Ordered logit

• Assessing Service Utilization by Ellen Dornelas, 5/2/97
  • Service Utilization by Elaine Kersten, 5/7/97
    
    

We wanted a model which would allow us to make use of the quantity of services utilized at post-discharge follow-up, rather than just the presence of absence. Ordinal logit allows you to look at the conditional probability of a single ordinal outcome variable, given a set of independent variables. It is based on a proportional odds model and has no parametric assumptions. It is useful when the assumption of of linearity for ordinary regression is violated. This method allows you to calculate individual and joint probabilities. Therefore you can try different combinations of independent variables to show the risk for a specified outcome (in our case, level of post-discharge services utilized measured on a 4 point scale) for a patient with a given set of characteristics.

• Using Service Utilization to Predict outcome by Elaine Kersten, 5/21/97