Create date_onset Property in cardio_metabolic_disorders.py

[thewati]

Creating date_onset property for conditions in the cardio_metabolic_disorders module. Fixes issue #1845. The nc_diabetes_date_onset property will be used in issue #1595

[thewati]

I thought it would be better to just capture the date_onset of all conditions. Then I could just access nc_diabetes_date_diagnosis in the Diabetic Retinopathy module later.
In the cardio_metabolic_disorders module, people may undergo remission or lose a condition through idx_loses_condition. Once this happens I am not resetting nc_diabetes_date_diagnosis because I thought preserving the historical onset date even after the condition resolves is important as it may still pose a risk

[thewati]

Hi @mnjowe and @tbhallett,
This is ready for review

[mnjowe]

This is well noted @thewati. I will take a look

[mnjowe]

Hi @thewati. Looking at the use case of this PR, I think adding a property that captures date_onset of all condition will not be a good idea. The model is already heavy and I think we should exhaust all possible alternatives before we decide on introducing a new property. Also I think adding this date_onset property to all conditions while we are interested only in one condition will accumulate some unnecessary computational costs

In this case, how about we create a dictionary that maps diabetes individual Ids to their date of diabetes onset? This should only look into diabetes as we are not interested in any other conditions. That way, if you capture their date of onset in Diabetic Retinopathy you can delete the keys in the diabetes dictionary to make it as smaller as possible. What do you think?

@tbhallett any thoughts?

[thewati]

Hi @thewati. Looking at the use case of this PR, I think adding a property that captures date_onset of all condition will not be a good idea. The model is already heavy and I think we should exhaust all possible alternatives before we decide on introducing a new property. Also I think adding this date_onset property to all conditions while we are interested only in one condition will accumulate some unnecessary computational costs

In this case, how about we create a dictionary that maps diabetes individual Ids to their date of diabetes onset? This should only look into diabetes as we are not interested in any other conditions. That way, if you capture their date of onset in Diabetic Retinopathy you can delete the keys in the diabetes dictionary to make it as smaller as possible. What do you think?

@tbhallett any thoughts?

Thanks for pointing this out. I had coded it this way so that anybody can use it in future if needed and also to maintain consistency... But I agree, this will cause unnecessary overhead. I can change this to a dictionary that only stores the date of onset as you suggested

[thewati]

@mnjowe this is now using a dictionary. I'm done