Trends and Emerging Research Areas in Postoperative Sleep Disturbances: A Bibliometric Analysis

Introduction

Sleep is a complex, naturally occurring physiological state essential for human health and survival. Sleep disorders are a variety of dysfunctions and disturbances that manifest during the sleep-wake process. According to the International Classification of Sleep Disorders-third edition (ICSD-3), sleep disorders are grouped into seven major categories: insomnia, sleep-related breathing disorders, central disorders of hypersomnolence, circadian rhythm sleep-wake disorders, parasomnias, sleep-related movement disorders, and other sleep disorders.¹ With disruptions in living environment, social rhythms, light patterns, and changes in biological factors, sleep disorders have become an essential public health problem and have attracted increasing interest over the past decades. It is estimated that up to 50% of primary care patients experience insomnia.² Due to personal and perioperative factors, sleep disturbances are more common among surgical patients.³

Sleep is controlled by specific brain regions and corresponding neurons, which are mainly regulated by endogenous sleep-wake cycle and homeostatic processes.⁴ Based on polysomnographic monitoring, normal sleep consists of non-rapid eye movement (NREM) and rapid eye movement (REM) sleep; NREM sleep can be subdivided into stages N1, N2, and N3 according to the depth of sleep.^5,6 Immediately after surgery (especially major surgery), sleep structure is characterized by sleep deprivation, sleep fragmentation, and loss of stage N3 (slow-wave sleep) and REM sleep.⁷ Patients usually report decreased sleep time, light sleep, increased awakenings, and frequent nightmares.⁸ Postoperative Sleep Disturbances (PSD) significantly impact surgical outcomes by prolonging hospital stays, delaying recovery, increasing the risk of complications such as delirium,³ delayed functional recovery,⁹ more chronic postsurgical pain,¹⁰ increased adverse cardiac events,¹¹ and worsened quality of life.¹² Individuals undergoing major surgeries, those with longer surgery durations,¹³ preoperative anxiety,¹⁴ postoperative pain,¹⁵ or older adults are at higher risk.¹⁵ Medications, particularly opioids and anesthetics,¹⁵ could worsen PSD by disrupting sleep architecture and increasing sleep latency. The underlying biological mechanisms involve changes in neurotransmitters and immunological processes. Alterations in neurotransmitters, such as decreased levels of serotonin and norepinephrine,¹⁶ contribute to disrupted sleep. Additionally, surgical trauma and sleep deprivation could trigger the release of pro-inflammatory cytokines and reduce melatonin secretion, further impairing sleep quality.¹⁷ These mechanisms highlight the need for targeted interventions to improve postoperative sleep and overall patient outcomes. Over the past 20 years, PSD has attracted increasing attention, resulting in significant progress in the understanding of its risk factors, underlying mechanisms, and preventive strategies.

Electronic documents are indispensable to scientific researchers. However, the massive amount of available literature can be overwhelming. Extracting relevant information without effective reading and analysis strategies is a significant challenge. Bibliometric analysis, a method that integrates statistical techniques and data visualization, has gained prominence in evaluating international collaborations, citation analysis, research trends, and focal areas within a particular area.^18,19 This scientific bibliometric approach is an effective way to address this challenge.

Despite previous reviews on PSD from different perspectives,^3,10 significant gaps remain in the visualization, analysis, and trends in research priorities in the field. Therefore, the main objective of this study was to provide a bibliometric and visual analysis of a comprehensive overview of publications, trends, emerging research areas, and potential future directions in PSD, providing valuable insights for future research in this field.

Materials and Methods

Data Sources and Search Strategies

This study was approved by the Ethics Committee of the Fourth Hospital of Hebei Medical University (2023KS121). The requirement for informed consent was waived. The raw data were extracted from the Web of Science Core Collection (WoSCC) on October 24, 2023. The search keyword query as follows: TS= (“sleep disturbance*” OR “sleep disorder*” OR “sleep architecture” OR “sleep quality” OR “sleep structure” OR “sleep wake” OR “non-rapid eye movement” OR “rapid eye movement”) AND TS= (postoperative OR after operation OR post-surgery OR postoperative period). To account for the database’s ongoing updates, two reviewers conducted separate data searches on the same day. This data was subsequently verified independently for retrieval and deliberated upon to reach a final agreement of 0.90.²⁰

Data Collection and Cleaning

The raw data were cleaned and de-duplicated with CiteSpace (Version 6.3.R1). A thesaurus file was used to correct erroneous elements before the bibliometric analysis with VOSviewer (Version 1.6.19). Finally, the cleaned data were imported into R software (Version 4.3.1), VOSviewer, and CiteSpace for bibliometric analysis.

