Annual Raking and Weighting of the Traumatic Brain Injury Model Systems Database

• Activity Number: 375
• Type: Contributed
• Date/Time: Tuesday, August 2, 2016 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Jessica McKinney Ketchum* and Jeffrey Cuthbert and Gale Whiteneck and C.B. Eagye and Cindy Harrison-Felix
• Companies: Craig Hospital and Sky Ridge Medical Center and Craig Hospital and Craig Hospital and Craig Hospital
• Keywords: Raking ; Weighting ; Trauma ; Brain Injury

Abstract

Over 20,000 adults receive inpatient rehabilitation for traumatic brain injury (TBI) annually in the US. The TBI Model Systems National Database (TBIMS NDB) longitudinally follows a cohort of such individuals, collecting long-term outcomes, but it is not representative of the larger population. The Uniform Data System for Medical Rehabilitation (UDS) and the American Medical Rehabilitation Providers Association's eRehabData (eRehab) collect and report data from at least 92% of all civilian inpatient rehabilitation facilities. Previous work by Corrigan et al. and Cuthbert et al. have used iterative proportional fitting (raking) and weight trimming to make population estimates from the TBIMS NDB by aligning the cohort with multi-year population characteristics from UDS/eRehab data. However, UDS/eRehab data show clear changes in population parameters over time which must be incorporated into these analyses. The research presented here extends the previous work by raking the TBIMS NDB using annual UDS/eRehab data (2002-2013). Incorporating annual changes will significantly improve the precision of population-level estimates derived from the TBIMS NDB cohort.