search

alistairewj / pao2-fio2-prediction

Predict the PaO2/FiO2 ratio from non-laboratory derived data
3 stars 1 forks source link

Negin's feedback on the initial ventdurations dataset #3

Closed tompollard closed 6 years ago

tompollard commented 6 years ago

Feedback from Negin on the ventilation dataset extracted in: https://github.com/alistairewj/pao2-fio2-prediction/blob/master/ventduration/ventilation-durations.ipynb

We do not have a codebook for this data but I am assuming that ventnum is type of vent and '1' is AC/MV and '2' is NIMV? So in patient with unique id 015-63008 she was intubated for 5093 minutes, then on NIMV for 442 minutes and then reintubated for 2675 minutes?

If that's the case it makes sense.

On average ppl are intubated for 2 days with the extreme being 28 days? Also makes sense - although not sure if this just for ventnum 1?

Approximately 0.4% come in with a trach if trach_chronic_offset means this - which could be correct

20% mortality seems about right

Two points:

What stands out as potentially incorrect is the bedside_trach_offset_1st --- does this mean the number of ppl who get a new trach? If so I would guess this to be around 5-10% of people... it currently has 0 people

Can we get place of discharge? This would add meaning to whether they survived

Also will need: Admitting and dc diagnoses A marker of severity of illness Comorbidities Medication list in hospital (eventually - )

tompollard commented 6 years ago

As venturations have moved to the Northwell repo, I'll close this issue.