OHDSI Home | Forums | Wiki | Github

How to handle missing values in EHR data analysis?


(Selva) #1

Hello Everyone,

I am currently working on a project where I would like to know the different measurements/lab tests that a person with specific condition should have underwent? I am not from healthcare background.so, your inputs would be helpful

I understand not all patients with condition A would have gone through the required measurement test. But what I would like to know is the list of measurement / lab tests for each condition? This is to find a way to handle missing values

For example,

  1. what are all the possible measurements/lab tests for T2DM? What are the measurements/lab tests that he has actually underwent for T2DM? in this I will be able to identify MACR, MNACR etc

  2. what are all the possible measurements/lab tests for cancer? What measurement tests did he actually undergo?

Once I have this,then I might be able to know about MACR, MNACR etc

Is there anyway in Atlas or SQL approach that I can use to get this info? I am sure it’s a common problem in EHR data analysis and you would have handled this. Can you share your experience of handling this in simplest way as possible?


(Christian Reich) #2


Doesn’t work that way. Diagnostic tests and figuring out the diagnosis is a many-to-many complex non-linear problem. Some tests results have a different meaning given other preconditions or treatments. Some tests are screening tests, while others are confirmation tests. Some tests give a generic indication to search in a certain direction.


What is “MACR” and “MNACR”?

(Selva) #3

@Christian_Reich - Hi, thanks for the response.

Exactly that’s why I would like to know how people usually handle it.

Apologies the short form should be as follows

MAR - missing at random

MCAR - missing completely at random

MNAR - missing not at random