Role of Attending Practice Variability in Prone Positioning Initiation: A Retrospective Cohort Study

Abstract

Background: Prone positioning is underused, despite mortality benefits. Prior studies highlight that patient-independent factors may influence prone positioning rates, but attending-specific contributions are unknown.

Research question: Does significant variability in prone positioning rates exist among attending physicians?

Study design and methods: This is a retrospective cohort study of 514 adults receiving mechanical ventilation in a tertiary-care medical or surgical ICU from January 1, 2015, through June 30, 2024. Inclusion criteria included Pao₂ to Fio₂ ratio of ≤ 150 with Fio₂ of ≥ 60% and positive end-expiratory pressure of ≥ 5 cm H₂O within 0 to 36 hours and 36 to 72 hours of intubation. The primary outcome was prone positioning within 72 hours of intubation or 24 hours of meeting prone positioning criteria. We hypothesized that attending variability was a significant predictor of prone positioning. We fit a mixed-effects logistic regression model to evaluate attending-level variability in prone positioning use, adjusting for 6 potential patient-centered prone positioning barriers and facilitators (age, BMI, COVID-19 status, code status, Pao₂ to Fio₂ ratio, and vasopressor use) and ICU location (medical or surgical).

Results: Among 514 patients eligible for prone positioning, 87 patients (17%) underwent prone positioning. Significant attending-level variability in prone positioning was noted among the 48 attendings included in the analysis, with risk- and reliability-adjusted rates ranging from 14.9% to 74.2% and a median OR among attending physicians of 2.6 (95% CI, 1.7-5.2). This effect size was associated more strongly with prone positioning than a 30-mm Hg decrease in Pao₂ to Fio₂ ratio. Even among patients with clinical documentation of ARDS on the day of prone positioning eligibility, the median OR among attending physicians was 2.4 (95% CI, 1.5-7.3). Additional patient factors predicting prone positioning included COVID-19 status, code status, and Pao₂ to Fio₂ ratio.

Interpretation: Our results show that large variation in prone positioning practices exists among attending providers, and future work should consider attending-focused and system-wide interventions as potential novel targets to improve prone positioning rates.

Keywords: ARDS; critical care delivery; mechanical ventilation; prone positioning; provider variability.

Figure 1 –. Diagram showing selection of the study cohort. ECMO = extracorporeal membrane oxygenation.

Figure 2 –. Graph showing the distribution of prone positioning use among 48 attending physicians. Prone positioning rate estimates are based on the mixed-effects logistic regression model with a random effect for attending physician on the day of meeting prone positioning criteria. Estimates were calculated for a hypothetical 59-year-old man with full code status, negative results for COVID-19, a BMI of 30 kg/m², Acute Physiology and Chronic Health Evaluation IV score of 76, and PaO₂ to FIO₂ ratio of 76; receiving pressors at a norepinephrine equivalent dose of ≤ 0.1 μg/kg/min; and cared for in a medical ICU at the time of meeting prone positioning criteria.

Figure 3 –. A, Graph showing the distribution of patients not placed in the prone position who met prone positioning criteria (red) and patients who were placed in the prone position (blue). B, Graph showing the quarterly rates of prone positioning across the study period, January 1, 2015 through June 30, 2024.