Related Experiment Videos
Expanded Longitudinal Validation of Adaptive Learning Stability in AI-Assisted Neurosurgical Residency Training
Baglan Mustafayev1, Alina Mustafayeva2, Askar Bakhtiyarov3
1Department of Neurosurgery and Neuropathology, National Centre for Neurosurgery, Astana, Kazakhstan.
Objective:
To evaluate the longitudinal stability, reproducibility, and robustness of Adaptive Learning Stability (ALS) within an expanded AI-assisted neurosurgical residency cohort during a 12-month observation cycle.
Design:
This was a prospective controlled educational study. Residents completed 4 consecutive quarterly checkpoints integrating cognitive, technical, and decision-making assessments over 12 months. The RCKaS-AI framework incorporated predictive analytics, individualized adaptive remediation, and mentor-validated feedback pathways. Primary analysis evaluated longitudinal entrustment trajectory stability across checkpoints. Secondary analyses included inter-resident variance dynamics, remediation frequency, faculty workload, and predictive agreement between forecasted and observed competency progression.
Setting:
Two accredited neurosurgical residency programs in Kazakhstan, from January 2025 to January 2026.
Participants:
Sixty-four neurosurgical residents randomized into standard RCKaS (n = 32) and AI-enhanced RCKaS-AI (n = 32) cohorts.
Results:
Both cohorts demonstrated progressive improvement in entrustment scores across sequential checkpoints. In the RCKaS-AI cohort, mean entrustment scores increased from 59.1 ± 12.8 at CP1 to 94.3 ± 6.9 at CP4, compared with 58.4 ± 13.1 to 85.7 ± 10.4 under standard RCKaS conditions (p = 0.001). Inter-resident variance decreased by 41% in the AI-assisted cohort versus 18% in the standard cohort. Residents within the RCKaS-AI framework achieved entrustment level 4 significantly earlier than controls (38 ± 7 weeks vs 50 ± 9 weeks, p = 0.003). Mean remediation frequency decreased from 3.4 ± 1.1 to 1.9 ± 0.7 sessions per resident, while faculty workload declined by 34%. Predictive analytics demonstrated strong agreement between forecasted and observed competency trajectories (r = 0.84; 95% CI, 0.77-0.89).
Conclusions:
Expanded longitudinal implementation of the RCKaS-AI framework demonstrated reproducible stabilization of competency trajectories during neurosurgical residency training. These findings support the robustness of adaptive AI-assisted feedback systems as a practical mechanism for longitudinal competency regulation within structured surgical education programs. Further multicenter validation is required to evaluate external reproducibility across broader residency environments.