Tricky when AI meets everyday life in primary care

Although AI tools are used for an initial assessment of patients seeking care, a process which is called triage, for more efficient and safe primary care, there is a lack of knowledge about how the practical implementation is done best. Researchers at Halmstad University have investigated strategies for introducing AI in connection with triage.

“For leaders in healthcare, the technology itself is not the problem, but the implementation. The study helps us understand in which areas it is relevant to develop strategies for implementation”, says Elin Siira, Associate Senior Lecturer at Halmstad University and one of the researchers behind the study.

Faster preparations

Triage is a process of assessing and prioritising patients based on symptoms and medical history. In this case, it’s a chat assisted by AI that asks follow-up questions depending on the applicant’s answers. For the staff, the AI tool generates a report with a summary of the patient’s responses, any images the patient has attached, and suggestions for diagnoses. With the help of the report, healthcare professionals can better and faster prepare for the meeting with the patient.

The study is based on interviews with operational managers and others in senior positions in primary care, at both more traditional health centres and digital health care institutions. At the time of the interviews, four Swedish regions used the system.

“Research in the field is often based on hypothetical reasoning. These interviews are based on practical experience of introducing AI systems”, says Elin Siira.

Positive but with obstacles

Overall, leaders were positive about the introduction of AI. However, they saw a number of obstacles, partly linked to the health care staff and partly to the organisation.

“Members of staff who were unaccustomed to working with new aids, smartphones and the like could be worried about how they would work with the tool and what it would do. Doctors also highlighted the central aspect of the clinical encounter with the patient. What will AI be able to help with there, they wondered”, says Elin Siira.

An important measure was to involve the staff to feel ownership of the technology. They worked to prevent fear and scepticism and had meetings between departments and regions to build experiences. Superuser networks were created where some colleagues had more responsiblity and knowledge.

“To meet frustration that the tool has been “added” to the staff from higher ups, the leaders pointed out that the staff have great power. The work is done at the far end of the care chain, and then management cannot micromanage how the tool is to be used. ”

Creating habits

Some health centres were unaccustomed to handling patients and triage digitally in general, and patients were unaccustomed to seeking care digitally.

“This led to more patients seeking care in several ways, both digitally and by calling and perhaps even visiting the health centre. One way to meet this was to adapt the introduction to the digital maturity. First, the care seekers needed to get a bit used to seeking information digitally, then perhaps the AI tool could be connected.”

At the same time, staff were encouraged to instruct and support patients in seeking care digitally instead of calling.

The technology itself

The leaders also addressed obstacles with the technology itself.

“The tool had an inability to deal with underlying diseases, such as linking pain in a joint with previous rheumatism or nausea with pregnancy”, says Elin Siira.

The AI tool also did not have access to the patient’s medical records in the compilation for the staff. It was a troublesome factor that the AI tool and existing systems were not interconnected. Elin Siira thinks it is surprising that the leaders talked so little about it:

“The staff were forced to work in complicated and outdated systems and had to cut and paste between the AI tool and the medical record system. This means duplication of work. However, the systems are a problem that the leaders have very little control over.”

The leaders encouraged the staff to make independent decisions and trust their own knowledge. Staff were also asked to report shortcomings to the company behind the AI tool. At the same time, the tool was constantly evolving:

“The AI tool is constantly learning, unlike other systems that are upgraded from time to time”, says Elin Siira.

The implementation is the big job

According to the study, the leaders seem to want the AI tool to create immediate value in order for the work to be better or more efficient. At the same time, they argue that the benefits will above all be noticeable later on.

“The leaders were very keen to be up to date with technological developments and AI, and need to resort to a lot of improvisation in the introduction. They can’t wait for the perfect AI system, because then they’ve missed the big job that is the implementation”, says Elin Siira.

Text: Kristina Rörström