Study Results
Results pending
The study team has not published outcome measurements, participant flow, or safety data for this trial yet. Check back later for updates.
Basic Information
Get a concise snapshot of the trial, including recruitment status, study phase, enrollment targets, and key timeline milestones.
Recruitment Status
COMPLETED
Total Enrollment
20000 participants
Study Classification
OBSERVATIONAL
Study Start Date
2019-08-01
Study Completion Date
2022-01-15
Brief Summary
Review the sponsor-provided synopsis that highlights what the study is about and why it is being conducted.
At present, there is no universal early warning system implemented in all Basque hospitals, but there are previous experiences, sometimes based on models generated in other health systems. In this project we intend to provide a robust model, based on the analysis of patient data from three Basque hospitals, i.e. generated in our population.
A three-phase study has been designed:
1. st phase: Derivation of the predictive model by means of a reprospective cohort study in which patients hospitalised at the Galdakao-Usansolo Hospital, Donostia University Hospital and Araba University Hospital will be recruited.
2. nd phase: Creation of an alarm system based on the probability of risk of clinical deterioration and implementation of the system in the electronic medical record (EHR) of the HGU, in the form of an "Action Guide".
3. rd phase: The model will be validated by comparing the percentages of clinical deterioration by means of a quasi-experimental intervention study, comparing the results of the HGU hospital where the system will be implemented, before and after the intervention and, on the other hand, with those of Hospital Universitario Donostia (HUD) and Hospital Universitario de Araba (HUA), where normal clinical practice will be followed, with an early warning system based on vital signs in HUD and clinical criteria in HUA.
Sociodemographic and clinical variables will be collected (patient's condition on arrival on the ward, main diagnosis, comorbidities, prescribed treatments and procedures performed during hospitalisation and prior to the onset of deterioration) and laboratory parameters.
This information will be extracted from the osabide global data exploitation system, Oracle Business Intelligence, and the laboratory data will be extracted from the information systems of the clinical laboratories of the participating centres.
Logistic regression models will be created with the dependent variable being clinical deterioration (cardiorespiratory arrest, death, admission to intensive care units) on a database of 10000 hospitalised patients. For external validation, at least 8000 admissions will be prospectively evaluated and multilevel modelling will be performed to see the influence of centre membership on the outcome variable. Confounding will be controlled for using propensity-score techniques.
A three-phase study has been designed:
1. st phase: Derivation of the predictive model by means of a reprospective cohort study in which patients hospitalised at the Galdakao-Usansolo Hospital, Donostia University Hospital and Araba University Hospital will be recruited.
2. nd phase: Creation of an alarm system based on the probability of risk of clinical deterioration and implementation of the system in the electronic medical record (EHR) of the HGU, in the form of an "Action Guide".
3. rd phase: The model will be validated by comparing the percentages of clinical deterioration by means of a quasi-experimental intervention study, comparing the results of the HGU hospital where the system will be implemented, before and after the intervention and, on the other hand, with those of Hospital Universitario Donostia (HUD) and Hospital Universitario de Araba (HUA), where normal clinical practice will be followed, with an early warning system based on vital signs in HUD and clinical criteria in HUA.
Sociodemographic and clinical variables will be collected (patient's condition on arrival on the ward, main diagnosis, comorbidities, prescribed treatments and procedures performed during hospitalisation and prior to the onset of deterioration) and laboratory parameters.
This information will be extracted from the osabide global data exploitation system, Oracle Business Intelligence, and the laboratory data will be extracted from the information systems of the clinical laboratories of the participating centres.
Logistic regression models will be created with the dependent variable being clinical deterioration (cardiorespiratory arrest, death, admission to intensive care units) on a database of 10000 hospitalised patients. For external validation, at least 8000 admissions will be prospectively evaluated and multilevel modelling will be performed to see the influence of centre membership on the outcome variable. Confounding will be controlled for using propensity-score techniques.
Related Clinical Trials
Explore similar clinical trials based on study characteristics and research focus.
Early Warning System for Clinical Deterioration on General Hospital Wards
NCT01280942
Escalation of Care
Cardiopulmonary Arrest
Respiratory Arrest
+2 more
COMPLETED
NA
Development of an Automatically Generated and Wearable-based Early Warning System
NCT05699967
Deterioration of Patient's State of Health
COMPLETED
Early Identification of Clinical Deterioration Using a Wearable Monitoring Device
NCT04220359
Clinical Deterioration
Hemodynamic Instability
COMPLETED
Prediction of Patient Deterioration Using Machine Learning
NCT05045742
Infection
Heart Failure
Chronic Obstructive Pulmonary Disease
+6 more
UNKNOWN
Machine Learning Prediction of Parameters of Early Warning Scores in General Wards
NCT06574906
Patient Safety
ACTIVE_NOT_RECRUITING
Detailed Description
Dive into the extended narrative that explains the scientific background, objectives, and procedures in greater depth.
