Predicting Disability Enrollment Using Machine Learning

Citation:

Layton, Tim, Helge Liebert, Nicole Maestas, Daniel Prinz, and Boris Vabson. Working Paper. “Predicting Disability Enrollment Using Machine Learning”.
PDF634 KB

Abstract:

We use data on enrollment in the Supplemental Security Income (SSI) and Social Security Disability Insurance (SSDI) program and data on health care spending by Medicaid beneficiaries to analyze the extent to which Medicaid spending is predictive of future disability insurance receipt among non-disabled teenagers and future disability insurance disenrollment among disabled teenagers. In our first set of analyses, we find that we currently do not have enough data to predict future SSI and SSDI enrollment among non-disabled teenagers. In our second set of analyses, we find that observed Medicaid spending among disabled teenagers can be used to predict SSI disenrollment. Our results indicate that machine learning models using information on healthcare spending may be useful for identifying current teenage SSI recipients who are more or less likely to be removed from SSI.

Notes:

NBER Disability Research Center Paper NB 18-Q4
Executive Summary