Search

Chronic postsurgical pain (CPSP) affects 10–50% of surgical patients and is a major contributor to long-term opioid use and reduced quality of life. Current predictive models treat patients independently and fail to capture how risk evolves over time or how postoperative opioid trajectories influence divergence in outcomes. We propose a dynamic graph neural network (GNN) framework in which patients are modeled as nodes and similarity-based edges evolve over time based on opioid prescription patterns, pain scores, and preoperative psychological factors. The model includes (1) a patient graph with static preoperative features, (2) a temporal edge update mechanism, (3) a GNN message-passing layer that aggregates information from dynamically connected patients, and (4) a prediction head estimating CPSP risk at 3, 6, and 12 months. By modeling changing patient relationships after surgery, the framework captures how similar patients may diverge or converge depending on postoperative management, enabling more accurate and personalized CPSP risk prediction using longitudinal electronic health record data.

Diagnostic services are central to clinical decision-making because imaging, laboratory testing, and cardiology diagnostics often determine the next step in diagnosis, treatment, or referral. Bottlenecks in these services can delay care pathways and increase wait times when demand rises faster than available capacity. Current forecasting approaches in diagnostic departments are often reactive and based on historical averages, recent appointment counts, or manual manager judgment. Such approaches may miss upstream signals such as referral surges, seasonal disease activity, and physician ordering behavior. This article proposes a predictive model for forecasting diagnostic service demand by integrating ambulatory referral volume, seasonal disease trends, physician ordering patterns, equipment availability, and historical appointment backlogs. The model is intended to support short- and medium-term capacity planning across diagnostic services. The proposed approach uses a supervised time-series forecasting framework, such as gradient boosting with temporal features or a recurrent neural architecture, trained on historical diagnostic order and scheduling data. Inputs would be engineered from referral streams, diagnostic ordering records, seasonal indicators, equipment schedules, and backlog measures. Conceptually, the model would generate daily or weekly demand forecasts for each diagnostic modality and service line. Forecasts would include uncertainty bounds and operational alerts when projected demand is expected to exceed available appointment capacity. The proposed predictive model could enable proactive capacity management in diagnostic departments. By anticipating demand before backlogs become severe, the model could support improved scheduling, better equipment utilization, and reduced patient waiting times.