Phase 1. Derivation of the predictive model Data from patients admitted to the participating centres during 2017 will be extracted from the OBI electronic medical record and the laboratory data management system. The data analyst contracted in this project, together with clinical collaborators, from the health management unit and the research unit, will be responsible for cleaning and cleaning the data and for the generation and internal validation of the predictive model.
Phase 2: Implementation in the electronic medical record: Once the model and the risk scale have been generated, it will be implemented as an alarm system in Osabide Global in the participating centres. To this end, an action guide will be created by the HGU's UGS, which, once validated, will then be exported to the other two centres. The system will operate on the basis of the RIC elements associated with vital signs, and will guide professionals through the process, so that the physician will be shown the analytical data that have been shown to be necessary to calculate the risk, or in the event that the patient has not been asked for that specific laboratory test, recommendations for extraction. Once the system has all the data, a warning will pop up on the screen with the results of the risk scale and the mild-moderate-severe risk category and a report with recommendations for action.Recommendations based on the ViEWS scale, from the physician informing the nurse in charge at the start of continuous monitoring and visit in less than 15 minutes by the responsible attending physician who will also inform the ICU physician to the transfer of the patient to a unit where he/she will be monitored and acted upon, are given in annex 1. The alarm system will be presented in the HGU by the clinical collaborators, GSU and data analyst. The data analyst will be, together with the co-researchers, the person in charge of generating the alarm system, of participating in the meetings in which it is presented in the centres and subsequently, will carry out a constant exploration of the use and adherence of the risk scale, will collect doubts and problems that may arise from the implementation of the system, and will also be responsible for the implementation of the risk scale.
Phase 3: External validation of the predictive model. Comparison of clinical outcomes: The predictive model will be validated once 100 events per centre have occurred (hopefully once about 8000 patients have been admitted). During this time, data will be collected during and after implementation in order to compare clinical deteriorations observed before and after implementation. The data analyst will be extracting the information, sorting and cleaning it in order to be able to carry out the corresponding statistical analyses later on.
STATISTICAL ANALSIS
The data processing procedure of the present project will be established following the following steps, divided into two sections:
A) Development and validation of predictive models of clinical deterioration, transfer to the critical care unit, cardiorespiratory arrest or death during hospitalisation.
Descriptive analysis of the model derivation/internal validation cohort consisting of patients from the three participating centres of the retrospective cohort. Information on losses in the variables recorded in the study will be collected. Information on excluded patients will also be collected. The possibility of applying imputation techniques to the missing data in the variables recorded will be assessed.
2.- The following points will be followed for the creation of the predictive models: a. The sample collected shall be divided into two subsamples: Derivation group 1: The total sample will be divided into 60% for the derivation of the predictive models for the main outcome variable - clinical deterioration; Group 2 of validation of the predictive rules: the models will be validated in this sample (40% of the sample). Analyses will also be performed disaggregated by gender. b. In Group 1, risk factors for clinical deterioration will be identified, as well as for transfer to the critical care unit, cardiorespiratory arrest or death during hospitalisation. A bivariate analysis will be performed to study which of the possible predictor variables are related to each outcome parameter. Those variables with a p-value \<0.20 will be identified as potential predictors to be entered into a multilevel multivariate logistic regression model. Those variables that are statistically significant will be chosen for the final scale. The categories of variables in this model will be assigned a score in relation to the parameter β obtained in that multivariate model. In this way, a total score and an ROC curve will be obtained for it. From this scale, X risk categories will be created.
3.- Goodness of fit and comparison of the predictive models developed. Parameters will be found to evaluate the discriminative and calibration capacity of the predictive models developed in section 2. On the one hand, in the case of dichotomous dependent variables, the area under the ROC curve (AUC, discriminative capacity) will be calculated, considering a value \>0.80 to be a robust predictive model. In addition to the AUC, the calibration of the model will be estimated through the Hosmer-Lemeshow test (good calibration for a p-value ≥0.05).
