A STATISTICAL METHOD FOR MONITORING NON-ACCEPTABLE DIAGN

点击翻译按钮获取中文摘要

The government under Medicare, private insurers, and some managed care organizations have adopted the (Diagnosis Related Group) DRG system as a basis for reimbursing hospitals for inpatient stays. The DRG code for a hospital stay is based on a complicated algorithm that uses patient medical records for determining health care reimbursement. A recent audit of the Medicare system estimates that billions of unnecessary dollars may have been spent due to incorrect coding by providers. Currently, a finding of incorrect DRG codes results from expensive audits of samples of patient medical records. As the audit relies on human capital which requires adequate staffing and coordination with the hospital, financial adjustments from the audit will lag from the time of the hospital stay. Also, only a sample of all medical records are examined, most records are not. This project will use data already available from one insurer to design and test a model to predict whether a claim is coded incorrectly. Estimates from the statistical model will be used as input to the statistical monitor to determine whether the process (percent of incorrect DRG codes) has changed. Strategies for application of the model and monitor will be determined. The effectiveness of the new system will be compared to that which is currently in use by examining the rate of incorrect DRG codes and the dollars of unnecessary payments. An inexpensive statistical control system to monitor the incorrect DRG coding for all claims would decrease the administrative costs and increase the precision of monitoring for unnecessary payments. A statistical model built on electronically available information could expedite the auditing process and provide adjustments in a more timely fashion. All claims could be included in such a system and those claims with a higher chance of being incorrect could be further examined. The results of the analysis could then be used to improve the predictive accuracy of the original model. The model and monitor will not be proprietary and the results of the research will be published in the open scientific literature.