Composite model

A composite model is a model that calculates multiple outcomes based on a single input. In this way, it is possible to use multiple models at once. This detailed guide demonstrates exactly how to add composite models and how to use them.
To create a new composite model, make sure that you're signed in. Then click on the tab Models and on New Composite Model.

This is the screen you always start at when creating new models or editting existing models.
Screenshot of the page:

Title:
Start your model with filling in the title. The title is the only required field necessary for saving a model.

Author:
After saving the model at least one time, the author of the model is the user that is signed in.

Status:
There are three possibilities to choose:

  • Draft
  • Private
  • Public

If the model is saved on Public, an Evidencio employee gets a review request for that model. If this was not the intention, the review request can be withdrawn by saving the model on either Draft or Private. Make sure that the composite model is working properly before it is saved on Public.

Sharing:
It is possible to share your model with connections on Evidencio, this way you can let connections (such as co-workers) help you with the model-editing or let a connection review your model before saving it to public. It is necessary to save your model at least one time before sharing is possible.

Description:
This field is meant to describe the use of the model. It will often be the first thing anyone sees when they open a model. This field can give some context regarding the model or give some instructions about how to use the model.

Language:
Select the language you use for the creation of the model.
note: Make sure that the chosen language corresponds with the language of the model you want to combine in the composite model.

Specialties:
Please select the specialty that is most related to the topic of your model since it is allowed to select only one specialty.

MeSH terms:
Start typing in the MeSH term field and select multiple terms that belong to the topic of your selected models.

If you filled in all the fields, make sure to save your model and then go on to the next screen.

Screenshot of the page: , screenshot of usting the search bar: after adding models:
This tab shows a search bar. Simply type the keywords of the models you want to add in the composition and click on Add. Once a model was added, it will be removed from the search engine to avoid duplicate models. Select the models and click on clear in the searchbar to see the list of models you've added.
Clicking on the bin button removes the model from your list.
Click and hold the drag button to drag a model and change the order of models that will be shown on the eventual model page. (For example: it is neater to show the order of survival over 1, 5 and 10 years in that order than to show the survival over 5 years, then 1 year and then 10 years.).

If all the desired models are added, save model and go to the next page.

Screenshot of the page: and when copying variables
In this tab, all the necessary variables can be added or copied from existing models. Make sure that all the variables in all added models are matched to the variables of the composite model to avoid errors in the calculation. Easiest way to add the variables is to start at the first model you've added in the Variables tab and click on the Copy all button. Then go to the next models step-by-step and match their variables with the onces you already copied. If you have a new variable, simply click on copy to add it to your composite model input.
In some cases, conversions of the variables are needed to make a single input compute multiple outcomes on different models. An example can be the length of an individual. If a model uses length in centimeters and another model uses length in meters, on of these two models needs a conversion. Also, some models use age as a categorical variable (i.e. <30 years or 30+ years) and others use it as a continuous variable. Make sure to convert it into a single input in your composite model to avoid duplicate input fields.

If all variables are copied and matched correctly, make sure to save model and go on to the last screen.

Screenshot of the page:
In this page you're able to add result interpretation to clarify the prediction(s) made in the composite model. For every added the model inside the composite model, there will be a link to the original model. So a very detailed explanation of all the separate models is not necessary.

The option Result range is possible to tick, this will show a range of minimum and maximum calculated risks. This is handy for composite models that predict the risk of a similar event on different models. There are for instance several models that all predict the probability of lymph node involvement in prostate cancer patients that are using different formulas. Therefore, the predicted risks might differ per model for the same individual. With the result range the estimation of the risk can be given clearly. (i.e. calculated risk for lymph node involvement is 5%-12%).

Conditional information can be given based on several input paramaters or outcomes. This is especially handy if input parameters can be contradictory. For instance if weight and BMI are inputs for two seperate models but are both in entered in the composite model as input variables. Then the input of 150kg for weight does not match a BMI of 20kg/m2, simply because the individual would have to be 2.74 meters long. Of course, this type of errors can be avoided using formula segments in the variables section to make sure only one input for either weight/length/BMI is necessary.

There are several other occasions in which conditional information can provide valuable context. Some models use thresholds to perform or omit surgical procedures or other treatment possibilities. These recommendations can be shown in the conditional information.

If everything is filled in to your satisfaction. Click on Save model and then Show model to try out your newly created composite model. If everything seems to work fine, you might want to save the model Publicly.