Comparison of Methods for Imputation of Paid Amounts in Medical Insurance Data

• Activity Number: 40
• Type: Contributed
• Date/Time: Sunday, August 11, 2002 : 4:00 PM to 5:50 PM
• Sponsor: Section on Health Policy Statistics*
• Author(s): Evguenia Jilinskaia*+ and Stanley Norton+ and Cathy Johnson
• Affiliation(s): Pharmetrics, Inc and Pharmetrics, Inc and Pharmetrics, Inc
• Keywords: imputation ; nonhomogenity ; robust regression ; skewed data ; cost and utilization ; least median squares

Abstract

Cost and Utilization metrics are frequently used in measuring effectiveness and cost differences among alternative treatment patterns. Data submitted by health plans to PharMetrics, Inc. exhibits, to a varying degree, missing attributes. Two important data elements in the analysis of healthcare expenditures, "Charged" amounts and "Paid" amounts, are highly correlated financial variables. Pharmaceutical companies and healthcare organizations analyze cost based upon actual expenditures rather than a potentially inflated charged amount. A significant number of claims, such as those resulting from capitated reimbursement relationships, are sometimes represented with zero paid amounts. To analyze homogeneous strata of claims, claim records are classified by type of plan, place of service, inpatient/outpatient, and physician procedure. Different methods of imputation for unknown paid amounts, based on robust regression--least median squares, least trimmed squares, and iteratively reweighted least squares (bi-weighted Tukey estimates)--are compared with robust median ratio estimates. In the presence of outliers, large skewness and nonhomogenity, robust ratio estimates yield a better quality of fit.