Comparative assessment of three standardized robotic surgery training methods

Objectives To evaluate three standardized robotic surgery training methods, inanimate, virtual reality and in vivo, for their construct validity. To explore the concept of cross-method validity, where the relative performance of each method is compared. Materials and Methods Robotic surgical skills were prospectively assessed in 49 participating surgeons who were classified as follows: 'novice/trainee': urology residents, previous experience <30 cases (n = 38) and 'experts': faculty surgeons, previous experience ≥30 cases (n = 11). Three standardized, validated training methods were used: (i) structured inanimate tasks; (ii) virtual reality exercises on the da Vinci Skills Simulator (Intuitive Surgical, Sunnyvale, CA, USA); and (iii) a standardized robotic surgical task in a live porcine model with performance graded by the Global Evaluative Assessment of Robotic Skills (GEARS) tool. A Kruskal-Wallis test was used to evaluate performance differences between novices and experts (construct validity). Spearman's correlation coefficient (ρ) was used to measure the association of performance across inanimate, simulation and in vivo methods (cross-method validity). Results Novice and expert surgeons had previously performed a median (range) of 0 (0-20) and 300 (30-2000) robotic cases, respectively (P < 0.001). Construct validity: experts consistently outperformed residents with all three methods (P < 0.001). Cross-method validity: overall performance of inanimate tasks significantly correlated with virtual reality robotic performance (ρ = -0.7, P < 0.001) and in vivo robotic performance based on GEARS (ρ = -0.8, P < 0.0001). Virtual reality performance and in vivo tissue performance were also found to be strongly correlated (ρ = 0.6, P < 0.001). Conclusions We propose the novel concept of cross-method validity, which may provide a method of evaluating the relative value of various forms of skills education and assessment. We externally confirmed the construct validity of each featured training tool.