Shift Work Intern Nurses Exhibit Four Distinct Sleep Inertia Symptom Networks

Intern nurses who undertake shift work experience sleep inertia across four distinct clinical profiles, with each profile characterized by a unique central symptom ranging from delayed physical reaction to slowed cognitive processing, according to a cross-sectional investigation published in BMC Psychiatry. Sleep inertia, defined as the transitional physiological state between sleeping and waking, is typically marked by transient drowsiness, reduced vigilance, and temporary impairment of cognitive and executive performance. In hospital environments that demand continuous twenty-four-seven operation, shift workers frequently face abrupt awakenings to respond to critical medical emergencies, where degraded execution ability can lead to occupational hazards and clinical safety incidents. Prior surveys have noted that approximately twenty-five percent of shift workers intentionally choose against taking naps during night shifts specifically to avoid the debilitating disorientation brought on by sleep inertia. By employing latent profile analysis alongside network analysis, the research team sought to map the complex interconnections among individual symptoms of sleep inertia and clarify how these symptoms interact across different levels of severity in early-career healthcare professionals.

Understanding individual differences in sleep inertia has historically posed an empirical challenge, as underlying physiological and psychological mechanisms remain poorly delineated despite established observations that people experience post-awakening grogginess with varying intensity. Previous laboratory investigations involving healthy young adults documented substantial interindividual variability in the magnitude of sleep inertia, while earlier self-report surveys among college students identified four discrete subgroups of severity. To advance beyond basic categorical classification, the authors adopted a symptom-network psychopathology framework, which conceptualizes psychiatric and psychological syndromes as emergent systems arising from direct mutual interactions between observable symptoms rather than unobserved latent diseases. Within this network modeling approach, individual symptoms are represented as nodes, while the pairwise correlational dependencies between them are represented as edges. By analyzing mathematical centrality indices such as node strength and expected influence, researchers can identify the central symptoms that exert the greatest cumulative activation across the broader symptom network, pinpointing potential leverage points for targeted clinical interventions.

The study cohort comprised 1,442 intern nurses engaged in shift work, recruited through convenience sampling from a nursing college located in Huizhou, China, in 2022. The research protocol commenced with an initial recruitment drive conducted half a month after the students started their clinical internships, reaching an initial sample of 1,636 intern nurses. Among those invited, 1,540 individuals successfully completed the administered electronic questionnaires, representing an effective response rate of 94.13 percent. From this responding cohort, ninety-eight participants who worked exclusively during conventional daytime hours were systematically excluded, yielding a final analytical sample of 1,442 shift-working intern nurses. Teachers assisted in distributing electronic questionnaires, and all participants or their guardians provided signed electronic informed consent prior to completing self-administered surveys on mobile phones or personal computers. The research design was formally reviewed and approved by the Human Research Ethics Committee of South China Normal University and adhered strictly to the ethical standards delineated in the 1989 Declaration of Helsinki.

Participant symptoms of sleep inertia were evaluated using the validated Chinese version of the Sleep Inertia Questionnaire, a comprehensive twenty-one-item psychometric instrument that assesses specific manifestations such as difficulty getting up, cognitive slowness, and emotional tension. Each item on the questionnaire is scored using a five-point Likert scale extending from one, representing not at all, to five, representing all the time, generating a total composite score between 21 and 105, with higher aggregate values reflecting greater overall severity of sleep inertia. The questionnaire has previously shown robust validity and reliability across general community samples, depressive populations, and Chinese university cohorts, where internal consistency reached a Cronbach’s alpha of 0.95. In the current investigation of shift-working intern nurses, the instrument exhibited excellent internal consistency, yielding a Cronbach’s alpha of 0.954. All data were maintained under strict confidentiality, and participating interns retained unconditional freedom to withdraw from the investigation at any point during questionnaire completion.

