Foresites patientcare reduces patient falls in hospitals using machine learning methods
Background: Patient falls are major problems in hospitals both for patients and the hospital. Falls dramatically reduce patient health and safety and dramatically increase the cost of healthcare. Foresite’s Patientcare is a passive monitoring system that uses infrared imaging and decision tree learning algorithms to drastically reduce patient falls, thus increasing patient health and safety while reducing the cost of healthhcare.
Patient falls are among the eight preventable conditions not reimbursed by CMS (Medicare and Medicaid) and insurance companies. There are millions of hospital patient falls every year costing hospitals billions of dollars in unreimbursed expenses. This does not take into consideration falls of elders in their living environment which are significantly larger issues and are greatly reduced by another Foresite product, namely Eldercare.
The goal of our study was to 1) identify which patients are most likely to fall in hospitals and 2) reduce the number of falls in hospitals.
Methodology: Foresite Patientcare uses an infrared depth sensor to capture image data. Patientcare then utilizes a two-stage system for fall detection. The first stage of the system characterizes the vertical state of a segmented 3D object, and then identifies on ground events through temporal segmentation of the vertical state time series of tracked 3D objects. The second stage utilizes an ensemble of decision trees and features extracted from an on ground event to compute a confidence that a fall preceded it. Fall risk assessment is based on gait speed of the segmented moving objects.
Results: Foresite installed Patientcare in rooms of a top 10 US hospital in August of 2014. For the previous two years, the same rooms of the hospital averaged 7.04 falls per month and .96 falls per month with an injury. For the study, Patientcare monitored patients for 3,975 patient days. Foresite alerted the hospital 66 times for patients having "Elevated Risk" and 11 times for patients having "High Risk" for a total of 77 times of a combined higher probability of falling.
During the study, patients assessed with an
Elevated Risk status fell 3.0% of the time within 24 hours, patients assessed
with a High Risk status fell 9.1% of the time within 24 hours and patients
assessed with a Decreased Risk/Normal status fell 0.00% of the time within 24
Accordingly, 100% of the patient falls were within the 1.9% of the patient population that Patientcare assessed as either Elevated Risk or High Risk. 0.0% of the patient falls were within the 98.1% of the patient population that Patientcare assessed as Decreased/Normal Risk.
In addition, patient falls reduced from 7.04 falls per month to 1.52 falls per month. Patient falls with injury decreased from 0.96 patient fall per month to 0.00 patient falls per month.
Conclusions: Foresite Patientcare is successfully able to determine which patients are most likely to fall. In addition, during the study, Patientcare dramatically reduced patient falls by 78.4%.