Artificial Intelligence Legal Literacy and Procedure-Concordant Use of an AI-Enabled Early-Warning System Among Hospital Nurses in Romania

Background and Objectives: Regulation (EU) 2024/1689, the Artificial Intelligence Act, requires organisations deploying AI systems to ensure appropriate AI literacy and human oversight. We examined whether AI legal literacy was associated with procedure-concordant use of an AI-enabled early-warning system among hospital nurses and whether the association of prior AI-governance training with concordant use persisted after literacy was taken into account. Materials and Methods: This single-centre cross-sectional study included 87 nurses in a tertiary university hospital. AI legal literacy was measured with a study-specific 24-item instrument, and procedure-concordant AI use, the primary outcome, was defined as meeting all five criteria of the institutional documentation audit in 10 audited interactions per nurse. The primary analysis was multivariable logistic regression, confirmed in three-level mixed-effects models of 870 audited interactions nested within nurses and wards. Secondary and exploratory analyses (inverse-probability weighting, mediation decomposition, latent class analysis, and an internally validated prediction model) are reported as hypothesis-generating. Results: Forty-four nurses (50.6%) were procedure-concordant. Concordance was higher among trained than untrained nurses (70.3% versus 36.0%; p = 0.002). Each 10-point increase in AI legal literacy was associated with higher adjusted odds of concordant use (adjusted odds ratio 5.16, 95% CI 2.04–13.06; p < 0.001). The association persisted in three-level mixed-effects models (aOR 3.61 per 10 points, 95% CI 1.98–6.58). The association with prior training was attenuated after adjustment for literacy and was not statistically significant after inverse-probability weighting (OR 2.01, 95% CI 0.82–4.90); an exploratory decomposition was consistent with, but could not establish, an indirect association through literacy. Conclusions: Higher AI legal literacy was strongly associated with procedure-concordant AI use in this single-centre study. The findings support further evaluation of literacy-focused governance education and institutional oversight in hospital nursing practice; causal and predictive claims await prospective, multicentre confirmation.