EarlGlynn / PhysioNet-Sepsis-Challenge

PhysioNet 2019 Challenge: Early Prediction of Sepsis from Clinical Data

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

PhysioNet 2019 Sepsis Challenge

Files related to Early Prediction of Sepsis from Clinical Data: the PhysioNet/Computing in Cardiology Challenge 2019.

When combined into a single dataset, the 5000 files contain 188,453 records with 42 fields.

Missingness

Fields always present: ICULOS, Age, Gender, HospAdmTime, SepsisLabel

Fields with < 15% missing values: HR, O2Sat, SBP, MAP, DBP

Fields with 20% to 90% missing: Temp, Resp, Glucose, Unit1, Unit2

All other fields have > 90% missing values.

The median patient percent missing is 100% for 16 of the quantities.

Demographics

Only about 1.4% of raw records indicate sepsis. [2623 / 188,453]

Only about 5.6% of the patients have sepsis. [279 / 5000]

Sepsis reported in first hour for 20.1% of sepsis patients. [56 / 279]

First Hour Sepsis Indicated

Vital Signs

Heart Rate and Sepsis

Laboratory Value

Blood Urea Nitrogen and Sepsis

About

PhysioNet 2019 Challenge: Early Prediction of Sepsis from Clinical Data


Languages

Language:HTML 98.8%Language:Jupyter Notebook 1.2%