Correcting Under-reporting in HCUP Inpatient Data

Medical University Of South Carolina, Charleston SC

Investigators

Linked publications & trials

Abstract

DESCRIPTION (Provided by the Applicant): The HCUP contains the largest all-payer inpatient database in the United States as well as a number of state specific inpatient databases (SID). It is thus an extremely valuable data source for research. While recognizing the important role that HCUP inpatient data have in investigating important health research topics, there is a paucity of academic literature exploring the quality of this data. Hospitals have two parallel systems of patient level information. The first, based on the medical record, contains ICD-9-CM codes recorded by professional coders. Many procedures have a corresponding ICD-9-CM code but not all procedures are required to be coded for reimbursement purposes. Consequently, there is no obvious incentive for medical record coders to register information for non-reimbursable procedures. The second comes from the hospital's internal accounting system. All hospital activity regarding a patient is automatically registered in revenue codes for billing purposes. Revenue codes are thus assumed to be a more accurate measure of utilization. HCUP national data only contain data from medical records (ICD-9-CM codes) while some of the state databases also contain revenue codes. Studies using HCUP inpatient data often measure utilization by the presence of an ICD-9-CM code for the procedure in the discharge. This common practice will lead to inaccurate results if the studies employ non-reimbursable procedures based on ICD-9-CM, due to under-reporting. Since it is possible to compare utilization for a limited number of procedures in both sources of information (ICD-9-CM and revenue) in some states, the extent and variability of under-reporting of two important non-reimbursable procedures: Computerized Tomography (CT) and Magnetic Resonance Imaging (MRI) will be analyzed. A correction method, based on detection control estimation, will be proposed and estimated using inpatient data for 15 DRGs for 9 states for the year 2001. Results from the correction model will be compared with those from a utilization model based on revenue codes to validate the applicability of the proposed correction to other states and procedures. It is expected that this method may serve to increase the value of HCUP inpatient data to health services researchers.