Kait O’Callahan, Clinical Coordinator at Unitec, is our guest author. She, along with Sharon Sitters, Lecturer at Unitec, and Mike Peterson, Clinical Tutor at Whanganui Hospital, have done a study through Unitec into the perceptions around self-rostering in radiographers. Rosterlab was fortunate enough to support their research with our AI rostering platform.

When is autonomy not autonomy? When it’s forced upon you in the name of research.

This was something Sharon and I were cognisant of when we approached Whanganui’s radiography department about self-rostering. However, it wasn’t until we finished our project that we realised just how many assumptions we had made and how naïve we had initially been.

It was Mike, clinical tutor at Whanganui and co-researcher in this project, that approached us with the idea of bringing self-rostering to the department. Mike’s motivation was admittedly self-centred – he wants to improve his work/life balance by jiggling his schedule such that whether he plays rugby or not is no longer determined by his roster. Sharon and I could see the potential benefits of self-rostering almost immediately; unlike Mike, as lecturers we both have flexible working practices that allow us to easily fit in our lives around work. This level of autonomy is something both Sharon and I value, and the lack of autonomy is what inhibits us from returning to radiography full-time. But, what if a self-rostering system could fit the needs of both the staff and the department? By having staff working shifts they predominantly want to work, and having service needs still met, our research team envisioned a radiology utopia, where happy radiographers come to work feeling fulfilled professionally and personally.

Obstacle 1: not everyone else could see our vision.

We greatly underestimated how entrenched traditional rostering practices were in radiology. How we underestimated something that obvious seems embarrassing, but I think we were so sold on our vision for a future that we forgot to ask syoutaff exactly what was and wasn’t working for them. It’s a learning we will take to our future research – first, suss out what is going well and isn’t going well with the roster, THEN try and help them solve their problems. Don’t assume problems and don’t assume you have the solutions. There were staff that were very happy with their current roster; part of this was because they had created ways over time of getting the roster to work for them. A staff member aptly called it ‘customs and practices’, and it was our first sign that some of Whanganui’s staff were self-rostering already. Naturally, these people had an aversion to self-rostering as they were already getting what they wanted; they already had the autonomy we thought we were bringing. Self-rostering would level the playing field, and, as Sharon said, ‘make unseen things seen’. Of course, there were plenty of staff who were unhappy with their current roster, although they were at pains to clarify that it wasn’t the fault of the person running the roster, who was universally liked. And this brings me to the next obstacle…

Obstacle 2: self-rostering ultimately means automating someone’s job.

Manual rostering might be an exhausting, thankless task, but behind each roster is a person who has dedicated time and energy to creating the best roster they can. For Whanganui, this was the job of the rostering manager (and radiographer) Melanie*. By bringing in Rosterlab’s AI, we could have been interpreted as saying to Melanie, ‘hey thanks for your time, but this can do your job better and faster’. I can see how staff felt they didn’t want to offend Melanie by ditching them for an AI based system. It’s also not hard to see how a staff member might see Melanie’s job being automated, and wonder if theirs is next. While we saw Rosterlab’s software as a way to free up Melanie’s valuable time, in hindsight it could be seen as a ‘the robots are coming’ move, and that fear is one almost as old as manual rostering itself.

Obstacle 3: the robots are coming… but they think like us, right?

Sharon, Mike, and I are no AI experts, but without preconceived ideas of Whanganui’s rostering practices (and in Mike’s case, extensive training on Rosterlab’s system), we could get our head around how to self-roster effectively in a relatively short amount of time. What we underestimated (drastically, may I add) was the amount of training required to help the staff to self-roster using Rosterlab’s technology. The main issue we had were that Whanganui’s staff were so used to their manual roster that they made assumptions that the AI would think just like Melanie. A classic example of this was when a staff member commented that they were unhappy the AI had put them on call the night before annual leave. From the AI’s perspective, that is efficient rostering as any call-break would run into the next day, (and the staff member’s leave – not costing the department anything), but from the staff member’s perspective it’s a disaster. When we explained to the staff member that she would have had to tell the AI to roster that night as ‘no-call’, she exclaimed, “I just thought it would think like Melanie.” Thankfully, this obstacle is a fairly easy one to overcome, it just requires more training.

Solution: education, education, education

It’s a cliché, but it’s true. For a self-rostering system to be effective, education needs to be prioritised. It needs to be bi-directional; learn first from the department, and then let them learn about the system from you. Work with staff to solve their problems, and don’t assume people hold the same values as you. Our research team did attempt this with our community-based participatory approach; we held training days and we listened to hours worth of informal discussions. But still, we could have done more. And we will, because our next step is to pilot this programme in another radiology department. Watch this space.

*names have been changed