dsi-icl / optimise

OptimiseMS
https://optimise.dsi.ic.ac.uk
MIT License
1 stars 1 forks source link

Separate "No" and "Continue with missing data" options #1435

Open ewadding opened 2 years ago

ewadding commented 2 years ago

Currently, several forms that users are shown when entering a new visit have the option "No; or Continue with missing data". The problem with this is that we cannot see which visits have incomplete data, as there is no way to distinguish this from a genuine "No" response. This is an important issue from a data quality point of view.

Please can the forms be changed to have two separate options, "No" and "Continue with missing data".

We also need to amend the database to record which visits have missing data and which are genuine "No"s. I would suggest adding binary indicator fields such as completeRelapses, completeSAE (etc) to each Visit document, which would be set to FALSE if "Continue with missing data" is selected and to TRUE otherwise. These fields would of course need to be updated if the user returns later to fill in the missing data. This is just a suggestion - there may be easier/ more efficient solutions.

This will need to be done for each of the relevant forms (pages 3-10 of the new visit dialogue).

fguitton commented 2 years ago

Hello @ewadding, the button "No; or Continue with missing data" is a single choice to a Yes/No question. The verbiage was chosen to "scare" clinicians into entering data. In this context "No" means "Continue with missing data" Every bit of questionnaire is optional, in the same way it would be outside of the setup screens. This was extensively discussed at the time.

For every fields, if they are not present in the mongo object then they are missing and were never set by the user, is they are present but set to null or otherwise unreasonable values, then they were looked at or measured but explicitly set to be invalid or null. Such is the case for antibody presence/measure where a value of -1 represents that the test was done, but no antibody was found. This was also the result of extensive discussions at the time.

This said, if you think there is indeed a case were a bug leads to use missing this distinction, please let me know which field so we can try reproducing the problem and find a way to address it 👍🏼

ewadding commented 2 years ago

Hi Florian,

The issue is that we need to be able to tell whether absence of a particular type of item in the visit record means there was no item, or whether the data simply hasn't been entered yet.

Take form 3, relapses, for example. The software gives users the opportunity to edit the data at a later point in time, so in this context the option "No; or Continue with missing data" might mean either "there was no relapse" or "I'll fill it in later". We need confirmation of whether it's a true negative response or we simply have no idea how complete our data is. We might be missing a huge number of relapses without knowing it, and this has real implications for the integrity of the study. This has been extensively discussed as something that needs to be changed.