Care demand networks in maternity care - an innovative approach exploring the complexity of care demands with routine data: Retrospective observational study

Background: Care demand is complex, variable, and intense. Yet, most staffing tools that aim to inform staffing requirements reduce it to overly simplified metrics. Electronic health records contain substantial information that could support decision-making as affected by care demands. Standardized care activity catalogs document nursing interventions and the associated care time spent, offering rich insights into care delivery. However, their combined potential to inform staffing models and for exploring the complexity of care remains untapped. Network analysis allows for the analysis and visualization of critical patterns in care demands. Objective: This study applies network analysis to explore maternity care demand and illustrates how to describe care demand networks according to network terminology. Design: Retrospective observational study using routinely collected data. Settings: Maternity units in a Swiss tertiary hospital. Participants: 2 346 maternal case records during the postnatal period. Methods: A correlation network analysis was conducted using 244 382 recorded care activities, with the minutes associated for each activity and following individual cases summed up over the hospital stay. Pearson correlations between activities were estimated to create a partial-correlation network, while edges with |correlations| ≥ 0.15 were retained. The final undirected, weighted graph was analyzed using standard network metrics to explore the features of care from the network structure. Results: A total of 113 different care activities were recorded, with an average duration of 3.7 min. The resulting network suggests dense subgroups of correlated care activities in the maternity care process. Conclusions: This study demonstrated the successful application of network methodology to visualize and enhance the understanding of maternity care using routinely collected data. The network approach can be further explored to understand day-to-day care demands and eventually to predict care demands for upcoming days, which fosters staff rostering based on previous care demands. Registration: #HealthServicesResearch #RoutineData #Workload.