Deep Learning Models for Prediction of Intraoperative Hypotension Using Non-invasive Parameters

Samsung Medical Center

Status: Completed

Conditions: Intraoperative Hypotension; General Anesthesia

Study type: Observational

Funder types: Other

Identifiers

NCT05762237

SMC 2022-09-096

Details and patient eligibility

About

The investigators aimed to investigate the deep learning model to predict intraoperative hypotension using non-invasive monitoring parameters.

Full description

Intraoperative hypotension is associated with various postoperative complications such as acute kidney injury. Therefore, precise prediction and prompt treatment of intraoperative hypotension are important. However, it is difficult to accurately predict intraoperative hypotension based on the anesthesiologists' experience and intuition. Recently, deep learning algorithms using invasive arterial pressure monitoring showed the good predictive ability of intraoperative hypotension. It can help the clinician's decisions. However, most patients undergoing general surgery are monitored by non-invasive parameters. Therefore, the investigators investigate the prediction model for intraoperative hypotension using non-invasive monitoring.

Enrollment: 5,175 patients

Sex: All

Volunteers: No Healthy Volunteers

Inclusion criteria

• The patients who are included in the open database, VtialDB.
• The patients who underwent inhaled general anesthesia for non-cardiac surgery.
• The patients who have non-invasive monitoring data including blood pressure, electrocardiography, pulse oximetry, bispectral index, and capnography.

Exclusion criteria

• The patient with missing data.

Trial design

5,175 participants in 1 patient group

Group

Description:

In the open source database (VitalDB, https://vitaldb.net), the patients who underwent general anesthesia with non-invasive monitoring including blood pressure, electrocardiography, pulse oximetry, bispectral index, capnography, and minimal alveolar concentration of inhalation agent.