one or more surrounding parameters from a third-party database at a given time;
input the accessed measurements to the respiratory disease risk model to predict a geographic regional risk for the given time; and
send a respiratory disease risk notification to a plurality of computing devices in the geographic region including the geographic regional risk.
13. The non-transitory computer-readable storage medium of claim 8 , further comprising:
access measurements of one or more surrounding parameters for a plurality of geographic regions;
generate, by the respiratory disease risk model, a geographic regional risk for each geographic region of the plurality based on the accessed measurements; and
generate a respiratory disease risk map by aggregating the geographic regional risk scores for each geographic region.
14. The non-transitory computer-readable storage medium of claim 13 , wherein the respiratory disease risk model is trained to generate an immediate expected respiratory disease risk or a time-lagged expected respiratory disease risk, wherein the time-lagged expected respiratory disease risk describes an impact of a time-lagged exposure on the expected incidence of medication usage events for the given user.
15. A respiratory disease management system comprising:
a computing device associated with a given user, the computing device comprising a processor; and
a non-transitory computer-readable storage medium storing instructions encoded thereon that, when execute by a processor, cause the processor to:
access a respiratory disease risk model comprising a logistic regression function trained to predict a respiratory disease risk for a user based on an expected incidence of medication usage events for the user, the respiratory disease risk model trained using a dataset of users, wherein each user of the training dataset is associated with an expected incidence of medication usage events and is labeled with a respiratory disease risk;
for a given user, input an expected incidence of medication usage events for the given user to the respiratory disease risk model;
generate, by the respiratory disease risk model, a respiratory disease risk for the given user; and
send a respiratory disease risk notification to a computing device associated with the given user including the respiratory disease risk and the expected incidence of medication usage events for the given user.
16. The system of claim 15 , further comprising instructions that cause the processor to:
determine a recommended action based on the respiratory disease risk for the given user; and
update the respiratory disease risk notification to include the recommended action; and
send the updated respiratory disease risk notification to the computing device associated with the given user.
17. The system of claim 15 , wherein the respiratory disease risk model is further trained to predict a respiratory disease risk for a population of users sharing common characteristics, the respiratory disease risk model trained based on a population dataset of users with the common characteristics.
18. The system of claim 15 , further comprising instructions that cause the processor to:
access a land use regression sub-model trained to predict air quality data for a location based on land use characteristics of the location;
determine a current location of the given user;
input one or more land use characteristics to the land user regression sub-model to predict air quality data for the current location of the given user; and
generate the respiratory disease risk for the given user by inputting the expected incidence of medication usage events for the given user and the predicted air quality data for the current location to the respiratory disease risk model.
19. The system of claim 15 , further comprising:
train the respiratory disease risk model to predict a geographic regional risk for a geographic region by assigning significance values to a plurality of surrounding parameters, wherein each significance value describes the effect of a surrounding parameter of the plurality on the geographic regional risk;
access measurements of one or more surrounding parameters from a third-party database at a given time;
input the accessed measurements to the respiratory disease risk model to predict a geographic regional risk for the given time; and
send a respiratory disease risk notification to a plurality of computing devices in the geographic region including the geographic regional risk.
20. The system of claim 15 , wherein the respiratory disease risk model is trained to generate an immediate expected respiratory disease risk or a time-lagged expected respiratory disease risk, wherein the time-lagged expected respiratory disease risk describes an impact of a time-lagged exposure on the expected incidence of medication usage events for the given user.