A comprehensive search of relevant databases was covered all available records from their establishment through November 10, 2024. The databases researchers searched were: PubMed, the Cochrane library, Web of Science, CINAHL, EMBASE, the Joanna Briggs Institute (JBI), China National Knowledge Infrastructure (CNKI), Wanfang Data, China Science and Technology Journal Database (VIP) and Chinese BioMedical Literature Service System (SinoMed) and a manual search was carried out for relevant literature sources (For more in-depth details, please refer to Supplementary Appendix A). To maximize study identification and ensure methodological rigor, we implemented a multi-faceted approach that included: systematic examination of previously published reviews, meticulous scrutiny of reference lists from all included studies, and critical analysis of existing meta-analyses to identify potentially eligible articles that might have been overlooked through conventional search methods.

The PICOS framework, which encompasses population, intervention, comparison, outcome, and study design, was employed to formulate stringent inclusion criteria for study selection (22). The inclusion criteria were established as follows: (1) population. ICUs patients (≥18 years old); (2) intervention. Experimental group received guide-based interventions; (3) comparison. Control group implemented nursing procedures as usual; (4) outcomes. Primary outcome is PR rate or PR time. Secondary outcome is to evaluate unplanned extubation rate, delirium incidence and other complications rate; (5) study design. RCT, published in peer-reviewed journals, with language restrictions limited to English and Chinese publications.

Exclusion criteria were established as follows: (1) studies in which the intervention was poorly described, making it impossible to determine if it was a “guide-based intervention.”; (2) outcome measures were incomplete and data could not be extracted for meta-analysis; (3) reviews, case reports, cohort studies, cross-sectional studies, etc; (4) abstract-only articles; (5) literature that is duplicated, incomplete or incorrect.

The citations of all the studies obtained from the search were imported into the reference management software, Endnote X9. Subsequently, two researchers, (the primary and the co-primary authors) independently evaluated the methodological quality and relevance of the retrieved studies. This evaluation was carried out in strict accordance with the pre-established inclusion and exclusion criteria. Data extracted from each study included the authorship, year of publication, country of origin, sample size, details of the guide-based intervention, study settings, outcome measures, and principal findings. Any discrepancies identified between the two researchers during the evaluation process were systematically addressed through iterative discussion, with unresolved disagreements being referred to a third senior researcher for final arbitration.

Two independent researchers (the primary and co-primary authors) conducted a rigorous assessment of bias risk and methodological quality in the included RCTs, following the standardized criteria established in the Cochrane Handbook for Systematic Reviews of Interventions (23). The evaluation encompassed seven critical domains of potential bias: randomization sequence generation, allocation concealment, participant blinding, outcome assessment blinding, insufficient outcome data, selective reporting, and other sources of bias.

The meta-analysis was conducted using Review Manager (RevMan) software, version 5.2. When dealing with dichotomous outcome measures, the effect magnitudes were represented as risk ratios (RR) along with the corresponding 95% confidence intervals (CI). In contrast, for continuous outcome variables, the analysis made use of weighted mean differences (WMD) together with their 95% CI. Statistical heterogeneity was assessed through multiple indicators, including chi-square test, I², and P value. A fixed-effects model was applied when heterogeneity was deemed acceptable (P > 0.1 and I² < 50%), whereas a random-effects model was implemented in cases of significant heterogeneity (P < 0.1 or I² ≥ 50%), except when studies demonstrated substantial clinical homogeneity. To evaluate the robustness of findings and identify potential sources of heterogeneity, sensitivity analyses were performed through sequential exclusion of individual studies. The assessment of publication bias was carried out by analyzing the symmetry of the funnel plot. Additionally, when the analysis involved more than 10 studies, formal statistical tests, namely Begg's and Egger's tests, were conducted using Stata statistical software (version 18.0; StataCorp LP, College Station, TX) to supplement the funnel plot analysis.

The systematic review ultimately included 14 studies (19, 20, 24–35), of which 13 were conducted in China and one in Colombia. Figure 1 presents a detailed flowchart of the search and selection process. The included studies mainly investigated adult patients in ICU, with patient recruitment and intervention implementation occurring across various ICU subtypes, including integrated ICUs, general ICUs, and neurological ICUs. Comprehensive characteristics of the included studies, along with their primary outcomes, are systematically presented in Table 1.

These guides of included studies took various forms, such as CPGs, ABCDEF bundle and syntheses of best evidence. The measurement instruments utilized during the intervention predominantly originate from guidelines available on official websites or are authored by experts, supplemented by some self-compiled assessment scales. Implementation was carried out by multidisciplinary healthcare teams comprising physicians, critical care specialist nurses, and clinical technicians. Intervention efficacy demonstrated significant variability based on program-specific configurations and their corresponding implementation parameters. Detailed characteristics of all guide-based interventions, including their specific components and implementation interventions, are presented in Table 2.

