Evaluating the impact of artificial intelligence-driven self-rostering: A dual-site pilot in medical imaging

Work-life balance is an increasing priority in healthcare. However, this presents a challenge for public hospitals, where shift work, weekend work and night shifts are essential to providing 24-h care. Self-rostering (SR) is a widely researched method that can increase work-life balance for health professionals. However, the introduction of artificial intelligence (AI) is crucial to overcome the complexity of multi-department scheduling common in medical imaging departments. Guided by previous findings, this study explores work-life balance, shift-swapping and sick-leave in relation to AI-SR.

This research was undertaken alongside an AI-SR pilot, in two medical imaging departments (one urban, one regional) in Aotearoa, New Zealand (NZ). There were 71 participants, 81 % were female, all aged between 20 and 70. An explanatory-sequential mixed methods design was utilised. Data of shift-swaps, sick leave and work-life balance was collected, and thematic analysis was undertaken.

Sick leave results varied between sites, linked qualitatively to improved staffing-levels and decreased work-life conflict, no change was noted in shift-swapping. Qualitative themes were agency, work fitting around life, fairness and loss aversion and receptivity.

AI-based self-rostering can help medical imaging staff balance work with their personal lives and may reduce sick leave, but more research is needed on its impact on wellbeing. Our findings also suggest that incorporating explainable AI could improve fairness and user acceptance, although guidance on XAI design for healthcare is scarce.

Our study shows that AI-SR systems improve work-life balance. However, those looking to implement AI-SR systems should consider system transparency, to help mitigate loss aversion and improve perceptions of fairness. This is crucial to improving user understanding and, ultimately, acceptance of the scheduling approach and AI technologies more generally.