NEW(-ish) PAPER
Ever wanted to harness repeatedly-measured predictors within a clinical prediction model but found the existing literature overwhelming?
Look no further, our recent methodological review can help!
#epitwitter #statstwitter #phdchat #epidemiology
Ever wanted to harness repeatedly-measured predictors within a clinical prediction model but found the existing literature overwhelming?
Look no further, our recent methodological review can help!
#epitwitter #statstwitter #phdchat #epidemiology
To make this field more accessible for applied researchers, the extracted modelling techniques were grouped based on similarity, and how they used repeated observations to enhance prediction.
The three main motivations to incorporate repeatedly-measured predictors were:
To improve model specification and applicability over time
To infer an error-free predictor value at a pre-specified time
To account for the effects of predictor change over time
To improve model specification and applicability over time
To infer an error-free predictor value at a pre-specified time
To account for the effects of predictor change over time
...and the following approaches have been discussed:
Time-dependent covariate modelling
Generalised estimating equations
Landmark analysis
Two-stage models
Trajectory classification
Joint models
Machine learning algorithms
Time-dependent covariate modelling
Generalised estimating equations
Landmark analysis
Two-stage models
Trajectory classification
Joint models
Machine learning algorithms
Unfortunately, there is no straight-forward method here, but the choice of methods can be reduced by considering:
the type and amount of data available at prediction time
how the model can benefit from longitudinal information
what is known before model development
the type and amount of data available at prediction time
how the model can benefit from longitudinal information
what is known before model development
Note: some of these approaches can harness repeated observations on a population for model development, and only SINGLE predictor measurements from a patient at the time of prediction. Others can also harness a patient's repeated observations at the time of risk prediction..
Also, key stages of prediction model development were often overlooked in this context & provide motivation for further methods research:
sample size
handling of missing data
variable selection
model validation
quantification of change in predictive performance
sample size
handling of missing data
variable selection
model validation
quantification of change in predictive performance
This methodological review was completed with @jamiecsergeant, Mark Lunt, @glen_martin1, and @khyrich, funded by @NIHRresearch, and supported by @CfE_UoM.
Thank you for taking the time to read all or even part of this thread, and I hope you enjoy reading the paper
Thank you for taking the time to read all or even part of this thread, and I hope you enjoy reading the paper
Apologies, I thought I had posted the link with image.
Please find the link to the paper here: https://diagnprognres.biomedcentral.com/track/pdf/10.1186/s41512-020-00078-z
Please find the link to the paper here: https://diagnprognres.biomedcentral.com/track/pdf/10.1186/s41512-020-00078-z