Development and Weighting of a Task-Nurse-Nursing Robot Fit Indicator System and a Three-Dimensional Fit Calculation Framework for Biosafety Emergency Response: A Delphi and Fuzzy Analytic Hierarchy Process Study (Preprint)

<sec> <title>BACKGROUND</title> <p>Biosafety emergency rescue scenarios feature high infection risks and extremely low operational error tolerance, making traditional routine nursing human-robot matching criteria inapplicable to on-site task allocation. This study therefore constructs a three-dimensional indicator system covering task, nurse and robot dimensions to quantitatively evaluate human-robot-task matching levels during biosafety emergency response.</p> </sec> <sec> <title>OBJECTIVE</title> <p>To develop a task-nurse-nursing robot fit indicator system for biosafety emergency response, determine the relative weights of indicators at each level, and propose an operational three-dimensional fit calculation framework as a basis for developing and validating a task-matching assessment tool.</p> </sec> <sec> <title>METHODS</title> <p>An initial indicator framework was developed through theoretical analysis, literature review, and qualitative interviews. A two-round Delphi survey of 15 experts was conducted to screen and revise the indicators, and Kendall's coefficient of concordance (W) was used to assess agreement among the experts. Fuzzy analytic hierarchy process was then used to construct aggregate judgment matrices, calculate indicator weights, and assess matrix compatibility. A conceptual fit calculation framework was further developed for three candidate execution modes: nurse-only, nursing robot-only, and nurse-robot collaborative execution. The framework was based on requirement-capability mapping, weighted capability deficits, and safety veto rules.</p> </sec> <sec> <title>RESULTS</title> <p>The initial framework comprised 3 first-level, 10 second-level, and 23 third-level indicators; after two Delphi rounds, the final system comprised 3 first-level, 9 second-level, and 36 third-level indicators. The valid response rate was 100% in both rounds, and the expert authority coefficient was 0.907. Kendall's W values for the first-, second-, and third-level indicators were 0.325, 0.313, and 0.338 in round 1 and 0.496, 0.386, and 0.386 in round 2, respectively (all P &lt; 0.01), indicating good expert agreement. The initial weights of the first-level indicators were 0.3822 for nursing task characteristics, 0.3522 for nurse capabilities, and 0.2656 for nursing robot performance. The compatibility indices of all aggregate judgment matrices were &lt;0.1. Based on the indicators and weights, a three-dimensional fit decision process comprising requirement-capability mapping, hard-constraint screening, capability-deficit calculation, and ranking of candidate execution modes was proposed.</p> </sec> <sec> <title>CONCLUSIONS</title> <p>This study developed a preliminary task-nurse-nursing robot fit indicator system, relative indicator weights, and a conceptual fit calculation framework. The aggregate judgment matrices demonstrated acceptable internal compatibility. The framework provides a computable pathway for developing a task-matching assessment tool; however, the mapping relationships, rating anchors, thresholds, measurement properties, and practical scheduling utility require further calibration and validation through simulation exercises and multicenter studies.</p> </sec>