Combining Generalized Linear Mixed Modeling and Random Effects Modeling to Investigate Probability of Outcome Over Time: a NIDRR Traumatic Brain Injury Model Systems Sponsored Presentation

• Activity Number: 188
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Christopher Pretz*+
• Companies: Craig Hospital
• Keywords: Random Effects Modeling ; Generalized Linear Mixed Modeling ; Longitudinal Analysis ; Growth Curve Modeling

Abstract

Many different species of outcome measures are used in Rehabilitation Medicine as well as other disciplines to catalogue change in response over time. In such settings it is often necessary to understand how the probability of an event increases or decreases at the patient (individual) level. To model time dependent probability of an event data at the patient level the following approach is proposed. The first step is to employ a generalized linear mixed model to estimate logits based on a temporally collected dichotomous response. Once a vector of logits is estimated for each respective patient, patient level profiles which represent temporal change can be generated. These profiles can then be modeled using random effects modeling upon which logits are transformed into probabilities. Once logits are transformed into probabilities, random effects modeling allows for understanding of how the probability of an event is not only related to time, but how the probability of an event is influenced by a series of patient level covariates. This presentation includes examples of using data from the Traumatic Brain Injury Model System National Database. Computer generated interactive