Weight change analysis

KNIME Analytics Platform
1

I have 30,000 patient visits and I am looking to find the weight change from the first to the last visit but I do not know how to remove patients with visits that a weight was not entered so it appears as a zero and the analysis reports that they lost all of their body weight. I am brand new at knime so this may seem like a simple question, but I have no idea how to manage it! Thank you in advance! 

I also have to remove patients with only one visit.

 

Karen

 

2

Hi Karen I used the Row Filter and set the "use range checking" lower bound and upper bound both to zero. 

I am attaching an example workflow 

Best Regards

 

 

3

Thank you! I was not able to see or Import the examples as I have a Mac and I have been having difficulty Importing examples.  If you can take a screenshot that would probably work better.  
I am also

trying to remove patients who have had only one visit so there would be only one uniqueID, the patients with multiple visits have multiple uniqueID's...any thoughts?

 

4

Hi Karen I uploaded a Power Point presentation in Google Drive which you can find in this

.

I am also uploading the workflow in other format

Let me know if it is ok

Best Regards

 

5

  Hi, I have installed KNIME 3.2, but I don't know how to download  installation guide  to see it.

"Double click" is not a way, Drag&Drop too as is written here: https://tech.knime.org/whats-new-in-knime-32#KNIMEFileExtension. Do anybody know what i do wrong?

6

If the weight was not entered, I assume there is a missing value in the corresponding field ? If yes, you can use the `Missing Value` node to filter them out.

Patients having only one visit (assuming granularity = 1 row per visit per patient): 

`GroupBy(patient id) to count the number of visits -> Joiner on patient id with the original table -> Row Filter excluding occurrence of 1 in the number of visits`

There's nothing wrong kicking out missing value observations, however, you may have other non-missing variables of interest for such visits. In that case it may be reasonable to assume that the weight compared to the last visit did not change (whispering: significantly), thus impute it using the previous visit of the same patient when available and mark the observation as imputed:

`Sorter (patient id, date) -> Group Loop Start (on patient id) -> Lag (weight) -> Loop End -> Rule Engine to mark the to be soon imputed values -> Rule Engine to impute weight`

As a bonus, the analysis will be able to report which (measurable) circumstances have lead to the weight of the patients not being recorded.

7