Bibliometric Analysis

We conducted the analysis using bibliometrix R package, VOSviewer, and CiteSpace. The Bibliometrix R package is an open-source tool for performing comprehensive bibliometric analyses, which was used to analyze the overview, annual, and cumulative numbers of publications, growth trends, countries/regions, institutions, authors, and source journals. Given the nonlinear relationship between the year and the cumulative number of publications, we constructed a polynomial model to capture the growth trends. Since it was clearly cyclical over time, time series were introduced into the model to predict future developments. The rationality and validity of the model were verified using residual analysis and goodness-of-fit tests. Bibliometric networks were constructed and visualized using VOSviewer and CiteSpace. VOSviewer is a software tool used to construct and visualize bibliometric networks. It was used to construct visualization analyses of countries/regions, source journals, and co-occurring keywords; to map scientific networks; and to display knowledge structures, collaborations, and the evolution of relationships. CiteSpace is citation visualization and analysis software developed by Prof. Chaomei Chen. Co-citation references were analyzed to evaluate the weight or influence of each reference through centrality. Different clusters of co-citation references were analyzed using the LLR algorithm. Centrality refers to the frequency with which a node acts as a link between two other nodes. A high centrality indicates greater weight or influence. Then, VOSviewer and CiteSpace were used to analyze the co-occurrence keywords, burst keywords, and co-citation reference networks.

Potential Biases Across Different Time Periods

To ensure the robustness of our analysis, we conducted a detailed assessment of potential biases that could influence the results. Publication bias occurs when studies with positive or significant results are more likely to be published than those with negative or inconclusive results are. To mitigate this bias, we used the WoSCC database, which is known for its high data quality and completeness. We then included both highly cited and less frequently cited articles to capture a broad spectrum of research. Citation bias can occur if certain studies are over-represented because of their popularity or the influence of their authors. To address this, we employed burst analysis to identify emerging research frontiers, rather than relying solely on citation counts. This approach helps to highlight novel and rapidly evolving research areas that may not yet have accumulated high citation counts. Additionally, to address self-citations and citation circles, which can artificially inflate citation counts. We detected and adjusted for excessive self-citations and citation circles using network analysis, which provides a more reliable assessment of research impact.

Sensitivity Analysis

To comprehensively understand the evolution of PSD research over time, we conducted a sensitivity analysis focusing on the period 2010–2023. This period was selected for the following reasons. Primarily, it captures the most recent trends, including the impact of the COVID-19 pandemic, which has significantly influenced the research directions and priorities. Moreover, this period coincided with a notable increase in PSD-related publications, as identified in our initial bibliometric analysis. The sensitivity analysis allowed us to concentrate on the most active and relevant research themes. Lastly, this period includes the introduction and evaluation of new interventions that have become central to PSD research. Consequently, we set the time period for the sensitivity analysis as 2010–2023. We introduced a time factor to analyze the organization and keyword overlay network mapping by using VOSviewer. Additionally, burst analysis was conducted to identify the most important references. This analysis revealed key nodes that spiked in frequency during specific periods, indicating their high value for researchers at those times. These findings help identify research frontiers and potential future trends.

Results

Overview of Publications, Analysis of Publication Output and Growth Trend Prediction

The search results from the WoSCC database totaled 1,748 publications. The number of publications is commonly used as a measure of productivity, with original articles and reviews serving as key indicators of the current state of research.²¹ Therefore, non-article and non-review publications, as well as irrelevant and non-English documents, were excluded (Figure 1). This resulted in a final dataset of 1,559 publications, including 1,370 articles and 189 reviews, originating from 650 journals and involving 7,438 authors. The first publication in this dataset dates back to 1991, with a steady increase in scientific output thereafter, reaching a peak of 205 publications in 2022. International collaboration accounted for 12.44% of the articles. From 1991 to 2022, scientific interest has progressively increased, with an average annual growth rate of 16.56% in published papers. A polynomial model was then fitted to the cumulative number of publications. The polynomial model was expressed as follow: y = 2.21x² - 35.85x + 148.89 (R² = 0.96) (Figure 2A).

