AI platform with wearable sensor patch may predict HF rehospitalizations
A wearable sensor predicted HF rehospitalizations with similar predictive accuracy to implantable devices, according to data from the LINK-HF study published in Circulation: Heart Failure.
“Advances in technology and in artificial intelligence make it possible to collect and analyze large amounts of data,” Josef Stehlik, MD, MPH, Christi T. Smith Professor of Medicine at University of Utah School of Medicine in Salt Lake City and co-chief of the advance heart failure program at University of Utah Hospital and Salt Lake City Veterans Affairs Medical Center, told Healio. “We used data from a wearable sensor and an artificial intelligence analytical platform and showed that we can accurately predict worsening of heart failure needing hospital admission.”
Researchers analyzed data from 100 patients (mean age, 68 years; 98% men) with a history of HF who were hospitalized for HF exacerbation at four Veterans Affairs hospitals. Patients were provided a wearable sensor (Vital Connect) that would adhere to their chest and collect data on heart rate, arrhythmia burden, heart rate variability, gross activity, respiratory rate, sleep, walking, body posture and body tilt. These data were continuously streamed to a phone via Bluetooth, and general machine learning was used to analyze the data.
An analytical platform developed a personalized baseline model of normal physiological values after discharge. A clinical alert would be triggered if there were differences between estimated vital signs and actual monitored values.
Patients were asked to wear the sensor 24 hours per day for at least 30 days and up to 90 days after hospital discharge. The clinical event of interest was readmission to the hospital after the initial discharge from the HF exacerbation hospitalization.
Compliance was high, as 87 patients completed 30 days of monitoring and 74 patients completed 90 days of monitoring.
During the study, there were 35 unplanned hospitalization events including 24 events related to worsening HF. Precursors of worsening HF hospitalizations were detected by the sensor with a sensitivity between 76% and 88% and a specificity of 85%.
The median time between the initial alert from the sensor to readmission was 6.5 days.
“This should provide sufficient time for a treatment intervention and hopefully stop the worsening and decrease the risk of admission to the hospital for many of the patients,” Stehlik said in an interview.
Stehlik added that more research is needed in the use of this technology. He said, “Now that we confirmed the accuracy of our prediction, we plan to conduct a larger study where the patient’s doctors will be alerted of the impending heart failure worsening and treatment will be changed accordingly. This study will hopefully confirm that the timely prognostic information can improve the health of heart failure patients.” – by Darlene Dobkowski
For more information:
Josef Stehlik, MD, MPH, can be reached at josef.stehlik@va.gov; Twitter: @josefstehlik.
Disclosures: Stehlik reports he is a consultant for Abbott and Medtronic. Please see the study for all other authors’ relevant financial disclosures.