Models for extraction of Social Determinants of Health from clinical notes

• RoBERTa NER: model-roberta-ner
• CNN NER: model-cnn-ner
• Hybrid BoW-GBM NER: model-hybrid-bow

Using models

It is advised to create a virtual environment for this package, e.g.: conda create -n sdoh python3.9 spacy

To install a model, e.g. model-hybrid-bow, run following in a shell:

pip install sklearn
python -m spacy download en_core_web_md

cd model-hybrid-bow/package/en_sdoh_bow_cui-0.0.2 && pip install . && cd -
# OR
cd model-cnn-ner/packages/en_sdoh_cnn_ner_cui-0.0.0 && pip install . && cd -

from within Python:

import spacy
nlp = spacy.load("en_sdoh_bow_cui")
text = "Home Environment:  Lives with family  Work: disability"
doc = nlp(text)

for ent in doc.ents:
    print(ent, ent.label_, ent._.score, ent._.cui, sep=" | ")

References

Automatic Extraction of Social Determinants of Health from Medical Notes of Chronic Lower Back Pain Patients Dmytro Lituiev, Benjamin Lacar, Sang Pak, Peter L Abramowitsch, Emilia De Marchis, Thomas Peterson

doi: https://doi.org/10.1101/2022.03.04.22271541