Health information system for searching, analyzing and annotating patient data

1. A method comprising:

at a server, preprocessing one or more patient documents to create a plurality of segments of the document at varying levels of granularity;

at a server, enriching the one or more patient documents by:

generating entity links between supporting evidence in the one or more patient documents and entities that are concepts in a semantic taxonomy having a graphical structure of related concepts by, for each respective document:

scoring a plurality of candidate entity links with a concept score, each link comprising a specification of a concept and one or more segments, wherein the concept score is based on (1) an occurrence score based on the number of occurrences of instances of a first concept in the respective document, and (2) a relationship score based on instances of additional concepts in the respective document that are not the first concept, wherein the additional concepts are each connected to the first concept in the graphical structure of related concepts; and

retaining the respective candidate entity links as entity links if the concept score is greater than a threshold;

correcting the one or more entity links using a word-sense disambiguation annotator by evaluating the relatedness of the respective segments to a library of homonyms and collections of words associated with respective members of homonym pairs, and removing respective entity link associations with concepts, terms, and/or extracted health data corresponding to a respective member if the relatedness of a respective segment to a respective collection of words is low, and adding additional entity link associations with concepts, terms, and/or extracted health data corresponding to the respective member if the relatedness to a respective collection of words is high; and

correcting the one or more entity links using a coordinate-expansion annotator by:

identifying conjunctions in the respective segments and creating separated terms based on the segments;

evaluating whether additional instances of entities exist in the segments by parsing the separated terms of the segments; and

generating additional entity links accordingly; and

at a server, generating a search index for the one or more patient documents based on the one or more entity links.

2. The method of claim 1 , wherein enriching the one or more patient documents further comprises:

generating entity links between supporting evidence in the one or more patient documents and entries in a customer-specific dictionary or a molecule database, by creating a respective entity link comprising a specification of a term, a term type, and one or more segments that satisfy a dictionary relationship rule, wherein the dictionary relationship rule is based on whether a respective entry is the same as a term in a respective segment.

3. The method of claim 1 , wherein one or more concepts of the semantic taxonomy have a synonyms attribute.

4. The method of claim 1 , wherein at least one entity link specifies segments located in two or more patient documents.

5. The method of claim 1 , further comprising:

for each of one or more entity links, inserting tags referencing entity link information around the associated segments in a markup language version of the one or more patient documents.

6. The method of claim 1 , wherein enriching the one or more patient documents further comprises:

generating entity links between supporting evidence in the one or more patient documents and terms in a dictionary by creating a respective entity link comprising a specification of a term, a term type, and one or more segments that satisfy a dictionary relationship rule, wherein the dictionary relationship rule is based on whether a respective term is the same as a term in a respective segment.

7. The method of claim 1 , wherein the occurrence score depends upon how often the concept occurs in a title, keywords, abstract, or body of the respective document.

8. The method of claim 1 , wherein whether an instance of a concept occurs in the respective document is based on whether an attribute of the concept is the same or similar to text in one or more segments of the respective document.

9. The method of claim 1 , wherein the relationship score is based on additional concepts that are each connected to the first concept in the semantic taxonomy with an edge distance of one.

10. The method of claim 1 , wherein the relationship score is based on additional concepts that have a positive relationship to the first concept.

11. The method of claim 1 , further comprising:

generating care plan information based on the one or more entity links, wherein the one or more patient documents are associated with a patient, and the one or more entity links include health conditions and socioeconomic data for the patient.

12. The method of claim 1 , further comprising:

identifying a plurality of entity link concepts as RAF concepts, wherein RAF concepts are concepts that affect a risk adjustment factor (RAF) determination;

generating one or more RAF scores for a patient associated with the patient documents based on one or more subsets of the plurality of entity link concepts.

13. A method comprising:

at a server, preprocessing one or more patient documents to create a plurality of segments of the document at varying levels of granularity;

at a server, enriching the one or more patient documents by:

generating entity links between supporting evidence in the one or more patient documents and entities that are concepts, terms, and/or extracted health data using a stack of annotators, wherein the respective entity links comprise a specification of a value, a value type, and one or more segments, by:

generating one or more document-section-specific entity links using a document-section-specific annotator by:

generating one or more semantic-type entity links using a semantic-type annotator by:

 generating one or more base-term-type entity links using a base-term-type annotator; and

 based on the presence of semantic elements of a semantic-type rule, wherein one or more of the semantic elements were generated as components of base-term-type entity links, evaluating the semantic-type rule, wherein the semantic-type rule evaluates vital signs or observations, laboratory tests or laboratory results, drugs or dosages, conditions, or treatment procedures; and

 associating the results of the semantic-type rule as semantic-type entity links; and

based on the presence of section-specific elements of a document-section-specific rule, wherein one or more of the section-specific elements were generated as components of base-term-type entity links or semantic-type entity links, evaluating the document-section-specific rule; and

associating the results of the document-section-specific rule as document-section-specific entity links;

correcting the one or more entity links using a word-sense disambiguation annotator by evaluating the relatedness of the respective segments to a library of homonyms and collections of words associated with respective members of homonym pairs, and removing respective entity link associations with concepts, terms, and/or extracted health data corresponding to a respective member if the relatedness of a respective segment to a respective collection of words is low, and adding additional entity link associations with concepts, terms, and/or extracted health data corresponding to the respective member if the relatedness to a respective collection of words is high; and

correcting the one or more entity links using a coordinate-expansion annotator by:

identifying conjunctions in the respective segments and creating separated terms based on the segments;

evaluating whether additional instances of entities exist in the segments by parsing the separated terms of the segments; and

generating additional entity links accordingly; and

at a server, generating a search index for the one or more patient documents based on the one or more entity links.

14. The method of claim 13 , wherein the semantic-type rule comprises matching a pattern in a respective segment using a regular expression.

15. The method of claim 13 , wherein the semantic-type rule comprises a machine-readable clinical guideline.

16. The method of claim 13 , wherein the base-term-type annotator creates base-term-type entity links that identify negation of a term in a segment.

17. The method of claim 13 , wherein the base-term-type annotator creates base-term-type entity links that identify ages, gender, and geography.

18. The method of claim 13 , wherein the base-term-type annotator creates base-term-type entity links that identify temporal values.

19. The method of claim 13 , wherein enriching the one or more patient documents further comprises:

generating entity links between supporting evidence in the one or more patient documents and entries in a customer-specific dictionary or a molecule database, by creating a respective entity link comprising a specification of a term, a term type, and one or more segments that satisfy a dictionary relationship rule, wherein the dictionary relationship rule is based on whether a respective entry is the same as a term in a respective segment.

20. The method of claim 13 , wherein at least one entity link specifies segments located in two or more patient documents.

21. The method of claim 13 , further comprising:

for each of one or more entity links, inserting tags referencing entity link information around the associated segments in a markup language version of the one or more patient documents.