Step-downs in wage replacement rates across workers' compensation jurisdictions in Australia: impact on scheme exit rate - public pre-registration of research questions and analytical approach
Date created: | Last Updated:
: DOI | ARK
Creating DOI. Please wait...
Description: Update: published in Occupational and Environmental Medicine: https://oem.bmj.com/content/early/2020/03/27/oemed-2019-106325 Background: Step-downs are scheduled reductions in the rate of compensation paid to injured workers. Their primary purpose is to encourage return to work. While there is almost no evidence for their effect, each Australian workers' compensation implements them, with variation in when they come into effect (timing) and degree of reduction (magnitude). We will examine whether step-downs affect the rate at which injured workers leave the compensation system. Methods: 1) Regression discontinuity Using the National Dataset for Compensation-based statistics, we will convert record-level claims data into benefit cessation rates, documenting the proportion of workers' compensation claims that leave the system each week. Using a regression discontinuity (RD) design, we will test whether step-downs increased the benefit cessation rate within each jurisdiction. Analyses will be limited to claims lodged between July 2008 and June 2015 (allowing a minimum of two years of follow up; the data were last updated to June 2017). Where jurisdictions have changed either the step-down rate or initial rate of compensation, we will only include claims lodged after the latest implementation date to ensure consistency in exposure. We will exclude claims that leave the system in the first four weeks and claims affected by the maximum wage replacement cap, which mitigate or preclude step-down effects. RD analyses will be conducted in R using the rddtools package. The models will be fit to the non-linear patterns of benefit cessation using polynomial terms. We will create models up to 10 polynomial terms (separate slopes pre- and post-step-down) and select the best fit model based on the Akaike Information Criterion. We have developed our approach using Victorian data, finding a slight but significant decrease in benefit cessation of -2.4 percentage points (p = .023). The attached plot suggests an increase in the week prior to step-down implementation, indicating an anticipatory effect of claimants leaving the system before their compensation was reduced. In addition to looking at each jurisdiction as a whole, we will also analyse the effect on six conditions: mental health, fractures, back and neck musculoskeletal, neurological conditions, other trauma, and all other conditions. 2) Meta-analyses We will conduct a meta-analysis of RD models at the jurisdiction and jurisdiction/injury level to derive a pooled effect estimate and determine the proportion of variance attributable to heterogeneity. If there is evidence of heterogeneity, we will look at several factors as possible causes including timing of step-down, the magnitude of reduction (difference between pre- and post-step-down rate), and the pre step-down wage replacement rate. Meta-analyses will be conducted in R using the metafor package.