Panel Paper:
A New Approach to Analyzing Opioid Use Among SSDI Applicants
*Names in bold indicate Presenter
Yet, little is known about the rates of opioid use among SSDI applicants because of data limitations. Although SSDI applicants are required to report medications, medications are recorded as a combination of coded and open-ended text fields. So a research team would have to recode free-text fields to make them useful for empirical studies. In the past, this required the training of coders and manual coding—a process that is labor intensive, time consuming, prone to error, and usually cost prohibitive.
This study addresses this major data issue that limits what is known about opioid use among SSDI applicants. The study is a proof of concept study for the proposition that machine learning can be used to classify free-form text of medication information in the Social Security Administration’s Stored Data Repository (SDR) and that the algorithm developed can be used in the future to produce statistics on trends of drug use of SSDI applicants. Specifically, the study uses an innovative supervised machine-learning algorithm to identify opioids recorded in free-form text and combine that information with opioids identified in populated medication codes. Using this information, we produce statistics on the prevalence of opioid use among SSDI applicants.
The study finds that 35 percent of applicants reported one or more opioid use when they applied to SSDI program in 2013, with about 30 percent who used opioid reported two or more opioid use. Most frequently reported opioids substance include Tramadol, Hydrocodone with APAP, Oxycodone, Percocet, and Vicodin. 47 percent of applicants mentioned pain as the reason for the use of medicine at least once; among those using opioids, 89 percent reported using the medicine for pain management.