Creation, Implementation and Validation of Intra- and Postoperative Risk Prediction Models

Postoperative Complications Clinical Trial

Official title: Creation, Implementation and Validation of Intra- and Postoperative Risk Prediction Models

Status Completed

Clinical Trial Summary

This project aims to create and validate surgical risk prediction models for the prediction of complications in patients pending surgery during the operation, in the immediate postoperative period and up to one month after discharge.

At present there is no risk assessment system in place, except for the ASA scale which is mainly based on the subjective impression of the facultative, who assesses it in the universal preoperative consultations that we have planned in the system. In this project we intend to provide robust models, based on the analysis of data from patients in 4/5 Basque hospitals, i.e. generated in our population.

Clinical Trial Description

A three-phase study has been designed:

1. st phase: Derivation and internal validation of the predictive model by means of a reprospective cohort study in which patients operated on at the Galdakao-Usansolo Hospital (HGU), Urduliz Hospital (HU), Basurto University Hospital (HUB), Donostia University Hospital (HUD) and Araba University Hospital (HUA) will be recruited. Hospital universitario de Donostia (HUD) and Hospital universitario de Araba (HUA) over XXX years and data will be obtained from the preoperative period until the month of discharge from the operation. For the identification and creation of these models, machine learning techniques will be used with the main purpose of identifying variables not described in the literature. Machine learning is the most important branch of Artificial Intelligence. Within Machine Learning, supervised learning is the most widely used area. Supervised learning allows computers to learn to perform tasks by discovering and exploiting complex patterns in large amounts of data. In the specific case of data from electronic medical records, Machine Learning algorithms allow us to use the historical data of each patient so that the computer learns to anticipate future events in a personalised way.

2. nd phase: External validation of the models created in the first phase in a cohort of patients operated on in 2020 in the same centres. The methodology proposed by Debray et al. will be applied.

3. rd phase: Evaluation of results after the implementation of the models in the EHR of the Galdakao-Usansolo Hospital in the form of an 'Action Guide'. Based on the risk stratification carried out in the previous phases, the anaesthesia department will create recommendations for action according to the level of risk. The percentages of mortality and intra- and postoperative complications will be compared by means of a quasi-experimental intervention study, comparing the results of the HGU hospital where the risk scale and the consequent recommendations will be implemented, before and after its implementation, and also comparing them with the percentages of patients who become complicated and/or die in HU, HUB, HUD and HUA, where the usual clinical practice will be followed, based on the ASA scale. This prospective cohort, once the risk scale has been implemented, will also be used for external validation (2020-2021).

Socio-demographic and clinical variables (main diagnosis, comorbidities, treatments, previous interventions, intraoperative data, post-operative data, procedures performed during hospitalisation, and complications up to one month after hospital discharge) and laboratory parameters will be collected.

This information will be extracted from osabide's global data exploitation system, Oracle Business Intelligence, and the laboratory data will be extracted from the information systems of the clinical laboratories of the centres involved. ;

Related Conditions & MeSH terms

• Postoperative Complications
• Risk Factors

• NCT number: NCT06411496
• Study type: Observational
• Source: Hospital Galdakao-Usansolo
• Status: Completed
• Start date: June 1, 2018
• Completion date: June 1, 2023