Abstract
-
Purpose
This study aimed to develop an artificial intelligence (AI)-based pediatric medication safety competence (PMSC) virtual simulation program and evaluate its effectiveness among nursing students.
-
Methods
This quasi-experimental study used a nonequivalent control group pretest-posttest design and was conducted in December 2025. The AI-based PMSC virtual simulation program was developed based on Kolb’s experiential learning theory, following its 4-stage cycle of concrete experience, reflective observation, abstract conceptualization, and active experimentation. Participants were 48 third-year nursing students in South Korea who were assigned to either the experimental group (n=24) or the control group (n=24) through a nonrandomized allocation procedure. The experimental group participated in the PMSC virtual simulation program. Outcomes included PMSC, medication calculation confidence (MCC), Simulation Task Load Index (SIM-TLX), and learning immersion (LI). Data were collected before and after the intervention using validated instruments, and statistical tests were used to evaluate program effectiveness.
-
Results
Compared with the control group, the experimental group showed significant improvement in PMSC (t=2.48, p=.017). No significant between-group differences were observed in MCC (t=−0.87, p=.389) or LI (t=0.97, p=.336). In addition, SIM-TLX scores decreased significantly from pretest to posttest in the experimental group (t=−2.61, p=.016).
-
Conclusion
The AI-based PMSC virtual simulation program may help strengthen PMSC among nursing students and may be suitable for use in pediatric nursing education. Future research should refine the program to better address individual learning needs and integrate complementary simulation-based strategies into pediatric medication nursing education.
-
Key words: Artificial intelligence; Nursing students; Patient safety; Pediatric nursing; Simulation training
INTRODUCTION
Pediatric medication administration is recognized as a complex, multidisciplinary, and multistage process that requires comprehensive professional knowledge, skill, clinical judgement and attitude across the medication process, including a thorough understanding of pharmacological properties [
1]. In particular, pediatric patients are more vulnerable to medication errors than adults because of weight-based dose calculations, the need for dilution due to limited standardized pediatric formulations, and communication differences associated with developmental stages [
2].
In addition, pediatric patients exhibit considerable pharmacokinetic variability [
3]. Their renal and hepatic functions are immature, which makes drug metabolism and excretion different from those of adults, complicating dosage adjustment [
2,
3]. Due to these characteristics, the incidence of medication errors in pediatric patients has been reported to be approximately 3 times higher than in adults, highlighting medication safety as a particularly critical issue in pediatric care settings [
1]. Medication safety competence in pediatric nursing extends beyond medication administration and requires clinical reasoning, critical thinking, and highly proficient technical skills [
4]. In particular, weight-based dosage calculation is the area prone to error in pediatric drug administration, and as a single calculation error can lead to a serious adverse drug event, which should be treated as a fundamental core competency [
5].
Nursing students develop medication competence through pediatric nursing theory and practice in the undergraduate curriculum. However, pediatric medication management is still perceived as a challenging task [
2]. Even when nursing students possess substantial theoretical knowledge, they often find it difficult to directly perform medication management tasks—such as medication preparation and administration—during clinical practicum, and they have limited opportunities to observe how medication administration is handled or reported [
5]. As clinical placements in pediatric wards are limited, students may have fewer opportunities to gain sufficient experience in administering medications to pediatric patients [
4]. In addition, limited opportunities for active participation may make it difficult for students to translate theoretical learning into practical competence [
4,
5].
Pediatric medication safety administration involves a complex nursing process that includes confirming prescriptions, calculating doses, preparing and diluting medications, administering drugs, monitoring, and communicating in a manner appropriate to the child’s developmental stage [
1,
4,
5]. In pediatric nursing practice, safe medication administration requires not only conceptual understanding but also the ability to apply knowledge in complex clinical situations. Therefore, multifaceted educational strategy with integration of theoretical knowledge and practical skills is needed for nursing students to systematically develop safety competencies in pediatric medication administration.
Medication-focused nursing education programs are increasingly incorporating digital instruction aligned with Generation Z (individuals generally born between the mid-1990s and the early 2010s) to enhance performance capabilities [
6]. In particular, programs incorporating virtual reality (VR), augmented reality, and gamification have been shown to promote learners’ immersion, motivation, and skill acquisition [
7,
8]. In pediatric nursing, various approaches—such as VR-based learning for injection skills and simulation of medication dose calculation and injection techniques have been developed and used to enhance clinical performance competence [
9,
10]. However, previous simulation-based education is often structured around predetermined scenarios, which may limit opportunities for interactive and customized learning that reflect learners’ competency levels and characteristics of pediatric patients.
To address these challenges, artificial intelligence (AI)–based education is gaining attention. AI is being applied across various fields, including education, and is introducing new approaches that can transform existing educational methodologies and enhance learning experiences [
11]. Recently, AI has been used in education through learning applications, virtual patients, AI chatbots, simulations, and platforms such as Chat Generative Pre-trained Transformer (GPT) [
11]. It can support learners by providing customized feedback and opportunities for repeated practice appropriate to their level. By analyzing learners’ input in real time, AI can identify patterns of errors in clinical performance and reasoning. It can also promote adaptive learning by providing educational content tailored to students’ individual needs and learning styles [
12]. Orkaby et al. [
13] also suggested that AI could be used as a verification tool to prevent medication errors. Therefore, simulations that incorporate AI tools could be advantageous for strengthening competencies in drug selection, dosage calculation, communication, and technical execution in pediatric medication administration, where the risk of errors is high [
12,
13].
