📄 medicine

Factors Influencing Clinical Nurses’ Intention to Use and Actual Use of Generative Artificial Intelligence in China: A Multicentre Cross-sectional Study

This multicentre cross-sectional study of 1,921 clinical nurses in China reveals that their intention to use and actual adoption of generative artificial intelligence are significantly driven by perceived usefulness, ease of use, social influence, and trust, while being hindered by perceived risks, highlighting the need for strategies that enhance value and address safety concerns to facilitate AI integration in nursing practice.

Original authors: Xiaoyan Gong, Lili Wu, Shan Li, Rongping Cha, Yijing Weng, Yeru Xia, Yunxia Shen, Jingbang Liu

Published 2026-06-25

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Xiaoyan Gong, Lili Wu, Shan Li, Rongping Cha, Yijing Weng, Yeru Xia, Yunxia Shen, Jingbang Liu

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a hospital ward as a busy kitchen. For years, the chefs (nurses) have been cooking with traditional tools. Now, a new, super-smart robot assistant (Generative AI) has arrived, promising to chop vegetables, write recipes, and even suggest new dishes. But before the chefs start using this robot, they have to decide: Is it actually helpful? Is it easy to use? Do I trust it not to burn the food? And am I even allowed to use it?

This study is like a giant survey sent to nearly 2,000 chefs in Chinese hospitals to answer exactly those questions. The researchers wanted to know what makes a nurse say "Yes, I'll use this AI" and what actually stops them from using it.

Here is what they found, broken down into simple concepts:

The Main Ingredients for Adoption

The researchers used a popular recipe for understanding new technology called UTAUT (Unified Theory of Acceptance and Use of Technology). Think of this as a checklist for whether someone will try a new gadget. They added a few extra ingredients specific to AI: Trust and Risk.

Here are the four main things that made nurses want to use the AI:

The "Does it help me?" Factor (Performance Expectancy):
- The Analogy: If the robot can chop onions in 2 seconds instead of 2 minutes, the chef is happy.
- The Finding: This was the biggest reason nurses wanted to use AI. If they believed the tool would actually make their work easier or better, they were much more likely to try it.
The "Is it hard to learn?" Factor (Effort Expectancy):
- The Analogy: If the robot has a complicated remote control with 50 buttons, the chef won't touch it. If it just has one "Start" button, they will.
- The Finding: Nurses wanted to know the tool was easy to use. If it felt simple and intuitive, they were more willing to adopt it.
The "Can I trust it?" Factor (Technology Trust):
- The Analogy: Would you let a robot cook your baby's food if you weren't sure it wouldn't add poison?
- The Finding: Nurses needed to believe the AI was reliable and safe. If they thought the AI might make up facts (hallucinations) or give wrong medical advice, they hesitated.
The "What do others think?" Factor (Social Influence):
- The Analogy: If the head chef and the whole kitchen staff are raving about the new robot, you're more likely to try it too.
- The Finding: If colleagues and managers encouraged using AI, nurses were more likely to follow suit, though this mattered less than usefulness or ease of use.

The "Stop Sign"

There was one major factor that acted like a brake pedal: Perceived Risk.

The Analogy: If the chef thinks the robot might leak data, steal secrets, or hurt a patient, they won't use it, no matter how fast it is.
The Finding: Fear of mistakes, privacy leaks, or ethical issues significantly lowered the intention to use the AI.

The "Push" from the Kitchen Manager

Even if a nurse wanted to use the AI, they couldn't do it without the right tools.

The Finding: Facilitating Conditions (having the right computers, internet, and training) were crucial. You can have the best recipe, but if you don't have a stove, you can't cook.

Who Uses It Differently?

The study noticed that different groups of nurses reacted differently, like different types of cooks:

Men vs. Women:
- Women seemed to care more about whether the tool was useful and what their colleagues thought.
- Men (though there were fewer of them in the study) seemed to care more about whether they could trust the machine.
Young vs. Old:
- Younger nurses (20–30s) were driven by how useful and easy the tool was.
- Older nurses (41+) were much more sensitive to risk. As they got older, the fear of making a mistake or the tool being unsafe became a bigger barrier. Interestingly, older nurses trusted the technology less than the younger ones did.
Willing vs. Unwilling:
- Nurses who were already willing to use AI responded to all the usual factors (usefulness, ease, trust).
- Nurses who were unwilling or neutral didn't seem to care about usefulness or trust at all. For them, the only thing that mattered was if they were forced to use it or if the system made it impossible to avoid.

The Bottom Line

The study concludes that for AI to become a regular part of a nurse's daily routine, hospitals need to do three things:

Show the value: Prove it actually saves time and helps patients.
Build trust: Make sure the AI is safe, reliable, and explainable.
Manage the fear: Address concerns about privacy and errors directly.

The researchers also noted that because they only asked nurses once (a "snapshot" in time), they can't say for sure that these factors cause the usage, only that they are strongly linked to it. They suggest future studies should watch nurses over a longer period to see how these feelings change.

In short: Nurses are ready to use AI, but only if it's helpful, easy, safe, and if they feel their hospital supports them while they learn to use it.

