UBC Theses and Dissertations
Identifying high-risk claims within the Workers' Compensation Board of British Columbia's claim inventory by using logistic regression modeling Urbanovich, Ernest
The goal of the project was to use the data in the Workers' Compensation Board (WCB) of British Columbia's data warehouse to develop a statistical model that could predict on an ongoing basis those short-term disability (STD) claims that posed a potential high financial risk to the WCB. We were especially interested in identifying factors that could be used to model the transition process of claims from the STD stratum to the vocational rehabilitation (VR) and long term disability (LTD) strata, and forecast their financial impact on the WCB. The reason for this focus is that claims experiencing these transitions represent a much higher financial risk to the WCB than claims that only progress to the health care (HC) and/or the short term disability (STD) strata. The sample used to investigate the conversion processes of claims consists of all STD claims (323,098) that had injury dates between January 1, 1989 and December 31, 1992. Although high-risk claims represent only 4.2 % of all STD claims, they have received 64.3% ($1.2 billion) of the total payments and awards ($1.8 billion) made to July 1999. Low-risk claims make up 95.8% of all the claims but only receive 35.7% ($651 million) of the payments and awards. Moreover, the average cost of high-risk claims ($86,200) is 41 times higher than the average cost of low-risk claims ($2,100). The main objective of the project was to build a reliable statistical model to identify high-risk claims that can be readily implemented at the WCB and thereby improve business decisions. To identify high-risk claims early on, we used logistic regression modeling. Since ten of the most frequently observed injury types make up 95.72% of all the claims, separate logistic regression models were built for each of them. Besides injury type, we also identified STD days paid and age of claimant as statistically significant predictors. The logistic regression models can be used to identify high-risk claims prior to or at the First Final STD payment date provided we know the injury type, STD days paid and age of claimant. The investigation showed that the more STD days paid and the older the injured worker, the higher the probability of the claim being high-risk.
Item Citations and Data