In particular, AI-based simulations enable repeated practice in a controlled environment, thereby enhancing clinical preparedness and understanding [
14]. They can promote learners’ critical thinking and evidence-based decision-making by posing clinical questions such as “Why did you choose this medication?” and “Why did you select this route of administration?” [
14]. In addition, virtual pediatric scenarios can support personalized learning and provide experiences that approximate real clinical settings while enhancing learner autonomy and immersion without time or space constraints [
9,
11]. Moreover, interactive communication between the AI system and learners enables repeated practice of medication administration across diverse situations [
15].
Thus, integrating AI with simulation-based instruction offers a flexible learning experience in which learners progress at their own pace [
16]. This approach enables repeated practice and targeted feedback on the steps in the pediatric medication administration process. In addition, the integration of AI-based simulation with subsequent real-world simulation may support the development of pediatric medication safety competence (PMSC) among nursing students by providing opportunities for both experiential practice and applied skill reinforcement [
15].
The experiential learning theory (ELT) by Kolb [
17] describes the process through which knowledge is constructed by transforming experience. Experiential learning is frames learning as an ongoing process rather than a fixed outcome, emerging through continuous interaction between the learner and the environment. This theory describes a 4-stage cycle comprising concrete experience, reflective observation, abstract conceptualization, and active experimentation. Learners may enter the cycle at any stage, and individual variation may occur [
17].
The effectiveness of pediatric medication safety education could be evaluated by concurrently measuring comprehensive PMSC (from prescription verification to monitoring and communication), medication calculation confidence (MCC) [
18], Simulation Task Load Index (SIM-TLX) [
19] and process indicators such as learning immersion (LI) [
20], which reflects the extent to which learners are cognitively and emotionally engaged with the learning environment. Guided by this framework [
17], this study aimed to develop and evaluate its effectiveness of an AI-based PMSC virtual simulation program for nursing students.
This study aimed to develop an AI-based PMSC virtual simulation program for nursing students and to evaluate its effectiveness.
2. Research Hypotheses
Hypothesis 1: Nursing students who participate in the AI-based PMSC virtual simulation program will demonstrate greater improvement in PMSC compared to those in the control group.
Hypothesis 2: Nursing students who participate in the AI-based PMSC virtual simulation program will demonstrate greater MCC compared to those in the control group.
Hypothesis 3: Nursing students who participate in the AI-based PMSC virtual simulation program will demonstrate lower SIM-TLX scores compared to preintervention.
Hypothesis 4: Nursing students who participate in the AI-based PMSC virtual simulation program will demonstrate greater LI compared to those in the control group.
3. Conceptual Framework
Grounded in Kolb’s ELT [
21], the program was organized into 4 sequential stages. In the abstract conceptualization stage, learners attended a lecture covering core theories and principles of pediatric medication safety and dosage calculation, intended to establish foundational knowledge and to cultivate appropriate attitudes toward safe nursing practice. In the concrete experience stage, learners worked through pediatric medication administration scenarios in the AI-based PMSC virtual simulation to apply lecture content within an authentic clinical context. AI-driven learning and real-time feedback were incorporated to promote PMSC, MCC, and LI. In the active experimentation stage, learners practiced the same scenario in an authentic simulation environment to internalize knowledge and strengthen practical confidence. Finally, in the reflective observation stage, learners reviewed their performance to integrate learning experiences and deepen their commitment to safe nursing practice. Given the cyclical linkage among the 4 stages, this study also examined whether the program effectively enhanced LI and lower SIM-TLX. On this basis, the study developed and evaluated an AI-based PMSC virtual simulation program designed to strengthen nursing students’ PMSC, MCC, SIM-TLX, and LI (
Figure 1,
Supplement 1).
METHODS
Ethical statements: This study was approved by the Institutional Review Board (IRB) of Chonnam National University (IRB No. 1040198-251124-HR-231-02). Informed consent was obtained from all participants.
1. Study Design
This study used a nonequivalent control group pretest–posttest design to develop and evaluate an AI-based PMSC virtual simulation program grounded in the ELT by Kolb [
17]. Reporting followed the Transparent Reporting of Evaluations with Nonrandomized Designs (TREND) Statement guidelines [
22].
Participants were third-year nursing students recruited from universities located in 2 cities in South Korea who had completed a pediatric nursing theory course. Volunteers were recruited through notices posted in nursing-student community channels and were assigned to the experimental or control group using a nonrandom allocation method after they understood the study purpose and voluntarily agreed to participate. The sample size was calculated using G*Power ver. 3.1.9.7 (Heinrich-Heine-Universität Düsseldorf) with a significance level of α=.05 and statistical power of .80. Assuming an effect size of Cohen’s d=0.85, the minimum required sample size was 22 participants per group [
23]. Allowing for an attrition rate of approximately 10%, 24 participants were recruited and allocated to each group at the outset, for a total of 48 participants.
Participants were recruited from December 12 to 14, 2025, through advertisements posted to online community for nursing studies and SNS (Social Network Service); nursing students KakaoTalk open chat rooms. With administrators’ permission, an online questionnaire (Google link; Google LLC) that included study information and an electronic consent form was posted to nursing students’ online communities. Interested students accessed the study summary via the Google link, which described the study purpose, procedures, right to withdraw, and personal information protection policies. Students who provided consent completed a self-administered preintervention questionnaire.
Participants were allocated to groups using a nonrandom convenience method. Following voluntary recruitment, the researcher allocated participants sequentially in the order of enrollment. After allocation, 48 students were assigned to the experimental and control groups (24 per group); no participants withdrew, and all were included in the final analysis. The survey was designed to be completed in approximately 15 minutes (
Figure 2).
Both groups completed the preintervention survey before participating in the study. Immediately after completing the AI-based PMSC virtual simulation session, the experimental group completed a follow-up questionnaire. The control group completed a follow-up questionnaire after their online lecture learning experience. Following the completion of data collection for the control group (December 16–17, 2025), data collection for the experimental group (December 18–23, 2025) was subsequently conducted. This sequential approach was adopted to minimize the risk of information contamination between groups. Specifically, by prioritizing data collection for the control group, we reduced the likelihood of participants being exposed to the AI-based PMSC virtual simulation content before receiving their assigned intervention.
4. Intervention Development
The Analysis, Design, Development, Implementation, and Evaluation (ADDIE) model of instructional design [
24] was followed in this study. The phases were as follows. (1) Analysis phase: A comprehensive literature review was conducted on medication nursing education and medication administration programs and simulation-based learning to guide the development of the program. (2) Design and development phase: The intervention consisted of 5 sequential sessions: a pediatric medication lecture (Session 1), AI-based virtual simulation (Sessions 2 and 3), a real-world simulation (Session 4), and a structured debriefing (Session 5) (
Figure 1).
The AI-based PMSC virtual simulation program was developed in 2 stages: (1) development of a scenario-based 3-dimensional virtual environment integrated with generative AI and (2) construction of an AI-driven learning and interaction system. First, pediatric medication focused clinical scenarios (e.g., intravenous antibiotic administration in a general ward and the administration of high-alert intravenous medications; anticonvulsants and vasopressors, in a pediatric intensive care unit [ICU]) was developed. Unity-based virtual environment was created to provide an authentic pediatric clinical setting (a pediatric ward and ICU) to mirror an authentic pediatric-nursing practice environment, which is used in both game development and medical simulation.
The AI-based PMSC virtual simulation program was developed to enable interactive, scenario-based learning through avatar-driven engagement. Within each scenario, learners interacted through an avatar and engaged in step-by-step decision-making process related to medication administration. The program was designed to allow learners to assess the clinical situation, solve medication-related tasks, and make sequential decisions to complete the medication administration process. To support this, children’s avatars in the simulation were implemented as Unity-based non-player characters such as “a 3-month-old infant and a 13-month-old infant with an electrocardiogram monitor and oxygen saturation meter attached,” using VARCO3D, a generative modeling tool (
https://3d.varco.ai/) (
Supplement 1).
In addition, an AI-based learning system integrated with a retrieval-augmented generation framework was developed to provide automated evaluation and individualized feedback. Learners could submit responses to clinical questions through the nurse avatar [
25]. Based on predefined criteria and provided individualized feedback, learners scoring below 80 received remediation, whereas those scoring 80 or higher progressed to the next stage. Further, an AI-based learner–parent interaction system was established incorporating speech recognition and natural language processing. When learners (the nurse avatar) communicated with parents avatars, the system generated context-appropriate verbal and nonverbal responses through Naver Cloud Platform CLOVA Speech API based on nursing scenarios, supporting pediatric communication practice [
26].
1) Implementation phase
The program began with an orientation session explaining its purpose and procedures. Two pediatric nursing professors delivered the intervention; each had more than 10 years of clinical and teaching experience in pediatrics. To minimize potential bias, the instructors did not grade or evaluate participants’ academic performance and were not responsible for their regular coursework.
The experimental group consisted of 24 nursing students who participated in both group-based and individual-based learning activities. AI based individual learning activities were completed independently, and for the real-world simulation session, students were organized into small groups of 3.
The intervention (5 sequential sessions) was delivered on the same day. In Session 1, classroom lecture covered pediatric medication administration topics, including fundamentals of pediatric pharmacology, the importance of medication safety, medication errors, high-alert medications, and medication dose calculation. Sessions 2 and 3 involved the AI-based PMSC virtual simulation with 2 sequential modules. Learners completed individually on personal computers in a self-directed learning environment within the classroom. In Session 4, for the real-world simulation, participants completed hands-on practice with a high-fidelity simulator in a clinical skills laboratory. The session was conducted in small groups under instructor supervision and used the same scenarios as the virtual simulation. Finally, Session 5 consisted of a structured debriefing in which participants reflected on pediatric medication administration processes and discussed their communication experiences with parents.
Participants in the control group (n=24) individually completed a 40‑minute pre-recorded online lecture that covering the fundamentals of pediatric pharmacology, the importance of medication safety, medication errors, high‑alert medications, and medication dose calculation. The lecture was uploaded on YouTube and distributed through a dedicated URL. Participants accessed the lecture independently at their preferred location (e.g., home or self-selected settings) and completed the learning at their own pace without instructor interaction. To ensure participation, students were required to post a comment on the YouTube video after viewing the lecture, and completion was confirmed by the research team.
2) Evaluation phase
Program effectiveness was evaluated by comparing pretest and posttest outcomes between the experimental and control groups A pretest including demographic characteristics and study variables was administered immediately before the intervention, followed by a posttest immediately after the intervention (
Figure 1).
1) PMSC
PMSC was measured using a researcher-developed instrument with 18 items across 5 subdomains: infection prevention and preparation; patient identification and verification; drug preparation and administration; patient communication and cooperation; and documentation and digital competency. The instrument was developed based on a comprehensive review of the literature on pediatric medication safety, clinical practice guidelines [
27,
28]. Items focused on safety performance behaviors that may occur in pediatric nursing practice.
Content validity was evaluated by a panel of 3 pediatric nurses with more than 10 years of clinical experience and education expertise in pediatric nursing. Each expert independently rated each item’s relevance on a 4-point scale. The item-level content validity index (I-CVI) was calculated as the proportion of experts rating each item as relevant. All items achieved an I-CVI of 1.00, and the scale-level content validity index (S-CVI/Ave) was 1.00, indicating satisfactory content validity. An example item reads: “Medications are administered via the correct route and at the correct rate; if an error occurs, administration can be stopped immediately and reported.” In this study, Cronbach’s α was .93. Higher scores on the 5-point Likert scale (1=not at all, 5=very much) indicate greater PMSC.
