Machine learning for hepatitis C
To predict which Hepatitis C patients are at high-risk for disease progression or adverse health outcomes, baseline characteristics are measured for patients as well as longitudinal data, including clinical, laboratory and/or biopsy results, which may be collected periodically in follow-up visits with a healthcare professional. A machine learning engine may predict whether a patient is at high-risk for disease progression or adverse health outcomes based on the baseline characteristics and the longitudinal data for the patient.
1. A computer-implemented method for identifying disease progression in Hepatitis C patients, the method executed by one or more processors programmed to perform the method, the method comprising:
obtaining, at one or more processors, a set of training data including a first subset having a first plurality of patient variables associated with a first set of patients having Hepatitis C who do not experience adverse health outcomes as a result of Hepatitis C and a second subset having a second plurality of patient variables associated with a second set of patients having Hepatitis C who do experience adverse health outcomes as a result of Hepatitis C;
receiving, at the one or more processors, a set of patient data for a patient collected over a period of time, wherein the set of patient data includes a first plurality of patient characteristics collected at a first time and a second plurality of patient characteristics collected at a second time, wherein at least some of the first and second plurality of patient characteristics are same patient characteristics collected at different points in time;
comparing, by the one or more processors, the set of patient data for the patient to the set of training data to determine a likelihood that the patient will experience adverse health outcomes as a result of Hepatitis C; and
causing, by the one or more processors, an indication of the likelihood that the patient will experience adverse health outcomes to be displayed on a user interface of a network-enabled device of a health care provider,
wherein the health care provider recommends a course of treatment to the patient according to the determined likelihood.
2. The computer-implemented method of claim 1 , wherein comparing the set of patient data for the patient to the set of training data includes:
for each of the patient characteristics, determining a slope and an acceleration for the patient characteristic based on a difference between the first patient characteristic collected at the first time and the second patient characteristic collected at the second time; and
comparing, by the one or more processors, the slope and the acceleration for each of the patient characteristics to the set of training data to determine the likelihood that the patient will experience adverse health outcomes.
3. The computer-implemented method of claim 1 , wherein obtaining a set of training data includes:
classifying, by the one or more processors, each patient variable in the set of training data as corresponding to the first set of patients having Hepatitis C who do not experience adverse health outcomes as a result of Hepatitis C or the second set of patients having Hepatitis C who do experience adverse health outcomes as a result of Hepatitis C;
assigning, by the one or more processors, a weight to each patient variable in the set of training data based on a level of importance associated with an amount in which the patient variable is indicative of whether the patient having Hepatitis C will experience adverse health outcomes; and
generating, by the one or more processors, a statistical model for determining whether the patient having Hepatitis C will experience adverse health outcomes based on the assigned weights and the classification for each patient variable.
4. The computer-implemented method of claim 1 , further comprising:
receiving, by the one or more processors, an indication of whether the patient experiences an adverse health outcome as a result of Hepatitis C; and
updating, by the one or more processors, the set of training data to include the set of patient data and the indication of whether the patient experiences the adverse health outcome.
5. The computer-implemented method of claim 1 , wherein an adverse health outcome includes at least one of: (i) fibrosis progression, (ii) liver-related death, (iii) liver decompensation, (iv) liver cancer, or (v) the patient requires a liver transplant.
6. The computer-implemented method of claim 1 , wherein the first and second plurality of patient characteristics includes fixed characteristics, clinical variables, laboratory variables, and biopsy variables, wherein the fixed characteristics and clinical variables include at least one of: (i) whether the patient has a history of alcohol and tobacco use, (ii) whether the patient has a history of diabetes, (iii) an estimated duration of hepatitis C infection, (iv) a body mass index, (v) a waist circumference, (vii) an amount of beta-blocker use, (viii) an amount of anti-hypertensive use, or (ix) an average amount of daily alcohol consumption.
7. The computer-implemented method of claim 1 , wherein comparing the set of patient data for the patient to the set of training data to determine a likelihood that the patient will experience adverse health outcomes as a result of Hepatitis C includes:
comparing, by the one or more processors, the set of patient data for the patient to the set of training data to determine a plurality of likelihoods that the patient will experience adverse health outcomes as a result of Hepatitis C for a plurality of predetermined threshold amounts of time; and
causing, by the one or more processors, indications of the plurality of likelihoods and the plurality of predetermined threshold amounts of time to be displayed on the user interface.