The methodological quality assessment of included studies, as presented in Figures 2, 3, revealed the following findings regarding risk of bias: among the 14 studies analyzed, random sequence generation methods varied significantly, with nine studies utilizing random number tables, two cluster RCTs employing drawing lots administered by ICU head nurses, one study using sealed envelopes, one study implementing a block randomization list, and one study applying stratified random sampling. Only a single study demonstrated appropriate allocation concealments. The majority of studies (n = 13) failed to implement blinding of participants and personnel, reflecting the practical challenges associated with masking guide-based interventions in clinical settings. Four studies reported adequate blinding procedures for outcome assessment. One study was classified as having high risk of bias due to incomplete outcome data. Regarding reporting bias, 13 studies showed low risk of selective outcome reporting, while 12 studies demonstrated low risk of other potential biases.

A sensitivity analysis was performed on studies demonstrating significant heterogeneity (P < 0.01, I² > 50%). Upon conducting a leave-one-out analysis, the heterogeneity of the remaining studies decreased (I² < 50%). Nonetheless, the PR time results were altered (P > 0.05; refer to Table 4). The sensitivity analyses indicated increased heterogeneity, primarily attributed to variations in sample size and the quality of the literature, with a focus on these particular studies (25, 28) (Table 4).

In this study, we constructed funnel plots to analyze the PR rate and the unplanned extubation rate. The funnel plot analysis revealed a relatively symmetrical distribution of effect sizes along the central axis, with most data points evenly dispersed on both sides of the plot, indicating a low likelihood of significant publication bias in the included studies. This is further supported by the Begg and Egger tests, which showed P > 0.05 for unplanned extubation rate (Supplementary Figure S9). However, evidence of publication bias was detected for the PR rate, necessitating cautious interpretation (Egger test, P = 0.004; Supplementary Figure S10). To further validate the stability of our findings, we performed a supplementary analysis using the trim-and-fill method to assess potential publication bias. The comparative analysis demonstrated consistent effect estimates between pre- and post-adjustment results, with no significant alterations in the direction or magnitude of the observed effects. This methodological validation confirms the stability and reliability of the pooled effect size, suggesting that our primary findings are not substantially influenced by potential publication bias (Supplementary Figure S11, Supplementary Table S1).

This study included 14 RCTs with 4,338 ICU patients to assess guide-based interventions for reducing the utilization of PR. While most studies couldn't achieve strict double-blindness due to the nature of the interventions, their overall quality is fair with clinical relevance. Main findings indicate that guide-based interventions can reduce the utilization of PR, delirium incidence and duration, unplanned extubation, other complications, and mechanical ventilation duration, while improving patient satisfaction. However, they don't shorten the length of stay in the ICU or improve patient agitation or anxiety. Multi-center and larger RCTs are needed for further validation.

