Developing and evaluating prediction models in rehabilitation populations

Model System:

Reference Type:

Accession No.:

Journal:

Year, Volume, Issue, Page(s):

Publication Website:

Abstract:

Article presents a 3-part framework for developing and validating prediction models in rehabilitation populations. The authors emphasize the inclusion of target user input in the process of prediction model development and results presentation, and discuss how study design and statistical decisions impact the internal and external validity of prediction model findings. Primary study design recommendations include selecting a relevant and reliable outcome measure, preselecting predictor variables (rather than the use of stepwise methods), and clarifying rules of thumb for the sample size-to-predictors ratio required for statistical power. The authors also provide a brief overview of recommended statistical model building approaches for multivariable prediction models, which includes the following 7 steps: data inspection, coding of predictors, model specification, model estimation, model performance, model validation, and model presentation. A template is offered to help stakeholders evaluate the scientific quality of a prediction model study. The article concludes with the authors’ perspectives on the future development and use of rehabilitation prediction models.

Author(s):

Participating Centers: