Granted Patent B2

US 11,923,074 · App. 17/174,705 · Granted Mar 5, 2024

Inventors: Ruoxin Li (Chapel Hill, NC); Yong Cai (Plymouth Meeting, PA); John H. Eichert, Jr. (Doylestown, PA); Emily Zhao (Wayne, PA); Yilian Yuan (North Wales, PA); Yanping Liu (Harleysville, PA); Steve Eichert (Jenkintown, PA); D. Bruce West (Doylestown, PA); William McClellan (Havertown, PA)

Assignee: IQVIA Inc.

G16H40/20G06N3/045G06N3/08

Abstract

Claims (68)

1. A method comprising:

generating a multi-relational network comprising links among a plurality of physicians based on peer network data, wherein each link indicates a first physician that influences a second physician, and a weight of the influence;

decomposing, using a deep learning engine, a treatment decision by a treating physician of the plurality of physicians into a magnitude of peer influence and a magnitude of control factor influence based on the multi-relational network and a plurality of control factors respectively, wherein the plurality of control factors includes a set of environmental factors and a set of patient factors, wherein decomposing the treatment decision comprises:

measuring a magnitude of the set of environmental factors in the magnitude of control factor influence, and

measuring a magnitude of the set of patient factors in the magnitude of control factor influence, wherein measuring the magnitude of the set of patient factors comprises:

training a model of the deep learning engine with patient longitudinal data to identify patient factors that are likely to affect the treatment decision by the treating physician; and

projecting, by an embedding process performed by the trained model of the deep learning engine, the set of patient factors from a first dimensional space into a second dimensional space, wherein the second dimensional space is smaller than the first dimensional space, and wherein the second dimensional space includes patient factors that are likely to affect the treatment decision by the treating physician;

distributing, by a processing device, the magnitude of peer influence among one or more physicians in the multi-relational network to determine a role that each of the one or more physicians in the multi-relational network has with respect to diffusion of information within the plurality of physicians; and

using the determined roles to optimize diffusion of products among the plurality of physicians in the multi-relational network.

2. The method of claim 1 , wherein

the set of environmental factors include one or both of (i) a set of characteristics of the treating physician, or (ii) a set of managed care factors; and

wherein the set of patient factors include one or both of (i) patient medical histories, or (ii) patient demographic data.

3. The method of claim 1 , wherein the trained model of the deep learning engine comprises a bidirectional encoder representations from transformers (BERT) engine configured perform the embedding process.

4. The method of claim 1 , wherein a tree structure represents the multi-relational network and the treating physician is located at a lowest level of the tree structure.

5. The method of claim 4 , wherein distributing the magnitude of peer influence among one or more physicians comprises:

calculating an influence score based on the magnitude of peer influence;

for each physician in a second level of the tree structure that influences the treating physician, distributing a portion of the influence score to the physician based on a weight of influence of the physician as indicated by the link between the physician and the treating physician; and

for each subsequent level of the tree structure, distributing a portion of the influence score to each physician in the subsequent level that influences one or more physicians in a previous level to whom a portion of the influence score was distributed based on a weight of influence of the physician as indicated by the link between the physician and the treating physician.

6. The method of claim 5 , wherein distributing a portion of the influence score to a physician in a subsequent level that influences one or more physicians in a previous level comprises:

distributing a portion of the influence score based on a weight of influence of the physician on the one or more physicians in the previous level.

7. The method of claim 1 , wherein generating the multi-relational network comprises:

identifying one or more missing links among the plurality of physicians; and

generating one or more link prediction models to complete the one or more missing links.

8. A system comprising:

a memory; and

a processing device, operatively coupled to the memory, the processing device configured to perform operations comprising:

generating a multi-relational network comprising links among a plurality of physicians based on peer network data, wherein each link indicates a first physician that influences a second physician, and a weight of the influence;

decomposing, using a deep learning engine, a treatment decision by a treating physician of the plurality of physicians into a magnitude of peer influence and a magnitude of control factor influence based on the multi-relational network and a plurality of control factors respectively, wherein the plurality of control factors includes a set of environmental factors and a set of patient factors, wherein decomposing the treatment decision comprises:

measuring a magnitude of the set of environmental factors in the magnitude of control factor influence, and

measuring a magnitude of the set of patient factors in the magnitude of control factor influence, wherein measuring the magnitude of the set of patient factors comprises:

training a model of the deep learning engine with patient longitudinal data to identify patient factors that are likely to affect the treatment decision by the treating physician; and

projecting, by an embedding process performed by trained model of the deep learning engine, the set of patient factors from a first dimensional space into a second dimensional space, the second dimensional space is smaller than the first dimensional space, wherein the second dimensional space includes patient factors that are likely to affect the treatment decision by the treating physician;

distributing the magnitude of peer influence among one or more physicians in the multi-relational network to determine a role that each of the one or more physicians in the multi-relational network has with respect to diffusion of information within the plurality of physicians; and

using the determined roles to optimize diffusion of products among the plurality of physicians in the multi-relational network.