2) MCC
MCC was assessed with the “Confidence in Drug Calculation” subscale of the Medication Calculation Skill Test (MCS test) developed by Grandell-Niemi et al. [
18] and subsequently modified and translated by Park et al. [
29]. An example of a measurement item is: “I have sufficient skills in medication dosage calculation.” This 7-item tool uses a 5-point Likert scale (1=not at all, 5=very much), and higher scores indicate greater confidence in drug calculation. The tool demonstrated good reliability, with Cronbach’s α of .89 in the study by Park et al. [
29] and .95 in the present study.
3) SIM-TLX
User workload in a simulation environment was evaluated using the SIM-TLX developed by Harris et al. [
19] and translated by Jeong et al. [
30]. SIM-TLX assesses the multidimensional cognitive and mental workload experienced by users in simulated environments [
19]. It comprises 9 items evaluating diverse aspects of task load: mental demand, physical demand, temporal demand, frustration, task complexity, stress, perceptual strain, task control, and presence/immersion. An example of a measurement item is: “How mentally exhausting was this assignment?” Each item was rated on a 7-point Likert scale (1=not at all, 7=very much). Lower scores indicate a lower level of perceived workload during the simulation task. The tool demonstrated good reliability, with Cronbach’s α of .94 in the study by Jeong et al. [
30] and .94 in the present study.
4) LI
LI was evaluated using the Learning Immersion Scale developed by Kim et al. [
20]. The LI Scale consists of 30 items across 9 dimensions: clear goals, immediate feedback, challenge-skill balance, concentration on the task at hand, merging of action and awareness, loss of self-consciousness, altered sense of time, sense of control, and autotelic experience. An example of a measurement item is: “Time passes very quickly while studying.” Items were rated on a 5-point Likert scale (1=not at all, 5=very much). Higher scores indicate a greater level of LI. The tool demonstrated good reliability, with Cronbach’s α of .93 in the study by Kim et al. [
20] and .95 in the present study.
Statistical analyses were conducted using IBM SPSS Statistics for Windows ver. 29.0 (IBM Corp.). All tests used 2-sided p-values with a 5% significance level (α=.05). Descriptive statistics (frequencies, means [M], standard deviations [SD], and percentages) summarized participant demographics and variable distributions. Chi-square tests and independent t-tests were used to assess the homogeneity of participant characteristics and variables. Independent t-tests compared groups on key variables at baseline and postintervention and on preintervention-to-postintervention change scores. Cronbach’s α was used to assess the internal consistency of the measurement instruments.
7. Research Ethics
The study adhered to IRB requirements and the Declaration of Helsinki throughout. Participants were assured that personal information would be handled confidentially. To uphold educational equity and ethical principles, control-group participants were informed at the start of the study that they could be eligible to participate in the same AI-based PMSC virtual simulation program as the experimental group after data collection had been completed. To this end, participants received a thorough pre-study explanation, gave their voluntary consent, and were free to withdraw from the study at any time. Participants received a small gift as a token of gratitude.
RESULTS
1. Homogeneity in the Characteristics of the Participants and Study Variables
Homogeneity testing showed no statistically significant between-group differences in general characteristics. Before the intervention, PMSC, MCC, and LI also did not differ significantly between groups (
Table 1). Thus, baseline homogeneity was confirmed.
Table 2 presents program effects on PMSC, MCC, SIM-TLX, and LI.
1) Hypothesis 1: PMSC
Posttest results showed significant improvement in PMSC in the experimental group (t=8.66,
p<.001) and the control group (t=4.93,
p<.001). Although both groups improved significantly, the experimental group’s gain in PMSC was significantly greater than that of the control group (t=2.48,
p=.017). Subdomain analysis revealed that infection prevention and preparation increased significantly only in the experimental group. Significant between-group differences were observed in patient identification and verification (t=2.10,
p=.041), patient communication and collaboration (t=2.46,
p=.018), and documentation and digital competency (t=2.11,
p=.040); the experimental group showed greater increases than the control group across all 3 subdomains. Hypothesis 1 was supported (
Table 2,
Figure 3).
2) Hypothesis 2: MCC
Pretest-to-posttest analysis revealed significant increases in MCC in both the experimental group (t=6.15,
p<.001) and the control group (t=9.82,
p<.001). However, the magnitude of the increase did not differ significantly between groups (t=−0.87,
p=.389). Hypothesis 2 was therefore rejected (
Table 2).
3) Hypothesis 3: SIM-TLX
Scores decreased significantly from pretest to posttest in the experimental group (t=−2.61,
p=.016). Hypothesis 3 was supported (
Table 2).
4) Hypothesis 4: LI
LI improved significantly after the intervention in both the experimental group (t=4.47,
p<.001) and the control group (t=3.87,
p=.001). However, the between-group difference was not significant (t=0.97,
p=.336). Hypothesis 4 was therefore rejected (
Table 2).
DISCUSSION
This study developed an AI-based PMSC virtual simulation program and evaluated its effectiveness of PMSC, MCC, SIM-TLX, and LI for nursing students.
In this study, the experimental group showed a significantly greater increase in PMSC than the control group. PMSC encompasses competencies including accurate medication calculation, safe administration practices, clinical decision-making, and effective communication with caregivers [
1,
4,
5]. As educational approaches that integrate those elements are essential for competency development, this intervention was designed to integrate these components through AI-based virtual simulation and experiential learning activities [
15,
31]. This AI-based program may have supported repetitive and adaptive learning in medication administration by delivering in-depth questions and learner-tailored quizzes when participants showed low accuracy on medication-related tasks.
