California Department of Health Services (CDHS) Selects iCapture

WALNUT CREEK, Calif.-- August 18, 2008

The California Vital Records Project……A Case History

In early 2008, the California Department of Health Services (CDHS) entered into a Contract with a team of Companies that included SolutionsWest, a Sacramento based consulting firm, FileNet Corporation and Impression Technology to design, develop and implement a new Vital Records Image Redaction and Statewide Access (VRIRSA) System in response to a California Legislature law, SB 247, which required that CDHS establish a single state database from which all informational-only copies of vital records would be issued and an array of other requirements associated with preventing vital records (Birth and Death certificates) use in establishing a fraudulent identity. In this Case History, we will present background information from the CDHS Request for Proposal, we will describe the architecture and components that form the VRIRSA System, and we will describe the extraordinary work that Impression Technology accomplished in developing an automated redaction capability within the complete solution.

Background (extracted from the CDHS RFP)

“Vital records have historically been used in establishing the lines of inheritance, establishing taxation, and eliminating people from voter registries. In the latter half of the 20th century, the administrative use of vital records increased substantially. Proof of identity, age, and/or the occurrence of an event (e.g. death) are a standard requirement for various socio-economic circumstances, including settling decedent estates, filing for death insurance benefits, and obtaining Social Security Numbers (SSN’s).

As these administrative uses have increased, so have the fraudulent uses of vital records. Vital records are considered “breeder documents”, which provide a window of opportunity for additional fraud. When a copy of a vital record is obtained, it may be used to forge other vital records or be used as proof of identification to falsely obtain credit cards, insurance policies, driver’s licenses, bank accounts and government benefits. This results in increased cost to the public. Personal costs include increases in service costs with credit cards, financial, and insurance institutes as attempts to identify and prosecute fraud are made. Social costs include increasing taxes and the inappropriate use of government funds through the fraudulent receipt of public programs and benefits.

The fraudulent use of vital records are proving more difficult to detect for a number of reasons including:

• Once a person obtains a certified copy of a vital record, it is difficult to ascertain if the holder is truly the person of record or an imposter.
• Since 1905, more than 100 different forms have been used (and are in circulation) for California birth certificates. The variations in the forms makes it difficult to determine if a birth certificate is real or a forgery.
• Some California Counties have issued certified abstracts of birth certificates. This contributes to the number of forms in circulation and increases the difficulty in determining the legitimacy of the record.
• Historically, proof of identification has not been required to obtain a birth certificate. This has created a ripe environment for fraud.”

While the CDHS RFP provides a significant amount of additional justification and supporting information, the following information from the VRIRSA final Design Document will give you an understanding of the scope and complexity of this project, clearly one of the most challenging efforts in the area of Vital Records performed to date:

• In the final analysis, it was determined that there were approximately 100 different variations of California Death certificate forms and 80 variations of Birth Certificates, with  records dating back to 1896.
• It was estimated that approximately 12,000,000 of the 45,000,000 paper certificates had been scanned and committed to a FileNet image repository. The remaining 30,000,000 plus paper documents would be scanned and committed to this same repository as a result of a separate project independent of the VRIRSA engagement
• Statewide access to VRIRSA required that each of the 58 Counties in California be able to send requests to the VRIRSA System that would reside in Sacramento for information-only copies of vital records and the system would be required to acknowledge and respond to each request within minutes.
• The original image of each record stored within the repository could not be altered in any way as a result of VRIRSA processing. All of the redaction and annotation described below had to be performed using a copy of the original image file.
• The information-only copy of the vital record required that selected fields on that record be automatically redacted prior to transmitting that image file back to the requesting County. Given approximately 180 variations of both Birth and Death certificates, the co-ordinates of the fields that required redaction was also variable and none of the certificates had the ‘target marks’ typically used to establish a form template in an automated ICR/OCR recognition system. In addition to the redaction of specified fields, the image returned to the County had to have the following notation ‘stamped’ on the image; “Informational. Not a Valid Document to Establish Identity”.
• With records dating back to 1896, the quality of the paper document and hence the scanned image of that record, varied widely. This added yet another degree of complexity when automatically examining a record that had been requested to determine the exact position of the fields that must be redacted.

Vital Records Image Redaction and Statewide Access System Workflow

The VRIRSA solution integrates and leverages proven products from both FileNet Corporation and Impression Technology. In general, FileNet products including Image Services, Content Services and Business Process Manager (BPM) were used to support the initiation of requests from the County Offices throughout the State, the storage of the original vital record image files and the associated index metadata that allowed search and retrieval of specified records, and the delivery of the redacted image files back to the requesting County.

