0_0_Drugs_ATC_Parser.ipynb 4.6 KB
# Challenge - [ED Lab Prediction]
_Nom à trouver_
## Objectif
Ce notebook récupère la classifications des médicaments (ATC-IV) à partir des API publiques de la `national library of medecine`.
```
This product uses publicly available data from the U.S. National Library of Medicine (NLM), National Institutes of Health, Department of Health and Human Services; NLM is not responsible for the product and does not endorse or recommend this or any other product.
```
# Récupération des données
import sqlite3
import pandas as pd
from xml.etree import ElementTree as ETconn = sqlite3.connect("./data/mimic-iv.sqlite")# Liste des traitements
drugs_gsn = pd.read_sql("""
SELECT gsn, ndc, name
FROM medrecon
GROUP BY gsn
""", conn)# Création de la liste pour utilisation de l'outils RxMix
# https://mor.nlm.nih.gov/RxMix/
## Utilisation de RxMix pour ne pas sur-solliciter les serveurs de la NML à travers des queries unitaires
drugs_gsn["gsn"].to_csv("./config/gsn_for_batch.txt", header=False, index=False)# Récupération des classes ATC associées
atccode = ET.parse("./config/atccode.xml")
atccode_parsed = [
(int(x[0].text),
[
y.text
for y in x[2].findall(".//classId")
if y.text not in ["-",""]
]) for x in atccode.getroot() if len(x) > 2
]n_found = len([x for x in atccode_parsed if len(x[1])>0])
n_total = drugs_gsn.shape[0]
print(f"Identification de {n_found} sur {n_total} ({n_found/n_total:.2f}) code NDC avec un code rxCUI associé.")Identification de 5906 sur 9262 (0.64) code NDC avec un code rxCUI associé.
atccode_parsed_filtered = dict([x for x in atccode_parsed if len(x[1])>0])
atccode_parsed_filtered_df = pd.DataFrame.from_dict(atccode_parsed_filtered, orient="index") \
.apply(lambda x: x.dropna().tolist(), axis=1) \
.reset_index() \
.rename(columns={"index":"gsn", 0:"atc"})
# Réunion de NDC et ATC
drugs_atc = atccode_parsed_filtered_df \
.explode("atc") \
.dropna() \
.drop_duplicates(["gsn","atc"])[["gsn","atc"]]drugs_atc.to_csv("./config/atc_items.csv", index=False)