Within the AI-based virtual simulation environment, learners encountered pediatric medication scenarios and made step by step medication-related decisions. This process may have provided opportunities for learners to apply medication-related knowledge in realistic clinical scenarios, while engaging in decision-making within given simulated situations. After the AI-based learning, participants engaged in real-world simulation, which may have facilitated the integration of theoretical knowledge with practical skills and procedural performance. This approach is consistent with prior research [
32] showing that simulation-based learning environments enhance the integration of knowledge and clinical skills while promoting medication safety competence. In addition, as in the work of Bayram and Ergün [
32], the iterative learning process—characterized by adaptive feedback, repeated attempts, and progression may have supported learners improvements of PMSC.
From a family-centered care perspective, communication with parents; a key subdomain of PMSC is a core competency. Structured communication education can help nursing students communicate more effectively with families, enabling them to better understand and participate in their child’s care. Previous studies also have suggested that patient-oriented simulation education could be associated with improvements in communication with patients and families [
33]. In this context, the AI-based PMSC virtual simulation program provided nursing students with opportunities to engage in medication-related question-and-answer interactions and communication with parents avatars. These interactions may have enhanced the quality of communication experiences by allowing learners to practice explaining medication-related information, responding to caregivers’ concerns, and engaging in family-centered communication.
Although MCC increased in both groups, no significant between-group difference was demonstrated. Although previous studies have reported improvements in medication confidence [
32], this study findings did not demonstrate an effect of the intervention on MCC. From the perspective of experiential learning, repeated engagement in the 4-step learning cycle may enhance learning by promoting reflection, internalization, and application [
18,
31].
Previous research has reported that satisfaction and confidence in medication could be increased when learners are provided with enough opportunities to adjust their learning pace and engage in repeated practice [
32]. In this study, the AI-based PMSC virtual simulation embedded medication calculation tasks within time-limited clinical scenarios. Although this program may have supported task engagement, it may not have provided sufficient opportunities for repeated practice, reflection, and error correction due to time-limited structure of the program. Furthermore, the single-day intervention may have limited learners’ opportunities to fully engage in the experiential learning cycle. Thus, future studies should examine whether extended intervention periods and flexible learnings could enhance medication confidence among nursing students.
SIM-TLX scores in the experimental group decreased significantly from pretest to posttest. This finding is consistent with Gohman et al. [
34], which suggests that VR-based simulations could provide a manageable learning environment without excessive cognitive load, and that repeated exposure to simulation may further reduce perceived workload during task performance. As SIM-TLX reflects perceived task load during simulation rather than comparative index of learning outcome, these findings may indicate that learners experienced a acceptable level of cognitive workload while participating in the simulation.
LI increased significantly in both groups; however, the between-group difference was not significant. This result may reflect the influence of multiple interacting factors including learners’ individual characteristics, adaptation to the learning environment, and contextual factors rather than a single determinant [
35]. Although AI technology can provide new opportunities for participation, learners’ immersion may be influenced by their familiarity with digital technologies and characteristics of the learning environment [
15]. Furthermore, in this study, the limited interactivity of non-immersive virtual simulations may have been insufficient to elicit the high level of immersion typically expected in immersive environments [
36].
Therefore, educational interventions aimed at enhancing nursing students’ LI should consider both the degree and quality of immersion. In addition, learning objectives and task difficulty should be tailored to learners’ competencies, and sufficient time and support should be provided to facilitate adaptation to AI-based learning environments [
15]. Future studies should examine how factors such as digital proficiency, familiarity with AI-based learning environments, and intervention duration influence learners’ immersion. Further future studies should explore how different types and levels of immersion influence learning outcomes in AI-based simulation environments.
This study is meaningful in that it highlights the potential of AI-based simulation as an educational approach for supporting PMSC among nursing students. However, the present study had several limitations. First, because the sample included nursing students from only 2 cities, the findings may have limited generalizability to nursing students across South Korea; nation-wide multi-center studies are needed to strengthen external validity in future study. In addition, although participants were recruited from multiple cities, information regarding their institutional affiliations was not collected. Consequently, the number and characteristics of participating universities could not be identified, limiting the assessment of sample representativeness across institutions, future studies should consider collecting institution- and region-level information, with appropriate ethical approval, to enhance the sample representativeness and generalizability of the findings. Second, data collection in the control and experimental groups was sequential rather than concurrent. The sequential design was adopted to minimize the risk of information contamination, given the nature of the AI-based simulation intervention. Despite this precaution, the sequential approach may have introduced temporal bias and uncontrolled external influences. These findings should therefore be interpreted with caution. Third, the study design may raise concerns regarding internal validity because of differential exposure between groups. The experimental group received a multi-component intervention, whereas the control group received only a lecture-based intervention. This difference in instructional intensity and modality may have introduced confounding. Although this design was intended to evaluate the added value of the AI-based simulation beyond conventional lecture-based instruction, the observed effects may have been influenced by differences in instructional exposure. Future studies should therefore consider more controlled comparisons with equivalent instructional intensity. Fourth, the primary outcome variable, PMSC, was measured using a researcher-developed instrument. Although content validity was assessed, construct validity was not formally analyzed and the use of a small number of experts may limit the rigor of the content validity evaluation. More rigorous validation procedures (such as exploratory or confirmatory factor analysis) are needed to strengthen the validity of the instrument. Fifth, participants were assigned using a nonrandom convenience method, which may have introduced selection bias and limited the internal validity of the study. Although baseline homogeneity was confirmed, unmeasured between-group differences cannot be excluded, and the findings should be interpreted with caution. Future studies should therefore employ randomized controlled designs. Finally, follow-up research is needed to evaluate longitudinal and broader competencies—such as communication with parents and multidisciplinary collaboration—in addition to the PMSC components administered in the present study.