As shown in the Figure below, once the requested vital record image was found within the FileNet repository, FileNet’s BPM application passed this image file and its associated metadata to the Impression Technology Automated Redaction Subsystem that was based on Impression’s iCapture suite of software applications. Following import, this subsystem provided automated image enhancement including de-skewing and de-speckle to improve image quality and through the use of the record metadata and an automated inspection of the image itself, the Redaction Subsystem was designed to determine which of the 180 variations of the vital record format the specific record represented. With this information, the ‘best fit’ template could be applied to the image and through further image processing, the location of the fields that would require redaction could be located with an extremely high accuracy rate. As shown on the workflow diagram below, this Subsystem was designed so that the system would automatically define a confidence level that it had indeed identified the vital record format and the fields for redaction, and this confidence level could be set and adjusted by CDHS staff during production operations. In the event that a ‘low confidence’ identification occurred, the subsystem would route that image file to a CDHS operator where, through the use of a custom application, the operator could manually redact the required fields and insert the redacted image file back into the standard workflow.

The Automated Redaction Subsystem provided many other features and capabilities that were required by CDHS. These included the ability to easily create new ‘best fit’ templates if required as new record formats were introduced by Counties, a Quality Assurance application that CDHS could utilize to QA any percentage of overall transactions it desired, extensive statistics and automated reporting, and complete audit trails on every transaction processed within VRIRSA.

A critical success factor for production operations and Statewide acceptance of this new solution was the ability of the system to perform automated redaction without operator intervention. With the potential for thousands of requests per day for informational copies of vital records across 58 Counties, manual redaction by multiple operators was simply not a viable option. The success of the Automated Redaction Subsystem is briefly described in the last section of this paper.

Automated Redaction----An Extraordinary Success

As noted in the early sections, the ability of the VRIRSA solution to perform automated redaction provided many unique challenges that went far beyond the technology typically used in implementing ICR or OCR recognition in document management or forms processing system. Not only did the project have to deal with up to 45,000,000 ‘non-image friendly’ images (e.g. no target or alignment marks, no form ‘dropout’, multiple lines/boxes, cursive handprint, etc), but the images themselves could be very poor quality and in any one of 180 formats!

Through the optimization of selected and proven iCapture applications including Image Processing, Form Identification, Form Edit for template definition, and the development of algorithms that allowed ‘fine tuning’ of where to apply redaction through the examination of an image at the pixel level, Impression Technology has been able to achieve extraordinary results in the percentage of vital records that can be automatically redacted with no operator intervention.

As an example of this, in the processing of 4,033 Vital Records from 1963, including multiple form template, automated redaction was successfully performed on 4,016 of these images, or 99.58%. As the quality of the image files from 1963 to the present is typically equal to, or better than, that obtained in 1963, we project automated redaction rates in the same 99%-plus range. Equally impressive, for 1,000 images of Vital Records from 1906 (103 year old documents, again with multiple formats) we achieved successful automated redaction on 717 records, or 71.7%.

The three pictures below are provided to give as example of the major steps within the redaction process. In Figure 1 below we can see the image file of a Vital Record as stored within the Image Repository. (Note: This is an example only as CDHS image files all contain sensitive information that may not be disclosed)

       Figure 1----Original Image File

In Figure 2 below, we see this same image following the Redaction Subsystem Image Processing. The image has been de-skewed, de-specked and (if required), it can also be shrunk or expanded to regain dimensional accuracy.

 Figure 2---Image after Redaction Subsystem Image Processing

In the final Figure below, we see the top 2/3 of this image file (the critical area of the document) showing that the specified signature fields, in this case three fields, have been redacted.   It is this type of performance and accuracy that significantly contributes to the success of the  California VRIRSA System

For additional information of this and other Impression Technology projects across the United States, please contact Bruce Lechner at 925 280 0010, ext. 20 or lechner@impression-technology.com.

About Impression Technology

Headquartered in Walnut Creek, California, Impression Technology specializes in the scalable design and delivery of automated data capture, remittance and e-File solutions for the public sector with customers in California, Michigan, Indiana, Oregon, Washington, Arizona, Wyoming, Puerto Rico and Louisiana. Impression specializes in the area Tax and Revenue, Labor, Law Enforcement, and Medical claims processing. Established in 1997, their professional services group is staffed with consultants formally with Teknekron and SHL Systemhouse with extensive experience in the delivery of world-class solutions including systems for Fidelity Investments, Merrill Lynch, Boeing, Lockheed, Mobil, General Motors, General Electric, Pratt and Whitney and many other fortune 100 companies. Impression Technology's flagship product iCapture® supports many key standards and government initiatives including Check21, SOA, MeF, FSET and assertion-based business rules engine enabling the automated processing of millions of transactions daily.

For more information, please contact Bruce Lechner or Mike Tokuyama, at 925.280.0010, ext. 20 or ext. 12 sales@impression-technology.com.