Figure 1 Flowchart of bibliometric analysis. Abbreviation: PSD, Postoperative sleep disturbances.

Figure 2 (A) Annual and cumulative number of publications. Let y represent the cumulative number of publications and x represent the time step. The time step is a sequence of integers starting from one, with each increment corresponding to a new year. (B) Network map of countries/regions (number of publications ≥ 5). Node size represents the number of publications, while the width of the connecting lines indicates the strength of collaboration links. China includes publications from the Chinese Mainland, Hong Kong, Macau, and Taiwan. The United Kingdom includes publications from England, Scotland, Northern Ireland, and Wales. (C) Top 10 authors by impact. As measured by the G index.

Model Validation

Model fit was assessed using residual analysis and goodness-of-fit tests. The plot of residuals versus fitted trends indicates nonlinearity. The residuals were randomly distributed around the zero line, confirming the homoscedasticity assumption. The residuals showed no significant trends over time, indicating independence. The model exhibited strong explanatory power with an R² value of 0.96. The F-test statistic of 480.8 and a p-value of 2.2×10⁻¹⁶ (P < 0.001) confirmed the model’s significance. The fitted polynomial regression curves were plotted along with the raw data to visually validate the fit of the model (Appendix). Overall, the polynomial model effectively captured the trends in cumulative publications over time.

Countries/Regions, Institutions and Authors

Considering that almost 95% of the papers were published post-2000 and earlier works did not meet the criteria for inclusion in the graphical network visualization, the analysis was confined to the period from January 2000 to October 2023. Consequently, 1,487 publications were included in the present study.

From 2000 to 2023, PSD-related papers were mainly published by 1,823 institutions in 61 countries/regions. Figure 2B shows a network of countries/regions with more than five publications. The intensity of international collaboration was the highest between the United States and Canada, as well as between the United States and China. Other countries/regions have been strengthening their collaborations over time.

We analyzed the major institutions with high output in this area. As shown in Table 1, it is among the top ten institutions with the highest global output. The most productive organization was the University of Toronto (n = 72, 4.8%). US institutions occupied seven places and made outstanding contributions to the area.

Table 1 Top 10 Institutions and Authors in PSD Research (01/2000 - 10/2023)

Overall, there were 7,164 authors in all publications. The most productive author was Frances Chung from the University Health Network and the University of Toronto. Table 1 displays the most productive 10 authors. It would be incomplete to assess the researchers’ academic output. Instead, we comprehensively assessed the authors’ impact through the H-index, G-index, and M-index. Essentially, the H-index is a mixed index of publications and citations, measuring the number of publications cited in a particular set of publications.²² These three indicators provide a comprehensive picture of academic productivity and citation quality. The top 10 influential researchers in the area of PSD (mainly referring to the G-index ranking) are shown in Figure 2C.

Source Journals

The academic journal with the most published papers on PSD was the International Journal of Pediatric Otorhinolaryngology, and the most co-cited journal was Anesthesiology. Source journal co-occurrence was categorized into the following clusters based on research direction: anesthesiology disciplinary journals (purple), sleep and circadian rhythm mechanisms (yellow), pain science-related journals (red), neuroscience journals (green) and ENT Head Neck Science and Pediatrics (blue) (Figure 3A).

Figure 3 (A) Analysis of source journal co-occurrence. (B) Author’s keywords network with occurrence ≥ 5. The size of each circle corresponds to the frequency of the keyword’s appearance in the literature. (C) Top 25 keywords with the strongest citation bursts. Citation bursts indicate keywords that have rapidly gained attention.

