WITHDRAWN: Estimating the Effect of Increasing the PF Ratio on Survival in ICU Patients with Acute Respiratory Distress: A Clinical Application of Longitudinal Modified Treatment Policies

• Activity Number: 89 - Nonparametric Methods for Modern Data
• Type: Contributed
• Date/Time: Monday, August 9, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Katherine Hoffman and Nicholas Williams and Ivan Diaz
• Companies: Weill Cornell Medicine and Weill Cornell Medicine and Weill Cornell Medical College
• Keywords: causal inference; longitudinal exposure; non parametric; critical care; time dependent; time to event

Abstract

The arterial partial pressure of oxygen to fraction of inspired oxygen (PaO2 to Fio2, or PF) ratio is a continuous measure of hypoxemia used to quantify the severity of Acute Respiratory Distress Syndrome (ARDS). Our goal was to estimate the overall effect on survival of an intervention that would increase the PF ratio by 50 among patients with clinically defined ARDS (PF ratio < 300). Our cohort contained 10,044 intubated intensive care unit (ICU) patients. The outcome variable was 14-day survival from the time of intubation. Baseline confounders include age, sex, race, number of Elixhauser comorbidities, and pneumonia status. Time-dependent confounders include daily mechanical ventilation status (invasive or non-invasive) and the Sequential Organ Failure Assessment (SOFA) score with the pulmonary component removed. We censored all patients at their day of hospital discharge. We estimated the effect of our Longitudinal Modified Treatment Policy (LMTP) non-parametrically using sequentially doubly robust (SDR) estimation with Super Learner candidate estimators. Our results estimated the effect of our LMTP to increase 14-day survival by 1.5% (95% CI: -0.1, 3.0%), from 88.3 to 89.8%.