CONCLUSION
This study developed and evaluated an AI-based PMSC virtual simulation program for pediatric medication safety among nursing students in South Korea. The results showed that the program improved nursing students’ PMSC, with significant reductions in SIM-TLX. The program can serve as an educational intervention in pediatric nursing education to strengthen medication safety competence, including communication with pediatric patients and their caregivers. Future research should refine the program to address individual learning needs and should evaluate its application across a wider range of nursing education settings.
ARTICLE INFORMATION
Supplementary material
Supplement 1.
Screenshots of the artificial intelligence-based pediatric medication safety competence virtual simulation program.
Figure 1.Overview of the study design and the artificial intelligence (AI)-based pediatric medication safety competence (PMSC) virtual simulation program. EG, experimental group; CG, control group; LI, learning immersion; MCC, medication calculation confidence; SIM-TLX, SimulationTask Load Index; IV, intravenous; PICU, pediatric intensive care unit.
Figure 2.Consort flow diagram. AI, artificial intelligence; CG, control group; EG, experimental group; PMSC, pediatric medication safety competence.
Figure 3.(A–D) Differences in outcome variables between the groups. CG, control group; EG, experimental group; LI, learning immersion; MCC, medication calculation confidence; PMSC, pediatric medication safety competence; SIM-TLX, Simulation Task Load Index.
Table 1.Homogeneity tests for participant characteristics and study variables (N=48)
|
Characteristic |
EG (n=24) |
CG (n=24) |
χ2 or t |
p
|
|
Sex |
|
|
0.00 |
1.000 |
|
Female |
18 (75.0) |
18 (75.0) |
|
|
|
Male |
6 (25.0) |
6 (25.0) |
|
|
|
Age (yr) |
|
|
0.08 |
.773 |
|
≤21 |
11 (45.8) |
12 (50.0) |
|
|
|
≥22 |
13 (54.2) |
12 (50.0) |
|
|
|
Region |
|
|
0.00 |
1.000 |
|
Gwang-ju |
23 (95.8) |
23 (95.8) |
|
|
|
Chonnam |
1 (4.2) |
1 (4.2) |
|
|
|
Religion |
|
|
0.12 |
.989 |
|
Christian |
1 (4.2) |
1 (4.2) |
|
|
|
Catholic |
5 (20.8) |
6 (25.0) |
|
|
|
Buddhism |
2 (8.3) |
2 (8.3) |
|
|
|
None |
16 (66.7) |
15 (62.5) |
|
|
|
Satisfaction with nursing studies |
|
|
1.30 |
.862 |
|
Very dissatisfied |
1 (4.2) |
0 (0.0) |
|
|
|
Dissatisfied |
1 (4.2) |
1 (4.2) |
|
|
|
Moderate |
7 (29.2) |
9 (37.5) |
|
|
|
Satisfied |
11 (45.8) |
10 (41.7) |
|
|
|
Very satisfied |
4 (16.7) |
4 (16.7) |
|
|
|
Interpersonal relationships |
|
|
2.95 |
.228 |
|
Difficult |
2 (8.3) |
0 (0.0) |
|
|
|
Average |
8 (33.3) |
12 (50.0) |
|
|
|
Good |
14 (58.3) |
12 (50.0) |
|
|
|
Experience with medication administration education |
|
|
0.40 |
.525 |
|
Yes |
6 (25.0) |
8 (33.3) |
|
|
|
No |
18 (75.0) |
16 (66.7) |
|
|
|
Perceived usefulness of school-based pediatric medication traininga)
|
|
|
1.56 |
.457 |
|
Helpful |
1 (16.7) |
1 (12.5) |
|
|
|
Unsure |
4 (66.7) |
7 (87.5) |
|
|
|
Not helpful |
1 (16.7) |
0 (0.0) |
|
|
|
Experience with AI-based education |
|
|
2.09 |
.149 |
|
Yes |
14 (58.3) |
9 (37.5) |
|
|
|
No |
10 (41.7) |
15 (62.5) |
|
|
|
Courses in which AI education was provideda)
|
|
|
2.85 |
.723 |
|
Fundamentals of nursing |
1 (7.1) |
1 (11.1) |
|
|
|
Adult nursing |
2 (14.3) |
2 (22.2) |
|
|
|
Pediatric nursing |
2 (14.3) |
0 (0.0) |
|
|
|
Women’s health nursing |
4 (28.6) |
1 (11.1) |
|
|
|
Nursing and AI |
3 (21.4) |
3 (33.3) |
|
|
|
Other |
2 (14.3) |
2 (22.2) |
|
|
|
Pediatric medication safety competence |
3.84±0.62 |
4.08±0.57 |
–1.35 |
.183 |
|
Infection prevention and preparation |
4.55±0.50 |
4.57±0.52 |
–0.14 |
.889 |
|
Patient identification and verification |
4.03±0.64 |
4.33±0.54 |
–1.71 |
.095 |
|
Medication preparation and administration |
3.33±0.75 |
3.55±0.81 |
–0.97 |
.336 |
|
Patient communication and collaboration |
3.78±0.78 |
4.11±0.71 |
–1.55 |
.129 |
|
Documentation and digital competency |
3.08±1.11 |
3.46±1.03 |
–1.21 |
.231 |
|
Medication calculation confidence |
2.59±0.79 |
2.38±0.91 |
0.87 |
.390 |
|
Learning immersion |
3.32±0.74 |
3.22±0.55 |
0.53 |
.597 |
Table 2.Comparison of dependent variables between the experimental and control groups (N=48)
|
Variable |
EG (n=24) |
CG (n=24) |
t (p) |
ES(d) |
|
Pretest |
Posttest |
Pre–post t (p) |
Pretest |
Posttest |
Pre–post t (p) |
|
Pediatric medication safety competence |
3.84±0.62 |
4.64±0.29 |
8.66 (<.001) |
4.08±0.57 |
4.55±0.47 |
4.93 (<.001) |
2.48 (.017) |
0.72 |
|
Infection prevention and preparation |
4.55±0.50 |
4.83±0.29 |
2.63 (.015) |
4.57±0.52 |
4.71±0.45 |
1.33 (.197) |
0.99 (.329) |
0.29 |
|
Patient identification and verification |
4.03±0.64 |
4.68±0.30 |
5.08 (<.001) |
4.33±0.54 |
4.64±0.39 |
3.55 (.002) |
