Development and Clinical Validation of an Entrustable Professional Activities Framework for Respiratory Nurses: A Longitudinal Workplace-Based Study

Background: Entrustable professional activities (EPAs) are increasingly used in competency-based assessment, but evidence remains limited for nursing workplace-based assessment (WBA) systems that can capture learning trajectories while providing defensible reliability for programmatic decisions. Objective: The objective of this study is to develop a respiratory nursing EPA framework and clinically validate core EPAs using longitudinal WBA data, including performance progression, attainment patterns, and reliability requirements based on generalizability (G) theory and decision (D) studies. Methods: This study included two phases: (1) EPA framework development using evidence synthesis and a two-round Delphi consultation and (2) longitudinal clinical validation using WBA data. In the validation phase, 30 respiratory nurses were assessed by five trained raters over 12 weeks on four core EPAs (EPA1, EPA2, EPA5, and EPA11) using a 1–5 entrustment scale (attainment defined as a score ≥ 4). Learning trajectories were modeled using a negative exponential function (learning speed indexed by β0). Reliability was evaluated using a crossed-facet G-study framework (person × rater × occasion) with mixed-effects modeling, and D-studies were used to identify minimal rater–occasion configurations meeting EP2 thresholds. Results: In the Delphi phase, 21 experts completed both rounds (100% response rate), with high authority coefficients (Cr = 0.841 and 0.843) and significant inter-expert agreement (Kendall’s W = 0.504 and 0.532; both p < 0.01), yielding a final 13-item respiratory nursing EPA framework. In the validation phase, 1392 core EPA assessments were analyzed, and event-level attainment ranged from 72.5% to 89.1%. Learning speeds were heterogeneous (β0 = 0.038–0.156), with the fastest improvement in EPA11 and the slowest in EPA2. In the G-study, residual variance accounted for 65.05% of the total variance, followed by person (18.21%) and occasion/week (10.13%) effects, and the main effect on rater was small (0.14%). Under the observed protocol (five raters and approximately four ratings/week), reliability reached acceptable levels (G = 0.792; EP2 = 0.712). The minimum D-study configurations were one rater × 11 ratings/week for EP2 ≥ 0.70 (EP2 = 0.703) and two raters × 12 ratings/week for EP2 ≥ 0.80 (EP2 = 0.812). Conclusions: The respiratory nursing EPA framework showed strong expert-consensus support and promising longitudinal clinical validation evidence. A core EPA-based WBA approach demonstrated developmental sensitivity and acceptable dependability under feasible sampling plans, supporting its use for structured competency assessment and progression monitoring in respiratory nursing training.