System and method for data sensitive filtering of patient demographic record queries

Patent metadata

Field	Value
Publication number	US-9336283-B2
Application number	US-44386206-A
Country	US
Kind code	B2
Filing date	May 31, 2006
Priority date	May 31, 2005
Publication date	May 10, 2016
Grant date	May 10, 2016

Abstract

A method for data sensitive filtering in a patient database searches includes providing a search criteria comprising one or more search locator fields, determining a retrieval formula from the search criteria that maximizes error tolerance in the search criteria while satisfying a predefined response time requirement, and retrieving candidate records from the database. If no retrieval formula can be found that satisfies the response time requirements, the method includes requesting additional search criteria, scoring each candidate record by comparing a search criteria locator field with a corresponding retrieved record field, and determining whether the score of the candidate record exceeds a predefined threshold. If the candidate score does exceed the threshold, the candidate record is added to a list of records to be returned in response to the search criteria.

First claim

What is claimed is:

1. A computer implemented method for data sensitive filtering in patient database searches, said method comprising the steps of:
providing a search criteria for searching a database, the search criteria comprising characters entered into multiple fields of an available plurality of search locator fields;
determining, with a processor, a retrieval formula based on said search criteria that maximizes error tolerance prior to the execution of the retrieval formula, wherein the error tolerance is maximized to achieve a candidate record range bounded by a maximum and a minimum number of records to be returned, wherein the candidate record range is achieved by establishing a probability of a subset of the characters entered into the multiple fields using a comparison of the subset of the characters to a predetermined collection of high probability character strings of said database, and wherein the error tolerance is maximized by determining a candidate record filtering condition that allows the maximum number of candidate records within the candidate record range to be retrieved within a response time requirement;
retrieving, by the processor and based on said retrieval formula, candidate records from said database, wherein the determination of the retrieval formula occurs prior to the retrieving;
scoring, by the processor, each said candidate record by comparing a search criteria locator field with a corresponding retrieved record field, wherein comparing comprises performing a field by field comparison of said locator field and said candidate record field pair to fill in components of a comparison result vector c j for a field pair j and using a field comparison method predefined for each field pair;
scoring said comparison result c 1 based on one or more probabilities using a formula
score( c j )=log( P 1j ( c j ))−log( P 0j ( c j ))=log( P 1j ( c j )/ P 0j ( c j )),

wherein P 0j (c j ) and P 1j (c j ) are probabilities that are functions of the number of matching characters in said pair of fields;
summing score(c j ) over all fields j where the both the locator field and the corresponding field in said candidate record are not blank to calculate a first score; and
determining, by the processor, whether said score of said candidate record exceeds a predefined threshold, and if said candidate score does exceed said threshold, adding said candidate record to a list of records to be returned in response to said search criteria.

2. The method of claim 1 , wherein said search criteria includes a first name and a last name, and wherein the method further comprises compiling a list of names equivalent to said first name in said search criteria.

3. The method of claim 1 , wherein if both a first name and a last name are specified in the search criteria, the method further comprises:
swapping said first and last names in the search criteria;
repeating said steps of performing a field by field comparison of the swapped locator field with the candidate record field, scoring said comparison result ci, and summing said score over all fields j to calculate a second score; and
selecting a maximum of said first score and said second score minus a swap penalty.

4. The method of claim 3 , further comprising transforming said score to a 0 to 100 scale according to a formula
transformedScore=(score−minScore)/(maxScore−minScore)%100,

wherein
maxScore=Σ j maxScore j ,
minScore=Σ j minScore j ,

wherein the locator field that is used to generate the j th component of the comparison vector is not blank in the search criteria, and wherein maxScore j =max(score(c j )) and minScore j =mi(score(c j )) over all possible values of c j .

5. The method of claim 1 , wherein said field comparison method for a field is one of an exact distance match, a Hamming distance, an edit distance, an edit distance with swap, a first name distance, and a last name distance.

6. The method of claim 1 , further comprising, for each candidate record in said list of records to be returned:
retrieving a most recent complete record from said database for said search locator field;
for each field that has a non-empty value in said candidate record, replacing the field value in the retrieved complete record with the corresponding value in the candidate record; and
adding the altered complete record to said database.

7. A computer implemented method for data sensitive filtering in patient database searches, said method comprising the steps of:
providing a search criteria for retrieving one or more records from a database comprising one or more search locator fields;
determining, with a processor, a maximum number of candidate records to be retrieved from said database;
determining, with the processor, a number of candidate records to be retrieved based on a number of characters in said search criteria being correct;
determining, with the processor, a retrieval formula from said search criteria that maximizes error tolerance in said search criteria while satisfying said maximum number of candidate records, wherein the retrieval formula comprises selecting a sub-string of said search field, searching a dictionary of high frequency strings of said database for said substring, and performing a database query when said search sub-string is not found in said dictionary;
retrieving, with the processor and based on the retrieval formula, said candidate records from said database;
scoring, by the processor, each said candidate record by comparing a search criteria locator field with a corresponding retrieved record field, wherein comparing comprises performing a field by field comparison of said locator field and said candidate record field pair to fill in components of a comparison result vector c j for a field pair j and using a field comparison method predefined for each field pair;
scoring said comparison result c j based on one or more probabilities using a formula
score( c j )=log( P 1j ( c j ))−log( P 0j ( c j ))=log( P 1j ( c j )/ P 0j ( c j )),

wherein P 0j (c j ) and P 1j (c j ) are probabilities that are functions of the number of matching characters in said pair of fields;
summing score(c j ) over all fields j where the both the locator field and the corresponding field in said candidate record are not blank to calculate a first score; and
determining, by the processor, whether said score of said candidate record exceeds a predefined threshold, and if said candidate score does exceed said threshold, adding said candidate record to a list of records to be returned in response to said search criteria.

8. The method of claim 7 , wherein if a number of candidate records retrievable using said retrieval formula falls below a minimum number of candidate records, reducing said number of characters in said search criteria assumed to be correct.

9. The method of claim 7 , wherein said maximum number of candidate records to be retrieved is determined from a frequency table of database values that represent probabilities of search criteria fields being included in said database, and wherein strings having probability values higher than a predefined threshold are stored in memory.

10. The method of claim 9 , wherein said threshold is based on the database size, required response time, and a speed of comparison between said search criteria and said candidate records.

11. The method of claim 9 , wherein said probabilities include probabilities of search criteria sub-fields being included in said database.

12. The method of claim 7 , further comprising:
comparing said search criteria with a first candidate record and storing a result of s

Assignees

• Giang Phan H
• Landi William A
• Sandilya Sathyakama
• Cerner Innovation Inc

Inventors

• Giang Phan H
• Landi William A
• Sandilya Sathyakama

Classifications

• G06F17/30533 Primary

  Physics · mapped topic

• G06F19/322

  Physics · mapped topic

• G06F17/30389

  Physics · mapped topic

• G06F17/30536

  Physics · mapped topic

• G16Z99/00

  Subject matter not provided for in other main groups of this subclass · CPC title