Machine Learning Monitoring for Clinical Deterioration

Recruiting at 1 trial location

Overseen By Borna Safabakhsh, MS, MBA

Age: 18+

Sex: Any

Trial Phase: Academic

Sponsor: AgileMD, Inc.

Disqualifiers: Under 18, Not in eCART unit

No Placebo Group All trial participants will receive the active study treatment (no placebo)

Approved in 1 Jurisdiction This treatment is already approved in other countries

Compensation: Varies

Number of Visits: Unknown

30000 Participants Needed

Trial Summary

What is the purpose of this trial?

In this study, the investigators will deploy a software-based clinical decision support tool (eCARTv5) into the electronic health record (EHR) workflow of multiple hospital wards. eCART's algorithm is designed to analyze real-time EHR data, such as vitals and laboratory results, to identify which patients are at increased risk for clinical deterioration. The algorithm specifically predicts imminent death or the need for intensive care unit (ICU) transfer. Within the eCART interface, clinical teams are then directed toward standardized guidance to determine next steps in care for elevated-risk patients. The investigators hypothesize that implementing such a tool will be associated with a decrease in ventilator utilization, length of stay, and mortality for high-risk hospitalized adults.

Do I need to stop my current medications for this trial?

The trial information does not specify whether you need to stop taking your current medications. It seems focused on monitoring rather than changing treatments, so you may not need to stop them.

What data supports the effectiveness of the eCARTv5 clinical deterioration monitoring treatment?

Research shows that machine learning models like MEWS++ can predict patient deterioration up to six hours before it happens, with better accuracy than traditional methods. This suggests that similar systems, like eCARTv5, could help healthcare providers make timely decisions to prevent worsening conditions. ^{1 2 3 4 5}

Is the eCARTv5 clinical deterioration monitoring system safe for humans?

The research articles do not provide specific safety data for the eCARTv5 system, but they focus on its ability to predict clinical deterioration using machine learning. These systems are designed to help healthcare providers identify patients at risk of worsening health, potentially improving patient outcomes by allowing for timely interventions. ^{1 3 4 5 6}

How is the eCARTv5 treatment different from other treatments for clinical deterioration?

The eCARTv5 treatment is unique because it uses machine learning to predict clinical deterioration up to six hours before it happens, allowing for timely interventions. Unlike traditional methods that rely on a limited set of variables, eCARTv5 analyzes a wide range of clinical data to provide more accurate predictions. ^{1 4 5 7 8}

Research Team

Dana P Edelson, MD, MS

Principal Investigator

AgileMD, Inc.

Eligibility Criteria

This trial is for adults over 18 years old who are admitted to specific hospital wards where the eCARTv5 monitoring system is used. It's not open to those under 18 or patients in wards without this technology.

Inclusion Criteria

I am admitted to a unit with eCART monitoring.

I am 18 years old or older.

Exclusion Criteria

You are not in a hospital unit where patients are monitored using eCART.

I am under 18 years old.

Timeline

Screening

Participants are screened for eligibility to participate in the trial

2-4 weeks

Intervention

Deployment of eCARTv5 into the EHR workflow to monitor patients and guide clinical teams

12 months

Follow-up

Participants are monitored for outcomes such as mortality, length of stay, and ventilator-free days

12 months

Treatment Details

Interventions

• eCARTv5 clinical deterioration monitoring

Trial Overview The study tests a machine learning tool, eCARTv5, integrated into hospitals' electronic health records. It predicts which patients might soon need intensive care by analyzing their vital signs and lab results.

Participant Groups

Group I: Intervention Arm Experimental Treatment 1 Intervention

Intervention Arm (experimental): eCARTv5 will monitor all adult medical-surgical (ward) patients at hospitals that implement the tool in their EHR. A pre vs. post analysis will be done to compare the impact of the tool at the intervention hospitals.

Group II: Control Arm Active Control 1 Intervention

Control Arm (active comparator): hospital sites that do not implement eCARTv5 will be active comparator.

eCARTv5 clinical deterioration monitoring is already approved in United States for the following indications:

Approved in United States as eCARTv5 for:

• Clinical deterioration monitoring in hospitalized ward patients

Who Is Running the Clinical Trial?

AgileMD, Inc.

Lead Sponsor

Trials

2

Recruited

60,000+

Department of Health and Human Services

Collaborator

Trials

240

Recruited

944,000+

University of Wisconsin, Madison

Collaborator

Trials

1,249

Recruited

3,255,000+

University of Chicago

Collaborator

Trials

1,086

Recruited

844,000+

BayCare Health System

Collaborator

Trials

5

Recruited

30,700+

Findings from Research

The eCART Lite machine learning tool, developed using data from 556,848 adult hospital admissions, effectively predicts clinical deterioration using just age, heart rate, and respiratory rate, outperforming existing models like MEWS and NEWS.

With an area under the curve (AUC) of 0.79 for ICU transfer prediction and 0.80 for combined outcomes, eCART Lite demonstrates high accuracy, making it a valuable tool for resource-limited inpatient settings.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Less is more: Detecting clinical deterioration in the hospital with machine learning using only age, heart rate, and respiratory rate. Akel, MA., Carey, KA., Winslow, CJ., et al.[2022]

The Intensive care Warning Index (I-WIN) model, developed using machine learning on data from 488 infants with congenital heart disease, can predict clinical deterioration up to 8 hours in advance with high accuracy (AUC of 0.92 at 4 hours before deterioration).