2.10 (.041) |
0.61 |
|
Medication preparation and administration |
3.33±0.75 |
4.55±0.39 |
9.33 (<.001) |
3.55±0.81 |
4.47±0.52 |
6.43 (<.001) |
1.56 (.125) |
0.45 |
|
Patient communication and collaboration |
3.78±0.78 |
4.60±0.45 |
6.63 (<.001) |
4.11±0.71 |
4.47±0.64 |
2.60 (.016) |
2.46 (.018) |
0.71 |
|
Documentation and digital competency |
3.08±1.11 |
4.44±0.66 |
7.66 (<.001) |
3.46±1.03 |
4.25±0.72 |
3.97 (.001) |
2.11 (.040) |
0.61 |
|
Medication calculation confidence |
2.59±0.79 |
3.81±0.91 |
6.15 (<.001) |
2.38±0.91 |
3.81±0.77 |
9.82 (<.001) |
−0.87 (.389) |
0.25 |
|
SIM-TLX |
2.94±0.94 |
2.45±0.97 |
–2.61 (.016) |
- |
- |
- |
- |
0.53 |
|
Learning immersion |
3.32±0.74 |
3.82±0.66 |
4.47 (<.001) |
3.22±0.55 |
3.58±0.59 |
3.87 (.001) |
0.97 (.336) |
0.28 |
REFERENCES
- 1. D'Errico S, Zanon M, Radaelli D, Padovano M, Santurro A, Scopetti M, et al. Medication errors in pediatrics: proposals to improve the quality and safety of care through clinical risk management. Front Med (Lausanne). 2022;8:814100. https://doi.org/10.3389/fmed.2021.814100
- 2. Öztürk Şahin Ö, Aközlü Z, Taşdelen Y. Pediatric nursing students’ self-efficacy regarding medication administration and clinical comfort and worry: a pre-posttest comparative study of nurse mentoring versus peer mentoring. Nurse Educ Pract. 2023;71:103712. https://doi.org/10.1016/j.nepr.2023.103712
- 3. Lu H, Rosenbaum S. Developmental pharmacokinetics in pediatric populations. J Pediatr Pharmacol Ther. 2014;19(4):262-276. https://doi.org/10.5863/1551-6776-19.4.262
- 4. Özgüven Öztornaci B, Karakul A, Ardahan Akgül E, Doğan P, Yildirim Sari H, Başbakkal Z. The effect of web-based double-eye control program on the prevention of pediatric medication errors of nursing students: a quasi-experimental study. Medicine (Baltimore). 2025;104(49):e46222. https://doi.org/10.1097/MD.0000000000046222
- 5. Lin FY, Wu WW, Lin HR, Lee TY. The learning experiences of student nurses in pediatric medication management: a qualitative study. Nurse Educ Today. 2014;34(5):744-748. https://doi.org/10.1016/j.nedt.2013.08.004
- 6. Dodson TM, Thompson-Hairston K. A scoping review of gen z nursing students: learning preferences and educator strategies. Teach Learn Nurs. 2025;20(1):85-90. https://doi.org/10.1016/j.teln.2024.11.007
- 7. Chang YY, Chao LF, Chang W, Lin CM, Lee YH, Latimer A, et al. Impact of an immersive virtual reality simulator education program on nursing students’ intravenous injection administration: a mixed methods study. Nurse Educ Today. 2024;132:106002. https://doi.org/10.1016/j.nedt.2023.106002
- 8. Gokalp MG, Yucel SC, Yilmaz O. The effect of game-based learning on the acquisition of intramuscular injection skills. Nurse Educ Pract. 2025;82:104172. https://doi.org/10.1016/j.nepr.2024.104172
- 9. Mahmoud F, Abdel-Salam A. The effect of interactively virtual reality eyeglasses on the of pediatric nursing students performance. IOSR J Nurs Health Sci. 2018;7(4):58-67. https://doi.org/10.9790/1959-0704095867
- 10. Marchi NM, Zimmermann E, Kelling CS, Wong S, Juniper KM. A medication dosage simulation strategy to improve patient safety. Am Nurse Today. 2019;14(9):1.
- 11. Ma J, Wen J, Qiu Y, Wang Y, Xiao Q, Liu T, et al. The role of artificial intelligence in shaping nursing education: a comprehensive systematic review. Nurse Educ Pract. 2025;84:104345. https://doi.org/10.1016/j.nepr.2025.104345
- 12. Lin CC, Huang AY, Lu OH. Artificial intelligence in intelligent tutoring systems toward sustainable education: a systematic review. Smart Learn Environ. 2023;10(1):41. https://doi.org/10.1186/s40561-023-00260-y
- 13. Orkaby B, Kerner E, Saban M, Levin C. Bridging generational gaps in medication safety: insights from nurses, students, and generative AI models. BMC Nurs. 2025;24(1):382. https://doi.org/10.1186/s12912-025-03034-8
- 14. Park SA, Kim HY. Development and effects of a scenario-based labor nursing simulation education program using an artificial intelligence tutor: a quasi-experimental study. Womens Health Nurs. 2025;31(2):143-154. https://doi.org/10.4069/whn.2025.06.18
- 15. Sengul T, Sarıköse S. Enhancing learning outcomes through AI-driven simulation in nursing education: a systematic review. Clin Simul Nurs. 2025;106:101797. https://doi.org/10.1016/j.ecns.2025.101797
- 16. Atkins R, Brown KM, Mudd SS, Ghobadi K, Baker DJ, Szanton S. Reimagining nursing education: leveraging competency-based education, artificial intelligence, and simulation for a diverse and practice-ready workforce. Nurs Outlook. 2025;73(6):102582. https://doi.org/10.1016/j.outlook.2025.102582
- 17. Kolb DA. Experiential learning: experience as the source of learning and development. FT Press; 2014.
- 18. Grandell-Niemi H, Hupli M, Leino-Kilpi H, Puukka P. Medication calculation skills of nurses in Finland. J Clin Nurs. 2003;12(4):519-528. https://doi.org/10.1046/j.1365-2702.2003.00742.x
- 19. Harris D, Wilson M, Vine S. Development and validation of a simulation workload measure: the simulation task load index (SIM-TLX). Virtual Real. 2020;24(4):557-566. https://doi.org/10.1007/s10055-019-00422-9
- 20. Kim YY, Kim SW, Koo KH. Development and validation of the learning flow scale for high school students. J Curric Eval. 2017;20(4):95-119. https://doi.org/10.29221/jce.2017.20.4.95
- 21. Alruwaili AN, Alshammari AM, Alhaiti A, Elsharkawy NB, Ali SI, Ramadan OME. Virtual reality simulation for high-risk neonatal emergency nursing training: a mixed-methods study on nurse competency and outcomes. BMC Nurs. 2025;24(1):367. https://doi.org/10.1186/s12912-025-03038-4
- 22. Des Jarlais DC, Lyles C, Crepaz N, TREND Group. Improving the reporting quality of nonrandomized evaluations of behavioral and public health interventions: the TREND statement. Am J Public Health. 2004;94(3):361-366. https://doi.org/10.2105/ajph.94.3.361
- 23. Faul F, Erdfelder E, Lang AG, Buchner A. G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behav Res Methods. 2007;39(2):175-191. https://doi.org/10.3758/bf03193146
- 24. Molenda M, Pershing JA, Reigeluth CM. Designing instructional systems. In: Craig RL, editor. ASTD training and development handbook. 4th ed. McGraw-Hill; 1996.
- 25. Lewis P, Perez E, Piktus A, Petroni F, Karpukhin V, Goyal N, et al. Retrieval-augmented generation for knowledge-intensive NLP tasks. Adv Neural Inf Process Syst. 2020;33:9459-9474.
- 26. Yang HJ, Oh EB, Kim JM. Comparison of automatic speech recognition system for school-aged children’s narratives: Naver Clova Speech and Google Speech-to-Text. Commun Sci Disord. 2023;28(1):30-38. https://doi.org/10.12963/csd.23952
- 27. World Health Organization. Medication without harm [Internet]. World Health Organization; c2026 [cited 2026 Feb 10]. Available from: https://www.who.int/initiatives/medication-without-harm
- 28. The Joint Commission. Home—The Joint Commission [Internet]. The Joint Commission; c2026 [cited 2026 Feb 10]. Available from: https://www.jointcommission.org/en/
- 29. Park HS, Cho GY, Kim DH, Kim SH, Kim MS. The mediating effect of drug calculation confidence in the relationship between interest in medication and drug calculation competency. J Korean Biol Nurs Sci. 2013;15(4):155-163. https://doi.org/10.7586/jkbns.2013.15.4.155
- 30. Jeong Y, Kim K, Ryu J. The impact of cognitive, affective, and psychomotor learning perception on learning outcomes in the extended reality-based nursing simulation. In: Auer ME, Tsiatsos T, editors. Immersive learning research network conference proceedings. Springer Nature Switzerland; 2024. p. 446-459. https://doi.org/10.1007/978-3-031-80475-5
- 31. Uppor W, Klunklin A, Viseskul N, Skulphan S. Effects of experiential learning simulation-based learning program on clinical judgment among obstetric nursing students. Clin Simul Nurs. 2024;92:101553. https://doi.org/10.1016/j.ecns.2024.101553
- 32. Bayram HÇ, Ergün S. Effect of screen-based simulation on pediatric drug administration among nursing students: a randomized controlled trial. Rev Esc Enferm USP. 2025;59:e20250249. https://doi.org/10.1590/1980-220X-REEUSP-2025-0249en
- 33. Asuncion AM, Quintos-Alagheband ML, Leavens-Maurer J, Akerman M, Janicke P, Cavanaugh S. Utilization of family as faculty: a patient directed simulation education to improve patient and family communication during Patient-Family Centered Rounds (PFCR). Pediatr Qual Saf. 2022;7(3):e551. https://doi.org/10.1097/pq9.0000000000000551
- 34. Gohman T, Nisar H, Gupta A, Javed MJ, Rau N. Development and usability of a virtual reality umbilical venous catheter placement simulator. Int J Comput Assist Radiol Surg. 2024;19(5):881-889. https://doi.org/10.1007/s11548-024-03072-8
- 35. Kim YS, Lee KS. Factors influencing learning flow of Korean nursing students: a systematic review and meta-analysis. J Korea Acad Ind Coop Soc. 2022;23(6):151-162. https://doi.org/10.5762/KAIS.2022.23.6.151
- 36. Choi HI. Nursing students’ learning immersion, learning self-efficacy, and learning satisfaction for web-based virtual clinical simulation and clinical case seminars [dissertation]. Seoul National University; 2025.
, Soohyeon Cho3
