Predictive Analytics For AI-Assisted Patient No-Show Management And Clinic Revenue Optimization: A Simulation-Based Research

Abstract

Patient non-attendance at planned appointments for outpatient care is a constant operational and financial challenge facing healthcare systems. Missed appointments lead to less capacity utilisation, less efficient workflow, and direct revenue loss. While predictive models have increasingly been used in no-show forecasting, many studies focus more on the statistical performance of models without integrating the predictions into operational decision-making and financial evaluation. This Research develops a simulation-based predictive analytics framework relating machine learning, threshold optimisation and revenue modelling in an outpatient clinic context. A synthetic data set of 10 000 records of appointments was created to model behavioural, demographic and operational determinants of attendance. Logistic Regression and Random Forest models were evaluated based on ROC-AUC, Precision, Recall and Cross validation. Although predictive discrimination was moderate (AUC approx. 0.58), threshold sensitivity analysis showed that lowering the classification threshold to 0.30 allowed for an increase in recall to 60%, which would allow for effective identification of high-risk patients in a cost-sensitive environment. A financial simulation was performed to compare baseline scheduling against an AI assisted risk-based intervention strategy. Under normal circumstances, no-shows incurred significant revenue loss. The combined use of predictive risk scoring and targeted reminder interventions decreased revenue loss, and produced positive net financial improvement, even when conservative assumptions of intervention effects were used. Sensitivity analysis confirmed the robustness of financial gain in multi effectiveness scenarios. The results show that predictive analytics can help to add efficiency to outpatient clinics as long as the findings are linked to economic decision logic. Rather than prioritizing statistical accuracy, value arises along the lines of combining model outputs with operational intervention strategies. This Research presents contribution of a decision-oriented framework for AI assisted no-show management and emphasizes the need for cost sensitive threshold calibration in healthcare analytics.