Prediction of Patient Deterioration Using Machine Learning
Study Details
Study Description
Brief Summary
This is a retrospective observational study drawing on data from the Brigham and Women's Home Hospital database. Sociodemographic and clinic data from a training cohort were used to train a machine learning algorithm to predict patient deterioration throughout a patient's admission. This algorithm was then validated in a validation cohort.
Condition or Disease | Intervention/Treatment | Phase |
---|---|---|
|
Study Design
Arms and Interventions
Arm | Intervention/Treatment |
---|---|
Training A subset of patients that are used to train the machine learning algorithm. |
Other: Traditional vital sign alarms versus the BioVitals Index vs the National Early Warning Score 2
We will retrospectively compare the alarms produced from traditional vital sign alarms (thresholds set by clinicians) versus the BioVitals Index vs the National Early Warning Score 2
|
Validation A subset of patients that are "held back" and used to validate the algorithm's accuracy. |
Other: Traditional vital sign alarms versus the BioVitals Index vs the National Early Warning Score 2
We will retrospectively compare the alarms produced from traditional vital sign alarms (thresholds set by clinicians) versus the BioVitals Index vs the National Early Warning Score 2
|
Outcome Measures
Primary Outcome Measures
- Alarm burden [From admission to discharge, measured in hours, on average 5 days]
The number of alarms fired per patient per hour
Secondary Outcome Measures
- Sensitivity for recognition of a safety composite [From admission to discharge, on average 5 days]
The sensitivity (true positives divided by condition positives) for detection of a safety composite (overnight visit, extra unplanned visit, transfer back to the hospital, death during admission, delirium, loss of consciousness, or other major event).
- Specificity for recognition of a safety composite [From admission to discharge, on average 5 days]
The specificity (true negatives divided by condition negatives) for detection of a safety composite (overnight visit, extra unplanned visit, transfer back to the hospital, death during admission, delirium, loss of consciousness, or other major event).
- Positive predictive value for recognition of a safety composite [From admission to discharge, on average 5 days]
The positive predictive value (true positives divided by the sum of true positives plus false positives) for detection of a safety composite (overnight visit, extra unplanned visit, transfer back to the hospital, death during admission, delirium, loss of consciousness, or other major event).
- Negative predictive value for recognition of a safety composite [From admission to discharge, on average 5 days]
The negative predictive value (true negatives divided by the sum of true negatives plus false negatives) for detection of a safety composite (overnight visit, extra unplanned visit, transfer back to the hospital, death during admission, delirium, loss of consciousness, or other major event).
- Rate of alarms with clinical utility [From admission to discharge, on average 5 days]
We will use general estimating equations (GEE) with three outcomes per patient (the number of clinically important alarms for BioVitals, NEWS2, and traditional vital signs); the GEE will account for the clustering between the three outcomes on a patient. The GEE will use a negative binomial marginal model with a log-link for the number of alarms with clinical utility and an offset for log length-of stay (in hours); with this model, we model the rate per hour of number of alarms with clinical utility with BI, NEWS2, and traditional vital signs. The main covariate in the negative binomial model will be a three-level covariate for method: BI vs NEWS2 vs traditional vital signs, and the exponential of the effect of this covariate will be a pair-wise rate ratio for BI vs NEWS2 vs traditional vital signs.
Eligibility Criteria
Criteria
Inclusion Criteria:
Cared for in the Brigham and Women's Home Hospital study
Exclusion Criteria:
Incomplete continuous monitoring data
Contacts and Locations
Locations
Site | City | State | Country | Postal Code | |
---|---|---|---|---|---|
1 | Brigham and Women's Hospital | Boston | Massachusetts | United States | 02115 |
2 | Brigham and Women's Faulkner Hospital | Boston | Massachusetts | United States | 02130 |
Sponsors and Collaborators
- Brigham and Women's Hospital
- Biofourmis Inc.
Investigators
- Principal Investigator: David Levine, MD MPH MA, Associate Physician
Study Documents (Full-Text)
None provided.More Information
Publications
None provided.- 2017P002583d