Linear model

The linear model is maybe the easiest standard model on Evidencio. It uses the formula P = β0 + xβ
Where β0 is an intercept or baseline. and xβ is the sum of the beta coefficients multiplied by their input.

Linear models can be often used for the implementation of online risk scores or clinical prediction rules. If certain (risk) criteria are present in the patient, they get points assigned. Eventual amount of points corresponds with a riskprobability or a threshold is set where a treatment might be indicated for the patient.

To add a linear model, go to Models and click on New Model. Go to the fourth tab called Model and the linear model is automatically selected. Make sure to enter an intercept or baseline and select if your variables are single values or multiplied. More info on how to edit or add the rest of a new model can be found here.

For this example a clinical prediction rule to identify patients that are most likely to benefit from spinal manipulation. The clinical prediction rule describes 5 criteria that have to be met. The criteria can be seen in the picture below:

In the picture we can see that 5 criteria are selected. In the paper describing the development of the clinical prediction rule, the authors state that at least 4 out of the 5 criteria should be met.

  • In this case, no baseline or intercept is needed, so we fill in 0. (this means that in the eventual calculation 0 will be added to the sum of the beta coefficients.
  • Go to the sub-tab variables and enter the five criteria that should be met. Set the beta coefficient of 1 if the criteria is met and 0 if it is not. (Example)
  • Go the fifth tab Result and add an outcome stratification where the outcome 4 or 5 shows the description that patients might benefit from spinal manipulation. Add another outcome stratification if the outcome lies between 0 and 3 for patient that might not benefit from spinal manipulation. (Example)
  • Make sure to add some pre- or post-result text so the user of the model can understand what the calculated outcome means. If everything is filled in, click on save model and show model to try out the model. A result might look like this
  • Always make sure to add a link to the description of the development of a model before saving a model to public. Reviewers will be able to verify if the model is working properly and result interpretation matches the original description.