Scientific clinical diffusion network
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a heterogenous network and determining diffusion weights of scientific evidence in the heterogenous network. The computer obtains clinical data associated with a clinical network of healthcare providers and scientific data associated with a scientific network of scientists. The computer generates a first graph network representing the clinical network and a second graph network representing the scientific network. The computer identifies a first set of target nodes in the first graph network representing clinical leaders, and a second set of target nodes in the second graph network representing scientific leaders. The computer predicts links connecting the first graph network to the second graph network and generates a heterogenous network with nodes from the first graph network and from the second graph network, and the links connecting the first graph network to the second graph network.
1 . A computer-implemented method performed by one or more computers comprising:
generating, from clinical data associated with a clinical network of healthcare providers, a first graph network representing the clinical network, the first graph network comprising first nodes representing healthcare providers and first edges connecting two first nodes and indicative of influence between the corresponding healthcare providers;
generating, from scientific data associated with a scientific network of scientists and using a machine learning engine, a second graph network representing the scientific network and indicative of influence between different scientists of the scientists, the second graph network comprising second nodes and second edges, each second node representing a corresponding scientist from the scientists, each second edge connecting two second nodes and indicative of influence between the corresponding scientists of the connected second nodes;
identifying, using survey data representing healthcare sociometric information of the clinical network and scientific network, a first set of target nodes in the first graph network and a second set of target nodes in the second graph network, wherein each node in the first set of target nodes represents a clinical leader in the clinical network, and wherein each node in the second set of target nodes represents a scientific leader in the scientific network;
predicting, using the first and second set of target nodes and by a link prediction model, links connecting the first graph network to the second graph network;
generating a heterogenous network comprising third nodes and third edges, wherein the third nodes comprise the first nodes from the first graph network and the second nodes from the second graph network, and wherein the third edges comprise the links connecting the first graph network and the second graph network;
determining, from the heterogenous network, a first time instance associated with scientific evidence, wherein the first time instance represents a time instance of a scientific publication being associated with at least a subset of the third nodes;
measuring, for each of one or more additional time instances of the heterogenous network and using the links of the heterogenous network associated with the first set of target nodes, an adoption volume of the scientific evidence for the corresponding healthcare providers from the first set of target nodes; and
determining, from the measured adoption volume, a diffusion weight of the scientific evidence for a link from the links connecting the first graph network to the second graph network.
2 . The computer-implemented method of claim 1 , wherein measuring adoption volume of the scientific evidence comprises measuring a number of treatment decisions for patients of the healthcare providers in the clinical network.
3 . The computer-implemented method of claim 1 , further comprising:
generating, by a model configured from the heterogenous network, an output comprising a time-series analysis of adoption of a piece of scientific evidence by the third nodes of the heterogenous network.
4 . The computer-implemented method of claim 3 , wherein the time period is a historical time period that occurs prior to the first time instance.
5 . The computer-implemented method of claim 3 , wherein the time period is a future time period that occurs after the first time instance.
6 . The computer-implemented method of claim 3 , wherein the model is configured as a simulator of adoption volume.
7 . The computer-implemented method of claim 3 , further comprising:
identifying, from the output, the subset of the third plurality of the heterogeneous network having a diffusion rate at least a threshold value.
8 . The computer-implemented method of claim 1 , wherein clinical data is obtained from a plurality of clinical data sources, the plurality of clinical data sources comprising (i) medical claims, (ii) prescription claims, (iii) healthcare provider demographic data, (iv) medical products, and (v) medical procedures.
9 . The computer-implemented method of claim 1 , wherein scientific data is obtained from a plurality of scientific data sources, the plurality of scientific data sources comprising (i) clinical trials, (ii) research studies, (iii) patents, (iv) grants, (v) conference presentations, (vi) publications, and (vii) treatment guidelines.
10 . The computer-implemented method of claim 1 , wherein the survey data is obtained from surveys of at least a first subset of the healthcare providers of the clinical network.
11 . The computer-implemented method of claim 1 , wherein the first graph network of the clinical network comprises one or more identified clinical subnetworks, each identified clinical subnetwork being a patient sharing network.
12 . The computer-implemented method of claim 1 , wherein the second graph network of the scientific network comprises one or more identified scientific subnetworks, each identified scientific subnetwork being a co-authorship network.
13 . The computer-implemented method of claim 1 , wherein the machine learning engine comprises a natural language processing model.
14 . The computer-implemented method of claim 1 , wherein the first graph network and the second graph network are bipartite graphs of the heterogenous network.
15 . The computer-implemented method of claim 1 , wherein the clinical leader is a healthcare provider from the clinical network with an amount of influence that exceeds a first threshold value.
16 . The computer-implemented method of claim 1 , wherein the scientific leader is a scientist from the scientific network with an amount of influence that exceeds a second threshold value.
17 . The computer-implemented method of claim 1 , wherein the scientific data is indicative of scientific publications, and wherein generating the second graph network comprises generating, by the machine learning engine, one or more mappings between the scientific publications and the healthcare providers of the clinical network.
18 . The computer-implemented method of claim 17 , wherein the machine learning engine is trained to perform an inference task comprising text classification and entity identification, wherein the machine learning engine is configured to perform a semantic analysis of the scientific publications by identifying and classifying entities in the scientific publications, and wherein the one or more mappings represent an influence of a respective scientific publication on the respective healthcare provider.
19 . A scientific clinical diffusion system comprising:
a computing device comprising at least one processor; and
a memory communicatively coupled to the at least one processor, the memory storing instructions which, when executed by the at least one processor, cause the at least one processor to perform operations comprising:
generating, from clinical data associated with a clinical network of healthcare providers, a first graph network representing the clinical network, the first graph network comprising first nodes representing healthcare providers and first edges connecting two first nodes and indicative of influence between the corresponding healthcare providers of the connected first nodes;
generating, from scientific data associated with a scientific network of scientists and using a machine learning engine, a second graph network representing the scientific network and indicative of influence between different scientists of the scientists, the second graph network comprising second nodes and second edges, each second node representing a corresponding scientist from the scientists, each second edge connecting two second nodes and indicative of influence between the corresponding scientists of the connected second nodes;
identifying, using survey data representing healthcare sociometric information of the clinical network and scientific network, a first set of target nodes in the first graph network and a second set of target nodes in the second graph network, wherein each node in the first set of target nodes represents a clinical leader in the clinical network, and wherein each node in the second set of target nodes represents a scientific leader in the scientific network;
predicting, using the first and second set of target nodes and by a link prediction model, links connecting the first graph network to the second graph network;
generating a heterogenous network comprising third nodes and third edges, wherein the third nodes comprise the first nodes from the first graph network and the second nodes from the second graph network, and wherein the third edges comprise the links connecting the first graph network and the second graph network;
determining, from the heterogenous network, a first time instance associated with scientific evidence, wherein the first time instance represents a time instance of a scientific publication being associated with at least a subset of the third nodes;
measuring, for each of one or more additional time instances of the heterogenous network and using the links of the heterogenous network associated with the first set of target nodes, an adoption volume of the scientific evidence for the corresponding healthcare providers from the first set of target nodes; and
determining, from the measured adoption volume, a diffusion weight of the scientific evidence for a link from the links connecting the first graph network to the second graph network.
20 . A non-transitory computer-readable storage device storing instructions that when executed by one or more processors of a computing device cause the one or more processors to perform operations comprising:
generating, from clinical data associated with a clinical network of healthcare providers, a first graph network representing the clinical network, the first graph network comprising first nodes representing healthcare providers and first edges connecting two first nodes and indicative of influence between the corresponding healthcare providers of the connected first nodes;
generating, from scientific data associated with a scientific network of scientists and using a machine learning engine, a second graph network representing the scientific network and indicative of influence between different scientists of the scientists, the second graph network comprising second nodes and second edges, each second node representing a corresponding scientist from the scientists, each second edge connecting two second nodes and indicative of influence between the corresponding scientists of the connected second nodes;
identifying, using survey data representing healthcare sociometric information of the clinical network and scientific network, a first set of target nodes in the first graph network and a second set of target nodes in the second graph network, wherein each node in the first set of target nodes represents a clinical leader in the clinical network, and wherein each node in the second set of target nodes represents a scientific leader in the scientific network;
predicting, using the first and second set of target nodes and by a link prediction model, links connecting the first graph network to the second graph network;
generating a heterogenous network comprising third nodes and third edges, wherein the third nodes comprise the first nodes from the first graph network and the second nodes from the second graph network, and wherein the third edges comprise the links connecting the first graph network and the second graph network;
determining, from the heterogenous network, a first time instance associated with scientific evidence, wherein the first time instance represents a time instance of a scientific publication being associated with at least a subset of the third nodes;
measuring, for each of one or more additional time instances of the heterogenous network and using the links of the heterogenous network associated with the first set of target nodes, an adoption volume of the scientific evidence for the corresponding healthcare providers from the first set of target nodes; and
determining, from the measured adoption volume, a diffusion weight of the scientific evidence for a link from the links connecting the first graph network to the second graph network.