This model represents a significant shift towards using data-driven approaches for risk prediction in critical care, potentially improving patient outcomes by allowing timely interventions based on routinely collected electronic health record data.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Early prediction of clinical deterioration using data-driven machine-learning modeling of electronic health records. Ruiz, VM., Goldsmith, MP., Shi, L., et al.[2022]

A systematic review of 18 studies on real-time automated alerts for clinical deterioration in hospitalized patients found that most did not show significant improvements in patient outcomes, such as ICU admissions or in-hospital deaths.

Only four studies reported positive outcomes, and these were the only ones that involved direct clinician engagement, suggesting that involving healthcare providers may be crucial for the effectiveness of such alert systems.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

A scoping review of real-time automated clinical deterioration alerts and evidence of impacts on hospitalised patient outcomes. Blythe, R., Parsons, R., White, NM., et al.[2022]

The MEWS++ machine learning model can predict patient deterioration or death up to six hours in advance, significantly improving early detection compared to traditional methods, which rely on a limited set of variables.

In a study of 96,645 patients, the random forest model outperformed the traditional Modified Early Warning Score (MEWS) with a sensitivity increase of 37% and an AUC-ROC improvement of 14%, demonstrating its potential for timely clinical intervention.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

MEWS++: Enhancing the Prediction of Clinical Deterioration in Admitted Patients through a Machine Learning Model. Kia, A., Timsina, P., Joshi, HN., et al.[2020]

The HAVEN system, developed using data from 230,415 patient admissions, significantly outperforms existing early warning systems in predicting in-hospital deterioration, achieving a c-statistic of 0.901 compared to scores ranging from 0.700 to 0.863.

HAVEN can identify 42% of patients at risk of cardiac arrest or unplanned ICU admissions up to 48 hours in advance, providing a crucial lead time for intervention, which is much higher than the 22% identified by the next best scoring system.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Detecting Deteriorating Patients in the Hospital: Development and Validation of a Novel Scoring System. Pimentel, MAF., Redfern, OC., Malycha, J., et al.[2022]

This systematic review aims to identify and evaluate the automated features, technologies, and algorithms used in electronic early warning/track and triage systems (EW/TTS) designed to predict clinical deterioration, highlighting their potential to improve patient safety.

The study will analyze data from various databases without time limitations, ensuring a comprehensive understanding of current EW/TTS capabilities, which could lead to advancements in fully automated systems for preventing clinical deterioration.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

A protocol for a systematic review of electronic early warning/track-and-trigger systems (EW/TTS) to predict clinical deterioration: Focus on automated features, technologies, and algorithms. Rostam Niakan Kalhori, S., Deserno, TM., Haghi, M., et al.[2023]

Machine learning (ML) techniques have been applied in 29 studies to predict patient clinical deterioration in hospitals, showing a wide range of predictive accuracy with area under the curve values from 0.55 to 0.99, depending on the model and input features used.

Despite the promising advancements in using ML for early identification of at-risk patients, further research is needed to assess the practical application and effectiveness of these models in real-world clinical settings.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Evaluation of machine learning-based models for prediction of clinical deterioration: A systematic literature review. Jahandideh, S., Ozavci, G., Sahle, BW., et al.[2023]

The study aims to implement a 'Just-in-Time Training' program for nurses using the National Early Warning Score 2 (NEWS2) to enhance early recognition of clinical deterioration in hospitalized patients, which is crucial for improving patient outcomes.

By training nurses to identify early signs of clinical deterioration, the study seeks to reduce mortality rates and the need for emergency interventions, highlighting the importance of timely nursing interventions in patient care.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Nursing Training for Early Clinical Deterioration Risk Assessment: Protocol for an Implementation Study. Lourenço, LBA., Meszaros, MJ., Silva, MFN., et al.[2023]

References

1. Ireland pubmed.ncbi.nlm.nih.gov

Less is more: Detecting clinical deterioration in the hospital with machine learning using only age, heart rate, and respiratory rate. [2022]

2. United States pubmed.ncbi.nlm.nih.gov

Early prediction of clinical deterioration using data-driven machine-learning modeling of electronic health records. [2022]

3. England pubmed.ncbi.nlm.nih.gov

A scoping review of real-time automated clinical deterioration alerts and evidence of impacts on hospitalised patient outcomes. [2022]

4. Switzerland pubmed.ncbi.nlm.nih.gov

MEWS++: Enhancing the Prediction of Clinical Deterioration in Admitted Patients through a Machine Learning Model. [2020]

5. United States pubmed.ncbi.nlm.nih.gov

Detecting Deteriorating Patients in the Hospital: Development and Validation of a Novel Scoring System. [2022]

6. United States pubmed.ncbi.nlm.nih.gov

A protocol for a systematic review of electronic early warning/track-and-trigger systems (EW/TTS) to predict clinical deterioration: Focus on automated features, technologies, and algorithms. [2023]

7. Ireland pubmed.ncbi.nlm.nih.gov

Evaluation of machine learning-based models for prediction of clinical deterioration: A systematic literature review. [2023]

8. Canada pubmed.ncbi.nlm.nih.gov

Nursing Training for Early Clinical Deterioration Risk Assessment: Protocol for an Implementation Study. [2023]