Neurocognitive Outcomes After Extracranial Surgery and General Anesthesia in Patients with a History of Mild-to-Moderate Traumatic Brain Injury: Systemic Review and Meta-Analysis

Figures

• 

  Figure 1. Screening of identified articles was conducted using the artificial intelligence (AI)-driven ASReview tool. (A) ASReview chart quantifying the number of reviewed records compared to the total records. (B) ASReview graph illustrates how many records were marked as relevant over the course of the review. Screening was discontinued after hundreds of records were searched without identifying additional relevant articles. (C) ASReview demonstrating how AI enhances the efficiency of finding relevant articles compared to the random ordering of searchable records.

• 

  Figure 2. The PRISMA 2020 flow diagram illustrates the process of identifying, screening, and including studies in a meta-analysis. A total of 27,139 articles were retrieved from our selected databases using our curated search strategy (Supplementary Table S1). Utilizing the Rayyan screening tool, 8010 studies were identified as duplicates and removed from the initial set of 27,139 articles. The remaining 19,129 studies were screened using the artificial intelligence tool ASReview [31,32]. Based on ASReview’s ranking of relevant articles, the reviewers manually screened the top 20% of the most relevant studies by title and abstract, totaling 3826 articles. Of these, 29 were selected for full-text review, while the remaining 15,303 articles (19,129 − 3826 = 15,303) were deemed ineligible without title and abstract screening. The full texts of the 29 selected articles were assessed against the selection criteria, leading to the exclusion of 12 studies due to unrelated populations, 2 due to irrelevant study design, 6 due to irrelevant outcomes, and 4 due to the absence of an appropriate control group. Ultimately, five studies were included in the review [41,42,43,44,45].

• 

  Figure 3. Newcastle–Ottawa coding quality assessment results were used to evaluate the risk of bias for each included study [41,42,43,44,45] (reference number are from top to bottom of column). Under each criterion, an article was awarded “*” if it met the criteria, or “/” if it felt short in that category. Scoring was based on the number of “*”s out of a possible 9. A higher number of “*”s indicated a lower risk of bias.

• 

  Figure 4. Forest plot comparing differences in (A) Trail-making test A (TMT-A), (B) Trail-making test B (TMT-B), and (C) Glasgow Outcome Scale (GOS) scores using odds ratios [41,45]. The prism represents the overall statistical results of the experimental data, squares indicate the weight of each study, and horizontal lines represent the 95% CIs for each study. The p-values next to I2 on the right side of the forest plot represent the significance of the heterogeneity, while the p-values on the left side show the significance between the treatment and control groups. Abbreviations: TBI, traumatic brain injury; I2, heterogeneity; MD, mean difference; CI, confidence interval; SD, standard deviation; OR, odds ratio.

• 

  Figure 5. Forest plot comparing length of stay in the (A) ICU [42,44,45] and (B) hospital [41,42,43,44] between TBI patients who underwent extracranial surgery/GA and those who did not undergo surgery/GA. The prism represents the overall statistical results of the experimental data, squares indicate the weight of each study, and horizontal lines represent the 95% CIs for each study. The p-values next to I2 on the right side of the forest plot represent the significance of the heterogeneity, while the p-values on the left side show the significance between the treatment and control groups. Abbreviations: TBI, traumatic brain injury; ICU, intensive care unit; HOS, hospital; I2, heterogeneity; MD, mean difference; CI, confidence interval; SD, standard deviation. The arrow denotes a capped upper bound of the 95% CI.

• 

  Figure 6. Funnel plot suggesting a low risk of publication bias in length of hospital stay analysis.