Innovative behavior among clinical nurses: empirical level stratification and associated factors in a cross-sectional study in Zhejiang Province, China

Innovation in nursing practice is closely related to quality of care, yet nurses working in high-demand clinical environments may show relatively low levels of innovative behavior. Previous studies have primarily described nurses’ innovative behavior using total scores, with less attention to individual heterogeneity. This study used latent profile analysis (LPA) to identify empirically derived levels of innovative behavior among clinical nurses and to examine their statistical associations with organizational justice, job embeddedness, and other resource-related factors. A convenience sample of 422 clinical nurses was recruited from 27 tertiary hospitals in 10 cities in Zhejiang Province, China, between December 2025 and March 2026. Data were collected using a demographic questionnaire, the Nurse Innovative Behavior Scale, the Organizational Justice Scale, and the Job Embeddedness Scale. LPA was conducted using the mean scores of the three dimensions of innovative behavior as observed indicators. Univariate analyses and exploratory multinomial logistic regression were used to examine factors associated with profile membership. The three-profile solution was retained for descriptive stratification: 61.85% ( n = 261) were classified as the relative low-innovation level (C1), 31.99% ( n = 135) as the relative moderate-innovation level (C2), and 6.16% ( n = 26) as the relative high-innovation level (C3). These labels are sample-relative descriptors and do not represent validated clinical or managerial cutoffs. The three profiles differed significantly in total and dimensional scores of organizational justice and job embeddedness (all P < 0.001). These between-profile comparisons are descriptive and do not establish temporal or causal relationships. In exploratory multinomial logistic regression, organizational justice remained associated with profile membership after adjustment: for each one-point increase in organizational justice, the odds of C2 versus C1 were 1.081 (95% CI 1.044–1.120, P < 0.001), and the odds of C3 versus C1 were 1.285 (95% CI 1.162–1.422, P < 0.001). Nurses without an administrative position had lower odds of C2 versus C1 than nurses at head nurse or above (OR = 0.497, 95% CI 0.259–0.952, P = 0.035). Job embeddedness, research training, and monthly income were not statistically associated with profile membership after adjustment. The continuous one-factor model had a lower BIC than the three-profile model, while the four-profile solution showed better information-criterion fit than the three-profile solution. Accordingly, the three-profile solution should be interpreted as empirical stratification along a continuous innovative-behavior dimension rather than as mutually exclusive discrete types. In this sample, nurses’ innovative behavior was predominantly at a relatively low level and could be empirically stratified into low, moderate, and high levels based on the joint information from three dimensions. However, comparisons with the continuous model and the four-profile solution indicate that this stratification should not be regarded as the unique or optimal discrete latent structure. Organizational justice showed the most consistent association with profile membership across both comparisons, although the estimates were based on hard class assignment and the high-innovation profile was small. Whether organizational justice can serve as a modifiable intervention target requires prospective or intervention-based evaluation. For nursing management, a fair and supportive organizational environment may warrant further evaluation as a potential management priority, while profile stratification may be useful for descriptive management and hypothesis generation.