8. The computer-implemented method of claim 1 , wherein comparing the set of patient data for the patient to the set of training data includes comparing, by the one or more processors, set of patient data for the patient to the set of training data using one or more machine learning techniques.
9. The computer-implemented method of claim 8 , wherein the one or more machine learning techniques include at least one of random forests or boosting.
10. A computing device for identifying disease progression in Hepatitis C patients, the computing device comprising:
a communication network,
one or more processors; and
a non-transitory computer-readable memory coupled to the one or more processors and storing thereon instructions that, when executed by the one or more processors, cause the computing device to:
obtain a set of training data including a first subset having a first plurality of patient variables associated with a first set of patients having Hepatitis C who do not experience adverse health outcomes as a result of Hepatitis C and a second subset having a second plurality of patient variables associated with a second set of patients having Hepatitis C who do experience adverse health outcomes as a result of Hepatitis C;
receive, via the communication network, a set of patient data for a patient collected over a period of time, wherein the set of patient data includes a first plurality of patient characteristics collected at a first time and a second plurality of patient characteristics collected at a second time, wherein at least some of the first and second plurality of patient characteristics are same patient characteristics collected at different points in time;
compare the set of patient data for the patient to the set of training data to determine a likelihood that the patient will experience adverse health outcomes as a result of Hepatitis C; and
cause, via the communication network, an indication of the likelihood that the patient will experience adverse health outcomes to be displayed on a user interface of a network-enabled device of a health care provider,
wherein the health care provider recommends a course of treatment to the patient according to the determined likelihood.
11. The computing device of claim 10 , wherein to compare the set of patient data for the patient to the set of training data, the instructions cause the computing device to:
for each of the patient characteristics, determine a slope and an acceleration for the patient characteristic based on a difference between the first patient characteristic collected at the first time and the second patient characteristic collected at the second time; and
compare the slope and the acceleration for each of the patient characteristics to the set of training data to determine the likelihood that the patient will experience adverse health outcomes.
12. The computing device of claim 10 , wherein to obtain a set of training data, the instructions cause the computing device to:
classify each patient variable in the set of training data as corresponding to the first set of patients having Hepatitis C who do not experience adverse health outcomes as a result of Hepatitis C or the second set of patients having Hepatitis C who do experience adverse health outcomes as a result of Hepatitis C;
assign a weight to each patient variable in the set of training data based on a level of importance associated with an amount in which the patient variable is indicative of whether the patient having Hepatitis C will experience adverse health outcomes; and
generate a statistical model for determining whether the patient having Hepatitis C will experience adverse health outcomes based on the assigned weights and the classification for each patient variable.
13. The computing device of claim 10 , wherein the instructions further cause the computing device to:
receive, via the communication network, an indication of whether the patient experiences an adverse health outcome as a result of Hepatitis C; and
update the set of training data to include the set of patient data and the indication of whether the patient experiences the adverse health outcome.
14. The computing device of claim 10 , wherein an adverse health outcome includes at least one of: (i) fibrosis progression, (ii) liver-related death, (iii) liver decompensation, (iv) liver cancer, or (v) the patient requires a liver transplant.
15. The computing device of claim 10 , wherein the first and second plurality of patient characteristics includes fixed characteristics, clinical variables, laboratory variables, and biopsy variables, wherein the fixed characteristics and clinical variables include at least one of: (i) whether the patient has a history of alcohol and tobacco use, (ii) whether the patient has a history of diabetes, (iii) an estimated duration of hepatitis C infection, (iv) a body mass index, (v) a waist circumference, (vii) an amount of beta-blocker use, (viii) an amount of anti-hypertensive use, or (ix) an average amount of daily alcohol consumption.
16. The computing device of claim 10 , wherein to compare the set of patient data for the patient to the set of training data to determine a likelihood that the patient will experience adverse health outcomes as a result of Hepatitis C, the instructions cause the computing device to:
compare the set of patient data for the patient to the set of training data to determine a plurality of likelihoods that the patient will experience adverse health outcomes as a result of Hepatitis C for a plurality of predetermined threshold amounts of time; and
cause, via the communication network, indications of the plurality of likelihoods and the plurality of predetermined threshold amounts of time to be displayed on the user interface.
17. The computing device of claim 10 , wherein to compare the set of patient data for the patient to the set of training data, the instructions cause the computing device to compare the set of patient data for the patient to the set of training data using one or more machine learning techniques.
18. The computing device of claim 17 , wherein the one or more machine learning techniques include at least one of random forests or boosting.