The goal of this research is to create software that uses mathematical algorithms to detect medical billing coding errors prior to payment. The well-publicized failure of current healthcare cost containment technologies to prevent improper payments in both the commercial healthcare market and the federal Medicare program highlights the urgent need for a new approach to the growing problem of out of control medical costs. A recent federal study by the GAO estimated that improper payments by Medicare alone were in excess of 21 billion dollars, a truly staggering 48.1 percent of all improper payments by federal programs. Like SPAM, whose dynamic nature makes static or post hoc remedies ineffective, effective cost containment in one area often merely leads to the creation of new areas of abuse. Clearly, the ideal solution is a system that can evaluate the fairness of payments before they are made, and that can respond to dynamic patterns of abuse. The first step in creating such a system is the creation of robust method for sorting bills for appropriate rule-based analysis on the basis of the type of bill. Currently neither Medicare nor major insurers are capable of making this classification reliably except through the use of inefficient, static rules and the use of manual sorting--a costly and inefficient approach to assuring timely payment to hospitals and medical providers. We propose a novel method for using mathematical algorithms that utilize machine-learning (ML) methods to address the problem of medical bill categorization, the first step in coding error detection. Specifically, we propose the evaluation of a variety of genetic algorithms that are well adapted to the problems of large, dynamic datasets and can be """"""""trained"""""""" using real world correctly coded datasets in healthcare claims. This work is particularly timely due to recent Medicare contracting reform. Using more than 50 contractors and carriers, bill classification is largely determined by the carrier's contract. Centralizing this process to only four payment centers will require the classification system we propose. [This research is directed toward the development of software applications that will detect billing errors and perform proper edits to payment of medical bills. Current anticipated changes and reforms in the Medicare system will require these systems, which do not currently exist in the public or private sector.] ? ? ? ?