To rigorously identify heterogeneous symptom profiles, the investigators executed latent profile analysis within Mplus version 8.3, progressively evaluating models ranging from one to five latent classes against empirical fit criteria. Model selection incorporated multiple standard fit metrics, including the Akaike Information Criterion, Bayesian Information Criterion, sample-size adjusted Bayesian Information Criterion, entropy classification accuracy, the Lo-Mendell-Rubin likelihood ratio test, and the bootstrap likelihood ratio test. Although the five-class model yielded marginally lower information criteria and significant bootstrap likelihood tests, its Lo-Mendell-Rubin test produced an insignificant p-value exceeding 0.05, indicating that five classes did not significantly improve upon four. Conversely, the four-profile model demonstrated substantially lower information criteria than one-, two-, or three-class solutions while maintaining high classification entropy superior to the five-class alternative. Consequently, the four-profile model was selected as the optimal representation of sleep inertia heterogeneity among shift-working intern nurses.

Classification across the 1,442 participants assigned 21.29 percent (n = 307) of the intern nurses to the Low sleep inertia profile, 32.32 percent (n = 466) to the Mild sleep inertia profile, 34.60 percent (n = 499) to the Moderate sleep inertia profile, and 11.79 percent (n = 170) to the Severe sleep inertia profile. This empirical distribution demonstrated considerable structural stability when contrasted with earlier findings in general college student populations, which had similarly identified four discrete subgroups. However, the proportion of shift-working intern nurses classified into the Severe profile was slightly lower than the sixteen percent previously observed among general undergraduates, a divergence that the authors hypothesized may stem from occupational adaptive practices such as scheduled prophylactic napping, planned caffeine consumption, and strategic exposure to bright ambient light during active shifts. Notably, across both student and healthcare cohorts, the combined proportion of participants experiencing moderate-to-severe sleep inertia approached fifty percent, underscoring the widespread prevalence of pronounced post-awakening grogginess among shift workers.

Following profile determination, network estimation was executed in R version 4.3.2 utilizing the Extended Bayesian Information Criterion graphical least absolute shrinkage and selection operator, paired with Spearman rank correlations through the qgraph package. Graph visualization applied the Fruchterman-Reingold layout algorithm, with identical symptom nodes anchored across networks using the averageLayout function to allow direct visual comparisons. Symptom importance was evaluated primarily through expected influence, alongside node strength, to capture the cumulative directional impact of connections in networks containing both positive and negative associations. In the Low sleep inertia network, item eighteen, reacting more slowly, emerged as the most central symptom. In contrast, item twelve, feeling tense, functioned as the primary central symptom in the Mild sleep inertia network. For the Moderate sleep inertia group, item six, feeling sleepy, served as the dominant central symptom, while item seventeen, thinking more slowly, demonstrated the highest centrality in the Severe sleep inertia network.

The structural topology of the four networks displayed notable variations in density and connectivity across the theoretical maximum of 210 possible edges connecting the twenty-one questionnaire items. Empirical estimation identified 57 non-zero edges in the Low sleep inertia network, all of which were positive connections with no negative associations retained. The Mild sleep inertia network exhibited 73 non-zero edges, comprising 49 positive and 24 negative edges, while the Moderate network contained 80 non-zero edges, consisting of 55 positive and 25 negative edges. The Severe sleep inertia network contained 71 non-zero edges, all of which were positive. The strongest single connection within the Low network occurred between item ten, getting winded more easily, and item twelve, feeling tense, with an edge weight of 0.474. In the Mild network, the strongest coupling emerged between item thirteen, feeling anxious about the upcoming day, and item fourteen, dreading starting the day, displaying an edge weight of 0.546. Within both the Moderate and Severe networks, the strongest positive association linked item seventeen, thinking more slowly, with item eighteen, reacting more slowly, showing edge weights of 0.544 and 0.637, respectively.

A critical topological distinction identified across the profiles was the retention of negative edge weights exclusively within the Mild and Moderate networks, contrasted with their complete shrinkage to zero in the Low and Severe networks. The authors suggested that negative edges in mild and moderate presentations may represent active compensatory or inhibitory regulatory mechanisms, wherein certain awakening symptoms actively dampen or suppress the activation of others during the physiological transition to wakefulness. Conversely, the uniform presence of purely positive associations in both the Low and Severe networks reflects mutually reinforcing co-activation cycles among symptoms. In the severe presentation, this pervasive positive reinforcement may accelerate symptom cascades into an escalating negative spiral, whereas in low-severity individuals, positive interconnections may reflect cohesive, non-pathological physiological synchronization, often characterized as a positive spiral among healthy individuals experiencing minimal functional impairment.