4.- Internal validation of the predictive models The validation of the predictive model will be carried out in validation group 2. The predictive model and the scale will be validated in this group by musing the predicted values obtained in the derivation sample (group 1). Sensitivity, specificity and area under the curve will be obtained by comparing them with the results obtained in the derivation sample. The calibration capability of the models will be assessed by the Hosmer-Lemeshow test and the discrimination capability by ROC curves. As a form of additional internal validation, a validation by bootstrapping methods will be performed.
5.-External validation of predictive models External validation of the models developed in the retrospective cohort will be applied to the prospective cohort. The linear predictor (LP) will first be found in the remaining cohorts by considering the regression coefficients obtained in the original models. Logistic regression models will be developed. Three main procedures will be considered for the external validation of these models: model fit, discrimination and calibration. First, to verify model fit, we will develop a logistic regression model on the remaining cohorts, considering LP as a covariate to estimate its beta regression coefficient. If the beta regression coefficient of the slope is 1, it will be considered a good fit.Secondly, to assess the discrimination of the model, the AUC will be calculated, by bootstrapping and cross-validation methods. An AUC value \>0.80 will be considered a robust predictive model. Thirdly, the calibration will be evaluated by calculating the overall calibration, the calibration slope (both derived from the calibration plots) and the Hosmer-Lemeshow test.
B) Evaluation of the EWS for clinical impairment 6.- Comparison of results: For the comparison of results after the implementation of the alarm system, a comparison will be made of percentages of patients suffering clinical deterioration before and after its implementation the HGU and also between the percentages appearing in the HGU, where the alarm system will be implemented and the HUDs and HUAs where they will continue with their usual practice during this evaluation period. In order to evaluate the possibility of bias occurring when forming the cohorts, a propensity score will be calculated to explain the probability of being assigned to one (HGU patients) or the other cohort (HUD+HUA) conditional on the independent variables observed and explained in the section "Exposure variables". This propensity score will be entered as an independent variable in a replication of the final models together with exposure and the coefficients obtained in both cases will be compared.
Phase 2: Implementation in the electronic medical record: Once the model and the risk scale have been generated, it will be implemented as an alarm system in Osabide Global in the participating centres. To this end, an action guide will be created by the HGU's UGS, which, once validated, will then be exported to the other two centres. The system will operate on the basis of the RIC elements associated with vital signs, and will guide professionals through the process, so that the physician will be shown the analytical data that have been shown to be necessary to calculate the risk, or in the event that the patient has not been asked for that specific laboratory test, recommendations for extraction. Once the system has all the data, a warning will pop up on the screen with the results of the risk scale and the mild-moderate-severe risk category and a report with recommendations for action.Recommendations based on the ViEWS scale, from the physician informing the nurse in charge at the start of continuous monitoring and visit in less than 15 minutes by the responsible attending physician who will also inform the ICU physician to the transfer of the patient to a unit where he/she will be monitored and acted upon, are given in annex 1. The alarm system will be presented in the HGU by the clinical collaborators, GSU and data analyst. The data analyst will be, together with the co-researchers, the person in charge of generating the alarm system, of participating in the meetings in which it is presented in the centres and subsequently, will carry out a constant exploration of the use and adherence of the risk scale, will collect doubts and problems that may arise from the implementation of the system, and will also be responsible for the implementation of the risk scale.
Phase 3: External validation of the predictive model. Comparison of clinical outcomes: The predictive model will be validated once 100 events per centre have occurred (hopefully once about 8000 patients have been admitted). During this time, data will be collected during and after implementation in order to compare clinical deteriorations observed before and after implementation. The data analyst will be extracting the information, sorting and cleaning it in order to be able to carry out the corresponding statistical analyses later on.
STATISTICAL ANALSIS
The data processing procedure of the present project will be established following the following steps, divided into two sections:
A) Development and validation of predictive models of clinical deterioration, transfer to the critical care unit, cardiorespiratory arrest or death during hospitalisation.
Descriptive analysis of the model derivation/internal validation cohort consisting of patients from the three participating centres of the retrospective cohort. Information on losses in the variables recorded in the study will be collected. Information on excluded patients will also be collected. The possibility of applying imputation techniques to the missing data in the variables recorded will be assessed.
