NCT05471193: PRECAIN

Prediction of Cardiac Instability in Intensive Care

Completed Last updated 17 August 2022

What this trial tests

trial testing Machine Learning Prediction in Hemodynamics in 3,069 participants. Completed in 31 July 2022.

Timeline

1 June 2022

Primary endpoint
31 July 2022

31 July 2022

Quick facts

Lead sponsor	Kepler University Hospital
Status	Completed
Study type	OBSERVATIONAL
Enrollment	3,069
Start date	1 June 2022
Primary completion	31 July 2022
Estimated completion	31 July 2022
Sites	1 location across Austria

Drugs / interventions tested

• Machine Learning Prediction

Conditions studied

• Hemodynamics — all drugs for Hemodynamics →

Sponsor

Kepler University Hospital

Who can join

18 and older, any sex, with Hemodynamics. Patients with the condition only — healthy volunteers not accepted.

Sponsor's own description

A large number of different organ functions are recorded in real time for patients who are monitored in an intensive care unit. On the one hand, the measured values collected in this way are used for continuous monitoring of vital parameters, but they are also evaluated several times a day in order to be able to make decisions regarding further diagnostics and therapy. In the first case, threshold values can be defined, and if these are exceeded or fallen short of, the treatment team is automatically alerted. If these limits are set too liberally, then the alert will only indicate an acute risk to the patient, where extensive pathophysiological changes have already occurred. If the limits are chosen too restrictively, then there are frequent false alarms, since the limits are exceeded in most cases due to natural fluctuation, without this having any pathological value. The consequence is a so-called "alarm fatigue", which in the worst case leads to ignoring correct alarms and thus endangers the patients. By design, all of these readings only show the status quo of a patient. It is the task of the treatment team to predict from the course of these readings whether a threatening situation is developing for the patient.

For daily clinical practice, it would be better if dangerous changes in vital signs could be predicted. In this case, it would be possible to intervene therapeutically not only when a dangerous situation has arisen, but to try to avert this situation through adequate measures by changing the therapy strategy. In such a case, the treatment team would no longer be confronted with emergency alarms, but could counteract an impending deterioration with a long lead time.

The first approaches for detecting a drop in blood pressure, for example, which are based on simple models, are already in clinical use.

Publications & conference data

1 peer-reviewed publication reference this trial (live from Europe PMC):

1. Mean arterial pressure is all you need in a machine learning model for mean arterial pressure prediction.
   Tschoellitsch T, Kaltenleithner S, Maletzky A, Moser P, et al · · 2025 · PMID 40726206 · DOI 10.1097/eja.0000000000002238