Systems and methods for coding health records using weighted belief networks
Systems and methods to code medical records using weighted belief networks are provided. Medical records are received, and may be subjected to pre-processing which includes deduplication of records, indexing the records, meta-tagging the records, and annotating the records. An entity extractor then generates entity dictionaries from public sources. A network creator generates a belief network based on medical relationships. An annotation aligner receives normalized annotations of historical medical records, and a network weighter assigns probability values to the belief network using the normalized annotations to generated a weighted belief network. A health care code classifier utilizes the weighted belief network to classify the medical records by comparing entities within the medical records.
1. A computerized method for coding medical records comprising:
receiving medical records;
generating entity dictionaries from public sources;
generating a belief network based on medical relationships;
receiving normalized annotations of historical medical records where the historical medical records include structured and unstructured data that has been previously coded;
calculating a probability that each association between two concepts in the belief network is true based upon the coded historical medical records;
weighting the belief network using the probabilities; and
classifying the medical records by comparing entities within the medical records to the weighted belief network, wherein an entity below a first threshold association with a medical concept will not be coded, an entity above the first threshold but below a second threshold association with the medical concept be provided to a human coder for review, and an entity above the second threshold association with the medical concept will be automatically coded.
2. The method of claim 1 , wherein the belief network is a Bayesian network.
3. The method of claim 2 , wherein the belief network is a cyclic directed graph with nodes of random variables and relationships between the nodes codify parent/child relationships.
4. The method of claim 3 , wherein the belief network has a domain for each random variable.
5. The method of claim 4 , wherein the belief network includes a set of conditional probability distributions for each variable X given by: P(X|parents(X)).
6. The method of claim 1 , wherein the belief network is a triple data structure comprising a subject-predicate-object.
7. The method of claim 1 , wherein the medical relationships used to generate the belief network is a Web Ontology Language (OWL) and Resource Description Framework (RDF) ontologies.
8. The method of claim 1 , further comprising preprocessing the medical records.
9. The method of claim 8 , wherein the pre-processing includes deduplication of records, indexing the records, meta-tagging the records, and annotating the records.
10. The method of claim 1 , further comprising outputting the classified medical records to at least one coder for review.
11. A computerized system for coding medical records comprising:
an interface for receiving medical records;
two or more servers;
wherein the two or more servers comprise at least a hardware processor, and wherein the two or more servers are configured for:
generating entity dictionaries from public sources;
generating a belief network based on medical relationships,
receiving normalized annotations of historical medical records where the historical medical records include structured and unstructured data that has been previously coded,
calculating a probability that each association between two concepts in the belief network is true based upon the coded historical medical records,
weighting the belief network using the probabilities, and
classifying the medical records by comparing entities within the medical records to the weighted belief network, wherein an entity below a first threshold association with a medical concept will not be coded, an entity above the first threshold but below a second threshold association with the medical concept be provided to a human coder for review, and an entity above the second threshold association with the medical concept will be automatically coded.
12. The system of claim 11 , wherein the belief network is a Bayesian network.
13. The system of claim 12 , wherein the belief network is a cyclic directed graph with nodes of random variables and relationships between the nodes codify parent/child relationships.
14. The system of claim 13 , wherein the belief network has a domain for each random variable.
15. The system of claim 14 , wherein the belief network includes a set of conditional probability distributions for each variable X given by: P(X|parents(X)).
16. The system of claim 11 , wherein the belief network is a triple data structure comprising a subject-predicate-object.
17. The system of claim 11 , wherein the medical relationships used to generate the belief network is a Web Ontology Language (OWL) and Resource Description Framework (RDF) ontologies.
18. The system of claim 11 , wherein one of the two or more servers is further configured for preprocessing the medical records.
19. The system of claim 18 , wherein the pre-processing includes deduplication of records, indexing the records, meta-tagging the records, and annotating the records.
20. The system of claim 11 , wherein one of the two or more servers is further configured for outputting the classified medical records to at least one coder for review.