Co-Occurrence Author’s Keywords and Burst Keywords

Author’s keywords are an important part of an article, which in turn can represent the research topic of a publication. To explore the research hotspots of keywords, co-occurrence analysis was performed by extracting keywords with more than five frequencies of occurrence using VOSviewer and clustering visualization maps. Using co-occurrence clustering analysis, six clusters were identified based on research themes (Figure 3B). These six clusters were represented by different colors: 1) Anesthesia (red): general anesthesia, sleep, propofol, neurons, mechanisms, etc.; 2) Melatonin (green): sleep quality, children, deprivation, polysomnography, stress, etc.; 3) Pain (blue): analgesia, chronic pain, preoperative anxiety, efficacy, inflammation, etc.; 4) Surgery (yellow): postoperative pain, opioid, ropivacaine, dexamethasone, block, etc.; 5) Sleep quality (purple):dexmedetomidine, intensive-care-unit, elderly-patients, critically-ill patients, etc.; 6) Recovery (light blue):sleep disturbance, general anesthesia, patient satisfaction, etc. It can be summarized into six major research trends: sleep mechanisms associated with anesthesia, melatonin and sleep disturbances, pain management, effects of analgesic drugs, dexmedetomidine and sleep quality, postoperative recovery, respectively.

Burst detection is used to identify the emergence or decline of specific keywords. Among the top 25 burst keywords, the top ten with the highest burst intensity were positive airway pressure, anesthesia, surgery, randomized controlled trial, expression, apnea syndrome, preoperative anxiety, efficacy, bariatric surgery, and apnea. These keywords have sustained high attention over multiple years, indicating strong scholarly interest in these research areas. In recent years, emerging keywords hotspots such as “validation”, “general anesthesia”, “neurons” and “functional connectivity” have gained increasing attention (Figure 3C).

Most Cited Reference and Co-Citation Reference Network

Citation frequency is a measure of the approval of the scientific literature. The top 10 cited references in PSD research were listed in Table 2. Co-citation references are research methods that measure the degree of the relationship between references. In other words, it refers to being cited by other publications. The co-citation references were analyzed by clustering to identify similar themes. Using the LLR algorithm, co-citations were grouped into 17 visual clusters. Statistical data showed a well-structured cluster with Q = 0.8174 (modularity), indicating considerable cluster structure. Silhouette (S = 0.9441), indicating that the clusters were highly convincing. Among these clusters, the “preoperative sleep quality” was the most prominent (Figure 4).

Table 2 Top 10 Most Cited References in PSD Research (01/2000 - 10/2023)

Figure 4 Co-citation analysis and visual clustering. The co-citations were visualized into 17 clusters, each representing a thematic or conceptual area within postoperative sleep disturbances research.

Sensitive Analysis

Over-Lay Network

Approximately 90% of the literature has been published after 2010. Accordingly, sensitivity analysis was performed from 2010 to 2023. Then, a time-based dimension was introduced to explore research trends and quantify the change over a given period. Therefore, we identify different institutions’ productivity and keywords that have emerged in different historical periods.

Studies from the University of Copenhagen and Stanford University were conducted around 2014. Institutions such as the University of Toronto and the University of Pennsylvania mainly produced studies around 2016. In recent years, more contributions have been made by the Mayo Clinic, the Cleveland Medical Center, and research institutes from China such as Peking University, China Medical University, Capital University of Medical Sciences, and Fudan University (Figure 5A). Trends in the publication journals from different sources were showed in Figure 5B. This visualization highlights citation dynamics and journal influence across disciplines. The citing papers were predominantly from journals in medicine, clinical medicine, molecular biology, immunology, and neurology. The cited papers were mainly from journals in health, nursing, medicine, molecular biology, psychology, education, and social sciences.

Figure 5 (A) Trends in institutional publication over time. The trends highlight the contributions of different institutions over the study period and reveal the evolving landscape of institutional involvement. (B) Dual-layer citation map of citing and cited journals. Disciplines of citing journals (left) and cited journals (right). Curves represent citation connectors. The vertical axis of the ellipse indicates the number of publications, while the horizontal axis indicates the number of authors. (C) Co-occurring keywords overlay visualization. The overlay feature allows for the identification of key topics and their evolution over time.

