PPMI Clinical Codes: Invitation to Collaborate and Request for Suggestions

Analyzing and Reusing Data
, , ,
1

Hello everyone!

Some of you already know the GitHub repository with Jupyter notebooks based on PPMI data. These notebooks provide (1) LEDD and medication specific LEDD calculations, (2) assessment of levodopa challenge response, (3) identification of medical conditions, and (4) identification of medication use. This work was my contribution to the Data Modality Taskforce this year, and I posted it on Discourse in February. For those who have not seen it, I am pleased to share it here.

Over this year and the next, I plan to keep expanding these notebooks. The goal is to add data transformations on top of existing PPMI clinical data, producing derived variables that are ready to use in analyses. Researchers can treat these outputs as outcomes or covariates of interest. For example, a transcriptomics study could test whether identified patient clusters are associated with levodopa responsiveness (a variable that needs to be calculated and transformed from raw PPMI data).

So I am posting for two reasons. First, to share the planned new features for these notebooks. Second, to invite anyone interested to contribute scripts for PPMI clinical data transformations they consider relevant and would like to share.

Here is what I plan to add over the next few months:

  1. Longitudinal tracking of DBS surgery and time to procedure
  2. Longitudinal tracking of death among patients
  3. Core clinical transformations and curation by milestone
  4. Longitudinal patient and clinician changes in global impression scales

So, my dear post reader, if you have scripts that fit this effort and are willing to share, or if you do not have scripts but have ideas for useful PPMI clinical transformations I should add, please reach out to me!

4 Likes
2

Me! I am your use case of a researcher analyzing transcriptomic data that would love to look at these covariates.

Thank you for the update. This is super useful! I would be happy to do some testing on the code and provide feedback.

3 Likes
3

Dear Daniel,

What about including clinically minimal important difference (MCID)? There are a lot of papers about that, and I’ve recently run an analysis on MCID and DBS follow-up.

When using clinical instruments, such as questionnaires and scales, MCID can better capture clinically significant changes instead of raw numerical data, thus projecting and explaining what’s meaningful from the patients’ perspective.

I’d love to hear what you think about it!

I could share my scripts on this, as well as some scripts for calculating motor subtype from MDS-UPDRS II & III :slight_smile:

I’ll keep thinking about which other scripts would be useful for this!

Ana

3 Likes
4

Hi Ana!

For sure, your suggestion is welcome! I think those are also important definitions and having them in those codes is good addition! Please, reach out to me for us to organize this and, of course, give you proper credit for your contribution!

1 Like
5

This is great @danieltds , I really enjoyed reviewing your notebooks, and I am happy to do it again. One minor suggestion: consider using notebook mainly as a demo, rather than placing all the codes there. You could save main python functions in separate scripts and simply import the into the notebook.. This design would make it easier for users to adapt and build on your work and also the notebooks will be shorter and simpler.

1 Like
6

Thanks for the suggestion, Hirotaka! I will look forward to this possibility as well!