August, 1996 Inspiration is only 2% By Jeff Forsman, National Director of Insurance Technology, Scan-Optics Thomas Edison said that "genius is 2% inspiration and the rest is perspiration." In health claims processing, this adage applies to achieving the level of efficiency that qualifies as genius. The traditional approach to health claims processing seems simple enough: acquire a good scanner and a good recognition engine, along with software to correct characters and edit errors. Depending on volume, such a system can be expected to save your organization hundreds of thousands of dollars annually. Generally speaking, this approach is effective, but only up to a point. Using this approach will, of course, provide a marked improvement over manual systems. But it doesn't address a number of specific areas which, when improved, can provide additional savings and gains in productivity well beyond those achieved by the initial move from a manual method to an automated one. To paraphrase Edison, inspiration is only 2% and the rest is simply diligent work--or, in the vernacular, sweating the details. The goal of most claim processing organizations is to achieve the highest degree of accuracy at the lowest overall cost. An effective system, therefore, should focus on this objective and then utilize whatever technology, tools and techniques are needed to maximize the efficiency of each step in the claim payment process. In addition to character recognition, there are at least four steps where careful system engineering can contribute to the efficiency of health claims processing: 1. Image capture and preparation 2. Automatic post-recognition processing 3. Character correction 4. Edit failure correction So, a more complete and effective solution for health claims processing focuses on the primary objective but then also addresses each step of the overall process. Improve each step of the scanning and recognition process and you significantly improve the overall effectiveness of the process itself. Putting it all together When a health claim recognition system focuses on overall organizational costs and proper attention is paid to each of these steps, the resulting improvement will be dramatic. Let's examine a case study: For purposes of this scenario, assume that a given organization currently processes 10,000 paper claims per day. See the table titled "Savings Comparison." Column 1: First Phase Automation. If the organization in this scenario applies basic scanning and recognition technology to its work, it can achieve a fairly high recognition rate (or raw read rate) on about 80% of the forms they scan. However, the remaining forms are considered unscannable because the average recognition rate is too low, causing most fields to appear for character correction. Nevertheless, the average number of claims handled by an operator jumps from 30 per hour to 51 per hour. Column 2: Image Preparation. When image preparation techniques (such as automatic thresholding, de-skewing, edge detection and trimming) are applied, the organization sees gains in two ways. Not only are more forms considered scannable, but there is also a higher degree of recognition accuracy on all forms. (Note that the figures in column 2 have actually been achieved in production environments.) Column 3: Automatic Character Correction with Improved Interface. By correcting additional fields in post-recognition processing, the organization can eliminate the redundant corrections which represent about 20% of the remaining fields. Additionally, improvements to the user interface can reduce the time spent correcting a field. Column 4: Automatic Edit Correction. A variety of techniques (database lookups for certain fields, etc.) can be applied to correct additional errors in data. These result in improvements in speed and accuracy, and provide further cost savings. Column 5: Automatic Patient and Provider Matching. Finally, automated matching of the patient and provider information for 50% of records (which represents the vast majority of the edit correction effort) creates the maximum benefit. Overall operator productivity soars, as do the annual savings. In the typical production environment, the quality of the documents to be processed varies considerably. An effective health claims scanning and recognition system must be able to handle the majority of incoming documents with little or no special effort on the part of the user. Regardless of document quality, the image presented to the recognition engine must be of high enough quality so that it is "in the zone" in which high-performance recognition occurs. This process begins with the scanner, which first must be able to handle the daily volume of claims in the desired number of working hours. Output image quality is the next most important factor affecting claim processing productivity. Automatic thresholding, deskew, edge detection and trimming are among the features available in certain scanners which can contribute measurably to the initial quality of the image presented to the system. Even with the highest performance scanners, however, there are many attributes of claim images that inhibit good character recognition such as: * Stray marks * Stamps * Faint dot matrix print * Form misalignment Because of such problems, personnel in many health claims scanning operations must first sort claims into various categories of "scannable" and "unscannable" forms. In addition to the sorting cost, this often means that no other cost savings can be achieved for as many as 20% of the incoming forms which cannot