2.- The following points will be followed for the creation of the predictive models: a. The sample collected shall be divided into two subsamples: Derivation group 1: The total sample will be divided into 60% for the derivation of the predictive models for the main outcome variable - clinical deterioration; Group 2 of validation of the predictive rules: the models will be validated in this sample (40% of the sample). Analyses will also be performed disaggregated by gender. b. In Group 1, risk factors for clinical deterioration will be identified, as well as for transfer to the critical care unit, cardiorespiratory arrest or death during hospitalisation. A bivariate analysis will be performed to study which of the possible predictor variables are related to each outcome parameter. Those variables with a p-value \<0.20 will be identified as potential predictors to be entered into a multilevel multivariate logistic regression model. Those variables that are statistically significant will be chosen for the final scale. The categories of variables in this model will be assigned a score in relation to the parameter β obtained in that multivariate model. In this way, a total score and an ROC curve will be obtained for it. From this scale, X risk categories will be created.
3.- Goodness of fit and comparison of the predictive models developed. Parameters will be found to evaluate the discriminative and calibration capacity of the predictive models developed in section 2. On the one hand, in the case of dichotomous dependent variables, the area under the ROC curve (AUC, discriminative capacity) will be calculated, considering a value \>0.80 to be a robust predictive model. In addition to the AUC, the calibration of the model will be estimated through the Hosmer-Lemeshow test (good calibration for a p-value ≥0.05).
4.- Internal validation of the predictive models The validation of the predictive model will be carried out in validation group 2. The predictive model and the scale will be validated in this group by musing the predicted values obtained in the derivation sample (group 1). Sensitivity, specificity and area under the curve will be obtained by comparing them with the results obtained in the derivation sample. The calibration capability of the models will be assessed by the Hosmer-Lemeshow test and the discrimination capability by ROC curves. As a form of additional internal validation, a validation by bootstrapping methods will be performed.
5.-External validation of predictive models External validation of the models developed in the retrospective cohort will be applied to the prospective cohort. The linear predictor (LP) will first be found in the remaining cohorts by considering the regression coefficients obtained in the original models. Logistic regression models will be developed. Three main procedures will be considered for the external validation of these models: model fit, discrimination and calibration. First, to verify model fit, we will develop a logistic regression model on the remaining cohorts, considering LP as a covariate to estimate its beta regression coefficient. If the beta regression coefficient of the slope is 1, it will be considered a good fit.Secondly, to assess the discrimination of the model, the AUC will be calculated, by bootstrapping and cross-validation methods. An AUC value \>0.80 will be considered a robust predictive model. Thirdly, the calibration will be evaluated by calculating the overall calibration, the calibration slope (both derived from the calibration plots) and the Hosmer-Lemeshow test.
B) Evaluation of the EWS for clinical impairment 6.- Comparison of results: For the comparison of results after the implementation of the alarm system, a comparison will be made of percentages of patients suffering clinical deterioration before and after its implementation the HGU and also between the percentages appearing in the HGU, where the alarm system will be implemented and the HUDs and HUAs where they will continue with their usual practice during this evaluation period. In order to evaluate the possibility of bias occurring when forming the cohorts, a propensity score will be calculated to explain the probability of being assigned to one (HGU patients) or the other cohort (HUD+HUA) conditional on the independent variables observed and explained in the section "Exposure variables". This propensity score will be entered as an independent variable in a replication of the final models together with exposure and the coefficients obtained in both cases will be compared.
Conditions
See the medical conditions and disease areas that this research is targeting or investigating.
Clinical Deterioration
Study Design
Understand how the trial is structured, including allocation methods, masking strategies, primary purpose, and other design elements.
Observational Model Type
COHORT
Study Time Perspective
OTHER
Study Groups
Review each arm or cohort in the study, along with the interventions and objectives associated with them.
PATIENTS ADMITTED IN THE HOSPITAL IN 2019
Inclusion criteria: patients over 18 years of age hospitalised on conventional hospital wards, in medical and surgical departments or awaiting a bed during their stay in the emergency department.
Exclusion criteria: patients with poor vital prognosis and DNR orders.
No interventions assigned to this group
PATIENTS ADMITTED IN THE HOSPITAL IN 2021
Inclusion criteria: patients over 18 years of age hospitalised on conventional hospital wards, in medical and surgical departments or awaiting a bed during their stay in the emergency department.
Exclusion criteria: patients with poor vital prognosis and DNR orders.
No interventions assigned to this group
Eligibility Criteria
Check the participation requirements, including inclusion and exclusion rules, age limits, and whether healthy volunteers are accepted.
Inclusion Criteria
* patients over 18 years of age hospitalised on conventional hospital wards, in medical and surgical departments or awaiting a bed during their stay in the emergency department.
Exclusion Criteria
* patients with a poor prognosis for life and a DNR order.