In the included 14 RCTs, the guide, for formulating a protocol to reduce the utilization of PR, includes CPGs, syntheses of best evidence, ABCDEF bundle and scoping review. The decision of PR with patients in ICU is often based on clinical experience, hence, the objectivity and standardization of PR decision-making are insufficient (39, 40). Thus, standardized assessment is crucial. Our findings indicate that the PR Decision Wheel and Assessment of PR framework is the most prevalent in different studies, with the corresponding interventions frequently exhibiting similarities. Most interventions have instituted PR decision-making teams (24, 28, 29, 31–33, 35). Assessment is essential for all intervention; however, the content and frequency of these assessments vary considerably. For comprehensive assessment of PR, certain studies implemented assessments at 8-h intervals (20, 27), whereas others implemented assessments at the conclusion of each shift, and the necessity for PR is re-evaluated within the same timeframe (30, 32). This meta-analysis specifically evaluated the effects of guide-based interventions on reducing the utilization of PR with patients in ICU. Firstly, this study found that guide-based interventions significantly reduce the utilization of PR with ICU patients. Interventions grounded in the PR Decision Wheel and the Assessment of PR entail the selection of diverse PR methodologies tailored to the severity of the patient's condition, with adjustments to the frequency of assessments as necessary (20, 26, 32). This approach has demonstrated efficacy in reducing the utilization of PR. Interventions based on the ABCEDEF bundle have demonstrated potential in reducing the utilization of PR (24, 28, 34). The bundle, however, consists of a multitude of interventions, incorporating diverse elements such as education and training (41), policy development (42), environmental modifications (43), and interdisciplinary collaboration (44). This complexity poses a challenge in determining the effectiveness of each individual component. Thus, additional research with factorial designs may be needed to identify the bundle's most effective components. While objective tools are available, their effective utilization by nurses necessitates a shift in their perception of PR. Research examining the perspectives of nursing staff on PR utilization reveals that, despite an awareness of its negative implications, entrenched practices and institutional norms frequently sustain its application (45, 46). This suggests a complex interaction between knowledge and practice, wherein even well-informed staff may encounter challenges in effectively implementing PR-reduction interventions due to systemic pressures and resource constraints. Consequently, numerous studies incorporate training programs for nurses to support this transition. Critical care nurses, who play a pivotal role in decision-making regarding the implementation of PR in ICUs. Through comprehensive training and the provision of clear guidelines, nurses can be better prepared to make informed decisions regarding the adjustment of assessment frequency as needed, potentially reducing reliance on these measures (45). Interventions based on syntheses of best evidence can be implemented to enhance nurses' decision-making skills, which has also been shown to effectively reduce the utilization of PR (25, 30). While this study provides evidence that guide-based interventions significantly reduce PR time, substantial heterogeneity was observed across studies. Subsequent subgroup analysis based on different PR time calculation methodologies revealed a marked reduction in heterogeneity, indicating that variations in time measurement formulas significantly influence the reported PR time outcomes. This finding suggests that standardization of PR time calculation methods is crucial for ensuring consistency and comparability across future studies in this field. Once the method for calculating PR time is standardized, all research findings will be based on a unified evaluative framework. This will enable researchers to clearly compare the core conclusions of different studies, accurately identify common patterns and distinct issues within the research, significantly enhance the efficiency of research utilization, and increase the academic value of the findings. Moreover, it will provide robust support for the transmission of knowledge and foster innovative breakthroughs within the field. Delirium represents a prevalent neuropsychiatric syndrome in ICU, demonstrating significant associations with multiple adverse clinical outcomes (11). PR has been identified as a modifiable risk factor and potential precipitant for delirium development (13). Therefore, reducing the utilization of PR is an important measure to prevent delirium from occurring. In our meta-analysis, we found that guide-based interventions were effective in reducing delirium incidence (19, 30, 34) and duration of delirium (19, 24, 28) and delaying delirium onset time (28). Guide-based interventions encompass the reduction of stress (e.g., family support, encourage preferences and identify and solve spiritual, social and environmental needs), the promotion of early activity (e.g., full range of joint motion, sitting exercise, bed-free activities, and walking exercises) and the titration of sedative and analgesic medications in accordance with the Richmond Agitation-Sedation Scale (RASS). Moreover, a study underscores the psychological effects of PR on family members during the COVID-19 pandemic, highlighting that involving them in care discussions and decisions can facilitate their understanding and coping with the utilization of PR, ultimately contributing to its reduction (47). Our study indicated that guide-based interventions can also shorten duration of mechanical ventilation. Guide-based interventions involve the systematic and timely assessment of pain, as well as the prompt removal of tubes. The reduction in the utilization of PR may contribute to a shortened duration of mechanical ventilation; however, this outcome is substantially affected by the patient's clinical condition. In clinical practice, it is imperative to develop a comprehensive, multi-dimensional evaluation framework. This framework should encompass, firstly, an assessment of individual patient variability, including factors such as age and history of underlying diseases, to ascertain the patient's tolerance to analgesic medications. Secondly, it should involve a thorough review of the intervention's implementation details, such as the precision of pain assessments and adherence to established protocols for the timing of tube removal. Concurrently, there is a need for dynamic monitoring of adverse reactions during the use of PR, integrating objective data to evaluate the feasibility and potential risks associated with reducing PR use. Therefore, a thorough evaluation of all relevant factors is imperative.

While PR is considered a preventive measure for unplanned extubation, relevant studies have pointed out that the longer the PR time, the greater the probability of unplanned extubation and the other complications rate also increases (13, 48). Meanwhile, high-quality meta-analyses have demonstrated that PR is an independent risk factor for unplanned extubation in ICU patients (49). The results of this study showed that the guide-based interventions can reduce unplanned extubation with patient in ICU. Guide-based interventions encompass the education and training regarding the knowledge of unplanned extubation, along with continuous assessment to ensure the timely removal of tubes. Interventions designed to minimize the application of PR may also be employed to reduce the incidence of unplanned extubations. Implementing these measures not only improve nursing quality but also enhance patient safety. In recent years, with the increasing attention of nursing managers to the systematic and normalized special training of nurses, the awareness of ICU nurses on PR has been strengthened. ICU nurses with strong PR awareness can accurately judge the timing and extent of PR reduction by dynamically assessing the patient's condition and the unplanned extubation rate has also decreased compared with the previous in routine care. Reduced utilization of PR can control the incidence of adverse events. Guide-based interventions are capable of evaluating the patient's PR level, timing, and release schedule through dynamic assessment. This approach can effectively minimize unnecessary PR and enhance the rationality and standardization of PR practices.

Based on the results of the meta-analysis, implementing the PR-reduction protocol in accordance with established guides and employing a multidisciplinary team approach are possible to reduce the utilization of PR. The frequency of patient assessments should be adjusted based on the specific type of PR involved. However, the variability in measurement tools, such as the PR Decision Wheel, RASS, CAM-ICU, and self-compiled ICU Patient PR Assessment Scale, can compromise the scientific rigor of the research. Consequently, the findings may be challenging to generalize. Therefore, there is a need to standardize the PR assessment scale.

Guide-based interventions show benefits in reducing the utilization of PR with patients in ICU. However, due to the diversity of guide-based interventions, determining the most effective guide-based protocol remains challenging. Further research utilizing factorial designs may be necessary to ascertain the most effective components of the guide. Furthermore, this study did not demonstrate improvement in patient agitation or anxiety and reduction length of stay in the ICU with guide-based intervention. Multi-center and larger RCTs are needed to validate these outcomes.