Worth the Wait? Improving Predictions of Prolonged Work Disability

DRC Brief 2018-02
Publisher: Washington, DC: Center for Studying Disability Policy, Mathematica Policy Research
Jul 09, 2018
Kara Contreary, Yonatan Ben-Shalom, and Brian Gifford
Prompt services and supports can help short-term disability insurance (STDI) claimants return to work after the onset of a medical condition that did not develop on the job. To efficiently target these early interventions to the right people, it is important to identify the claimants who would, without intervention, exhaust their STDI benefits and transition to longer-term support. We use a large database of STDI claims to estimate models that predict who will exhaust STDI and transfer to long-term disability insurance (LTDI). We first assess the potential advantages of waiting for some claims to resolve on their own as a way to narrow the set of claims used to predict who will exhaust STDI benefits or transition to LTDI. We then estimate predictive models four times: first on the full set of claims (as benefits begin), then on those claims that remained after two, four, and six weeks. We estimate the predictive models using half of our data, then generate predictions for the other half to assess the models’ performance. Even without modeling, waiting just a few weeks for some claims to resolve on their own can substantially increase targeting accuracy. Modeling after the waiting period narrows the target population even further, thereby improving efficiency. Before adopting a waiting strategy, however, it is important to consider the trade-offs involved in delaying the delivery of special services to claimants who could benefit from early interventions.