Truace
Subscribe

AI model predicts constipation in stroke ICU patients on tube feeding, but only with full-stay data

Study title: Clinical Phenotypes, Machine-Learning Risk Stratification and Treatment Outcomes of Constipation in Stroke Patients Receiving Enteral Nutrition: A Retrospective Study

Researchers retrospectively analysed records of 362 stroke patients treated in a neurological intensive care unit between 2019 and 2023, all receiving enteral nutrition. About 64% had documented constipation. Six algorithms were compared, and the TabPFN-based model discriminated best.

Read the study abstract

Background: Stroke-related complications and frequent feeding interruptions often compromise nutritional intake, making enteral nutrition (EN) a cornerstone therapy to improve nutritional status and support recovery. Nevertheless, constipation is highly prevalent and clinically consequential among stroke patients receiving EN, particularly in critically ill populations, where enteral nutrition-associated constipation can impede rehabilitation, increase complications and prolong intensive care unit (ICU) stay. Ai…

By the Truvace desk1 min read1 sourceEvidence: Moderate · 60/100 — Observational study · n = 362 · 1 independent sourcePeer-reviewedTRV-2026-2313· cite & verify
Clinical Phenotypes, Machine-Learning Risk Stratification and Treatment Outcomes of Constipation in Stroke Patients Receiving Enteral Nutrition: A Retrospective Study
Construction site tour of the Skilled Nursing Residence, in Edgartown, on the island of Martha's Vineyard, in Dukes County, Maryland on August 5, 2024 - 12 by USDAgov. Public domain

In brief

The findings suggest bowel risk should be reassessed repeatedly during intensive care rather than judged once, and they support developing dynamic, time-updated tools. The model was only internally validated on retrospective data, so its value for real-time bedside use has not been shown.

Main points

  1. Of 362 stroke patients, 233 (64.4%) had documented constipation.
  2. Table PFN-3 performed best; ROC-AUC was 0.629 with temporally restricted data and 0.771 with full-hospitalisation data.
  3. ROC-AUC rose by 0.142 once completed hospital-course variables were included.
  4. K-means clustering found three subgroups, with constipation frequencies of 67.35%, 56.98% and 88.89%.
  5. Observed relief was 80.3% with glycerin enema, 60.0% with lactulose, 72.7% with combined treatment and 65.7% with neither; these comparisons were descriptive.

The gain

With full-hospitalisation data, a machine-learning model (Table PFN-3) predicted constipation in stroke ICU patients on tube feeding with good discrimination (ROC-AUC 0.771) and close calibration.

The rundown

Patients came from a single neurological ICU. Six models were compared: logistic regression, random forest, gradient boosting tree, XGBoost, LightGBM and Table PFN-3. Evaluation used five repeats of nested five-fold cross-validation, with discrimination, precision, calibration and decision-curve analysis.

Models were run under a temporally restricted feature set and an expanded retrospective set. Discrimination was modest early (ROC-AUC 0.629) and much better with data from the whole hospital stay, showing that predictive information accumulates as admission progresses.

Clustering identified three clinical phenotypes, one with an 88.89% constipation frequency, described as a prolonged, treatment-intensive trajectory. Treatment results were summarised only descriptively, without controlled comparison.

Sources

  1. Peer-reviewedNursing in Critical Care2026-11-01

The debate