Instructions:

• From Perioperative Data Warehouse, insert data into the Data subfolder
• Run Notebook 1 then Notebook 2 and lastly Notebook 3
• Notebook 3 may require a GPU to make training quicker. Google Colab is a great option.

Structure of each Perioperative Data Warehouse file:

• flowsheets.csv:
  • Columns: HSH_ADMSN_ID, FLO_MEAS_NAME, NORMALIZED_NAME, NFLO_MEAS_ID, GROUPED_FLO_ID, MEAS_VALUE, UNITS, RECODED_TIME
• flowsheets pressures.csv:
  • Columns: HSH_ADMSN_ID, FLO_MEAS_NAME, NORMALIZED_NAME, NFLOW_MEAS_ID, GROUPED_FLO_ID, MEAS_VALUE, UNITS, RECODED_TIME, SYSTOLIC, DIASTOLIC, MAP
• Labs.csv:
  • Columns: HSH_ADMSN_ID, NORMALIZED_NAME, COMPONENT_ID, COMPONENT_NAME, ORD_VALUE, ORD_NUM_VALUE_CORRECTED, RESULT_TIME
• main.csv:
  • Columns: HSH_ADMSN_ID, ADMSN_MINUTES, ICU_MINS, DISCHARGE_MINUTES, IN_ICU_TIME, VERAPAMIL_TAKEN_TIME, BMI, AGE_LT90, SEX, ETHNICITY, RACE

Description of columns:

• HSH_ADMSN_ID: String with unique patient identifier
• FLO_MEAS_NAME: String with name of the flowsheet measure (from Epic)
• NORMALIZED_NAME: String with name of flowsheet measure that is standardized for groups of measures that should be averaged together
• NFLO_MEAS_ID: Integer ID uniquely identifying FLO_MEAS_NAME
• GROUPED_FLO_ID: Integer ID uniquely identifying NORMALIZED_NAME
• MEAS_VALUE: String raw value of output measure
• UNITS: String raw value of output units
• RECODED_TIME, RESULT_TIME: Time relative to ICU admission in minutes (integer)
• SYSTOLIC, DIASTOLIC: Integer for systolic and diastolic of pressure values
• ORD_VALUE: Value of lab
• ORD_VALUE_CORRECTED: Value of lab if there are any corrections that need to take place
• ADMSN_MINUTES: Integer time patient was admitted (should be 0)
• ICU_MINS: Integer time patient was admitted to ICU (relative to admission minutes)
• DISCHARGE_MINUTES: Integer time patient was discharged from ICU (relative to admission minutes)
• IN_ICU_TIME: Length of time patient was in ICU
• VERAPAMIL_TAKEN_TIME: Time (relative to admission) patient was administered verapamil. NaN/empty if not given
• BMI: BMI of patient (integer)
• AGE_LT90: 1 if patient is less than 90 years old, 0 otherwise.
• SEX: 'M' or 'F'
• ETHNICITY: String Ethnicity of patient
• RACE: String Race of patient