Minimum Eligible Age
18 Years
Eligible Sex
ALL
Accepts Healthy Volunteers
No
Sponsors
Meet the organizations funding or collaborating on the study and learn about their roles.
Biosistemak Institute for Health Systems Research
OTHER
Sponsor Role
collaborator
Biobizkaia
UNKNOWN
Sponsor Role
collaborator
Hospital Galdakao-Usansolo
OTHER_GOV
Sponsor Role
lead
Responsible Party
Identify the individual or organization who holds primary responsibility for the study information submitted to regulators.
Susana García Gutiérrez
Researcher
Responsibility Role
PRINCIPAL_INVESTIGATOR
Principal Investigators
Learn about the lead researchers overseeing the trial and their institutional affiliations.
Naia Mas, PhD
Role: PRINCIPAL_INVESTIGATOR
Hospital Galdakao-Usansolo
Locations
Explore where the study is taking place and check the recruitment status at each participating site.
Hospital Galdakao Usansolo
Galdakao, Bizkaia, Spain
Site Status
Countries
Review the countries where the study has at least one active or historical site.
Spain
Other Identifiers
Review additional registry numbers or institutional identifiers associated with this trial.
PI2019030
Identifier Type: -
Identifier Source: org_study_id
More Related Trials
Additional clinical trials that may be relevant based on similarity analysis.
The New Golden Standard: the Early Warning Score Algorithm
NCT04292158
COMPLETED
Early Detection of Health Deterioration in Elderly Patients After Hospitalization for Heart Failure Decompensation
NCT06126848
NOT_YET_RECRUITING
NA
Examination of the Impact of Better Surveillance and Communication of Patient Deterioration on Patient Related Outcomes
NCT01692847
COMPLETED
International Multi-center Study to Validate an Early Warning Algorithm for Worsening Heart Failure
NCT04758429
UNKNOWN
Evaluation of Home-based Sensor System to Detect Health Decompensation in Elderly Patients With History of CHF or COPD
NCT05865184
UNKNOWN
Evaluation of Home-based Sensor System to Detect Health Decompensation in Elderly Patients With History of CHF
NCT05865197
RECRUITING
Construction of Early Warning Model for Pulmonary Complications Risk of Surgical Patients Based on Multimodal Data Fusion
NCT06057688
UNKNOWN
Individual Early Warning Score (I-EWS)
NCT03690128
COMPLETED
NA
Pulmonary Artery DenerVation Clinical Study Using the Gradient Denervation System in Heart Failure Patients With Pulmonary Hypertension Group 2 (PreVail-PH2 OUS Study)
NCT06495970
RECRUITING
NA
SENSING-AI Cohort for Long COVID Patients
NCT05204615
COMPLETED
CHroniSense National Early Warning Score Study
NCT03448861
UNKNOWN
NA
Remote Symptom Review in Patients With Implantable Diagnostic Holter
NCT06431828
RECRUITING
NA
Developing Enhanced Prediction Models
NCT02983812
COMPLETED
Real-time Symptom Monitoring Using ePROs to Prevent Adverse Events During Care Transitions
NCT05282654
RECRUITING
NA
ALgorithms Adapted From Remote Monitoring
NCT05782140
RECRUITING
Smart Watch Insights for Prevention of Exacerbations and Enhance Rehabilitation - Movement Study
NCT06011356
ACTIVE_NOT_RECRUITING
Outcomes of Digital Alerting Systems in Secondary Care
NCT04638738
UNKNOWN
AI-Driven Prediction of Hospital-Acquired Infections With EHR
NCT06791382
RECRUITING
Life Monitoring for Rural Older Adults
NCT06974006
RECRUITING
A Rapid Diagnostic of Risk in Hospitalized Patients Using Machine Learning
NCT05893420
ACTIVE_NOT_RECRUITING
NA
Maintaining and imPrOving the intrinSIc capaciTy Involving Primary Care and caregiVErs
NCT04592146
UNKNOWN
NA
Development of Voice Biomarkers of Frequent COVID-19 and Long Covid-related Symptoms Based on Data From Users of the Long COVID Companion App
NCT07156994
NOT_YET_RECRUITING
SENSING-AI in Patients With Long COVID (SENSING-AI)
NCT06247332
COMPLETED
Elderly Chronic Diseases Online Biometric Analysis Home Living Technology
NCT02155686
COMPLETED
NA
Remotely Monitoring Patients With Chronic Conditions in Their Home
NCT03207802
COMPLETED