SHIFT-Hospital in Motion (Validation Study)

SHIFT-Hospital in Motion: Validation Study of a Newly Developed Algorithm for Movement Detection Using Activity Sensors in Acute Care Hospital Patients

The goal of this monocentric observational study involving acute hospitalised patients is to validate the accuracy of classification algorithms for the detection of various movements parameters.

Study Overview

• Status: Completed
• Conditions: Physical Inactivity; Hospitalized Patients

Detailed Description

Patients in hospitals spend the majority of their time inactive, sitting or lying down. Not being active is a common problem for patients in hospitals, often causing complications and impairing recovery, as it can lead to issues such as reduced blood volume, unsteady blood pressure when standing, weaker muscles, and a higher risk of infections, blood clots, and other health issues. The inactivity-related changes in the body in combination with the natural ageing process, the stress of being in the hospital, a poor nutritional status, and possibly troubles with thinking, memory, and understanding or depression diminish the ability to regenerate with overall compromised physiological resilience.

In order to quantify the amount of physical activity of hospitalised patients, the ability of activity sensors to distinguish between lying, sitting, standing and walking is an important requirement. A pilot study (NCT06403826) involving 40 patients examined the feasibility and effectiveness of using activity sensors in clinical settings. The study focused on detecting specific activities and movement patterns using sensors worn on various body parts, with the ankle identified as the preferred location for long-term monitoring. Most participants found wearing the sensors tolerable, indicating the practicality of this approach for extended patient activity monitoring.

The primary objective of this observational, single center study is to validate the accuracy of newly developed algorithms for the detection of various movements parameters. Two different sensors worn on the ankle are used to record the duration of lying, sitting/standing, and the number of steps taken when walking and climbing stairs (up/down). The validation is based on the movement data collected from this and the pilot study and will be conducted in two phases:

• phase 1 (model improvement): Increasing the robustness of the algorithms using training data
• phase 2 (model validation): Validation of the algorithms using test data

The secondary objectives are:

• to determine if altitude data enhances the algorithms' accuracy in distinguishing between stair climbing and walking
• to evaluate the comfort of wearing the sensors

The results of this study will advance healthcare by developing an algorithm that accurately determines the activity patterns of hospitalized patients, thereby enhancing monitoring and understanding of patient mobility in hospital settings.

• Study Type: Observational
• Enrollment (Actual): 101

Contacts and Locations

Study Locations

• Switzerland
  • Basel-Stadt
    • Basel, Basel-Stadt, Switzerland, 4031
      • Universitiy Hospital Basel, Division of Internal Medicine

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Probability Sample

Study Population

Acute hospitalized patients are continuously recruited by the project management team within the context of daily clinical practice from both medical and surgical wards at the University Hospital Basel. Recruitment takes place as soon as possible after admission to the hospital.

Description

Inclusion Criteria:

• patient must have been able to walk before hospitalisation (with or without aids)
• patient must be cognitively able to follow instructions (if a cognitive assessment has been carried out, this cut-off value counts, if no assessment is available, no cognitive impairment is assumed)
• ≥ 18 years
• signed informed consent

Exclusion Criteria:

• patient unable to move prior to hospital admission
• prior inclusion in the study
• discharge on the same day
• inability or contraindications to participate in the study or to follow the study procedures, e.g. due to certain neurological disorders (such as Parkinsonism, hemiplegia, severe Multiple Sclerosis), speech problems, mental disorders, or cognitive impairments
• isolated patient (unable to complete the test battery completely)

Study Plan

How is the study designed?

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Assessment of the accuracy of the classification algorithm for the detection of movements parameters	During both study phases, patient movement is recorded using two sensors on the non-affected/dominant ankle side. Lower extremities are filmed for precise data annotation. In phase 1 (model improvement), patients undergo a test battery to improve the accuracy of the algorithm developed in the pilot study for motion and activity detection. The test battery contains a fixed sequence of movements, including lying, sitting, standing, walking and stair climbing. In phase 2 (model validation), a less restrictive test battery is used to validate algorithm accuracy. Patients perform movements like lying, sitting, standing, walking, and stair climbing for up to 15 minutes. Sequence and duration are not predefined. Algorithm accuracy is assessed via a multi-class confusion matrix. Rows show actual classes, columns show predicted classes. The diagonal contains observations where predicted matches actual (true positive). Accuracy [%]= (Sum of diagonal elements/Total observations)*100.	during hospitalization (up to 1 hour)

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Comparison of the accuracy of the algorithm with and without the use of altimeter data	In this study two different sensors are used. One of these sensors records altimeter data in addition to gyroscope and accelerometer data. The aim of the comparison is to determine if the altimeter data improves the algorithm' accuracy in distinguishing between stair climbing and walking. The accuracy of the algorithm is calculated using a multi-class confusion matrix. The rows are the actual classes and the columns are the predicted classes. The diagonal of the matrix contains the observations where the predicted class matches the actual class (true positive). Accuracy [in %]= Sum of the diagonal elements / Total number of observations * 100. At the end, the accuracy of the algorithm with and without the use of altimeter data is compared.	during hospitalization (up to 1 hour)
Assessment of the comfort level associated with wearing the sensors	The comfort of wearing the sensors is evaluated after both study phases by a questionnaire. The responses from patients are being collected regarding the discomfort of wearing the sensors or any problems with the attachment.	during hospitalization (up to 10 min)

Collaborators and Investigators

• Sponsor: University Hospital, Basel, Switzerland
• Collaborators: Zurich University of Applied Sciences
• Investigators: Principal Investigator: Joris Kirchberger, University Hospital, Basel, Switzerland; Study Chair: Jens Eckstein, Prof. Dr. med., University Hospital, Basel, Switzerland

Study record dates

Study Major Dates

• Study Start (Actual): April 29, 2024
• Primary Completion (Actual): July 22, 2024
• Study Completion (Actual): July 22, 2024

Study Registration Dates

• First Submitted: April 23, 2024
• First Submitted That Met QC Criteria: April 29, 2024
• First Posted (Actual): May 2, 2024

Study Record Updates

• Last Update Posted (Actual): July 30, 2024
• Last Update Submitted That Met QC Criteria: July 29, 2024
• Last Verified: July 1, 2024

More Information

Terms related to this study

• Keywords: Movement detection; Activity sensors; Algorithm development and validation
• Other Study ID Numbers: 2024-00594; am24Eckstein

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No