Methods to estimate effectiveness of a medical treatment

Abstract

Implementations of the present disclosure include methods, systems, and computer-readable storage mediums for estimating treatment effect of medical treatments.

1 . A non-transitory, computer-readable medium storing one or more instructions executable by a computer system to perform operations comprising:
 receiving, from a database, a dataset comprising:
 a set of covariate vectors, each covariate vector including clinical covariates of a respective patient, and 
 a set of response indicators, each response indicator being associated with a respective covariate vector, wherein response indicators of the set of response indicators vary over a range, wherein each range includes a plurality of response classes for the respective clinical covariate, each response class being a portion of the respective range; and 
   estimating a treatment effect of a medical treatment by:
 determining a first subset of covariate vectors for patients who received the medical treatment, and a second subset of covariate vectors for patients who did not receive the medical treatment, 
 obtaining a first normalized probability by:
 applying a neural network model on the first subset of covariate vectors to obtain a first set of probabilities, wherein the neural network model when applied on a covariate vector, provides probabilities of the covariate vector falling in each response classes in the plurality of response classes associated with the covariate vector, and 
 calculating a normalized aggregation of the first set of probabilities to obtain the first normalize probability, 
 
 obtaining a second normalized probability by:
 applying the neural network model on the second subset of covariate vectors to obtain a second set of probabilities, and 
 calculating a normalized aggregation of the second set of probabilities to obtain the second normalize probability, and 
 
 comparing, for one or more response classes, (i) the first normalized probability to (ii) the second normalized probability; and 
   providing the estimated treatment effect for display on a graphical user interface of a computing device.   
     
     
         2 . The computer-readable medium of  claim 1 , wherein the neural network model utilizes a non-linear activation function and a loss function. 
     
     
         3 . The computer-readable medium of  claim 1 , wherein in estimating the treatment effect, the one or more response classes comprises all of the plurality of response classes. 
     
     
         4 . The computer-readable medium of  claim 1 , wherein at least two response classes in the plurality of response classes are disjoint. 
     
     
         5 . The computer-readable medium of  claim 1 , wherein the neural network model utilizes a softmax activation function and a cross-entropy loss function. 
     
     
         6 . The computer-readable medium of  claim 1 , wherein the neural network is a feedforward neural network. 
     
     
         7 . A non-transitory, computer-readable medium storing one or more instructions executable by a computer system to perform operations comprising:
 receiving, from a database, a first dataset comprising:
 a set of covariate vectors, each covariate vector including clinical covariates of a respective patient, and 
 a set of response indicators, each response indicator being associated with a respective covariate vector, wherein response indicators of the set of response indicators vary over a range, wherein each range includes a plurality of response classes for the respective clinical covariate, each response class being a portion of the respective range; 
   for each covariate vector:
 obtaining n+1 predicted responses for the covariate vector, each predicted response being obtained by applying a respective one of n+1 neural network models (NNMs) to the covariant vector, and 
 for each response class, estimating an association probability indicating that the covariate vector is associated with the response class based on the n+1 predicted responses; 
   estimating a treatment effect of a medical treatment by:
 determining a first subset of covariate vectors for patients who received the medical treatment, and a second subset of covariate vectors for patients who did not receive the medical treatment, and 
 comparing, for one or more response classes, (i) a first normalized aggregation of association probabilities for the first subset, to(ii) a second normalized aggregation of association probabilities of the second subset; and 
   providing the estimated treatment effect for display on a graphical user interface of a computing device.   
     
     
         8 . The computer-readable medium of  claim 7 , wherein the one or more response classes comprises all of the plurality of response classes. 
     
     
         9 . The computer-readable medium of  claim 7 , wherein the operations further comprise estimating an uncertainty on the treatment effect by:
 bootstrapping the first dataset for m times to obtain m third datasets;   calculating treatment effect for each third dataset in the m third datasets to obtain m treatment effects; and   calculating the uncertainty based on a distribution of the m treatment effects.   
     
     
         10 . The computer-readable medium of  claim 9 , wherein the uncertainty on the treatment effect includes at least one of a confidence interval and a standard deviation of the m treatment effects. 
     
     
         11 . The computer-readable medium of  claim 7 , wherein the NNM utilizes a non-linear activation function and a least square error loss function. 
     
     
         12 . The computer-readable medium of  claim 11 , wherein the non-linear activation function is a rectified linear unit (ReLU) function. 
     
     
         13 . A computer-implemented method executed by one or more processors, the method comprising:
 receiving, from a database, a dataset comprising:
 a set of covariate vectors, each covariate vector including clinical covariates of a respective patient, and 
 a set of response indicators, each response indicator being associated with a respective covariate vector, wherein response indicators of the set of response indicators vary over a range, wherein each range includes a plurality of response classes for the respective clinical covariate, each response class being a portion of the respective range; and 
   estimating a treatment effect of a medical treatment by:
 determining a first subset of covariate vectors for patients who received the medical treatment, and a second subset of covariate vectors for patients who did not receive the medical treatment, 
 obtaining a first normalized probability by:
 applying a neural network model on the first subset of covariate vectors to obtain a first set of probabilities, wherein the neural network model when applied on a covariate vector, provides probabilities of the covariate vector falling in each response classes in the plurality of response classes associated with the covariate vector, and 
 calculating a normalized aggregation of the first set of probabilities to obtain the first normalize probability, 
 
 obtaining a second normalized probability by:
 applying the neural network model on the second subset of covariate vectors to obtain a second set of probabilities, and 
 calculating a normalized aggregation of the second set of probabilities to obtain the second normalize probability, and 
 
 comparing, for one or more response classes, (i) the first normalized probability to (ii) the second normalized probability; and 
   providing the estimated treatment effect for display on a graphical user interface of a computing device.   
     
     
         14 . The method of  claim 13 , wherein in estimating the treatment effect, the one or more response classes comprises all of the plurality of response classes. 
     
     
         15 . The method of  claim 13 , wherein at least two response classes in the plurality of response classes are disjoint. 
     
     
         16 . The method of  claim 13 , wherein the neural network model utilizes a softmax activation function and a cross-entropy loss function. 
     
     
         17 . The method of  claim 13 , wherein the neural network is a feedforward neural network. 
     
     
         18 . The method of  claim 13 , wherein the neural network model utilizes a non-linear activation function and a loss function.