a b/docs/utilities/connectors/omop.md 

  1
  
# OMOP Connector





  
  

  2
  





  
  

  3
  
We provide a connector between OMOP-formatted dataframes and spaCy documents.





  
  

  4
  





  
  

  5
  
## OMOP-style dataframes





  
  

  6
  





  
  

  7
  
Consider a corpus of just one document:





  
  

  8
  





  
  

  9
  
```





  
  

  10
  
Le patient est admis pour une pneumopathie au coronavirus.





  
  

  11
  
On lui prescrit du paracétamol.





  
  

  12
  
```





  
  

  13
  





  
  

  14
  
And its OMOP-style representation, separated in two tables `note` and `note_nlp` (here with selected columns) :





  
  

  15
  





  
  

  16
  
`note`:





  
  

  17
  





  
  

  18
  
| note_id | note_text                                     | note_datetime |





  
  

  19
  
| ------: | :-------------------------------------------- | :------------ |





  
  

  20
  
|       0 | Le patient est admis pour une pneumopathie... | 2021-10-23    |





  
  

  21
  





  
  

  22
  
`note_nlp`:





  
  

  23
  





  
  

  24
  
| note_nlp_id | note_id | start_char | end_char | note_nlp_source_value | lexical_variant |





  
  

  25
  
| ----------: | ------: | ---------: | -------: | :-------------------- | :-------------- |





  
  

  26
  
|           0 |       0 |         46 |       57 | disease               | coronavirus     |





  
  

  27
  
|           1 |       0 |         77 |       88 | drug                  | paracétamol     |





  
  

  28
  





  
  

  29
  
## Using the connector





  
  

  30
  





  
  

  31
  
The following snippet expects the tables `note` and `note_nlp` to be already defined (eg through PySpark's `toPandas()` method).





  
  

  32
  





  
  

  33
  
```{ .python .no-check }





  
  

  34
  
import spacy





  
  

  35
  
from edsnlp.connectors.omop import OmopConnector





  
  

  36
  





  
  

  37
  
# Instantiate a spacy pipeline





  
  

  38
  
nlp = spacy.blank("eds")





  
  

  39
  





  
  

  40
  
# Instantiate the connector





  
  

  41
  
connector = OmopConnector(nlp)





  
  

  42
  





  
  

  43
  
# Convert OMOP tables (note and note_nlp) to a list of documents





  
  

  44
  
docs = connector.omop2docs(note, note_nlp)





  
  

  45
  
doc = docs[0]





  
  

  46
  





  
  

  47
  
doc.ents





  
  

  48
  
# Out: [coronavirus, paracétamol]





  
  

  49
  





  
  

  50
  
doc.ents[0].label_





  
  

  51
  
# Out: 'disease'





  
  

  52
  





  
  

  53
  
doc.text == note.loc[0].note_text





  
  

  54
  
# Out: True





  
  

  55
  
```





  
  

  56
  





  
  

  57
  
The object `docs` now contains a list of documents that reflects the information contained in the OMOP-formatted dataframes.