Technical Summary: Factors Influencing Clinical Nurses' Intention to Use and Actual Use of Generative Artificial Intelligence in China

Problem Statement
While generative artificial intelligence (GenAI) holds transformative potential for nursing practice—ranging from documentation automation to clinical decision support—empirical evidence regarding the specific factors driving its adoption among clinical nurses remains limited. Existing literature often addresses AI adoption in general healthcare contexts or mixed populations, failing to account for the unique professional characteristics of nurses, who constitute the largest group of healthcare professionals. Furthermore, the integration of GenAI introduces specific challenges, including risks of information hallucinations, bias, and ethical concerns, which may influence user acceptance differently than traditional technologies. There is a critical need to understand the determinants of both behavioral intention and actual use behavior among clinical nurses to facilitate successful implementation.

Methodology
This study employed a multicentre cross-sectional survey design conducted in China between July and August 2025.

Participants: A total of 1,921 valid responses were collected from registered nurses across various healthcare institutions using convenience sampling.
Theoretical Framework: The research utilized an extended Unified Theory of Acceptance and Use of Technology (UTAUT) model. The original UTAUT constructs (Performance Expectancy, Effort Expectancy, Social Influence, and Facilitating Conditions) were augmented with Technology Trust and Perceived Risk to better reflect the specific context of GenAI in nursing.
Instrument: A structured questionnaire was developed based on validated UTAUT instruments and refined through a two-round Delphi consultation with 15 nursing experts. It measured seven dimensions: Performance Expectancy (PE), Effort Expectancy (EE), Social Influence (SI), Facilitating Conditions (FC), Technology Trust (TR), Perceived Risk (PR), and Behavioral Intention (BI). Actual Use Behavior (UB) was measured via self-reported frequency and scope of GenAI use.
Analysis: Data were analyzed using Structural Equation Modelling (SEM) to test hypothesized relationships. Reliability and validity were assessed via Cronbach's alpha, Composite Reliability (CR), Average Variance Extracted (AVE), and the Fornell-Larcker criterion. Multi-group analyses were performed to examine moderating effects of gender, age, and voluntariness of use.

Key Contributions

Extended UTAUT Model for GenAI: The study integrates Technology Trust and Perceived Risk into the standard UTAUT framework, providing a more nuanced model for understanding the adoption of generative AI specifically within clinical nursing.
Dual Focus on Intention and Behavior: Unlike many studies that focus solely on intention, this research examines the pathway from behavioral intention to actual use behavior, identifying Facilitating Conditions as a critical bridge.
Demographic Heterogeneity: The study provides granular insights into how adoption drivers vary across gender, age, and voluntariness, challenging the assumption of uniform technology acceptance among nurses.

Results

Determinants of Behavioral Intention:
- Performance Expectancy (PE) was the strongest positive predictor of behavioral intention ( $\beta = 0.302, p < 0.001$ ).
- Effort Expectancy (EE) ( $\beta = 0.235, p < 0.001$ ) and Technology Trust (TR) ( $\beta = 0.207, p < 0.001$ ) also showed significant positive associations.
- Social Influence (SI) had a smaller but significant positive effect ( $\beta = 0.094, p = 0.004$ ).
- Perceived Risk (PR) was significantly negatively associated with behavioral intention ( $\beta = -0.134, p < 0.001$ ).
Determinants of Actual Use Behavior:
- Behavioral Intention (BI) was the primary driver of actual use ( $\beta = 0.409, p < 0.001$ ).
- Facilitating Conditions (FC) (e.g., organizational support, resources) were also significantly associated with actual use ( $\beta = 0.289, p < 0.001$ ), independent of intention.
Moderating Effects:
- Gender: For female nurses, PE, EE, SI, and TR were significant predictors. For male nurses, only EE and TR were significant; PE and SI were not. Technology trust appeared more salient for male nurses.
- Age: Perceived risk became a more significant negative predictor with increasing age. Conversely, the influence of Technology Trust weakened among nurses aged 41 and older.
- Voluntariness: In the "willing" group, all hypothesized paths were significant. In "neutral" and "unwilling" groups, most antecedent variables lost statistical significance, suggesting that intrinsic motivation alters the adoption mechanism.

Significance and Claims
The authors claim that the adoption of GenAI by clinical nurses is a multifaceted process driven primarily by perceived usefulness, ease of use, and trust, while being hindered by perceived risks. The study posits that successful integration requires strategies that go beyond simple training; they must actively address safety concerns and build institutional trust.

The paper emphasizes that organizational support (Facilitating Conditions) is crucial for translating intention into actual practice. Furthermore, the authors assert that a "one-size-fits-all" approach is insufficient; implementation strategies must be tailored to specific subgroups, such as prioritizing risk management for older nurses and enhancing perceived value for younger cohorts.

Limitations
The authors modestly acknowledge several limitations:

Causality: The cross-sectional design precludes causal inferences.
Sampling: The use of convenience sampling may introduce selection bias, potentially over-representing nurses with a pre-existing interest in AI.
Self-Reporting: Data reliance on self-reported measures may introduce common method bias, despite mitigation efforts.
Generalizability: The study was conducted within a single national context (China), which may limit the applicability of findings to other healthcare systems.

The paper concludes by calling for longitudinal research to better understand causal relationships and the evolution of AI adoption over time.