*** Source code for paper: Early prediction of level-of-care requirements in patients with COVID-19 ***
Due to the a restrictive data use agreement and HIPAA rules, we do not provide the data or part of the code for data pre-processing.


1. Hospitalization models training/testing

"Main_Hospitalization_RandomSplit_BWH.py":
Hospitalization models. By setting the parameter:

is_BWH = 1: using patients in BWH as test set;
is_BWH = 0: using random splitting to generate the test set.


2. ICU/Intubation models training/testing

"icu669_v04_20200912_4hour_binary_bwh.py":
ICU/Intubation models. By setting the parameters:

is_BWH = True: using patients in BWH as test set;
is_BWH = False: using random splitting to generate the test set;
col_y = 'flag_ventilator': intubation predictive model;
col_y = 'flag_ICU': ICU predictive model.


3. NLP for generating patients' profiles

"MGH_hnp(vis)_output_pmh(psh).py":
Extract past medical/surgical history from H&P/visit notes.

"MGH_hnp(vis)_output_med1~10.py":
Extract medication history using 10 formats from H&P/visit notes.

"MGH_hnp(vis)_output_symptom":
Extract symptoms from H&P/visit notes.

"MGH_hnp(vis)_output_vitals1~2":
Extract vitals using 2 strategies from H&P/visit notes.

"MGH_rad_output_find(imp).py":
Extracting radiology findings/impressions.