9. The system of claim 8 , wherein

the set of environmental factors include one or both of (i) a set of characteristics of the treating physician, or (ii) a set of managed care factors; and

wherein the set of patient factors includes one or both of (i) patient medical histories, or (ii) patient demographic data.

10. The system of claim 8 , wherein the trained model of the deep learning engine comprises a bidirectional encoder representations from transformers (BERT) engine configured perform the embedding process.

11. The system of claim 8 , wherein a tree structure represents the multi-relational network and the treating physician is located at a lowest level of the tree structure.

12. The system of claim 11 , wherein distributing the magnitude of peer influence among one or more physicians comprises:

calculating an influence score based on the magnitude of peer influence; and

for each physician in a second level of the tree structure that influences the treating physician, distributing a portion of the influence score to the physician based on a weight of influence of the physician as indicated by the link between the physician and the treating physician; and

for each subsequent level of the tree structure, distributing a portion of the influence score to each physician in the subsequent level that influences one or more physicians in a previous level to whom a portion of the influence score was distributed based on a weight of influence of the physician as indicated by the link between the physician and the treating physician.

13. The system of claim 12 , wherein distributing a portion of the influence score to an entity in a subsequent level that influences one or more entities in a previous level comprises:

distributing a portion of the influence score based on a weight of influence of the entity on the one or more physicians in the previous level.

14. The system of claim 8 , wherein generating the multi-relational network comprises:

identifying one or more missing links among the plurality of physicians; and

generating one or more link prediction to complete the one or more missing links.

15. A non-transitory computer-readable medium having instructions stored thereon which, when executed by a processing device, cause the processing device to perform operations comprising:

generating a multi-relational network comprising links among a plurality of physicians based on peer network data, wherein each link indicates a first physician that influences a second physician, and a weight of the influence;

decomposing, using a deep learning engine, a treatment decision by a treating physician of the plurality of physicians into a magnitude of peer influence and a magnitude of control factor influence based on the multi-relational network and a plurality of control factors respectively, wherein the plurality of control factors includes a set of environmental factors and a set of patient factors, wherein decomposing the treatment decision comprises:

measuring a magnitude of the set of environmental factors in the magnitude of control factor influence,

measuring a magnitude of the set of patient factors in the magnitude of control factor influence, wherein measuring the magnitude of the set of patient factors comprises:

training a model of the deep learning engine with patient longitudinal data to identify patient factors that are likely to affect the treatment decision by the treating physician; and

projecting, by an embedding process performed by a trained model of the deep learning engine, the set of patient factors from a first dimensional space into a second dimensional space, the second dimensional space is smaller than the first dimensional space, wherein the second dimensional space includes patient factors that are likely to affect the treatment decision by the treating physician;

distributing, by the processing device, the magnitude of peer influence among one or more physicians in the multi-relational network to determine a role that each of the one or more physicians in the multi-relational network has with respect to diffusion of information within the plurality of physicians; and

using the determined roles to optimize diffusion of products among the plurality of physicians in the multi-relational network.

16. The non-transitory computer-readable medium of claim 15 ,

wherein the set of environmental factors include one or both of (i) a set of characteristics of the treating physician, or (ii) a set of managed care factors; and

wherein the set of patient factors include one or both of (i) patient medical histories, or (ii) patient demographic data.

17. The non-transitory computer-readable medium of claim 15 , wherein the trained model of the deep learning engine comprises a bidirectional encoder representations from transformers (BERT) engine configured perform the embedding process.

18. The non-transitory computer-readable medium of claim 15 , wherein a tree structure represents the multi-relational network and the treating physician is located at a lowest level of the tree structure.

19. The non-transitory computer-readable medium of claim 18 , wherein distributing the magnitude of peer influence among one or more physicians comprises:

calculating an influence score based on the magnitude of peer influence; and

for each physician in a second level of the tree structure that influences the treating physician, distributing a portion of the influence score to the physician based on a weight of influence of the physician as indicated by the link between the physician and the treating physician; and

for each subsequent level of the tree structure, distributing a portion of the influence score to each physician in the subsequent level that influences one or more physicians in a previous level to whom a portion of the influence score was distributed based on a weight of influence of the physician as indicated by the link between the physician and the treating physician.

20. The non-transitory computer-readable medium of claim 19 , distributing a portion of the influence score to a physician in a subsequent level that influences one or more physicians in a previous level comprises:

distributing a portion of the influence score based on a weight of influence of the physician on the one or more physicians in the previous level.

Assignments (8)

Cited By (1)

US 12,603,185