Back to Healthcare Contents

Healthcare Data Capture:

High Quality & Low Cost

BY JUNE EVA PEOPLES

THE PROBLEM

More than 3 billion Health Care Finance Administration (HCFA) claim forms are processed annually in the U.S. The HCFA-1500 and the UB-92 forms represent the critical documents for billing and insurance reimbursements nationwide. Since data cannot be accessed as paper--it must be in a computer-readable, database-ready format--HCFA increasingly is mandating that all form data arrive exclusively as EDI. Insurance companies must also turn paper data into computer data to verify a claim and proceed with reimbursement. This means that, at some point in its lifecycle, each HCFA form is re-keyed by expensive human key entry operators.

Domestic and offshore key operators are employed to convert paper data because automated capture technologies, such as OCR, have not been able to cope effectively with the range of quality problems these forms present.

1. The forms are printed in red ink and are designed to drop out completely, leaving only printed data. The forms do not have registration marks or anchor points that allow the system to locate the data. Data effectively can occur anywhere. Since both forms are total drop out, there is no way to identify which form is which when you have scanned a mixed batch.
2. Data is typically printed in low-quality dot matrix fonts.
3. Faint marks, hand-print noise and hand stamps often obscure data fields.
4. Multiple variants exist for both form types.
5. People incorrectly fill in these forms. Essential fields are left blank, data is misplaced and "free-text comment" lines are often written in procedure fields.
6. The provider block field is unstructured (field #33 HCFA 1500). Effective automated data capture must not only read what is there, but make intelligent decisions about what characters constitute name, address, PIN#, and GRP#, so the correct data elements can be exported to the database.
7. Finally, anything a human typed on a keyboard contains errors. Paper and EDI original errors must first be caught and then corrected!

THE SOLUTION

RAF's Cartouche-Medical is a production-proven solution for capturing and perfecting data from HCFA-1500 and UB-92 forms and EDI streams. The end result is far more cost-effective and accurate data than double-key entry can produce. Cartouche-Medical is ready-to-run with full form templates and complete rules and dictionaries, including all available CPT, HCPCS and ICD-9 codes (International Classification of Diseases). A simple-to-use GUI (graphical user interface) allows the operator to select which fields to read, flag, and export. Service bureaus use this interface to quickly tailor which fields to capture for each data conversion customer or batch.

Cartouche-Medical is based on the only all new recognition technology designed expressly for forms processing. Our proprietary recognition technology uses field structure and dictionary information (Intelligent Contextual Analysis) inside the recognition process. This improves total capture accuracy, internally validates data, catches errors in original data and dramatically reduces the cost to convert each form. A flexible GUI presents flagged characters and fields for correction or confirmation.

Cartouche-Medical has efficiently solved the HCFA-1500 and UB-92 quality problems listed above:

1. CM identifies and registers on the data. No form lines are required to find data fields. A mixed stream of HCFA and UB-92 forms are quickly identified and processed.
2. CM's recognition algorithms have been trained on the largest group of fonts extant--the United States mail stream--so CM does the best job possible reading low quality characters.
3. It detects faint marks, hand print noise or stamps ONLY when they interfere with essential data. Those fields are then flagged for review in the Verify module.
4. Internal registration techniques handle all common HCFA and UB-92 form variants.
5. The user can define "must have" fields. Those that are incorrectly left blank are flagged and displayed in Verify module. The operator has the option to fix it or reject the whole page for exception processing.
6. A unique built-in feature understands and separates procedure line data (field #24 HCFA 1500) from creative "comments." Only relevant procedures are exported.
7. It automatically reads field #33 (HCFA 1500) and separates data into correct text elements. It flags the field when it is uncertain how to break up the data.
8. It catches errors in original data. This is done when data is read at high confidence, but the information is not consistent with the rules for that field or does not agree with related information in other fields. City-State-Zip information that does not agree or a non-valid procedure code in field #24 are two examples.

Figure A. Original image scanned after drop-out. Figure B. Form overlay re-applied to data for Verify viewing or printing. The HCFA 1500 and UB-92 are printed in red ink so that the form can be completely dropped out during scan time. A red bulb inside the scanner can't "see" the red ink, so only the black printed data is picked up in the scanned image. There are two benefits: the lines and noise of the form are removed so they don't interfere with the data characters yielding much more accurate recognition results; the file size of each image is much smaller (15-30KB versus 50-80KB) so storage costs and time to load each image are lower. With most capture products, once you scan drop out you've lost the form forever. Cartouche-Medical has a form overlay feature that lets users have the full benefit of drop-out and the ability to get the form back when it's needed.

EDI PERFECTION

HCFA paper forms and HCFA data coming directly as EDI both contain ERRORS in the original data. Simply turning what is on the page into computer text does not always yield accurate or usable data. HCFA forms processors estimate that at least one out of three HCFA-1500 forms contains one or more errors in the original. When these errors occur in key fields like patient number, name, or provider number, the database may not correctly recall the record. CM can take multiple inputs, such as scanned paper, electronic files, and raw key entry, and merge them all into a single validated stream of data. By taking advantage of all the rules and context available, we can separate good data from bad, catch errors in originals, and deliver only known good data for export.

RAF Technology, Inc. provides our customers with the world's most accurate and cost-effective solutions for capturing and perfecting all types of data, regardless of their source. Our intelligent data capture and perfection products bridge the transition from hard copy data to an electronic data world and simplify the management of multiple data types. RAF is proud of the fact that our technology is currently being deployed by the U.S. Postal Service at over 900 mail sorting sites nationwide. By October 31, 1997 our software will be processing 600 million mail pieces per day, 75% of all the world's mail. This is the largest data capture installation in the world.

June Eva Peoples is the Vice President of Marketing for RAF Technology, Inc. She is responsible for marketing the Cartouche product line, which includes Cartouche-Medical. She can be reached at 425-867-0700, 425-882-7370 (fax) or email at june@raf.com.