# Analysis of the anti-scalping mechanism of hospital appointment registration based on Bayesian theory

## TL;DR

A Bayesian-based system in hospital appointments improved fairness and efficiency by reducing speculative bookings and boosting patient access.

## Contribution

A novel Bayesian anti-scalping mechanism was developed and validated to enhance hospital appointment fairness and efficiency.

## Key findings

- The appointment completion rate increased from 64.6% to 78.0% after implementation.
- The Bayesian model achieved an AUC of 0.91 with high sensitivity and specificity in detecting abnormal registrations.
- The proportion of first-time patients increased by 18.7%, indicating improved access fairness.

## Abstract

To evaluate the clinical effectiveness of a Bayesian-based anti-scalping mechanism integrated into a hospital appointment system in improving fairness, efficiency, and patient accessibility while reducing speculative bookings.

A retrospective analysis was conducted from January 2019 to December 2022 (defined as the core observation period, with the intervention occurring in January 2021). Results up to mid-2023 are provided solely to demonstrate long-term trend stability. The Bayesian model identified “abnormal behaviors” based on a validated ground truth set, defined as accounts meeting at least two of the following criteria: (1) more than three cancelations within a 48 h window, (2) a single device ID associated with more than five medical card IDs, or (3) registration completion speeds faster than the 99th percentile of manual operation times (e.g., <2 s). These labels were manually verified by a cross-departmental audit team to minimize misclassification in the training set. Statistical comparisons were made using χ2 tests, trend analyses, and bootstrap resampling to verify robustness.

After the anti-scalping mechanism was implemented, the appointment completion rate increased from 64.6% (2019) to 78.0% (2022) (χ2 = 46.27, p < 0.001), while the no-show rate decreased from 35.4 to 22.0%. Online registration rose from 65.6% (2020) to 94.5% (2023), and the proportion of first-time patients increased by 18.7%, indicating improved fairness in access to medical services. To explicitly measure this, we utilized the Equal Opportunity metric, ensuring the model’s true positive rate was consistent across different age and socio-economic groups. Furthermore, a sensitivity check on false positives (FPs) revealed that only 0.4% of legitimate users were flagged; for these cases, a secondary SMS-based challenge-response mechanism was implemented to ensure that legitimate access remained unblocked, thus mitigating the risk of systemic bias. While this trend aligns with the global shift toward digital healthcare prompted by the COVID-19 pandemic, the synchronized reduction in suspicious behavior patterns suggests the algorithm’s specific contribution to system integrity. The Bayesian model achieved an AUC of 0.91 (95% CI 0.89–0.93) with 89.2% sensitivity and 85.1% specificity, accurately identifying abnormal registrations. Unlike general increases in online volume caused by external factors like the pandemic, the model’s discriminatory power is rooted in specific behavioral features (e.g., millisecond-level operation intervals and abnormal device-ID diversity) that distinguish automated scalping from legitimate patient behavior, regardless of the overall digital environment. Bootstrap analysis confirmed the reliability of the improvements, showing a mean +12.9% increase in completion rate and −13.4% reduction in no-show rate (95% CIs not crossing zero).

Integrating Bayesian inference into outpatient appointment systems can effectively enhance fairness, efficiency, and patient trust by reducing speculative bookings and improving accessibility. The findings demonstrate that data-driven management frameworks can bridge the gap between algorithmic modeling and real-world hospital operations, providing a replicable strategy for intelligent, equitable healthcare resource allocation.