Storing structured and unstructured clinical information for information retrieval

• US 10,445,378 B2
• Filed: 06/14/2016
• Issued: 10/15/2019
• Est. Priority Date: 12/12/2011
• Status: Active Grant

• Current Assignee: Cleveland Clinic Foundation
• Original Assignee: Cleveland Clinic Foundation
• Inventors: Sasidhar, Madhu
• Primary Examiner(s): Gilligan, Christopher L
• Assistant Examiner(s): Gilligan, C. Luke
• Application Number:
  US15/181,661
• Publication Number:
  US 20160371375A1
• Time in Patent Office: 1,218 Days
• CPC Class Codes: G06F 16/248 Presentation of query results
• CPC Class Codes: G06F 16/258 Data format conversion from or to a database
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/319 Inverted lists
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G16H 10/60 for patient-specific data, e.g. for electronic patient records
• CPC Class Codes: G16H 40/63 for local operation

Abstract

A computer-implemented method can include acquiring data from at least one data source, the acquired data including health data for a patient. The acquired data can be transformed into episode model data according to a context-specific data model and the episode model data can be stored in a database. The method also includes generating at least one inverted index document for at least a portion of an episode for the patient based on the episode model data.

Claims

1. A computer-implemented method, comprising:

acquiring data from at least one data source, the acquired data including health data comprising a plurality of data objects for at least one patient, the at least one data source comprises at least one of an electronic health record repository, a laboratory information systems, or an analytic systems;

transforming, by a processor, the acquired data into episode model data according to a context-specific data model and storing the episode model data in a database;

generating, by a processor, at least one inverted index document for at least a portion of an episode for the patient by a processor based on the episode model data, the generating comprising;

analyzing selected data objects in the episode model data;

determining a respective qualified name for selected data objects in the episode model data according to the context-specific data model, wherein the respective qualified name comprises at least two of a temporal identifier, a field name, and a data type; and

adding the respective qualified name to each of the selected data objects in the inverted index document;

searching, by a processor, the at least one inverted index document by a processor based on a query that includes at least a portion of a corresponding qualified name; and

storing, in memory, results data that includes search results in response to the searching.

2. The method of claim 1, wherein the at least one inverted index is generated at a scheduled time interval.

3. The method of claim 1, wherein the searching further comprising executing one or more queries periodically at a predetermined time interval or in response to an event trigger.

4. The method of claim 3, further comprising

evaluating the results data;

and

providing an alert based on the evaluation indicating the occurrence of a predetermined alert condition.

5. The method of claim 1, wherein the transforming further comprises:

segmenting the acquired data into temporal segments of the episode relative to a predefined base time, each of the temporal segments spanning a time period; and

adding the temporal identifier to selected data objects in each inverted index document to indicate which of the temporal segments each of the selected data objects belongs.

6. The method of claim 5, wherein the temporal identifier comprises a temporal prefix of a qualified name that is pre-pended to each of the selected data objects, according to a predefined schema, to identify a time segment for the selected data objects relative to a time when the inverted index document is generated.

7. The method of claim 1, wherein the temporal identifier specifies one of a plurality of time segments for the episode of care in which each of the selected data objects has been determined to reside, the field name provides a descriptor to represent information content for each of the selected data objects, and the data type indicates a predefined characteristic for data values stored in each of the selected data objects.

8. The method of claim 1, wherein the generating at least one inverted index document further comprises removing stop words from predetermined types of data objects in the episode model data and semantically expanding the predetermined types of the data objects in the episode model data to provide modified versions thereof in the inverted index document.

9. The method of claim 1, further comprising:

computing, by a processor, a score for the search results in the results data, and

ranking, by a processor, the search results in an order based on the computed score for each of the search results.

10. The method of claim 9, wherein the query comprises a plurality of search terms selected according to the data model, the method further comprising assigning a relative weight to each of the plurality of search terms such that the order of the search results varies depending on the relative weighting of the search terms.

11. The method of claim 10, wherein the relative weight is assigned in response to a user input.

12. The method of claim 1, wherein the at least one data source comprises an electronic health record repository.

13. A non-transitory machine readable medium having instructions executable by a processing resource, the instructions comprising:

a data converter programmed to access health data comprising a plurality of data objects from at least one data source and transform the accessed data to episode data for a given patient based on a data model that defines at least one of structure and content for storing the episode data in a database for an episode of care for the given patient, the at least one data source comprising one or more of an electronic health record repository, a laboratory information system, or an analytic system; and

an index generator programmed to;

generate an inverted index document based on the episode data;

analyze selected data objects in the episode model data;

determine a name for each of a plurality of selected data objects in the episode data according to the data model, wherein the name comprises at least two of a temporal identifier, a field name, and a data type; and

add the name to each of the plurality of selected data objects in the inverted index document;

a search engine programmed to;

query the inverted index document to search the at least one inverted index document, the query comprising at least one search term corresponding to at least a portion of the name associated with at least one of the plurality of selected data objects; and

provide search results in response to query.

14. The medium of claim 13, wherein the data converter is programmed to determine the name that is assigned to each of the plurality of data objects as a qualified name, the qualified name being added to each of the selected data objects in the inverted index document according to a schema.

15. The medium of claim 14, wherein the data converter further comprises:

a temporal segmenting function to segment the accessed data into time segments relative to an event; and

a prefix generator to generate a temporal prefix, corresponding to a portion of the qualified name, to specify one of the time segments for each of the selected data objects, wherein the temporal identifier comprises the temporal prefix.

16. The medium of claim 15, wherein the event is a time when the inverted index document is generated.

17. The medium of claim 13, wherein the search engine is programmed to execute one or more queries periodically at a predetermined time interval or in response to an event trigger.

18. The medium of claim 13, further comprising an alert generator to evaluate the results data and provide an alert based on the evaluation indicating the occurrence of a predetermined alert condition.

19. The medium of claim 13, wherein the index generator further comprises a semantic expansion function to expand to include an expanded set of values in the inverted index document that is generated, and a stop word removal function to remove predetermined stop words from of data objects that store a predetermined type of data in the database such that the predetermined stop words are omitted from the inverted index document that is generated.

20. A non-transitory machine readable medium having instructions executable by a processing resource, the instructions comprising:

a data converter programmed to;

access health data comprising a plurality of data objects from at least one data source and transform the accessed data to episode data for a given patient based on a data model that defines at least one of structure and content for storing the episode data in a database for an episode of care for the given patient, the at least one data source comprising one or more of an electronic health record repository, a laboratory information system, or an analytic system;

determine a field name for each selected data object according to the data model; and

determine a data type for each selected data object according to the data model,

an index generator programmed to;

generate an inverted index document based on the episode data;

analyze each selected data object in the episode model data;

determine a name for each selected data object in the episode data according to the data model; and

add name to each selected data object in the inverted index document based on a temporal prefix, the field name and the data type determined for each respective data object, wherein the temporal prefix specifies a respective time segment for each selected data object; and

a search engine programmed to;

query the inverted index document to search the at least one inverted index document, the query comprising at least one search term corresponding to at least a portion of the name associated with at least one selected data object; and

provide search results in response to query.