Friday - Aetiological vs prognostic modelling

8 important questions on Friday - Aetiological vs prognostic modelling

What are 4 types of epidemiologic studies?

1. Etiology - can the disease be caused by ...
2. Diagnostic - what disease suits the symptoms of the patient
3. Prognosis - what is the expected future course of the disease in the patient
4. Intervention - can this treatment influence the course of the disease

Is confounding an issue in prognosis research?

No, only in ethiologic research

How can you use the results of a prediction model?

- inform patients on their prognosis
  • create awareness
  • increase compliance
  • promote life-style changes

(results of prediction models should never be used for causality!)
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

What are 'valkuilen' of prediction studies?

Overfitting/optimism, when too much variables are included in the model

In prognostic research there are two types of validation, which two?

1) Internal validation - evaluate how well the model predicts for patients in the data that you used for model development
2) External validation - evaluate how well the model predicts in an external (new) set of patients

What are examples of internal validation precedures?

- apparent validation (complete development dataset)
- split-sample validation (randomly divide dataset in two parts, develop model for part 1 and determine quality with part 2)
- cross-validation (same as above but now in more parts

What are examples of external validation procedures?

- temporal validation = new dataset with more recent patients
- geographic validatoin - dataset with patients from a different hospital/other location
- completely independent validation = new researchers, different patient population, variable definitions may be slightly different

How to assess quality of model?

- callibration - how well do the predicted probabilities agree with the observed outcome frequencies
- discrimination = how well can the model distinguish between individuals with and without disease

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo