Uncovering Cognitive Strategies for ICU Parallel Nursing Multitasks: An Eye-tacking-based Hidden Markov Modeling Study

Uncovering Cognitive Strategies for ICU Parallel Nursing Multitasks: An Eye-tacking-based Hidden Markov Modeling Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Uncovering Cognitive Strategies for ICU Parallel Nursing Multitasks: An Eye-tacking-based Hidden Markov Modeling Study Tingting Feng, Jia-qi Cai, Sheng-feng Wang, Hao Wu, Xue Wu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7975457/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract With the influx of medical devices, parallel multitasking rapidly grows in clinic and may contribute to increasing human error and higher risks. Since it is impossible to avoid parallel multitasking in clinic, it is imperative to explore nurses’ cognitive strategy under this complex situation to ensure nursing quality and patient safety. However, there is a little research towards cognitive strategies of parallel multitasking in nursing. Thus, our study aimed at investigating nurses’ cognitive strategies in decision-making under parallel multitasking through observational research.30 eligible nurses completed the trials programmed by PsychoPy 2021.2.3 in the lab. Participants’ behavioral- and eye- movement data were recorded by PsychoPy and EyeLink 1000 respectively. Hidden Markov Modeling was applied to analyze the spatio-temporal dynamics of participants’ eye movements and uncovered the cognitive strategies. Results show that participants adopted two cognitive strategies under parallel multitasking: holistic- and chunk- cognitive strategy. The holistic cognitive strategy yielded a longer reaction time and heavier cognitive workload, but the differences in task accuracy between two cognitive strategies were insignificant. Besides no strategy switching and little transition between the last two ROIs was observed in both cognitive strategies, indicating low cognitive flexibility and strong tendency to take cognitive short-cuts. In the future more researches should be done to explore cognitive strategies in diverse nursing groups and provide suggestions to nurse education and management, therefore to better equip nurses in newly working environment. cognitive strategy parallel multitasking task performance cognitive workload human-machine interaction Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 1 Introduction Nursing practice is complex and labor-intensive, particularly in high-acuity settings such as the ICU and emergency department. Parallel multitasking is defined as the simultaneous execution or rapid switching between two or more tasks by an individual (Kim et al., 2022 ; Lu et al., 2025 ; Olin et al., 2022 ). Extensive research has demonstrated that parallel multitasking can compromise task performance, prolong response times, and elevate cognitive workload (Liu & Wang, 2012 ; Yuan & Zhong, 2024 ). Despite these documented challenges, parallel multitasking occurs frequently in nursing practice, which may result in interruption and distraction as well as a higher probability of human error. One study indicated that nurses spend over 50.1% of their working hours engaged in multitasking and make hundreds of ad hoc decisions per shift (Abu Arra et al., 2023 ). Considering the persistent nursing shortage and rising application of medical devices as well as artificial intelligence in the clinic, it is impossible to avoid multitasking. Thus, enhancing nurses’ decision-making capabilities under parallel multitasking especially for ICU nurses is of great urgency (Mathew et al., 2025 ). In accordance with the cognitive theory, human cognitive capacity is fundamentally limited, with working memory typically constrained to approximately four chunks of information at one time (Cowan, 2001 ; Sweller, 1988 ). Therefore, with the increasing of simultaneous task the likelihood of error will increase. Cognitive strategies refer to structured mental procedures employed to achieve specific goals or solve problems(Cameron & Jago, 2013 ). By exploring the cognitive strategies nurses employed in parallel multitasking, we can find out the barriers and facilitators in decision-making performance in this complex situation which is significant in safeguarding decision quality and improving patient safety. Although extensive studies have been conducted in exploring cognitive strategy in parallel multitasking in safety-critical industries—such as aviation and traffic control, researches towards cognitive strategies in nursing under parallel multitasking context remains limited (Li, 2025 ; Stasch & Mack, 2025). Currently the most commonly applied approaches for identifying cognitive strategies include subjective reporting methods (e.g. self-report questionnaire, think-aloud), EEG, fMRI and eye movement tracking. Although subjective evaluation methods are easier to implement, they are susceptible to participants' subjective biases and exhibit temporal delays, making it difficult to capture real-time cognitive processes (Jarosz et al., 2019 ; Jastrzębski et al., 2018 ). EEG and fMRI can identify cognitive strategy under parallel multitasking in real time with high accuracy, but they require participants to wear corresponding equipment which is costly and may interfere daily work, thus limited its application. Whereas eye movements can reflect participants’ thought process and mental states in real time, offering insights into their understanding of the problem, level of attention, information processing, and cognitive workload (Laurence et al., 2023 ; Wang & Zhan, 2025 ). Eye-tracking research results are especially beneficial in clinical settings with intensive human-machine interactions, including the ICU, emergency department and operating room. Thus, considering the accuracy, cost-effectiveness and future promotion, we selected the eye movement tracking technology to investigate the cognitive strategy in parallel multitasking. Hidden Markov Model (HMM) is a statistical model particularly well-suited for analyzing time-series data, which includes observable states, such as eye movement trajectories and unobservant hidden states like cognitive processes (Grewal et al., 2019 ). While traditional eye-tracking analyses, such as fixation duration and heatmaps can only provide static descriptions, HMM can map the complex eye movement sequences into a series of hidden cognitive states, thereby revealing an individual’s cognitive strategies during visual tasks (Griffin et al., 2024 ; Wang & Zhan, 2025 ). Thus, considering the increasing incidence of parallel multitasking in nursing work and its potential harmfulness, especially for high-risk department, we conducted this research to investigate ICU nurses’ cognitive strategy and their corresponding performance as well as cognitive workload in parallel multitasking. The research questions are: 1) What cognitive strategies do nurses employ during parallel multitasking? 2) How do different cognitive strategies perform?3) What is the cognitive workload associated with different cognitive strategies? 2 Materials and methods 2.1 Design This study utilized an observational research design. We required subjects to perform simulated ICU parallel multitask covering patients with different disease severity: the more severe illness (MS group) and the less severe illness (LS group). These two experimental blocks presented in random order, with each block consisting of 20 trials. 2.2 Participants The inclusion criteria were: (a) being a registered nurse; (b)at least six months of clinical work experience in an ICU or emergency department; (c) being in good physical and mental health, capable to cooperate fully with the experimental procedures; and (d) having normal or corrected-to-normal vision. Exclusion criteria included: (a) experiencing a recent cold or any form of eye discomfort; and (b) fatigue, particularly following a night shift. Participants were recruited from 3 tertiary hospitals in Beijing, China, between April and June using poster advertisements. Interested individuals contacted the research team via WeChat and underwent an initial eligibility screening. Those who met all the criteria were subsequently provided with an online demographic questionnaire and a self-developed risk perception survey (Appendix A). The risk perception survey was developed based on the latest definition of risk, including perception of the probability of risk in their nursing work, the potential consequences and risk tolerance. Each item uses a 7-point Likert scale, with one item being reverse-scored(Roszkowski & Davey, 2010).All participants received detailed information regarding the study procedures and provided informed consent using an IRB-approved form (Approval No. IRB00001052-25007). 2.3 Parallel Multitasking experiment 2.3.1 Eye-tracking apparatus Participants’ eye movement were recorded with the Eyelink 1000. 9 Point of gaze was sampled at 1000 Hz with an accuracy of 0.5~1.0 visual degrees. The experiment was conducted in a laboratory with constant control of sound and light conditions in Institute of Psychology, Chinese Academy of Sciences. Participants were seated in a chair approximately 70 cm from an flat screen monitor (1920 x 1080). 2.3.2 Stimulus Participants completed two experimental blocks, each comprising 20 trials. Each block represented a distinct clinical scenario: one featured patient with more severe conditions (MS group), while the other involved patients with less severe conditions (LS group). In each trial, a set of three patient scenarios was presented simultaneously (Fig.1). Each scenario consisted of a simulated medical device alarm interface accompanied by a corresponding diagnostic description (Fig.1). The MS group included critical conditions such as cardiogenic shock, extensive burns, severe sepsis, multiple fractures, massive pulmonary embolism, aortic dissection, and sepsis with renal failure. The LS group comprised less acute diagnoses, including chronic heart failure, status post thyroidectomy, metabolic acidosis, and ascites due to liver cirrhosis. The device alarm interfaces were modeled based on common ICU equipment, including ECG monitors, haemodialysis machines, blood glucose monitors, infusion pumps, enteral feeding pumps, and ventilators. All experimental materials were grounded in real clinical practice and subjected to three rounds of evaluation by a panel of four experts in clinical nursing and engineering psychology. The stimuli were further refined based on the results of a pilot study( N =2) before the final versions were implemented. 2.3.3 Procedure Prior to the formal experiment, participants completed five practice trials under the guidance of the researcher to familiarize themselves with the task requirements. The formal experiment began only after participants confirmed their understanding and readiness. Participants then completed two blocks of trials in a randomized order. In each trial, after observing the stimulus, they were required to type in the patient care sequence (eg.321, 123 or 312). After entering their decision, participants pressed the "Space" key to proceed to the next trial, the whole research methodology flow and experiment scenario are shown in Fig.2. 2.4 Data Analysis 2.4.1 Eye-movement Based Hidden Markov Modeling We analyzed eye-tracking data using a Hidden Markov Model approach implemented with EMHMM toolbox (v0.80) by MATLAB 2018b. This method captured both spatial and temporal dynamics of fixations, allowing researchers to identify dynamic visual processing strategies. The analysis involved three steps. Step 1: Eye-Movement Data Cleaning First of all, we thoroughly screened the eye-movement data and deleted invalid data points based on quality flags from the eye tracker. This data cleaning step can ensure data reliability for subsequent modeling. Step 2: Establishment of Individualized Hidden Markov Model Researchers developed an individualized HMM for each participant based on their sequence of fixation coordinates recorded during the trial. Model parameters including initial probabilities, transition probabilities between states and Gaussian emission probabilities defined each hidden state (i.e., region of interest, ROI), and were estimated using a variational Bayesian approach. In accordance with previous literature, the models were compared with 1 to 6 hidden states and then researchers selected the most suitable hidden states. Step 3: Hidden Markov Model Clustering The Variational Hierarchical Expectation-Maximization (VHEM) algorithm was applied to cluster individualized HMMs by parameter similarity and identify a small set of representative HMMs that captured common eye movement patterns across the sample. Each representative HMM summarized the spatial distribution of ROIs and transition patterns characteristic of a particular visual strategy. In this research, the Holistic-Chunk score (H-C score) is used to quantify the similarity between each participant’s eye movement pattern and the representative strategies. This score was derived from the log-likelihoods of the individual’s data being generated by each representative HMM, a positive value indicates closer alignment with the holistic process strategy, while a negative value reflects greater similarity to the chunk-based process strategy. 2.4.2 Analysis of task performance and cognitive workload of different cognitive strategies To investigate the impact of cognitive strategy on task performance and cognitive workload, we firstly conducted the statistical description for the demographic characteristics, risk perception, reaction time and task accuracy of all the participants and trials. Then, we analyzed the correlation between cognitive strategy represented by H-C score with the task performance and cognitive workload through linear regression. In this research, the saccadic duration and pupil size were selected as parameters to measure cognitive workload in this research (Liu et al., 2022; Liu et al., 2022). 2.5 Ethical approval The study was approved by the Ethics Committee of Peking University Health Science Center (IRB No.00001052-25007, February 19, 2025). Informed consent was obtained from all participants prior to participation. 3 Results 3.1 Demographics, and risk perception scores of participants A total of 30 participants were recruited from 3 tertiary hospitals in Beijing, China from April to June. 3 participants’ data were excluded in the data cleaning phase due to missing eye-tracking data and the included participants’ demographic information can be seen in table 1. Table 1 Demographic of the participants and their risk perception score ( N =27) Characteristics N (%)/M±SD/Median (IQR) Age (year) 25.48±3.08 Gender Female 17 (62.96%) Male 10 (37.04%) Working experience(month) 36 (12, 60) Education background Associate degree 8 (29.63%) Bachelor's degree 18 (66.67%) Graduate degree 1 (3.7%) Professional title Registered Nurse 18 (66.67%) Senior Nurse 8 (29.63%) Supervisor Nurse 1 (3.7%) Risk perception score 4.49±0.75 Probability of risk 5.00±1.21 Potential Consequences of Risk 5.56±1.16 Risk tolerance 5.07±0.99 3.2 Clustering of eye movement strategies using Hidden Markov Models Eye movement data from 27 participants were included in the analysis. Based the coordinate-based eye movement data, we identified two distinct cognitive strategies, they are the holistic cognitive process strategy and the chunk-based cognitive process strategy (Fig. 3 & 4). We analyzed the model fitness of models with 1 to 6 ROIs with both cognitive strategies, therefore to determine the optimal number of ROIs for each individual HMM. The Akaike information criterion (AIC) and the Bayesian information criterion (BIC) were used as model-fit indices to identify the optimal number of hidden states for the eye movement-based HMM (E-HMM)(Kuha, 2004; Wagenmakers & Farrell, 2004). A lower AIC and BIC values indicate a better fit of the E-HMM to the data. And we finally decided that the optimal number of ROIs was three based on AIC, BIC and the ROI performance. Among the participants who completed the MS group task( n =24), 11 participants (45.8%) applied holistic process strategy and 13 participants (54.2%) applied the chunk-based process strategy. Among the participants who completed the LS group task( n =26), 9 participants (34.6%) applied the holistic cognitive process strategy and 17 participants (65.4%) applied the chunk-based process strategy. 3.2.1 Holistic cognitive process strategy In the Holistic Process Strategy (Fig. 3), each ROI encompassed both the machine alert interface and the diagnose for the patient. This pattern suggests that nurses tended to integrate machine alert and diagnostic information into a coherent perceptual unit, which they compared in a systematic sequence later. The initial fixation was most frequently directed toward ROI1 (MS: 0.71; LS: 0.74), indicating a tendency to first attend to the central position patient. Note:The circular area with numbers indicates the region of interest, and the number represents the identifier of the region of interest. Each ROI integrates both the machine alert interface and diagnostic text for each patient. Markov transition probability matrix reflects systematic comparisons between patients. 3.2.2 Chunk-Based cognitive process strategy The Chunk-Based Process Strategy was characterized by a clear spatial separation of visual attention: ROI1 encompassed the machine alert interfaces across all three patients, while ROI 2 and ROI 3 contained their corresponding diagnose information (Fig.4). This pattern suggests that nurses employing this strategy firstly compared machine interface information across all cases before shifting attention to diagnostic details. The initial fixation distribution was highly concentrated, with a probability of 0.91 directed toward ROI1 (machine alert interface), indicating a strong prioritization of machine alert information during initial encoding. 3.3 Attention allocation characteristics under different experiment conditions 3.3.1 Transition between different rois under different cognitive strategy and condition We also analyzed the transition probability between different ROIs under different experiment conditions, which can reflect participants’ attention allocation on different ROIs and reveal their cognitive strategies (Fig.5).Through Fig.5 we can notice that the transition probability between ROI 1 and ROI 2, ROI 1 and ROI 3 are much higher than that between ROI 2 and ROI 3, indicating that there is a high probability that nurse made clinical decision under parallel multitask without careful comparing between the last two ROIs which may explain their unsatisfying task performance. 3.3.2 Gaze trajectories and stationary distributions of the two cognitive strategies To visualize the characteristics of the two cognitive process strategies, we further selected the most typical samples from each pattern and plotted their gaze trajectories (Fig. 6). For the holistic process strategy (ID=29), the stationary distributions for MS were 0.360, 0.312, and 0.318 across the three ROIs, and for LS, 0.377, 0.342, and 0.281, reflecting nearly equal attention allocation across all three regions. In contrast, the stationary distributions of the Chunk-Based Process Strategy revealed differential attention allocation: for MS, the values were 0.391, 0.439, and 0.170 across the three ROIs, and for LS, 0.340, 0.473, and 0.187, indicating that patient diagnose matters more than that of the holistic process strategy. A comparison between the LS and MS groups showed that the MS group exhibited more extensive gaze trajectories, suggesting greater attention engagement. When comparing the holistic process and chunk-based process strategy, it was observed that the holistic process strategy showed no difference in gaze trajectories between the machine alarm interface and the diagnosis interface, implying relatively uniform attention distribution. In contrast, the chunk-based strategy demonstrated significantly fewer gaze trajectories on the machine alarm interface compared to the diagnosis interface, indicating that this group allocated substantially more attention to diagnosis than to machine alerts. 3.4 Cognitive strategies and their task performance In our research, we applied the reaction time and task accuracy to measure task performance under different cognitive strategies, and the details can be seen in table 2. Table 2 Cognitive Strategies and their task performance Cognitive Strategies Experiment condition Task performance Decision Accuracy（%） /M±SD Reaction Time(ms) /M±SD Holistic MS 51.11 ± 2.17 292.36 ± 29.67 LS 40.63 ± 1.99 312.41 ± 48.72 Chunk-Based MS 44.64 ± 3.38 263.21 ± 23.79 LS 36.67 ± 2.70 267.42 ± 16.33 3.4.1 Regression analysis between cognitive strategy and task accuracy Regression analysis was performed with the H-C score as the independent variable and task accuracy as the dependent variable (Fig.7). The results revealed that in the MS group, the regression model was statistically significant. Specifically, the H-C score demonstrated a significant positive predictive effect ( β = 0.554, p = 0.006). In contrast, in the LS group, the regression model was not significant. The predictive effect of the H-C score on decision accuracy was not significant ( β = 0.290, p =0.179). A higher H-C score indicates a more typical holistic process strategy. These results suggest that holistic process strategy positively predicts decision accuracy in the MS group, whereas this relationship does not approach statistical significance in the LS group. 3.4.2 Regression analysis between cognitive strategies and response time Regression analysis was also conducted with the H-C score as the independent variable and reaction time as the dependent variable (Fig.8). The results showed that in the LS group, the relationship between H-C score and reaction time did not reach statistical significance (F=3.436, p = 0.078). Similarly, in the MS group, the relationship between H-C score and reaction time was not significant as well (F = 3.190, p = 0.089). These results indicate that the relationship between cognitive strategy and reaction time did not reach statistical significance in either group, though both showed similar trend effects that with the increasing likelihood of holistic process strategy application the reaction time will increase as well. 3.5 cognitive strategies and their correspondence cognitive workload 3.5.1 Regression analysis between cognitive strategies and saccadic duration A regression analysis was conducted with the H-C score as the independent variable and saccade duration as the dependent variable (Fig.9). The results revealed that in the LS group, the regression model was statistically significant (F= 17.105, p < 0.001), with an adjusted R² of 0.392, indicating that the H-C score accounted for 39.2% of the variance in saccadic duration. Specifically, the H-C score had a highly significant positive predictive effect on saccadic duration ( β = 0.645, p < 0.001). In contrast, the predictive effect in the MS group did not reach statistical significance ( p = 0.062). These findings suggest that a more typical holistic process strategy was associated with longer saccadic duration. This effect was particularly strong and significant in the LS group, whereas the trend in the MS group was relatively weaker and statistically insignificant. 3.5.2 Regression analysis between cognitive strategies and pupil size Considering the skewed distribution of pupil size data, this study employed Mann-Whitney U tests to compare differences in pupil diameter between the two cognitive strategies within the LS and MS groups separately(Fig.10).In the MS group, the pupil diameter of participants who applied the holistic process strategy (Median=2052.74) was larger than that of the participants who applied the chunk-based process strategy((Median=1729.06). However, the Mann-Whitney U test indicated that this difference was not statistically significant ( p = 0.139). In the LS group, the difference in pupil diameter between the two visual processing styles was smaller (holistic process strategy : (Median = 1841.81; chunk-based process strategy: (Median = 1807.76), and no significant difference was found by statistical testing (U = 65.000, p = 0.861). 4 Discussion 4.1 Cognitive strategies and their characteristics under parallel multitasks Our study employed an eye-tracking-based Hidden Markov Modeling approach to investigate the cognitive strategies in response to simulated ICU nursing parallel multitask characterized with nurse-machine interaction. Participants primarily employed two distinct strategies: the holistic process strategy and the chunk-based process strategy. These strategies align with the classical cognitive frameworks of constructive matching and response elimination, respectively (Federico & Montague, 1980; Zhang et al., 2012). The constructive matching strategy is characterized by participants deducing the rules that govern the graphics’ arrangement, using these rules to mentally construct a response, and subsequently searching for the matching option. In our study, the holistic process strategy aligns with this approach: participants treated the machine alert interface and each patient’s diagnosis as an integrated whole, retrieved information from long-term working memory, analyzed and compared the three patient situations, and finally entered the caring priority (Fig.3). Conversely, the response elimination strategy entails participants sequentially eliminating unreliable alternatives to arrive at the correct answer. The chunk-based process strategy corresponds to this approach: participants first processed the three machine alert interfaces to form an initial answer, then analyzed the patients’ diagnoses to eliminate incorrect options before determining the final answer (Fig.4).Additionally, the research results showed that all participants remained the same strategy in both group tasks, indicating that none of them adjusted their cognitive strategies in response to task characteristics, highlighting a limitation in cognitive flexibility that merits further attention. 4.2 Attention allocations of different cognitive strategy under parallel multitasks The study revealed that nurses employing the holistic process strategy showed a strong tendency to focus on the center patient, with attention distributed almost equally across all patients (Fig.3 & Fig.5). In contrast, participants using the chunk-based process strategy tended to first compare alarm information across all cases before shifting their attention to diagnostic details. The initial fixation was highly concentrated, with 0.91 probability directed toward ROI 1 (machine alert interface), indicating a strong initial prioritization of machine information during encoding. Research findings also indicated that nurses tend to make clinical decision under parallel multitask without careful comparing between the last two ROIs (Fig.5). 4.3 Task performance of different cognitive strategies under parallel multitasks In our study, task performance under different cognitive strategies was evaluated using two outcome indicators: task accuracy and response time. For trials involving more severe diseases, nurses adopting the holistic process strategy demonstrated higher task accuracy compared to those using the chunk-based process strategy (51.11% vs. 44.64%), though with longer response times (14.61s vs. 13.16s). Similarly, in trials with less severe diseases, the holistic process strategy also led to higher accuracy (40.63% vs. 36.67%), but again with longer response times (15.62s vs. 13.37s). These findings align with established cognitive theories and prior research. Thus, nurses working in high-risk clinical settings should weigh patient needs when deciding which cognitive strategy to employ. When caring for critically ill patients, where accuracy outweighs speed, it is advisable to adopt a holistic process strategy to enhance the synthesis of multifaceted information and support accurate clinical judgment. However, since sustained use of the holistic process strategy may be cognitively demanding for nurses, a chunk-based strategy is recommended for cases of lower severity, as it helps conserve time and reduce cognitive load. 4.4 Cognitive workload of different cognitive strategies under parallel multitasks Generally, the constructive matching strategy imposes a higher cognitive workload than the response elimination strategy and therefore influence their task performance (Wang & Zhan, 2025). The positive association between holistic process and saccade duration—particularly strong in the MS group—indicates that this strategy demands greater cognitive effort, potentially reflecting deeper information integration or more effortful comparison processes (Solomon et al., 2011). Although pupil size differences between strategies did not reach statistical significance, the trend toward larger pupils in the participants who applied holistic process strategy aligns with the interpretation that this strategy is cognitively demanding. But the increased cognitive workload may be justifiable in high-severity contexts where accurate prioritization is critical but less efficient in lower-severity scenarios. 4.5 Implications for research and practice This study represents a pioneering effort to investigate the cognitive strategies employed by nurses when performing parallel multitask in high-risk hospital departments. The experimental stimulus was developed based on real clinical cases and incorporated the most common medical devices found in ICU, thereby ensuring a high ecological validity. Using an eye-tracking-based Hidden Markov Modeling approach, this research identified two cognitive strategies adopted by nurses during parallel multitasking therefore revealed how nurses interact with multiple medical machines instantaneously. Besides we evaluated and compared the task performance, reaction time and cognitive workload associated with each strategy. Based on these findings, first we should develop medical devices with adaptive interfaces that can adjust information layout based on nurse-machine's interaction patterns and patients' clinical conditions. Second, nurses should weigh the pros and cons of different cognitive strategy and made proper clinical decisions under parallel multitask. The holistic process strategy should be applied when caring for critically ill patients to enhance patient safety, whereas the chunk-based process strategy is more suitable for less severe cases to improve efficiency, reduce cognitive load, and prevent exhaustion. Furthermore, we emphasized the importance of cognitive flexibility and recommend the development of school or on-the-job training programs to enhance this ability among nurses. 4.6 Research strength and limitations This study utilized eye movement data and a Hidden Markov Model to investigate the unobservant cognitive processes employed by nurses during parallel multitask characterized with multiple nurse-machine interaction, offering an objective and real-time depiction of cognitive dynamics with high accuracy. However, several limitations should be acknowledged. First, although the experiment was conducted in a sound-attenuated laboratory at the Institute of Psychology, Chinese Academy of Sciences—which helped minimize interference from extraneous variables—the controlled setting may reduce simulation fidelity compared to real clinical environments, thereby limiting the generalizability of the findings. Second, the modest sample size may have constrained the statistical power to detect subtle effects. Third, the majority of participants were novice nurses with fewer than five years of experience, meaning their cognitive strategies may differ from those of more experienced nurses. Future studies should include a more diverse range of participants with varying levels of experience and consider multi-center collaborations. Additionally, immersive virtual simulation presents a promising avenue in cognitive flexibility training for nursing students or on-job nurses. 5 Conclusion With the influx of various medical devices and persistent global shortage in nursing, making timely and accurate decision under multitask scenarios become an essential ability for nurses. And cognitive strategy plays a significant role in ensuring quality of decision-making. Our research identified two cognitive strategies nurses applied when managing multitasking conflicts. The holistic process strategy enhances decision accuracy in high-severity scenarios but may incur higher cognitive costs, whereas the chunk-based process strategy may suffice for lower-complexity tasks. These findings not only advance our theoretical understanding of nursing cognition under parallel multitask, but also provide an evidence-based foundation for developing targeted training and interface design solutions aimed at supporting optimal clinical decision-making and improving patient safety outcomes. Declarations Author Contribution X.W.: conceptualization, data curation, formal analysis, funding acquisition, project administration, resources, supervision, writing – review & editing.T.T. F: conceptualization, data curation, formal analysis, investigation, project administration, supervision, writing – original draft, writing – review & editing.J.Q. C:data curation, formal analysis, methodology, writing – original draft, writing – review & editing. S.F. W:data curation, formal analysis, writing – original draft.H.W:data curation, formal analysis, writing – original draft. Acknowledgement First and foremost, we would like to express our sincere gratitude to all the participants for their valuable contributions to this study. We are deeply grateful to Mr. Wei Chu-guang and Ms. Yuan Yi-ran, senior laboratory technicians at the Institute of Psychology, Chinese Academy of Sciences, for their expert assistance in the design and implementation of the experiments.We also extend our gratitude to Ms. Yuan Cui, Deputy Director of the Nursing Department at Peking University First Hospital; Ms. Ma Li, Head Nurse of the Emergency Department at Peking University Third Hospital; and Mr. Liu Feng-gang, Head Nurse of the ICU at the First Affiliated Hospital of University of South China, for their insightful review of the experimental stimuli. References Abu Arra AY, Ayed A, Toqan D, Albashtawy M, Salameh B, Sarhan AL, Batran A (2023) The Factors Influencing Nurses’ Clinical Decision-Making in Emergency Department. INQUIRY , 60 , 469580231152080. https://doi.org/10.1177/00469580231152080 Cameron LD, Jago L (2013) Cognitive Strategies. In: Gellman MD, Turner JR (eds) Encyclopedia of Behavioral Medicine. Springer, New York, NY. https://doi.org/10.1007/978-1-4419-1005-9_1120 Coviello E, Chan AB, Lanckriet GRG (2014) Clustering hidden Markov models with variational HEM. J Mach Learn Res 15(1):697–747 Cowan N (2001) The magical number 4 in short-term memory: a reconsideration of mental storage capacity. Behav Brain Sci 24(1):87–185. https://doi.org/10.1017/s0140525x01003922 Federico PA, Montague WE (1980) Aptitute, Learning, and Instruction: Cognitive process analysis of aptitude .Routledge Grewal J, Krzywinski M, Altman N (2019) Markov models - hidden Markov models. Nat Methods 9(16):795–796. https://doi.org/10.1038/s41592-019-0532-6 Griffin JW, Webb SJ, Keehn B, Dawson G, McPartland JC (2024) Autistic Individuals Do Not Alter Visual Processing Strategy During Encoding Versus Recognition of Faces: A Hidden Markov Modeling Approach. J Autism Dev Disord . Advance online publication. https://doi.org/10.1007/s10803-024-06259-9 Jarosz A, Raden M, Wiley J (2019) Working memory capacity and strategy use on the RAPM. Intelligence , (77), 101387. https://doi.org/10.1016/j.intell.2019.101387 Jastrzębski J, Ciechanowska I, Chuderski A (2018) The strong link between fluid intelligence and working memory cannot be explained away by strategy use. Intelligence 66:44–53. https://doi.org/10.1016/j.intell.2017.11.002 Kim Y, Ryu GW, Lee MJ, Choi M, Cho E (2022) Multitasking of Nursing Activities: A Time to Motion Study to Capture Nursing Workflow Using Eye Tracker. SSRN Electron J. https://doi.org/10.2139/ssrn.4027254 Kuha J (2004) AIC and BIC: Comparisons of Assumptions and Performance. Sociol Methods Res 33(2):188–229. https://doi.org/10.1177/0049124103262065 Laurence PG, Jana TA, Bunge SA, Macedo EC (2023) Eye Gaze Patterns during Reasoning Provide Insights Regarding Individual Differences in Underlying Cognitive Abilities. J Intell 11(4):75. https://doi.org/10.3390/jintelligence11040075 Li JX (2025) Toward Enhancing Multitasking Performance: Modeling, Prediction, and Intervention Design .[Doctoral dissertation, University of Washington].University Libraries,University of Washington Liu JC, Li KA, Yeh SL, Chien SY (2022) Assessing Perceptual Load and Cognitive Load by Fixation-Related Information of Eye Movements. Sens (Basel) 22(3):1187. https://doi.org/10.3390/s22031187 Liu K, Wang A (2012) Does media multitaskinging always hurt? A positive correlation between multitaskinging and multisensory integration. Psychon Bull Rev 19:647–653. https://doi.org/10.3758/s13423-012-0245-7 Lu Y, Wang J, Dai W, Zou B, Lu X, Hong W (2025) Research on risk factors of human stampede in mass gathering activities considering organizational patterns. Front Public Health 13:1555735. https://doi.org/10.3389/fpubh.2025.1555735 Mathew M, John A, Vazhakkalayil Ramachandran R (2025) Nurse stress and patient safety in the ICU: physician-led observational mixed-methods study. BMJ Open Qual 14(2):e003109. https://doi.org/10.1136/bmjoq-2024-003109 Oladele IO, Okoro CJ, Falana SO, Olajesu FO, Adelani SO, Onuh LN (2024) Advancement of human-machine collaboration in manufacturing: A review on industry 1.0–6.0. J Mech Eng Sci 18(4):10303–10329. https://doi.org/10.15282/jmes.18.4.2024.7.0813 Olin K, Göras C, Nilsson U, Unbeck M, Ehrenberg A, Pukk-Härenstam K, Ekstedt M (2022) Mapping registered nurse anaesthetists' intraoperative work: tasks, multitasking, interruptions and their causes, and interactions: a prospective observational study. BMJ Open 12(1):e052283. https://doi.org/10.1136/bmjopen-2021-052283 Snow RE, Federico P, Montague WE (1980) Aptitute, Learning, and Instruction. Cognitive process analysis of aptitude. Routledge Solomon D, Albert N, Sun Z, Bowers AM, Molnar M (2011) Complexity of Care Is Associated with Distressing Environmental Factors. Clin Nurse Specialist 25(4):186–192. https://doi.org/10.1097/nur.0b013e318221f2d3 Stasch SM, Mack W Exploring Task Prioritization in VR Flight Environments: Can Eye-Tracking Uncover Cognitive Control? In Proceedings of the 2025 Symposium on Eye Tracking Research and Applications (ETRA '25). Association for Computing Machinery, New York, NY, USA, Article (2025) 76, 1–7. https://doi.org/10.1145/3715669.3725898 Sweller J (1988) Cognitive Load during Problem Solving: Effects on Learning. Cognit Sci 2(12):257–285. https://doi.org/10.1016/0364-0213(88)90023-7 Wagenmakers E-J, Farrell S (2004) AIC model selection using Akaike weights. Psychon Bull Rev 11(1):192–196. https://doi.org/10.3758/BF03206482 Wang Z, Zhan P (2025) Eye-tracking-based hidden Markov modeling for revealing within-item cognitive strategy switching. Behav Res 175(57). N/A. https://doi.org/10.3758/s13428-025-02678-3 Yuan X, Zhong L (2024) Effects of multitaskinging and task interruptions on task performance and cognitive load: considering the moderating role of individual resilience. Curr Psychol 43(28):23892–23902. https://doi.org/10.1007/s12144-024-06094-2 Zhang P, Jiang Y, He S (2012) Voluntary Attention Modulates Processing of Eye-Specific Visual Information. Psychol Sci 23(3):254–260. https://doi.org/10.1177/0956797611424289 Additional Declarations No competing interests reported. Supplementary Files AppendixA.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 31 Dec, 2025 Editor assigned by journal 29 Oct, 2025 Submission checks completed at journal 29 Oct, 2025 First submitted to journal 29 Oct, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7975457","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":567854571,"identity":"133ceb68-d802-47f0-b086-4da4f0d0a0ee","order_by":0,"name":"Tingting Feng","email":"","orcid":"","institution":"Peking University","correspondingAuthor":false,"prefix":"","firstName":"Tingting","middleName":"","lastName":"Feng","suffix":""},{"id":567854572,"identity":"f6e71b1a-26f6-48cd-8e73-ee03e1a37e4f","order_by":1,"name":"Jia-qi Cai","email":"","orcid":"","institution":"Peking University","correspondingAuthor":false,"prefix":"","firstName":"Jia-qi","middleName":"","lastName":"Cai","suffix":""},{"id":567854573,"identity":"eba94264-c670-4aea-bc31-46fdc9af5f63","order_by":2,"name":"Sheng-feng Wang","email":"","orcid":"","institution":"Peking University","correspondingAuthor":false,"prefix":"","firstName":"Sheng-feng","middleName":"","lastName":"Wang","suffix":""},{"id":567854574,"identity":"4c17cfe4-ba86-4af3-b1b0-3c2d93c0b9cb","order_by":3,"name":"Hao Wu","email":"","orcid":"","institution":"Peking University","correspondingAuthor":false,"prefix":"","firstName":"Hao","middleName":"","lastName":"Wu","suffix":""},{"id":567854576,"identity":"e57afa3b-da03-48a0-aeb5-03a4c4f635e1","order_by":4,"name":"Xue Wu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAuklEQVRIiWNgGAWjYNCCigPMYJqHeC1nSNbC2HaAgXgtBjdyzCQ+zrvDrjsjgfHB2zYGeXPCWtLSJGdue8ZsdiOB2XBuG4PhzgYCWsxuJB+T5t12GKSFTZq3jSHB4ABBLYlt0n/ngLWw/yZSC9AWxgaILcxEabE/8yzZsucY0C9nHjZLzjknYbiBkBbJ9hzDGz9q7iSbHU8++OFNmY08QVtgIBkYOw1AWoJI9UBgR7zSUTAKRsEoGHEAABjVQbndGkUlAAAAAElFTkSuQmCC","orcid":"","institution":"Peking University","correspondingAuthor":true,"prefix":"","firstName":"Xue","middleName":"","lastName":"Wu","suffix":""}],"badges":[],"createdAt":"2025-10-29 04:38:04","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7975457/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7975457/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100691408,"identity":"ed0bbd77-f470-45e3-9f6f-eff5e87dfd53","added_by":"auto","created_at":"2026-01-20 14:06:32","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":125698,"visible":true,"origin":"","legend":"\u003cp\u003eIllustration of experiment stimulus\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote: Each trial presents three patients’ machine alert interface and diagnose simultaneously. In Fig.1(a), the first patient is diagnosed with acute lymphoblastic leukemia and the hemofiltration machine is alarming. The second patient is diagnosed with acute pancreatitis complicated by septic shock, now the nutrition pump is alarming. The third patient is experiencing respiratory failure following a stroke, the oxygen therapy machine is now alarming. In Fig.1(b), the first patient is now experiencing an acute myocardial infarction complicated by epilepsy. The second patient is a patient just finished an abdominal aortic aneurysm surgery. The third patient is a postoperative patient following pituitary adenoma surgery. Their ECG monitors are alarming at the same time.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/7decc745fe60605c79d51a54.png"},{"id":100691159,"identity":"35eff5b3-3eb4-44c7-be43-fab5c7035e95","added_by":"auto","created_at":"2026-01-20 14:04:25","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":191617,"visible":true,"origin":"","legend":"\u003cp\u003eResearch procedure and the experiment scenario\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/e944f0e4d02233cc389f53dc.png"},{"id":100691218,"identity":"e953589c-2086-4c53-aaf8-6d9fd2998846","added_by":"auto","created_at":"2026-01-20 14:05:10","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":259556,"visible":true,"origin":"","legend":"\u003cp\u003eHolistic cognitive process strategy\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote:The circular area with numbers indicates the region of interest, and the number represents the identifier of the region of interest. Each ROI integrates both the machine alert interface and diagnostic text for each patient. Markov transition probability matrix reflects systematic comparisons between patients.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/ca29b8f380bc15312d54b7cb.png"},{"id":100691387,"identity":"c3774432-bb0b-4f89-a898-bac0122cf560","added_by":"auto","created_at":"2026-01-20 14:06:20","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":266340,"visible":true,"origin":"","legend":"\u003cp\u003eChunk-based cognitive process strategy\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote: The circular area with numbers indicates the region of interest, and the number represents the identifier of the region of interest. ROI1 contains machine alert interfaces across all three patients; ROI2 and ROI3 contain diagnostic information. Markov transition probability matrix reflect systematic comparisons between patients.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/c71758ca64f0c78cbdb81211.png"},{"id":100691202,"identity":"af7c7ed6-1a9e-4287-8fb2-81224d9cbb3c","added_by":"auto","created_at":"2026-01-20 14:04:53","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":102499,"visible":true,"origin":"","legend":"\u003cp\u003eTransition probability across ROIs under different cognitive strategy in various condition\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote: Circle with number (1 or 2 or 3) represent different ROIs in various experiment condition, the number above the arrow indicate the transition probability between different ROIs.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/7b03f374743d541622464a4e.png"},{"id":100691058,"identity":"0ac8fc27-cb75-4070-a606-f91cd8162c31","added_by":"auto","created_at":"2026-01-20 14:02:56","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":523456,"visible":true,"origin":"","legend":"\u003cp\u003eGaze trajectories for Holistic and Chunk-Based Strategies in LS and MS Groups\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/cec8ade0ef381c444970c399.png"},{"id":100691076,"identity":"affa2e13-263d-469d-a317-f0e56b8fc340","added_by":"auto","created_at":"2026-01-20 14:03:19","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":110689,"visible":true,"origin":"","legend":"\u003cp\u003eScatter plot between the H-C score and task accuracy\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/d5c24983f91ed04bef7fc175.png"},{"id":100691802,"identity":"f494d1b5-d870-473d-8fd7-7d4c57dfb60d","added_by":"auto","created_at":"2026-01-20 14:07:35","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":119151,"visible":true,"origin":"","legend":"\u003cp\u003eScatter plot between the H-C score and response time\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/88563a671aa1d41c306a8180.png"},{"id":100691214,"identity":"cb62d350-668a-4d68-b71b-4daab1f4d937","added_by":"auto","created_at":"2026-01-20 14:05:08","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":101975,"visible":true,"origin":"","legend":"\u003cp\u003eScatter plot between the H-C score and saccadic duration\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/adedfab152e72c89e339a5fe.png"},{"id":100691309,"identity":"a66a5559-19dc-40a3-8d57-761777739419","added_by":"auto","created_at":"2026-01-20 14:05:34","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":92741,"visible":true,"origin":"","legend":"\u003cp\u003ePupil size of different cognitive strategies in different experiment condition\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/6cfcb5f98c83ff38c8b4cd72.png"},{"id":100796479,"identity":"157cd155-9b08-4896-9af6-35013a3df7a5","added_by":"auto","created_at":"2026-01-21 13:43:29","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2392699,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/e1c0772d-e08f-4640-81c5-e5bcebc39601.pdf"},{"id":100691082,"identity":"40c12e97-b6ef-4e76-a92d-ffada538f74d","added_by":"auto","created_at":"2026-01-20 14:03:30","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":11905,"visible":true,"origin":"","legend":"","description":"","filename":"AppendixA.docx","url":"https://assets-eu.researchsquare.com/files/rs-7975457/v1/7bccdfdb7c6e7b2f5d886773.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Uncovering Cognitive Strategies for ICU Parallel Nursing Multitasks: An Eye-tacking-based Hidden Markov Modeling Study","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eNursing practice is complex and labor-intensive, particularly in high-acuity settings such as the ICU and emergency department. Parallel multitasking is defined as the simultaneous execution or rapid switching between two or more tasks by an individual (Kim et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Lu et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Olin et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Extensive research has demonstrated that parallel multitasking can compromise task performance, prolong response times, and elevate cognitive workload (Liu \u0026amp; Wang, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Yuan \u0026amp; Zhong, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Despite these documented challenges, parallel multitasking occurs frequently in nursing practice, which may result in interruption and distraction as well as a higher probability of human error. One study indicated that nurses spend over 50.1% of their working hours engaged in multitasking and make hundreds of ad hoc decisions per shift (Abu Arra et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Considering the persistent nursing shortage and rising application of medical devices as well as artificial intelligence in the clinic, it is\u003c/p\u003e \u003cp\u003eimpossible to avoid multitasking. Thus, enhancing nurses\u0026rsquo; decision-making capabilities under parallel multitasking especially for ICU nurses is of great urgency (Mathew et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn accordance with the cognitive theory, human cognitive capacity is fundamentally limited, with working memory typically constrained to approximately four chunks of information at one time (Cowan, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Sweller, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e1988\u003c/span\u003e). Therefore, with the increasing of simultaneous task the likelihood of error will increase. Cognitive strategies refer to structured mental procedures employed to achieve specific goals or solve problems(Cameron \u0026amp; Jago, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). By exploring the cognitive strategies nurses employed in parallel multitasking, we can find out the barriers and facilitators in decision-making performance in this complex situation which is significant in safeguarding decision quality and improving patient safety. Although extensive studies have been conducted in exploring cognitive strategy in parallel multitasking in safety-critical industries\u0026mdash;such as aviation and traffic control, researches towards cognitive strategies in nursing under parallel multitasking context remains limited (Li, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Stasch \u0026amp; Mack, 2025).\u003c/p\u003e \u003cp\u003eCurrently the most commonly applied approaches for identifying cognitive strategies include subjective reporting methods (e.g. self-report questionnaire, think-aloud), EEG, fMRI and eye movement tracking. Although subjective evaluation methods are easier to implement, they are susceptible to participants' subjective biases and exhibit temporal delays, making it difficult to capture real-time cognitive processes (Jarosz et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Jastrzębski et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). EEG and fMRI can identify cognitive strategy under parallel multitasking in real time with high accuracy, but they require participants to wear corresponding equipment which is costly and may interfere daily work, thus limited its application. Whereas eye movements can reflect participants\u0026rsquo; thought process and mental states in real time, offering insights into their understanding of the problem, level of attention, information processing, and cognitive workload (Laurence et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Wang \u0026amp; Zhan, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Eye-tracking research results are especially beneficial in clinical settings with intensive human-machine interactions, including the ICU, emergency department and operating room. Thus, considering the accuracy, cost-effectiveness and future promotion, we selected the eye movement tracking technology to investigate the cognitive strategy in parallel multitasking.\u003c/p\u003e \u003cp\u003eHidden Markov Model (HMM) is a statistical model particularly well-suited for analyzing time-series data, which includes observable states, such as eye movement trajectories and unobservant hidden states like cognitive processes (Grewal et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). While traditional eye-tracking analyses, such as fixation duration and heatmaps can only provide static descriptions, HMM can map the complex eye movement sequences into a series of hidden cognitive states, thereby revealing an individual\u0026rsquo;s cognitive strategies during visual tasks (Griffin et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Wang \u0026amp; Zhan, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003eThus, considering the increasing incidence of parallel multitasking in nursing work and its potential harmfulness, especially for high-risk department, we conducted this research to investigate ICU nurses\u0026rsquo; cognitive strategy and their corresponding performance as well as cognitive workload in parallel multitasking. The research questions are: 1) What cognitive strategies do nurses employ during parallel multitasking? 2) How do different cognitive strategies perform?3) What is the cognitive workload associated with different cognitive strategies?\u003c/p\u003e"},{"header":"2 Materials and methods","content":"\u003cp\u003e2.1 Design\u003c/p\u003e\n\u003cp\u003eThis study utilized an observational research design. We required subjects to perform simulated ICU parallel multitask covering patients with different disease severity: the more severe illness (MS group) and the less severe illness (LS group). These two experimental blocks presented in random order, with each block consisting of 20 trials.\u003c/p\u003e\n\u003cp\u003e2.2 Participants\u003c/p\u003e\n\u003cp\u003eThe inclusion criteria were: (a) being a registered nurse; (b)at least six months of clinical work experience in an ICU or emergency department; (c) being in good physical and mental health, capable to cooperate fully with the experimental procedures; and (d) having normal or corrected-to-normal vision. Exclusion criteria included: (a) experiencing a recent cold or any form of eye discomfort; and (b) fatigue, particularly following a night shift. Participants were recruited from 3 tertiary hospitals in Beijing, China, between April and June using poster advertisements. Interested individuals contacted the research team via WeChat and underwent an initial eligibility screening. Those who met all the criteria were subsequently provided with an online demographic questionnaire and a self-developed risk perception survey (Appendix A). The risk perception survey was developed based on the latest definition of risk, including perception of the probability of risk in their nursing work, the potential\u0026nbsp;consequences\u0026nbsp;and risk tolerance. Each item uses a 7-point Likert scale, with one item being reverse-scored(Roszkowski \u0026amp; Davey, 2010).All participants received detailed information regarding the study procedures and provided informed consent using an IRB-approved form (Approval No. IRB00001052-25007).\u003c/p\u003e\n\u003cp\u003e2.3 Parallel Multitasking experiment\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e2.3.1 Eye-tracking apparatus\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eParticipants\u0026rsquo; eye movement were recorded with the Eyelink 1000. 9 Point of gaze was sampled at 1000 Hz with an accuracy of 0.5~1.0 visual degrees. The experiment was conducted in a laboratory with constant control of sound and light conditions in Institute of Psychology, Chinese Academy of Sciences. Participants were seated in a chair approximately 70 cm from an flat screen monitor (1920 x 1080).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e2.3.2 Stimulus\u003c/p\u003e\n\u003cp\u003eParticipants completed two experimental blocks, each comprising 20 trials. Each block represented a distinct clinical scenario: one featured patient with more severe conditions (MS group), while the other involved patients with less severe conditions (LS group). In each trial, a set of three patient scenarios was presented simultaneously (Fig.1). Each scenario consisted of a simulated medical device alarm interface accompanied by a corresponding diagnostic description (Fig.1). The MS group included critical conditions such as cardiogenic shock, extensive burns, severe sepsis, multiple fractures, massive pulmonary embolism, aortic dissection, and sepsis with renal failure. The LS group comprised less acute diagnoses, including chronic heart failure, status post thyroidectomy, metabolic acidosis, and ascites due to liver cirrhosis. The device alarm interfaces were modeled based on common ICU equipment, including ECG monitors, haemodialysis machines, blood glucose monitors, infusion pumps, enteral feeding pumps, and ventilators. All experimental materials were grounded in real clinical practice and subjected to three rounds of evaluation by a panel of four experts in clinical nursing and engineering psychology. The stimuli were further refined based on the results of a pilot study(\u003cem\u003eN\u003c/em\u003e=2) before the final versions were implemented.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e2.3.3 Procedure\u003c/p\u003e\n\u003cp\u003ePrior to the formal experiment, participants completed five practice trials under the guidance of the researcher to familiarize themselves with the task requirements. The formal experiment began only after participants confirmed their understanding and readiness. Participants then completed two blocks of trials in a randomized order. In each trial, after observing the stimulus, they were required to type in the patient care sequence (eg.321, 123 or 312). After entering their decision, participants pressed the \u0026quot;Space\u0026quot; key to proceed to the next trial, the whole research methodology flow and experiment scenario are shown in Fig.2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.4 Data Analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e2.4.1 Eye-movement Based Hidden Markov Modeling\u003c/p\u003e\n\u003cp\u003eWe analyzed eye-tracking data using a Hidden Markov Model approach implemented with EMHMM toolbox (v0.80) by MATLAB 2018b. This method captured both spatial and temporal dynamics of fixations, allowing researchers to identify dynamic visual processing strategies. The analysis involved three steps.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eStep 1: Eye-Movement Data Cleaning\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eFirst of all, we thoroughly screened the eye-movement data and deleted invalid data points based on quality flags from the eye tracker. This data cleaning step can ensure data reliability for subsequent modeling.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eStep 2: Establishment of Individualized Hidden Markov Model\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eResearchers developed an individualized HMM for each participant based on their sequence of fixation coordinates recorded during the trial. Model parameters including initial probabilities, transition probabilities between states and Gaussian emission probabilities defined each hidden state (i.e., region of interest, ROI), and were estimated using a variational Bayesian approach. In accordance with previous literature, the models were compared with 1 to 6 hidden states and then researchers selected the most suitable hidden states.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eStep 3: Hidden Markov Model Clustering\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe Variational Hierarchical Expectation-Maximization (VHEM) algorithm was applied to cluster individualized HMMs by parameter similarity and identify a small set of representative HMMs that captured common eye movement patterns across the sample. Each representative HMM summarized the spatial distribution of ROIs and transition patterns characteristic of a particular visual strategy. In this research, the Holistic-Chunk score (H-C score) is used to quantify the similarity between each participant\u0026rsquo;s eye movement pattern and the representative strategies. This score was derived from the log-likelihoods of the individual\u0026rsquo;s data being generated by each representative HMM, a positive value indicates closer alignment with the holistic process strategy, while a negative value reflects greater similarity to the chunk-based process strategy.\u003c/p\u003e\n\u003cp\u003e2.4.2 Analysis of task performance and cognitive workload of different cognitive strategies\u003c/p\u003e\n\u003cp\u003eTo investigate the impact of cognitive strategy on task performance and cognitive workload, we firstly conducted the statistical description for the demographic characteristics, risk perception, reaction time and task accuracy of all the participants and trials. Then, we analyzed the correlation between cognitive strategy represented by H-C score with the task performance and cognitive workload through linear regression. In this research, the saccadic duration and pupil size were selected as parameters to measure cognitive workload in this research (Liu et al., 2022; Liu et al., 2022). \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e2.5 Ethical approval\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe study was approved by the Ethics Committee of Peking University Health Science Center (IRB No.00001052-25007, February 19, 2025). Informed consent was obtained from all participants prior to participation.\u003c/p\u003e"},{"header":"3 Results","content":"\u003cp\u003e3.1 Demographics, and risk perception scores of participants\u003c/p\u003e\n\u003cp\u003eA total of 30 participants were recruited from 3 tertiary hospitals in Beijing, China from April to June. 3 participants\u0026rsquo; data were excluded in the data cleaning phase due to missing eye-tracking data and the included participants\u0026rsquo; demographic information can be seen in table 1.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1\u003c/strong\u003e Demographic of the participants and their risk perception score (\u003cem\u003eN\u003c/em\u003e=27)\u003c/p\u003e\n\u003ctable width=\"567\"\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd colspan=\"2\" width=\"381\"\u003e\n\u003cp\u003eCharacteristics\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003eN (%)/M\u0026plusmn;SD/Median (IQR)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd colspan=\"2\" width=\"381\"\u003e\n\u003cp\u003eAge (year)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e25.48\u0026plusmn;3.08\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd colspan=\"2\" width=\"381\"\u003e\n\u003cp\u003eGender\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"19\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"362\"\u003e\n\u003cp\u003eFemale\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e17 (62.96%)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"19\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"362\"\u003e\n\u003cp\u003eMale\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e10 (37.04%)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd colspan=\"2\" width=\"381\"\u003e\n\u003cp\u003eWorking experience(month)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e36 (12, 60)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd colspan=\"2\" width=\"381\"\u003e\n\u003cp\u003eEducation background\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"19\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"362\"\u003e\n\u003cp\u003eAssociate degree\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e8 (29.63%)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"19\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"362\"\u003e\n\u003cp\u003eBachelor's degree\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e18 (66.67%)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"19\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"362\"\u003e\n\u003cp\u003eGraduate degree\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e1 (3.7%)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd colspan=\"2\" width=\"381\"\u003e\n\u003cp\u003eProfessional title\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"19\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"362\"\u003e\n\u003cp\u003eRegistered Nurse\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e18 (66.67%)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"19\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"362\"\u003e\n\u003cp\u003eSenior Nurse\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e8 (29.63%)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"19\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"362\"\u003e\n\u003cp\u003eSupervisor Nurse\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e1 (3.7%)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd colspan=\"2\" width=\"381\"\u003e\n\u003cp\u003eRisk perception score\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e4.49\u0026plusmn;0.75\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"19\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"362\"\u003e\n\u003cp\u003eProbability of risk\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e5.00\u0026plusmn;1.21\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"19\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"362\"\u003e\n\u003cp\u003ePotential Consequences of Risk\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e5.56\u0026plusmn;1.16\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"19\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"362\"\u003e\n\u003cp\u003eRisk tolerance\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"186\"\u003e\n\u003cp\u003e5.07\u0026plusmn;0.99\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e3.2 Clustering of eye movement strategies using Hidden Markov Models\u003c/p\u003e\n\u003cp\u003eEye movement data from 27 participants were included in the analysis. Based the coordinate-based eye movement data, we identified two distinct cognitive strategies, they are the holistic cognitive process strategy and the chunk-based cognitive process strategy (Fig. 3 \u0026amp; 4). We analyzed the model fitness of models with 1 to 6 ROIs with both cognitive strategies, therefore to determine the optimal number of ROIs for each individual HMM. The Akaike information criterion (AIC) and the Bayesian information criterion (BIC) were used as model-fit indices to identify the optimal number of hidden states for the eye movement-based HMM (E-HMM)(Kuha, 2004; Wagenmakers \u0026amp; Farrell, 2004). A lower AIC and BIC values indicate a better fit of the E-HMM to the data. And we finally decided that the optimal number of ROIs was three based on AIC, BIC and the ROI performance.\u003c/p\u003e\n\u003cp\u003eAmong the participants who completed the MS group task(\u003cem\u003en\u003c/em\u003e=24), 11 participants (45.8%) applied holistic process strategy and 13 participants (54.2%) applied the chunk-based process strategy. Among the participants who completed the LS group task(\u003cem\u003en\u003c/em\u003e=26), 9 participants (34.6%) applied the holistic cognitive process strategy and 17 participants (65.4%) applied the chunk-based process strategy.\u003c/p\u003e\n\u003cp\u003e3.2.1 Holistic cognitive process strategy\u003c/p\u003e\n\u003cp\u003eIn the Holistic Process Strategy (Fig. 3), each ROI encompassed both the machine alert interface and the diagnose for the patient. This pattern suggests that nurses tended to integrate machine alert and diagnostic information into a coherent perceptual unit, which they compared in a systematic sequence later. The initial fixation was most frequently directed toward ROI1 (MS: 0.71; LS: 0.74), indicating a tendency to first attend to the central position patient.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote:The circular area with numbers indicates the region of interest, and the number represents the identifier of the region of interest. Each ROI integrates both the machine alert interface and diagnostic text for each patient. Markov transition probability matrix reflects systematic comparisons between patients.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e3.2.2 Chunk-Based cognitive process strategy\u003c/p\u003e\n\u003cp\u003eThe Chunk-Based Process Strategy was characterized by a clear spatial separation of visual attention: ROI1 encompassed the machine alert interfaces across all three patients, while ROI 2 and ROI 3 contained their corresponding diagnose information (Fig.4). This pattern suggests that nurses employing this strategy firstly compared machine interface information across all cases before shifting attention to diagnostic details. The initial fixation distribution was highly concentrated, with a probability of 0.91 directed toward ROI1 (machine alert interface), indicating a strong prioritization of machine alert information during initial encoding.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.3 Attention allocation characteristics under different experiment conditions \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e3.3.1 Transition between different rois under different cognitive strategy and condition\u003c/p\u003e\n\u003cp\u003eWe also analyzed the transition probability between different ROIs under different experiment conditions, which can reflect participants\u0026rsquo; attention allocation on different ROIs and reveal their cognitive strategies (Fig.5).Through Fig.5 we can notice that the transition probability between ROI 1 and ROI 2, ROI 1 and ROI 3 are much higher than that between ROI 2 and ROI 3, indicating that there is a high probability that nurse made clinical decision under parallel multitask without careful comparing between the last two ROIs which may explain their unsatisfying task performance.\u003c/p\u003e\n\u003cp\u003e3.3.2 Gaze trajectories and stationary distributions of the two cognitive strategies\u003c/p\u003e\n\u003cp\u003eTo visualize the characteristics of the two cognitive process strategies, we further selected the most typical samples from each pattern and plotted their gaze trajectories (Fig. 6). For the holistic process strategy (ID=29), the stationary distributions for MS were 0.360, 0.312, and 0.318 across the three ROIs, and for LS, 0.377, 0.342, and 0.281, reflecting nearly equal attention allocation across all three regions. In contrast, the stationary distributions of the Chunk-Based Process Strategy revealed differential attention allocation: for MS, the values were 0.391, 0.439, and 0.170 across the three ROIs, and for LS, 0.340, 0.473, and 0.187, indicating that patient diagnose matters more than that of the holistic process strategy.\u003c/p\u003e\n\u003cp\u003eA comparison between the LS and MS groups showed that the MS group exhibited more extensive gaze trajectories, suggesting greater attention engagement. When comparing the holistic process and chunk-based process strategy, it was observed that the holistic process strategy showed no difference in gaze trajectories between the machine alarm interface and the diagnosis interface, implying relatively uniform attention distribution. In contrast, the chunk-based strategy demonstrated significantly fewer gaze trajectories on the machine alarm interface compared to the diagnosis interface, indicating that this group allocated substantially more attention to diagnosis than to machine alerts.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.4 Cognitive strategies and their task performance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn our research, we applied the reaction time and task accuracy to measure task performance under different cognitive strategies, and the details can be seen in table 2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2\u003c/strong\u003e Cognitive Strategies and their task performance\u003c/p\u003e\n\u003ctable width=\"579\"\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd rowspan=\"2\" width=\"140\"\u003e\n\u003cp\u003eCognitive Strategies\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" width=\"85\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eExperiment condition\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd colspan=\"2\" width=\"354\"\u003e\n\u003cp\u003eTask performance\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"169\"\u003e\n\u003cp\u003eDecision Accuracy（%）\u003c/p\u003e\n\u003cp\u003e/M\u0026plusmn;SD\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"185\"\u003e\n\u003cp\u003eReaction Time(ms)\u003c/p\u003e\n\u003cp\u003e/M\u0026plusmn;SD\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd rowspan=\"2\" width=\"140\"\u003e\n\u003cp\u003eHolistic\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"85\"\u003e\n\u003cp\u003eMS\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"169\"\u003e\n\u003cp\u003e51.11 \u0026plusmn; 2.17\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"185\"\u003e\n\u003cp\u003e292.36 \u0026plusmn; 29.67\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"85\"\u003e\n\u003cp\u003eLS\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"169\"\u003e\n\u003cp\u003e40.63 \u0026plusmn; 1.99\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"185\"\u003e\n\u003cp\u003e312.41 \u0026plusmn; 48.72\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd rowspan=\"2\" width=\"140\"\u003e\n\u003cp\u003eChunk-Based\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"85\"\u003e\n\u003cp\u003eMS\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"169\"\u003e\n\u003cp\u003e44.64 \u0026plusmn; 3.38\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"185\"\u003e\n\u003cp\u003e263.21 \u0026plusmn; 23.79\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"85\"\u003e\n\u003cp\u003eLS\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"169\"\u003e\n\u003cp\u003e36.67 \u0026plusmn; 2.70\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"185\"\u003e\n\u003cp\u003e267.42 \u0026plusmn; 16.33\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e3.4.1 Regression analysis between cognitive strategy and task accuracy\u003c/p\u003e\n\u003cp\u003eRegression analysis was performed with the H-C score as the independent variable and task accuracy as the dependent variable (Fig.7). The results revealed that in the MS group, the regression model was statistically significant. Specifically, the H-C score demonstrated a significant positive predictive effect (\u003cem\u003e\u0026beta;\u003c/em\u003e = 0.554,\u003cem\u003e p \u003c/em\u003e= 0.006). In contrast, in the LS group, the regression model was not significant. The predictive effect of the H-C score on decision accuracy was not significant (\u003cem\u003e\u0026beta; \u003c/em\u003e= 0.290, \u003cem\u003ep\u003c/em\u003e =0.179). A higher H-C score indicates a more typical holistic process strategy. These results suggest that holistic process strategy positively predicts decision accuracy in the MS group, whereas this relationship does not approach statistical significance in the LS group.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.4.2 Regression analysis between cognitive strategies and response time\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRegression analysis was also conducted with the H-C score as the independent variable and reaction time as the dependent variable (Fig.8). The results showed that in the LS group, the relationship between H-C score and reaction time did not reach statistical significance (F=3.436, \u003cem\u003ep \u003c/em\u003e= 0.078). Similarly, in the MS group, the relationship between H-C score and reaction time was not significant as well (F = 3.190, \u003cem\u003ep\u003c/em\u003e = 0.089). These results indicate that the relationship between cognitive strategy and reaction time did not reach statistical significance in either group, though both showed similar trend effects that with the increasing likelihood of holistic process strategy application the reaction time will increase as well.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.5 cognitive strategies and their correspondence cognitive workload\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e3.5.1 Regression analysis between cognitive strategies and saccadic duration\u003c/p\u003e\n\u003cp\u003eA regression analysis was conducted with the H-C score as the independent variable and saccade duration as the dependent variable (Fig.9). The results revealed that in the LS group, the regression model was statistically significant (F= 17.105, \u003cem\u003ep \u003c/em\u003e\u0026lt; 0.001), with an adjusted R\u0026sup2; of 0.392, indicating that the H-C score accounted for 39.2% of the variance in saccadic duration. Specifically, the H-C score had a highly significant positive predictive effect on saccadic duration (\u003cem\u003e\u0026beta;\u003c/em\u003e = 0.645, \u003cem\u003ep \u003c/em\u003e\u0026lt; 0.001). In contrast, the predictive effect in the MS group did not reach statistical significance (\u003cem\u003ep \u003c/em\u003e= 0.062). These findings suggest that a more typical holistic process strategy was associated with longer saccadic duration. This effect was particularly strong and significant in the LS group, whereas the trend in the MS group was relatively weaker and statistically insignificant.\u003c/p\u003e\n\u003cp\u003e3.5.2 Regression analysis between cognitive strategies and pupil size\u003c/p\u003e\n\u003cp\u003eConsidering the skewed distribution of pupil size data, this study employed Mann-Whitney U tests to compare differences in pupil diameter between the two cognitive strategies within the LS and MS groups separately(Fig.10).In the MS group, the pupil diameter of participants who applied the holistic process strategy (Median=2052.74) was larger than that of the participants who applied the chunk-based process strategy((Median=1729.06). However, the Mann-Whitney U test indicated that this difference was not statistically significant (\u003cem\u003ep \u003c/em\u003e= 0.139). In the LS group, the difference in pupil diameter between the two visual processing styles was smaller (holistic process strategy : (Median = 1841.81; chunk-based process strategy: (Median = 1807.76), and no significant difference was found by statistical testing (U = 65.000, \u003cem\u003ep \u003c/em\u003e= 0.861).\u003c/p\u003e"},{"header":"4 Discussion","content":"\u003cp\u003e4.1 Cognitive strategies and their characteristics under parallel multitasks\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOur study employed an eye-tracking-based Hidden Markov Modeling approach to investigate the cognitive strategies in response to simulated ICU nursing parallel multitask characterized with nurse-machine interaction. Participants primarily employed two distinct strategies: the holistic process strategy and the chunk-based process strategy. These strategies align with the classical cognitive frameworks of constructive matching and response elimination, respectively (Federico \u0026amp; Montague, 1980; Zhang et al., 2012). The constructive matching strategy is characterized by participants deducing the rules that govern the graphics’ arrangement, using these rules to mentally construct a response, and subsequently searching for the matching option.\u0026nbsp;In our study, the holistic process strategy aligns with this approach: participants treated the machine alert interface and each patient’s diagnosis as an integrated whole, retrieved information from long-term working memory, analyzed and compared the three patient situations, and finally entered the caring priority (Fig.3). Conversely, the response elimination strategy entails participants sequentially eliminating unreliable alternatives to arrive at the correct answer. The chunk-based process strategy corresponds to this approach: participants first processed the three machine alert interfaces to form an initial answer, then analyzed the patients’ diagnoses to eliminate incorrect options before determining the final answer (Fig.4).Additionally, the research results showed that all participants remained the same strategy in both group tasks, indicating that none of them adjusted their cognitive strategies in response to task characteristics, highlighting a limitation in cognitive flexibility that merits further attention.\u003c/p\u003e\n\u003cp\u003e4.2 Attention allocations of different cognitive strategy under parallel multitasks\u003c/p\u003e\n\u003cp\u003eThe study revealed that nurses employing the holistic process strategy showed a strong tendency to focus on the center patient, with attention distributed almost equally across all patients (Fig.3 \u0026amp; Fig.5). In contrast, participants using the chunk-based process strategy tended to first compare alarm information across all cases before shifting their attention to diagnostic details. The initial fixation was highly concentrated, with 0.91 probability directed toward ROI 1 (machine alert interface), indicating a strong initial prioritization of machine information during encoding. Research findings also indicated that nurses tend to make clinical decision under parallel multitask without careful comparing between the last two ROIs (Fig.5).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4.3 Task performance of different cognitive strategies under parallel multitasks\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn our study, task performance under different cognitive strategies was evaluated using two outcome indicators: task accuracy and response time. For trials involving more severe diseases, nurses adopting the holistic process strategy demonstrated higher task accuracy compared to those using the chunk-based process strategy (51.11% vs. 44.64%), though with longer response times (14.61s vs. 13.16s). Similarly, in trials with less severe diseases, the holistic process strategy also led to higher accuracy (40.63% vs. 36.67%), but again with longer response times (15.62s vs. 13.37s). These findings align with established cognitive theories and prior research. Thus, nurses working in high-risk clinical settings should weigh patient needs when deciding which cognitive strategy to employ. When caring for critically ill patients, where accuracy outweighs speed, it is advisable to adopt a holistic process strategy to enhance the synthesis of multifaceted information and support accurate clinical judgment. However, since sustained use of the holistic process strategy may be cognitively demanding for nurses, a chunk-based strategy is recommended for cases of lower severity, as it helps conserve time and reduce cognitive load.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4.4 Cognitive workload of different cognitive strategies under parallel multitasks\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGenerally, the constructive matching strategy imposes a higher cognitive workload than the response elimination strategy and therefore influence their task performance (Wang \u0026amp; Zhan, 2025). The positive association between holistic process and saccade duration—particularly strong in the MS group—indicates that this strategy demands greater cognitive effort, potentially reflecting deeper information integration or more effortful comparison processes (Solomon et al., 2011). Although pupil size differences between strategies did not reach statistical significance, the trend toward larger pupils in the participants who applied holistic process strategy aligns with the interpretation that this strategy is cognitively demanding. But the increased cognitive workload may be justifiable in high-severity contexts where accurate prioritization is critical but less efficient in lower-severity scenarios.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4.5 Implications for research and practice\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study represents a pioneering effort to investigate the cognitive strategies employed by nurses when performing parallel multitask in high-risk hospital departments. The experimental stimulus was developed based on real clinical cases and incorporated the most common medical devices found in ICU, thereby ensuring a high ecological validity. Using an eye-tracking-based Hidden Markov Modeling approach, this research identified two cognitive strategies adopted by nurses during parallel multitasking therefore revealed how nurses interact with multiple medical machines instantaneously. Besides we evaluated and compared the task performance, reaction time and cognitive workload associated with each strategy. Based on these findings, first we should develop medical devices with adaptive interfaces that can adjust information layout based on nurse-machine's interaction patterns and patients' clinical conditions. Second, nurses should weigh the pros and cons of different cognitive strategy and made proper clinical decisions under parallel multitask. The holistic process strategy should be applied when caring for critically ill patients to enhance patient safety, whereas the chunk-based process strategy is more suitable for less severe cases to improve efficiency, reduce cognitive load, and prevent exhaustion. Furthermore, we emphasized the importance of cognitive flexibility and recommend the development of school or on-the-job training programs to enhance this ability among nurses.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4.6 Research strength and limitations\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;This study utilized eye movement data and a Hidden Markov Model to investigate the unobservant cognitive processes employed by nurses during parallel multitask characterized with multiple nurse-machine interaction, offering an objective and real-time depiction of cognitive dynamics with high accuracy. However, several limitations should be acknowledged. First, although the experiment was conducted in a sound-attenuated laboratory at the Institute of Psychology, Chinese Academy of Sciences—which helped minimize interference from extraneous variables—the controlled setting may reduce simulation fidelity compared to real clinical environments, thereby limiting the generalizability of the findings. Second, the modest sample size may have constrained the statistical power to detect subtle effects. Third, the majority of participants were novice nurses with fewer than five years of experience, meaning their cognitive strategies may differ from those of more experienced nurses. Future studies should include a more diverse range of participants with varying levels of experience and consider multi-center collaborations. Additionally, immersive virtual simulation presents a promising avenue in cognitive flexibility training for nursing students or on-job nurses.\u003c/p\u003e"},{"header":"5 Conclusion","content":"\u003cp\u003eWith the influx of various medical devices and persistent global shortage in nursing, making timely and accurate decision under multitask scenarios become an essential ability for nurses. And cognitive strategy plays a significant role in ensuring quality of decision-making. Our research identified two cognitive strategies nurses applied when managing multitasking conflicts. The holistic process strategy enhances decision accuracy in high-severity scenarios but may incur higher cognitive costs, whereas the chunk-based process strategy may suffice for lower-complexity tasks. These findings not only advance our theoretical understanding of nursing cognition under parallel multitask, but also provide an evidence-based foundation for developing targeted training and interface design solutions aimed at supporting optimal clinical decision-making and improving patient safety outcomes.\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eX.W.: conceptualization, data curation, formal analysis, funding acquisition, project administration, resources, supervision, writing \u0026ndash; review \u0026amp; editing.T.T. F: conceptualization, data curation, formal analysis, investigation, project administration, supervision, writing \u0026ndash; original draft, writing \u0026ndash; review \u0026amp; editing.J.Q. C:data curation, formal analysis, methodology, writing \u0026ndash; original draft, writing \u0026ndash; review \u0026amp; editing. S.F. W:data curation, formal analysis, writing \u0026ndash; original draft.H.W:data curation, formal analysis, writing \u0026ndash; original draft.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eFirst and foremost, we would like to express our sincere gratitude to all the participants for their valuable contributions to this study. We are deeply grateful to Mr. Wei Chu-guang and Ms. Yuan Yi-ran, senior laboratory technicians at the Institute of Psychology, Chinese Academy of Sciences, for their expert assistance in the design and implementation of the experiments.We also extend our gratitude to Ms. Yuan Cui, Deputy Director of the Nursing Department at Peking University First Hospital; Ms. Ma Li, Head Nurse of the Emergency Department at Peking University Third Hospital; and Mr. Liu Feng-gang, Head Nurse of the ICU at the First Affiliated Hospital of University of South China, for their insightful review of the experimental stimuli.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAbu Arra AY, Ayed A, Toqan D, Albashtawy M, Salameh B, Sarhan AL, Batran A (2023) The Factors Influencing Nurses\u0026rsquo; Clinical Decision-Making in Emergency Department. \u003cem\u003eINQUIRY\u003c/em\u003e, \u003cem\u003e60\u003c/em\u003e, 469580231152080. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/00469580231152080\u003c/span\u003e\u003cspan address=\"10.1177/00469580231152080\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCameron LD, Jago L (2013) Cognitive Strategies. In: Gellman MD, Turner JR (eds) Encyclopedia of Behavioral Medicine. Springer, New York, NY. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-1-4419-1005-9_1120\u003c/span\u003e\u003cspan address=\"10.1007/978-1-4419-1005-9_1120\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCoviello E, Chan AB, Lanckriet GRG (2014) Clustering hidden Markov models with variational HEM. J Mach Learn Res 15(1):697\u0026ndash;747\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCowan N (2001) The magical number 4 in short-term memory: a reconsideration of mental storage capacity. Behav Brain Sci 24(1):87\u0026ndash;185. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/s0140525x01003922\u003c/span\u003e\u003cspan address=\"10.1017/s0140525x01003922\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFederico PA, Montague WE (1980) \u003cem\u003eAptitute, Learning, and Instruction: Cognitive process analysis of aptitude\u003c/em\u003e.Routledge\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrewal J, Krzywinski M, Altman N (2019) Markov models - hidden Markov models. Nat Methods 9(16):795\u0026ndash;796. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41592-019-0532-6\u003c/span\u003e\u003cspan address=\"10.1038/s41592-019-0532-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGriffin JW, Webb SJ, Keehn B, Dawson G, McPartland JC (2024) Autistic Individuals Do Not Alter Visual Processing Strategy During Encoding Versus Recognition of Faces: A Hidden Markov Modeling Approach. \u003cem\u003eJ Autism Dev Disord\u003c/em\u003e. Advance online publication.\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10803-024-06259-9\u003c/span\u003e\u003cspan address=\"10.1007/s10803-024-06259-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJarosz A, Raden M, Wiley J (2019) Working memory capacity and strategy use on the RAPM. \u003cem\u003eIntelligence\u003c/em\u003e, (77), 101387. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.intell.2019.101387\u003c/span\u003e\u003cspan address=\"10.1016/j.intell.2019.101387\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJastrzębski J, Ciechanowska I, Chuderski A (2018) The strong link between fluid intelligence and working memory cannot be explained away by strategy use. Intelligence 66:44\u0026ndash;53. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.intell.2017.11.002\u003c/span\u003e\u003cspan address=\"10.1016/j.intell.2017.11.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim Y, Ryu GW, Lee MJ, Choi M, Cho E (2022) Multitasking of Nursing Activities: A Time to Motion Study to Capture Nursing Workflow Using Eye Tracker. SSRN Electron J. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2139/ssrn.4027254\u003c/span\u003e\u003cspan address=\"10.2139/ssrn.4027254\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKuha J (2004) AIC and BIC: Comparisons of Assumptions and Performance. Sociol Methods Res 33(2):188\u0026ndash;229. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/0049124103262065\u003c/span\u003e\u003cspan address=\"10.1177/0049124103262065\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLaurence PG, Jana TA, Bunge SA, Macedo EC (2023) Eye Gaze Patterns during Reasoning Provide Insights Regarding Individual Differences in Underlying Cognitive Abilities. J Intell 11(4):75. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/jintelligence11040075\u003c/span\u003e\u003cspan address=\"10.3390/jintelligence11040075\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi JX (2025) \u003cem\u003eToward Enhancing Multitasking Performance: Modeling, Prediction, and Intervention Design\u003c/em\u003e.[Doctoral dissertation, University of Washington].University Libraries,University of Washington\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu JC, Li KA, Yeh SL, Chien SY (2022) Assessing Perceptual Load and Cognitive Load by Fixation-Related Information of Eye Movements. Sens (Basel) 22(3):1187. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/s22031187\u003c/span\u003e\u003cspan address=\"10.3390/s22031187\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu K, Wang A (2012) Does media multitaskinging always hurt? A positive correlation between multitaskinging and multisensory integration. Psychon Bull Rev 19:647\u0026ndash;653. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3758/s13423-012-0245-7\u003c/span\u003e\u003cspan address=\"10.3758/s13423-012-0245-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLu Y, Wang J, Dai W, Zou B, Lu X, Hong W (2025) Research on risk factors of human stampede in mass gathering activities considering organizational patterns. Front Public Health 13:1555735. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fpubh.2025.1555735\u003c/span\u003e\u003cspan address=\"10.3389/fpubh.2025.1555735\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMathew M, John A, Vazhakkalayil Ramachandran R (2025) Nurse stress and patient safety in the ICU: physician-led observational mixed-methods study. BMJ Open Qual 14(2):e003109. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1136/bmjoq-2024-003109\u003c/span\u003e\u003cspan address=\"10.1136/bmjoq-2024-003109\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOladele IO, Okoro CJ, Falana SO, Olajesu FO, Adelani SO, Onuh LN (2024) Advancement of human-machine collaboration in manufacturing: A review on industry 1.0\u0026ndash;6.0. J Mech Eng Sci 18(4):10303\u0026ndash;10329. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.15282/jmes.18.4.2024.7.0813\u003c/span\u003e\u003cspan address=\"10.15282/jmes.18.4.2024.7.0813\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOlin K, G\u0026ouml;ras C, Nilsson U, Unbeck M, Ehrenberg A, Pukk-H\u0026auml;renstam K, Ekstedt M (2022) Mapping registered nurse anaesthetists' intraoperative work: tasks, multitasking, interruptions and their causes, and interactions: a prospective observational study. BMJ Open 12(1):e052283. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1136/bmjopen-2021-052283\u003c/span\u003e\u003cspan address=\"10.1136/bmjopen-2021-052283\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSnow RE, Federico P, Montague WE (1980) Aptitute, Learning, and Instruction. Cognitive process analysis of aptitude. Routledge\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSolomon D, Albert N, Sun Z, Bowers AM, Molnar M (2011) Complexity of Care Is Associated with Distressing Environmental Factors. Clin Nurse Specialist 25(4):186\u0026ndash;192. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1097/nur.0b013e318221f2d3\u003c/span\u003e\u003cspan address=\"10.1097/nur.0b013e318221f2d3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStasch SM, Mack W Exploring Task Prioritization in VR Flight Environments: Can Eye-Tracking Uncover Cognitive Control? In Proceedings of the 2025 Symposium on Eye Tracking Research and Applications (ETRA '25). Association for Computing Machinery, New York, NY, USA, Article (2025) 76, 1\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/3715669.3725898\u003c/span\u003e\u003cspan address=\"10.1145/3715669.3725898\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSweller J (1988) Cognitive Load during Problem Solving: Effects on Learning. Cognit Sci 2(12):257\u0026ndash;285. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/0364-0213(88)90023-7\u003c/span\u003e\u003cspan address=\"10.1016/0364-0213(88)90023-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWagenmakers E-J, Farrell S (2004) AIC model selection using Akaike weights. Psychon Bull Rev 11(1):192\u0026ndash;196. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3758/BF03206482\u003c/span\u003e\u003cspan address=\"10.3758/BF03206482\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Z, Zhan P (2025) Eye-tracking-based hidden Markov modeling for revealing within-item cognitive strategy switching. Behav Res 175(57). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003eN/A. https://doi.org/10.3758/s13428-025-02678-3\u003c/span\u003e\u003cspan address=\"N/A. 10.3758/s13428-025-02678-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuan X, Zhong L (2024) Effects of multitaskinging and task interruptions on task performance and cognitive load: considering the moderating role of individual resilience. Curr Psychol 43(28):23892\u0026ndash;23902. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s12144-024-06094-2\u003c/span\u003e\u003cspan address=\"10.1007/s12144-024-06094-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang P, Jiang Y, He S (2012) Voluntary Attention Modulates Processing of Eye-Specific Visual Information. Psychol Sci 23(3):254\u0026ndash;260. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/0956797611424289\u003c/span\u003e\u003cspan address=\"10.1177/0956797611424289\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"cognition-technology-and-work","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ctwo","sideBox":"Learn more about [Cognition, Technology \u0026 Work](http://link.springer.com/journal/10111)","snPcode":"10111","submissionUrl":"https://submission.nature.com/new-submission/10111/3","title":"Cognition, Technology \u0026 Work","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"cognitive strategy, parallel multitasking, task performance, cognitive workload, human-machine interaction","lastPublishedDoi":"10.21203/rs.3.rs-7975457/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7975457/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eWith the influx of medical devices, parallel multitasking rapidly grows in clinic and may contribute to increasing human error and higher risks. Since it is impossible to avoid parallel multitasking in clinic, it is imperative to explore nurses\u0026rsquo; cognitive strategy under this complex situation to ensure nursing quality and patient safety. However, there is a little research towards cognitive strategies of parallel multitasking in nursing. Thus, our study aimed at investigating nurses\u0026rsquo; cognitive strategies in decision-making under parallel multitasking through observational research.30 eligible nurses completed the trials programmed by PsychoPy 2021.2.3 in the lab. Participants\u0026rsquo; behavioral- and eye- movement data were recorded by PsychoPy and EyeLink 1000 respectively. Hidden Markov Modeling was applied to analyze the spatio-temporal dynamics of participants\u0026rsquo; eye movements and uncovered the cognitive strategies. Results show that participants adopted two cognitive strategies under parallel multitasking: holistic- and chunk- cognitive strategy. The holistic cognitive strategy yielded a longer reaction time and heavier cognitive workload, but the differences in task accuracy between two cognitive strategies were insignificant. Besides no strategy switching and little transition between the last two ROIs was observed in both cognitive strategies, indicating low cognitive flexibility and strong tendency to take cognitive short-cuts. In the future more researches should be done to explore cognitive strategies in diverse nursing groups and provide suggestions to nurse education and management, therefore to better equip nurses in newly working environment.\u003c/p\u003e","manuscriptTitle":"Uncovering Cognitive Strategies for ICU Parallel Nursing Multitasks: An Eye-tacking-based Hidden Markov Modeling Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-20 11:42:12","doi":"10.21203/rs.3.rs-7975457/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-12-31T15:20:47+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-29T10:30:48+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-10-29T10:25:25+00:00","index":"","fulltext":""},{"type":"submitted","content":"Cognition, Technology \u0026 Work","date":"2025-10-29T04:22:01+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"cognition-technology-and-work","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ctwo","sideBox":"Learn more about [Cognition, Technology \u0026 Work](http://link.springer.com/journal/10111)","snPcode":"10111","submissionUrl":"https://submission.nature.com/new-submission/10111/3","title":"Cognition, Technology \u0026 Work","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"3af0e9dc-cb6d-442d-a789-3470fb8ce2ad","owner":[],"postedDate":"January 20th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-01-20T11:42:13+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-20 11:42:12","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7975457","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7975457","identity":"rs-7975457","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}