be processed automatically. Including an image preparation step in the process addresses these problems. By performing such pre-processing, the percentage of claims that can be effectively processed (considered "scannable") can often be raised to over 98% and the average recognition performance can be improved as well. If such image preparation is done, the raw performance of the recognition engine, while still important, carries much less significance (or approximately 2.5 million claims per year). Labor cost is figured at $15.38 per hour. Operators enter an average of 48 fields per claim form with an average field length of seven characters. Using the current manual process, they are able to enter and correct 30 claims per operator productive hour. Additionally, there are an average of seven edit data fields per form, with an average of six seconds per edit correction. Each column in the Savings Comparison table represents a detail-level step in forms processing. The numbers included in each column provide strong evidence of the incremental dollar savings and gains in productivity that occur at each step of health claims processing as various optimizing techniques are applied. Many recognition engines will provide good performance if the quality of the image is sufficiently high. Automatic post-recognition processing After recognition but before operators get involved in character correction, there are many things that can be done to correct and prepare fields that contain low confidence characters. These include: * Stripping unnecessary characters and formatting the field * Character substitution (if the field has a defined format) * Table lookups and database substitutions * Intelligent transformation of fields * Leveraging redundant or related data on the form to help raise confidence or perform field substitutions Such techniques can significantly reduce the number of fields presented for character correction, directly impacting processing costs. Character correction The labor cost for character correction is a function of the time it takes on average to correct a field and the number of fields the operator has to correct per form. A big factor is motivation and incentives. Operations that focus on operator productivity and put in place management processes to measure and motivate high performance inevitably get better results. Good system design is also very important. Addressing human factors can enhancing motivation and thereby increase operator speed. Such factors include: * Easily visible displays * Screen layouts that minimize eye movement * Function key arrangements that minimize hand movement * Queued batches of work * Visual displays of operator output rates * Effective use of audible tones Another goal in effective health claims processing is minimizing the amount of repetitive processing. For example, a typical claim may have several service lines that contain the same dates of service and other data. By eliminating time spent on repetitive keying, operator output improves. Edit failure correction target for automation The edit failure correction step deserves a lot of attention because the amount of time spent getting the data in the format needed by the mainframe is often several times that of the character correction effort. Most doctors don't fill the forms out the way systems want them. Additionally, a doctor's office may deal with over 100 different insurance companies and other payers, many of which require different data. Healthcare providers and their systems simply cannot keep track of all the different requirements. The consequence is that, on most claims, some data must be changed in order to get it to pass edits. So, the goals of the system should be to correct as much data as possible and to facilitate the operator's ability to correct those edit problems that cannot be dealt with automatically. The system can often be designed to perform simple edit corrections without human involvement. By mapping known but outdated codes into correct codes, overlaying name and address information with entries from a database and the like, operator labor can be reduced. For many group insurers and other companies, as much as two-thirds of the edit correction time is spent getting the patient and provider information extracted from the claim to match up with the related information stored in the mainframe system's databases. Expert logic can be used to automate much of this effort and substantially speed up the correction process. When organizations focus on each step of health claims processing, instead of just dealing with the broad issues involved, they can reap rewards directly commensurate with the amount of attention paid to these details. As is true with most endeavors, the devil is in the details. But so, too, are significant cost and efficiency savings. Jeffrey A. Forsman is national director of insurance technology for Scan-Optics (East Hartford, CT) and is responsible for ImageEMC, a product designed specifically for processing health claims and related forms. He can be reached at 510-647-1063, fax 510-647-0171, E-mail jforsman@scanoptics.com.
	TOP OF PAGE		HOME ++ SEARCH IW ++ DAILY NEWSWIRE ++ CURRENT PUBLICATIONS ++ ABOUT IW BUYER'S GUIDE ++ DIIME ++ SUBSCRIBE ++ COMMENTS
	© 1995, 1996, 1997 Cardinal Business Media, Inc. All Rights Reserved. The names, logos and icons identifying CBMÆs products and services are proprietary marks of Cardinal Business Media, Inc. CBM has no liability for content or goods on the Internet except as set forth in the Terms and Conditions of Service.