In recent years, hot research keywords include dexmedetomidine, neuroinflammation, sleep quality, acupuncture, electroacupuncture, cancer, inflammation, enhanced recovery, and perioperative management. This demonstrated that researchers gradually focused on the deep-rooted mechanisms of PSD, preventive medications, and preventive practices (Figure 5C).

Burst References

CiteSpace was used to analyze the burst references. The top 25 references with the strongest burst over 2010–2023 were presented (Figure 6). Among the top 25 references, the main research areas were sleep-related neuroscience and clinical research on PSD. All references had burst times from 2010 to 2019. In terms of the time period, 10 publications became burst references starting in 2019. Most of these publications were clinical studies or reviews, with the highest burst strength of 10.49.

Figure 6 Top 25 references with the strongest citation bursts. Year: The year of the publication first appeared. Begin: The year of the citation burst began. End: The year of the citation burst ended. Strength: The intensity of the increase in citations.

Discussion

This study provided a comprehensive bibliometric and visual analysis of publications related to PSD, identifying key trends, emerging research areas, and potential future directions. The findings reveal a substantial increase in research output over the past two decades, reflecting growing interest in this field. By clarifying the current landscape of PSD research, this analysis offers valuable insights to guide future investigations and inform the development of targeted interventions.

By analyzing publication growth, the steady increase in the cumulative number of publications has attracted considerable interest and is progressing rapidly. The body of knowledge has expanded notably, with a particular surge observed post-COVID-19 pandemic. PSD research peaked in 2022. In this study, a polynomial model was chosen to fit the publication growth trend due to its ability to capture non-linear patterns and complex growth dynamics. Other models—such as linear (limited to constant growth rates), exponential (assuming constant percentage growth), and logarithmic (suitable for initial rapid growth followed by stabilization) — were less suitable for the data’s varying growth rates and non-linear nature. The polynomial model’s residual analysis and high goodness-of-fit confirm its effectiveness in capturing the data’s structure. It predicts that the cumulative number of publications will reach 1,608 in 2025 and 1,729 in 2026.

The United States leads in PSD research publications, largely due to contributions from institutions like Harvard University, the University of California System, and the University of Michigan. These institutions have a long-standing foundation and strong influence in the field. The University of Toronto was also prominent, with a focus on sleep disorders influenced by environmental, social, and biological factors. Mayo Clinic, Cleveland Clinic, and Chinese institutes showed rising trends in this area. International collaborations, especially between the US and Canada, are increasing, potentially driven by the 2017 Nobel Prize in Medicine for circadian rhythm research. Frances Chung, a top author in PSD research, has extensively studied sleep disorders in surgical patients. Her work highlights that preoperative apnea-hypopnea index (AHI), age, and opioid dosage were positively correlated with postoperative AHI.²⁸ Sleep architecture is disrupted postoperatively, particularly in the N1 stage, with the greatest breathing disruption in the N3 stage.⁷ A high preoperative Pittsburgh Sleep Quality Index (PSQI) score is a risk factor for postoperative sleep disturbances.³

The influence of journals correlates with the impact of their articles. The top three journals by publication count were International Journal of Pediatric Otorhinolaryngology, Anesthesia & Analgesia, and Otolaryngology-Head and Neck Surgery. The most cited journals were Anesthesiology, Sleep, and Anesthesia & Analgesia, highlighting anesthesia and neuroscience fields’ significant contributed to PSD research. Journals like Chest, Laryngoscope, and Otolaryngology-Head and Neck Surgery also published high-quality studies. Multidisciplinary journals, with their broad readership, promoted interdisciplinary collaboration.

