A validation study for using iterative proportional fitting to weight the traumatic brain injury model systems national database: An NIDRR-sponsored study

Model System:

Reference Type:

Accession No.:

Journal:

Year, Volume, Issue, Page(s):

Publication Website:

Abstract:

Study examined whether iterative proportional fitting (IPF) has the desired effect of aligning estimates and parameters when weighting data from the National Institute on Disability and Rehabilitation Research (NIDRR)-funded Traumatic Brian Injury Model Systems (TBIMS) national database. IPF, also known as raking, was first proposed in the 1940s as a method of estimating individual cell probabilities in a contingency table using fixed marginal row and column totals as constraints. The TBIMS has adopted IPF so that weights generated on the basis of known parameters can be used in analyses with the expectation that results are more representative of all late teens and adults receiving inpatient rehabilitation for a primary diagnosis of traumatic brain injury (TBI). These parameters include age at injury, race, sex, marital status, rehabilitation length of stay, payer source, and motor and cognitive Functional Independence Measure scores. This study demonstrates the utility of applying IPF to weight the TBIMS national database so that results of ensuing statistical analyses better reflect those in the United States who are 16 years and older with a primary diagnosis of TBI and are receiving inpatient rehabilitation. In general, IPF aligns population estimates on the basis of weighted TBIMS data and known population parameters. It is reasonable to assume that IPF has the same effect on unknown variables. This provides confidence to researchers wishing to use IPF for making population projections in analyses.

Author(s):

Participating Centers: