Prospective Multicentre Validation of a Landmark Machine Learning Model for ICU Mortality in Neurocritical Care (LANDMARK-ICU)

November 11, 2026 updated by: Han Chen, Fujian Provincial Hospital

A Prospective Multicentre Observational Study Validating a Landmark-Based Machine Learning Model for Dynamic Prediction of Mortality in Patients Receiving Continuous Intracranial Pressure Monitoring

Mortality prediction in neurocritical care relies on scores computed at admission, which cannot update as the clinical course evolves. We previously developed a landmark-based machine learning model that converts hourly physiological data, including intracranial pressure, into an updating risk of death. The model was developed in a large public database and validated retrospectively in a second, independent one. This prospective multicentre study will test the model in consecutive patients at participating hospitals in Fujian Province, China. Physiological data will be collected as part of routine care. The model will generate risk estimates that are recorded for analysis but not shown to treating clinicians, and no treatment decision will be influenced by the study. The primary analysis will assess discrimination and calibration of the model for death during the intensive care stay. Secondary analyses will compare the model with severity scores in routine use and examine performance by aetiology.

Study Overview

Status

Not yet recruiting

Conditions

Intervention / Treatment

Detailed Description

Model development used the Medical Information Mart for Intensive Care IV database and external validation used the eICU Collaborative Research Database. Both analyses were retrospective and are not part of this registration. This registration covers the prospective multicentre validation only. Inclusion requires age 18 years or older, continuous intracranial pressure monitoring for at least 24 hours, and a diagnosis of traumatic brain injury, aneurysmal subarachnoid haemorrhage or intracerebral haemorrhage. The model uses seven hourly physiological channels with observation masks and elapsed-time features. The primary endpoint is death during the intensive care stay; 28-day mortality, length of stay and Glasgow Outcome Scale at 6 months are secondary. The sample size of 100 patients follows the approved protocol, which assumed area under the curve 0.80 with alpha 0.05 and beta 0.20.

Study Type

Observational

Enrollment (Estimated)

100

Contacts and Locations

This section provides the contact details for those conducting the study, and information on where this study is being conducted.

Study Locations

    • Fujian
      • Fuzhou, Fujian, China
        • Fujian Provincial Hospital
      • Fuzhou, Fujian, China
        • First Affiliated Hospital of Fujian Medical University
      • Quanzhou, Fujian, China
        • Hui'an County Hospital
      • Quanzhou, Fujian, China
        • Second Affiliated Hospital of Fujian Medical University
      • Quanzhou, Fujian, China
        • Shishi City Hospital
      • Zhangzhou, Fujian, China
        • Zhangzhou Municipal Hospital

Participation Criteria

Researchers look for people who fit a certain description, called eligibility criteria. Some examples of these criteria are a person's general health condition or prior treatments.

Eligibility Criteria

Ages Eligible for Study

18 years to 90 years (Adult, Older Adult)

Accepts Healthy Volunteers

No

Genders Eligible for Study

All

Sampling Method

Probability Sample

Study Population

Adults in intensive care units of participating hospitals who require continuous intracranial pressure monitoring after acute brain injury.

Description

Inclusion Criteria:

  • Age 18 years or older
  • Continuous intracranial pressure monitoring for at least 24 hours
  • Diagnosis of traumatic brain injury, aneurysmal subarachnoid haemorrhage or intracerebral haemorrhage
  • Written informed consent from the patient or a legal representative

Exclusion Criteria:

  • Incomplete or unusable intracranial pressure recordings
  • Loss to follow-up or unavailable outcome information
  • Pre-existing severe neurological disease that would confound assessment

Study Plan

This section provides details of the study plan, including how the study is designed and what the study is measuring.

How is the study designed?

Design Details

  • Observational Models: COHORT
  • Time Perspectives: PROSPECTIVE

Cohorts and Interventions

Group / Cohort
Intervention / Treatment
Neurocritical care cohort

Consecutive patients receiving continuous intracranial pressure monitoring, contributing to the risk set at each landmark.

Observational application of a fixed prediction model to hourly physiological data; no treatment is assigned or withheld.

What is the study measuring?

