MOTION DETECTION AND ARTIFICIAL INTELLIGENCE IN THE AUTOMATIC DOCUMENTATION OF KEY FIGURES FOR HAND HYGIENE COMPLIANCE

Topic area: 4. Special issues in infection control
Topic: 4a. Hand hygiene
Abstract N°: ICPIC19-ABS-1349
Ehsan Khaljani, MBA* 1, Simon C. Slama1, Theresa Ebeling1
1HygNova GmbH, Berlin, Germany

Introduction:

Direct observation is the gold standard in the detection of hand hygiene compliance, although known biases influence quality of data. In this study, data of motion detection sensors was examined with mathematical algorithms to assess the value of improved technological opportunities in hand hygiene monitoring.

Objectives:

The goal of the study was to detect four properties concerning relevant parameters for hand hygiene monitoring in a laboratory setup:
1. Are patients’ beds occupied by patients or not?
2. If patients are present, is it possible to observe direct contact of healthcare professionals (HCPs) with patients as a surrogate
for a moment of hand hygiene?
3. Can information be gathered without wearable devices?
4. Can data be gathered without personalization of patients and healthcare professionals?

Methods:

Motion detection sensors were mounted over patients’ beds. Presence of patients and the treatment of healthcare professionals were recorded in an operationalized setup. 67 recordings took place which lasted between 2 to 5 minutes. In some recordings, there were duplications of observed cases (bed occupied and direct contact with patient). When HCPs were closer to patients than 10 cm, direct contact was noted. No wearable devices were used to gather data. Data was analyzed with an algorithm, which was developed for the purpose of the study. Direct observation took place
simultaneously to the recordings to validate the results of the algorithm.

Results:

In 53 recordings, the presence of the patient in the bed had to be defined. In 52 out of 53 recordings, the presence of the patient could be detected (98,1% of cases compared to 100% by direct observation). There was one false negative case. In 26 recordings, direct contacts of HCPs were examined. In all 26 cases, direct contacts of HCPs could be determined (100% compared to 100% by direct observation). In no case patients’ or HCPs could be identified out of the recorded data (0% compared to 100% by direct observation). No wearable devices had to be used in the study.

Conclusion:

The results show that algorithmic analysis of motion detection data can lead to information which can be used in the evaluation of hand hygiene compliance. Further studies will be necessary to identify if Artificial Intelligence can detect the 5 moments of hand hygiene out of motion detection data with an acceptable deviation.

Disclosure of Interest:

E. Khaljani, MBA Shareholder of: The author is a shareholder of the HygNova GmbH which uses motion detection technology to approximate the 5 WHO moments of hand hygiene in hospitals., S. Slama Shareholder of: The author is a shareholder of the HygNova GmbH which uses motion detection technology to approximate the 5 WHO moments of hand hygiene in hospitals., T. Ebeling Shareholder of: The author is a shareholder of the HygNova GmbH which uses motion detection technology to approximate the 5 WHO moments of hand hygiene in hospitals.

2019-09-10T12:08:47+00:00