AI video surveillance care home trial produced too many false alarms, say auditors

A federally funded CCTV trial in two South Australian care homes – where artificial intelligence was used to detect falls and screams – produced more than 12,000 false incidents over a 12-month period, as Ron Alalouff reports.

An audit of the trial by PwC found that the AI technology was “not yet sufficiently accurate at detecting incidents in a residential aged care setting”. It went on to say that while the accuracy of the system improved over time (as it was designed to), it did not achieve a level that would be acceptable to staff and management.

The report into the AU $785,000 (c. £467,000) trial found that in the final months of the trial, staff could not respond to every alert, resulting in at least one instance where staff did not respond to an actual fall.

The AI surveillance system was designed to capture video and audio and alert staff when excessive sounds or movements were detected. But it mistakenly set off alerts when staff would crouch or bend down to care for residents.

Video and audio recording

The surveillance system used in the trial consisted of cameras and audio devices with built-in recording, AI-programmed to detect specific movements and sounds – including falls, calls for help or screams – and an independent offsite monitoring centre, where operators did not view any footage, but who contact care staff when they received text alerts of incidents. Cameras and audio recorders were installed in common areas (inside and outside) and in residents’ bedrooms. The cameras and recorders in bedrooms were only used with the consent of individual residents.

The system provides that where an event is detected, the 90 seconds prior to the event and five minutes following it are captured. Footage is retained for 60 days, after which it is transferred to a Network Attached Storage system, where it is stored for seven years, providing an audit trail for all alerts received.

The trial included a four-month commissioning phase to calibrate the system. Several adjustments were made during the course of the pilot study, including:

• Reductions in the sensitivity of audio recording devices to accommodate high ambient noise levels
• Ongoing system training and testing, including having the technology provider on site to demonstrate movements for the system to detect and learn from
• Installation of additional cameras to reduce blind spots
• A series of technological patches to help the AI to distinguish inanimate objects (e.g. a coat draped over a chair) from people, and to refine the AI algorithm to improve accuracy

The audit found that the number of false alerts experienced at the sites was unexpected and unacceptable to staff. The number of false alerts in the first few months of the pilot study meant that staff at both sites were overwhelmed by the workload associated with responding to alerts.

In the final months of the study, staff were no longer able to respond to every alert. There was at least one instance where staff did not respond to an alert that turned out to be a genuine fall.

AI ‘matured’ during trial

Despite the high number of false alerts, the system had sufficiently matured to be able to detect some real incidents, including falls by residents. In the final three months of the study (January-March 2022), 22% of actual events were detected by the CCTV technology (compared to less than two per cent between July and December 2021).

The report found that a number of aspects of residential care impacted on the feasibility of the system, including:

• The behaviour of the residents: Those who are physically active around their rooms, for example, and those who vocalise often and loudly are more likely to be detected
• The number of people in a space: Several people within a space impacts the ability of the AI to accurately detect the pre-programmed movements and sounds, for example, where a staff member’s body blocks the view of a camera
• Environmental factors: Minimal furniture and clutter in a room was likely to reduce the number of alerts, while increased clutter, furniture or moving the location of furniture was likely to lead to more alerts

The report concluded that the CCTV system had neither negatively or positively influenced the quality and safety provided at the sites. It said: “We can expect to see the increased digitisation of residential care and it is important to learn from trials like the CCTV pilot project to determine what role technology has to play, and how and when it will be appropriate to support quality and safety in residential care. While the pilot did not achieve the aspirations held for it, there are many pilot insights that can help governments and the aged care sector in Australia when they are considering the use of surveillance technologies in residential care.”

Crying ‘wolf’

South Australia’s Health Minister, Chris Picton, told ABC News that 12,000 false alerts was “completely unacceptable”.

“That meant that staff had to respond time after time after time to false reports alerting from this system; that meant that they were taking time away from caring for patients at the bedside,” he said.

“The report notes that where there were some cases of actual true reports that were noted, it meant that staff weren’t responding to them because it became a case of the boy who cried ‘wolf’.”

South Australia Health has published a discussion paper on the role of surveillance and monitoring in the care of elderly people, including the different types of technology available.

Free Download: The Video Surveillance Report 2023

Discover the latest developments in the rapidly-evolving video surveillance sector by downloading the 2023 Video Surveillance Report. Over 500 responses to our survey, which come from integrators to consultants and heads of security, inform our analysis of the latest trends including AI, the state of the video surveillance market, uptake of the cloud, and the wider economic and geopolitical events impacting the sector!

Download for FREE to discover top industry insight around the latest innovations in video surveillance systems.

Download The Video Surveillance Report