Primary Outcome Measures

Outcome Measure
Measure Description
Time Frame
Area under the receiver operating characteristic curve (range 0 to 1) for death during the intensive care stay at the 24-hour landmark
Time Frame: 24 hours after intensive care admission, with the outcome followed to intensive care unit discharge or day 28

Discrimination of the locked landmark model, quantified as the area under the receiver operating characteristic curve, where 0.5 indicates no discrimination and 1.0 perfect discrimination. Death is ascertained from the medical record and the curve is estimated with a 95% confidence interval

24 hours after intensive care admission, with the outcome followed to intensive care unit discharge or day 28

Secondary Outcome Measures

Outcome Measure
Measure Description
Time Frame
Area under the receiver operating characteristic curve (range 0 to 1) for death during the intensive care stay at the 48-hour and 72-hour landmarks
Time Frame: 48 and 72 hours after intensive care admission, with the outcome followed to intensive care unit discharge or day 28

Same measurement as the primary outcome, estimated separately at the 48-hour and 72-hour landmarks with 95% confidence intervals

48 and 72 hours after intensive care admission, with the outcome followed to intensive care unit discharge or day 28
Calibration slope (dimensionless) of the landmark model at the 24-, 48- and 72-hour landmarks
Time Frame: 24, 48 and 72 hours after intensive care admission

Slope obtained by logistic regression of the observed outcome on the logit of the predicted risk; a value of 1 indicates perfect calibration, below 1 indicates over-fitting and above 1 under-fitting.

24, 48 and 72 hours after intensive care admission
Brier score (range 0 to 1) of the landmark model at the 24-, 48- and 72-hour landmarks
Time Frame: 24, 48 and 72 hours after intensive care admission

Mean squared difference between predicted risk and observed outcome, where lower values indicate better overall performance.

24, 48 and 72 hours after intensive care admission
Difference in area under the receiver operating characteristic curve (range -1 to 1) between the landmark model and the centre-recorded APACHE II score
Time Frame: 24, 48 and 72 hours after intensive care admission

Paired difference in discrimination, computed as model area under the curve minus APACHE II area under the curve on the same patients, with 95% confidence interval; negative values indicate the model discriminates worse than APACHE II

24, 48 and 72 hours after intensive care admission
Net benefit (dimensionless) of the landmark model from decision-curve analysis at the 24-hour landmark
Time Frame: 24 hours after intensive care admission

Net benefit computed across threshold probabilities from 0.05 to 0.50, compared with the strategies of treating all patients and treating none

24 hours after intensive care admission
Area under the receiver operating characteristic curve (range 0 to 1) by primary diagnosis at the 24-, 48- and 72-hour landmarks
Time Frame: 24, 48 and 72 hours after intensive care admission

Discrimination estimated separately for traumatic brain injury, aneurysmal subarachnoid haemorrhage and intracerebral haemorrhage, reported descriptively because stratum sizes are expected to be small

24, 48 and 72 hours after intensive care admission
28-day all-cause mortality, as specified in the approved protocol
Time Frame: 28 days after intensive care admission
28 days after intensive care admission
Length of intensive care stay
Time Frame: From intensive care admission to intensive care unit discharge, up to 28 days

Duration of intensive care stay in days, right-censored at day 28 for patients still in intensive care

From intensive care admission to intensive care unit discharge, up to 28 days
Glasgow Outcome Scale score (range 1 to 5, higher is better) at 6 months
Time Frame: 6 months after intensive care admission

Glasgow Outcome Scale, an ordinal scale from 1 (death) to 5 (good recovery), with higher scores indicating better outcome; assessed by telephone or outpatient follow-up

6 months after intensive care admission

Collaborators and Investigators

This is where you will find people and organizations involved with this study.

Sponsor

Collaborators

Investigators

  • Principal Investigator: Han Chen, MD, Fujian Provincial Hospital

Study record dates

These dates track the progress of study record and summary results submissions to ClinicalTrials.gov. Study records and reported results are reviewed by the National Library of Medicine (NLM) to make sure they meet specific quality control standards before being posted on the public website.

Study Major Dates

Study Start (Estimated)

December 1, 2026

Primary Completion (Estimated)

December 12, 2027

Study Completion (Estimated)

January 1, 2028

Study Registration Dates

First Submitted

November 7, 2026

First Submitted That Met QC Criteria

November 11, 2026

First Posted (Actual)

November 17, 2026

Study Record Updates

Last Update Posted (Actual)

November 17, 2026

Last Update Submitted That Met QC Criteria

November 11, 2026

Last Verified

November 1, 2026

More Information

Terms related to this study

Other Study ID Numbers

  • FJPH-ICP-01

Plan for Individual participant data (IPD)

Plan to Share Individual Participant Data (IPD)?

NO

IPD Plan Description

De-identified individual participant data will not be shared publicly because small centre volumes could permit re-identification. The model specification, analysis code and summary results will be published.

Drug and device information, study documents

Studies a U.S. FDA-regulated drug product

No

Studies a U.S. FDA-regulated device product

No

This information was retrieved directly from the website clinicaltrials.gov without any changes. If you have any requests to change, remove or update your study details, please contact register@clinicaltrials.gov. As soon as a change is implemented on clinicaltrials.gov, this will be updated automatically on our website as well.