Truace
Subscribe

Machine learning prediction of weaning and ICU discharge in prone-position ventilation for ARDS secondary to pneumonia

Source article: Interpretable Machine Learning Predicts Successful Weaning and ICU discharge alive Following Prone Positioning Ventilation in Moderate-to-Severe Acute Respiratory Distress Syndrome (ARDS) Secondary to Pneumonia

Objective We developed and validated interpretable machine learning models predicting successful weaning and survival to ICU discharge in patients with moderate-to-severe acute respiratory distress syndrome (ARDS) secondary to pneumonia receiving prone position ventilation (PPV). Methods We included 131 patients with moderate-to-severe ARDS secondary to pneumonia undergoing PPV for internal, temporal, and external validation. Models were developed using least absolute shrinkage and selection operator (LASSO) reg…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1655· cite & verify
Interpretable Machine Learning Predicts Successful Weaning and ICU discharge alive Following Prone Positioning Ventilation in Moderate-to-Severe Acute Respiratory Distress Syndrome (ARDS) Secondary to Pneumonia
US Navy 030415-N-4182M-001 Intensive Care Unit (ICU) nurse Lt. Cmdr. Mary Ann Brantley, Nurse Crops, works on a patient in the ICU aboard the Military Sealift Command (MSC) hospital ship USNS Comfort (T-AH 20) by U.S. Navy photo by Journalist Seaman Erica Mater.. Public domain
Trace impact readingContested
P 68The P score combines the specificity and measured human impact of the grounded problem claim with the strength of this Trace’s cited sources.

Both sides are scored from claims and sources, not community votes.

G 70The G score combines the specificity and measured human impact of the grounded gain claim with the strength of this Trace’s cited sources.

In brief

Researchers built machine learning models using LASSO feature selection to predict two outcomes in 131 patients with moderate-to-severe ARDS from pneumonia who received prone position ventilation: successful weaning and survival to ICU discharge. The support vector machine performed best internally, and the weaning model kept similar discrimination when tested over time and at an external site.

The work matters because it points toward transparent, individualized risk estimates for a critically ill group, with SHAP explaining which factors drive predictions. The results are exploratory, however. The discharge model was much less stable, the explanations are not causal evidence, and prospective multi-center validation has not yet been done.

Main points

  1. The study included 131 patients with moderate-to-severe ARDS secondary to pneumonia who underwent prone position ventilation.
  2. The best internal AUCs for the SVM model were 0.715 for weaning and 0.807 for ICU discharge.
  3. Weaning predictors included septic shock, platelet count, PaO2/FiO2 ratio at PPV, neutrophil-to-lymphocyte ratio, and peak airway pressure at PPV.
  4. ICU discharge predictors included APACHE II, peak airway pressure at PPV, multidrug-resistant organism infection, and NLR.
  5. Decision curve analysis suggested potential utility across broad thresholds, and calibration was good for weaning but only acceptable for ICU discharge.

The gain

A support vector machine model for weaning held up across temporal and external validation and may help individualize risk stratification during ventilation and prone positioning, alongside clinical judgement.

The problem

The ICU discharge model was unstable in temporal and external validation, so its predictions need cautious interpretation and cannot yet support reliable use.

The rundown

The models were developed with LASSO regression and assessed with AUC, calibration, and decision curve analysis. Validation was internal, temporal, and external, and SHAP was used to assess interpretability.

For weaning, the temporal AUC was 0.841 and the external AUC was 0.707, which the authors describe as promising but exploratory. The model is framed as an adjunct for individualized risk stratification during the conventional ventilation and PPV phase, complementing rather than replacing clinical judgement.

For ICU discharge, temporal and external AUCs were 0.657 and 0.665, so the model was considered preliminary. The authors say SHAP improved transparency and generated hypotheses rather than causal evidence, and they call for multi-center prospective validation.

What this doesn’t fix

The ICU discharge model showed limited temporal and external stability, SHAP findings are hypothesis-generating rather than causal, and the authors state that multi-center prospective validation is still needed.

Sources

  1. Peer-reviewedRespiratory Medicine2026-10-09

The debate