Using machine learning to predict patient risk associated with respiratory conditions

A system for predicting patient risk, the system comprising:

• a processor;
• a memory storing data associated with a machine learning model;
• and
• a non-transitory computer-readable memory interfaced with the processor and the memory, and storing instructions thereon that, when executed by the processor, cause the processor to:
• access a dataset identifying a plurality of patients, the dataset comprising (i) a first portion comprising first prescription information indicating a plurality of adherence levels respectively by a first portion of the plurality of patients to a plurality of medications prescribed in association with a respiratory condition, where each patient in the first portion of the plurality of patients needed additional or intervention care, and (ii) a second portion comprising second prescription information indicating a second plurality of adherence levels respectively by a second portion of the plurality of patients to the plurality of medications prescribed in association with the respiratory condition, where each patient in the second portion of the plurality of patients did not need additional or intervention care,
• train the machine learning model using the dataset comprising the first portion and the second portion,
• access an additional dataset identifying an additional plurality of patients, the additional dataset comprising prescription information indicating an additional plurality of adherence levels respectively by the additional plurality of patients to an additional plurality of medications prescribed in association with the respiratory condition,
• input the additional dataset into the machine learning model,
• after inputting the additional dataset into the machine learning model, output, by the machine learning model, an output comprising a set of respective likelihoods of whether at least a portion of the additional plurality of patients will need additional or intervention care,
• on a continuous basis:
• evaluate the output in combination with an outreach effort associated with at least the portion of the additional plurality of patients, and
• retrain the machine learning model based on evaluating the output in combination with the outreach effort.

A non-transitory computer-readable storage medium having stored thereon a set of instructions, executable by a processor, for predicting patient risk, the instructions comprising:

• instructions for accessing a dataset identifying a plurality of patients, the dataset comprising (i) a first portion comprising first prescription information indicating a plurality of adherence levels respectively by a first portion of the plurality of patients to a plurality of medications prescribed in association with a respiratory condition, where each patient in the first portion of the plurality of patients needed additional or intervention care, and (ii) a second portion comprising second prescription information indicating a second plurality of adherence levels respectively by a second portion of the plurality of patients to the plurality of medications prescribed in association with the respiratory condition, where each patient in the second portion of the plurality of patients did not need additional or intervention care;
• instructions for training a machine learning model using the dataset comprising the first portion and the second portion;
• instructions for accessing an additional dataset identifying an additional plurality of patients, the additional dataset comprising prescription information indicating an additional plurality of adherence levels respectively by the additional plurality of patients to an additional plurality of medications prescribed in association with the respiratory condition;
• instructions for inputting the additional dataset into the machine learning model;
• instructions for, after inputting the additional dataset into the machine learning model, outputting a result from the machine learning model, the result comprising a set of respective likelihoods of whether at least a portion of the additional plurality of patients will need additional or intervention care;
• and
• instructions for, on a continuous basis:
• evaluating the output in combination with an outreach effort associated with at least the portion of the additional plurality of patients, and
• retraining the machine learning model based on evaluating the output in combination with the outreach effort.