Keyword analyses provide valuable insights into emerging research directions, continuing research interests, clinical significance of keywords and disease epidemiology. In the early days, the main keywords focused on basic and clinical studies related to the subject headings of surgery,¹³ melatonin,¹⁴ tramadol,³¹ postoperative analgesia,^15,32 and pain.³³ Basic studies demonstrated that changes in sleep arousal may serve as an indicator of chronic neuropathic pain following peripheral nerve injury.³⁴ Meanwhile, clinical studies revealed that sleep disturbances following general anesthesia were positively correlated with increased opioid consumption.³¹ Different anesthetics may influence various endogenous sleep-wake substances,³⁵ with their regulatory mechanisms potentially depending on the specific type of anesthetic. Moreover, both surgery and regional anesthesia have been shown to reduce postoperative REM sleep.³⁶ With improved understanding from basic and clinical research, research themes have shifted significantly, driven by evolving clinical needs and scientific advancements. Although exogenous melatonin has shown some efficacy in reducing sleep latency in primary insomnia and delayed sleep phase syndrome, as well as regulating sleep-wake patterns in patients compared to placebo,³⁷ its impact on PSD remains modest and inconsistent across diverse patient populations. Interestingly, 27.5% of melatonin users reported insomnia as a reason for taking the supplement.³⁸ However, melatonin supplementation may moderately increase sleep propensity and is generally less effective than prescription sleep medications.³⁹ Given its limited ability to address the multifactorial nature of PSD, research on melatonin has declined in favor of more targeted interventions. Conversely, recent high-frequency co-occurrence keywords include “dexmedetomidine”, “neuroinflammation” and “acupuncture”. Dexmedetomidine has garnered increasing research attention due to its multifaceted benefits in improving postoperative outcomes. It effectively reduced pain, anxiety, and delirium while enhancing sleep quality.^26,40,41 Acting as a highly selective α2 receptor agonist, it interacts with the locus coeruleus—a brain region central to sleep regulation⁴²—thereby promoting endogenous sleep pathways and inducing a sleep state similar to natural N2 sleep.⁴³ Additionally, dexmedetomidine has been shown to attenuate pro-inflammatory cytokines, which are implicated in neuroinflammation and PSD.⁴⁴ However, the underlying mechanism of PSD may be related to neuroinflammation.⁴⁵ Pro-inflammatory and anti-inflammatory cytokines played important roles in circadian rhythm regulation, with TNF-α levels correlating with decreased sleep quality and IL-1β as a potential biomarker for circadian dysregulation.¹⁷ Neuroinflammation studies suggest that targeting cytokine pathways could mitigate PSD, offering a novel therapeutic approach. Acupuncture, an ancient Chinese technique, has shown promise in improving postoperative sleep quality through modern modalities like electroacupuncture (EA) and transcutaneous electrical acupoint stimulation (TEAS).⁴⁶ These interventions modulate the autonomic nervous system and reduce pain. While neuroinflammation-targeting therapies are still in early stages, acupuncture has demonstrated significant potential in reducing delirium and enhancing sleep quality in postoperative patients.⁴⁷ The rise of acupuncture reflects a growing interest in non-pharmacological interventions, driven by the need for holistic and patient-centered care.

The findings of this study align with existing literature in several key aspects. The progressive increase in PSD-related publications in research output are consistent with prior study,^3,10 which have also identified a growing interest in PSD over recent years. Additionally, the identification of major research trends, such as the impact of anesthesia and analgesic drugs on sleep,³⁵ aligns with well-established themes in the field. However, this study also revealed several novel insights. The emerging focus on neuroinflammation⁴⁸ as a potential mechanism underlying PSD represents a significant shift from earlier research, which primarily concentrated on pharmacological and clinical aspects. Furthermore, the study emphasizes the potential of non-pharmacological interventions, advocating for integrative approaches that combine both pharmacological and non-pharmacological strategies. These findings highlight the evolving nature of PSD research and the need to integrate multidisciplinary perspectives.

Our study had some limitations. First, the data were obtained from the Web of Science Core Collection, a widely used and recommended database for bibliometric analysis because of its high data quality and completeness. However, this limited the scope of bibliometric analysis, potentially resulted in an incomplete view of the research landscape. Second, many studies from non-English databases may not be indexed. Future research could benefit from the inclusion of additional databases to improve comprehensiveness. Finally, the bibliometric analysis method itself is based on publications. It is inherently subject to the effects of potential publication and citation bias. Researchers could enhance the accuracy by incorporating preprints, accounting for indexing delays, combining qualitative assessments, and refining search strategies.

Conclusion

Scientific publications on PSD have increased annually over the past two decades. Future research should focus on integrating multidisciplinary approaches, exploring neuroinflammation, evaluating non-pharmacological interventions, and long-term outcomes, which will advance scientific knowledge, enhance clinical practice, and ultimately improve patient outcomes and quality of life.