The goal of this methodological study was to use machine learning algorithms to improve predictions of patient-related nursing workload requirements. Researchers used nursing activity data, collected from LEP, the “documentation of nursing activities”, conducting this retrospective study at the University Hospital of Zurich across eight wards. Machine Learning models were trained by variables, including historical shift workload averages and overall ward workload trends. They found that ML models routinely outperformed baseline across the eight wards. Notably, the lasso regression model was the top performer, improving MAE (mean absolute error) prediction accuracy by 25%. Analysis indicated that early hospital shifts are consistently the most demanding across all wards; most specifically, the urology and maternity wards. On the other hand, night shifts are the least demanding, with low patient volumes and mild workloads. Overall, ML did reliably predict routine shifts; however, it still struggles with sudden changes in workload levels. It is important to note that the “best” model is dependent on the specific objectives, and the need for workload prediction varies across wards.