Statistical comparison of network structures via the NetworkComparisonTest package revealed significant differences in global connectivity that conformed to a distinct U-shaped trajectory across severity profiles. Using the Low sleep inertia network as a baseline reference, global strength values measured 6.215 for Low, dropped to 5.000 for Mild, rose to 7.546 for Moderate, and peaked at 7.605 for Severe sleep inertia. Permutation testing confirmed that the global strength of the Low network differed significantly from the Mild network (S = 1.215, p < 0.001), the Moderate network (S = 1.332, p < 0.001), and the Severe network (S = 1.391, p < 0.01). However, permutation tests evaluating network structure invariance indicated no statistically significant differences in the overall distribution of edge weights across profiles (Low versus Mild: M = 0.234, p = 0.210; Low versus Moderate: M = 0.259, p = 0.076; Low versus Severe: M = 0.318, p = 0.282). Furthermore, formal tests of individual edge invariance revealed that no individual edge weights differed significantly between networks after applying Holm-Bonferroni corrections for multiple comparisons.

The authors contextualized these central symptoms within established physiological and occupational frameworks. The prominence of bodily tension in mild presentations likely reflects musculoskeletal strain and soreness across the neck, shoulders, and back resulting from prolonged shift hours, physical patient handling, and constrained rest periods, exacerbated by fragmented sleep and frequent transitions between day and night duties. For the moderate group, the primacy of sleepiness aligns with neurophysiological evidence showing that awakening from deep sleep involves a transitional shift from high-amplitude low-frequency delta activity toward low-amplitude high-frequency wake rhythms, a process hindered when chronic sleep debt delays cortical activation. In severe cases, slowed thinking reflects documented metabolic dynamics during awakening, wherein cerebral blood flow initially remains lower than standard waking baselines and functional reactivation from subcortical structures to the prefrontal cortex requires measurable time, producing acute cognitive deficits that directly hinder rapid clinical decision-making during medical emergencies.

The authors underscored several methodological limitations that temper the immediate generalizability of their conclusions. First, the data relied entirely on subjective self-report questionnaires, leaving measurements vulnerable to reporting bias and recall inaccuracies that warrant replication through objective clinical assessments. Second, the cross-sectional survey design captured symptom networks at a single observational point, precluding definitive conclusions regarding directional or causal relationships among sleep inertia symptoms. Third, because the investigation sampled intern nurses from a single educational institution in Huizhou, the findings cannot be generalized directly to clinical patient cohorts, such as individuals diagnosed with idiopathic hypersomnia. To overcome these constraints, the researchers advocated for future prospective studies employing ecological momentary assessments and ambulatory actigraphy to capture dynamic, real-time symptom fluctuations across broader shift-working populations, such as attending physicians and disaster emergency responders, to rigorously evaluate targeted interventions designed to alleviate post-awakening cognitive impairment.

Subject of Research: Heterogeneous symptom network structures of sleep inertia among shift-working intern nurses

Article Title: Differences in network structures for distinct profiles of sleep inertia among intern nurses with shift work

Article References: Yu, Z., Wang, D., Fan, Y., Wu, W., Chen, H., Wang, X., Li, Y., Ma, Z., & Fan, F. (2026). Differences in network structures for distinct profiles of sleep inertia among intern nurses with shift work. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08879-0

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08879-0

Keywords: sleep inertia, shift work, intern nurses, network analysis, latent profile analysis, sleepiness, cognitive performance, occupational health, symptom networks, EBICglasso, expected influence, node centrality, psychometrics, Sleep Inertia Questionnaire, Huizhou China, musculoskeletal tension, cerebral blood flow, clinical safety