Early PT Consult Order Target Trial Emulation Project

CLIF VERSION

2.1.0

Objective

• Determine the effect of early PT consults on ICU outcomes for mechanically ventilated adults. Early is defined as within 48 hours of invasive mechanical ventilation or ICU admission (if transferred already intubated).
• Created using MIMIC IV Database converted to CLIF Format
• With early mobilization critaria algorithm taken from Eligibility for Mobilization Algorithm

Required CLIF tables and fields

1. patient: patient_id, race_category, ethnicity_category, sex_category, death_dttm
2. hospitalization: patient_id, hospitalization_id, admission_dttm, discharge_dttm,admission_category,discharge_category, age_at_admission
3. adt:hospitalization_id, in_dttm, out_dttm, location_category, location_type
4. vitals: hospitalization_id, recorded_dttm, vital_category, vital_value
   • vital_category = 'heart_rate', 'resp_rate', 'sbp', 'dbp', 'map', 'spo2', 'weight_kg'
5. labs: hospitalization_id, lab_result_dttm, lab_order_dttm, lab_category, lab_value, lab_value_numeric
   • lab_category = 'lactate', 'creatinine', 'bilirubin_total', 'po2_arterial', 'platelet_count'
6. medication_admin_continuous: hospitalization_id, admin_dttm, med_name, med_category, med_dose, med_dose_unit, med_group
   • med_category = 'norepinephrine', 'epinephrine', 'phenylephrine', 'vasopressin','dopamine', 'angiotensin', 'nicardipine', 'nitroprusside','clevidipine','cisatracurium','vecuronium','rocuronium','metaraminol','dobutamine'
7. respiratory_support: hospitalization_id, recorded_dttm, device_category, mode_category, tracheostomy, fio2_set, lpm_set, resp_rate_set, peep_set, resp_rate_obs
8. patient_assessments: hospitalization_id, recorded_dttm, assessment_category, numerical_value, categorical_value
   • assessment_category = 'braden_mobility', 'RASS', 'cam_total', 'gcs_total',
9. key_icu_orders: hospitalization_id,'order_dttm', 'order_category'

Cohort Identification

• Adults (age >= 18)
• On invasive mechanical ventiulation for at least 4 hours.
• Without a tracheostomy.
• Without a PT consult order in 24 hours prior to intubation.

Detailed Instructions for running the project

1. Requirements

The project requires Python 3.11+ with uv installed and R 4.x. The Jupyter notebooks are converted to just Python so Jupyter itself is not required. Uses UV and renv, respectively, for dependencies.

2. Download This Repository

3. Update Config:

Follow the instructions in config/README.md to set your site name, the path to your CLIF tables, the file type, and the time zone.

4. Run Pipeline

Run the entire pipeline using the commands.

chmod +x run_pipeline.sh   # make it executable (one time only)
bash run_pipeline.sh

Run it with bash (not source): the script uses set -euo pipefail, so if you source it, any failing step will exit your interactive shell/terminal. These scripts install the required Python and R dependencies.

Pipeline steps

Step	Script	Language	Description
1	1_cohort.py	Python	File organization, Cohort identification, STROBE diagram
2	2_data_gathering.py	Python	Gathers and aggregates data from multiple CLIF tables, creates "time_bin" and "hourly" data sets.
3	3_calculations.py	Python	Mobilization analysis, outcomes definitions
4	4_table_one.R	R	Table 1, Graphs, setup for CCW
5	5_ccw.R	R	Clone censor weight, outcomes models, bootstrapping

Output

We want the output saved to output/final and output/logs.

Authors

Giulio C. Rottaro Castejon

Jinping Liang

Haidong Lu

Fan Li

Snigdha Jain

Yale University