Applying graph theory to improve the quality of scientific evidence from textual information: Neural injuries after gynaecologic pelvic surgery for genital prolapse and urinary incontinence

Abstract

To provide the overall rate for all types of neurologic iatrogenic injuries during urogynaecologic surgery from textual data. Methods: Systematic research focused on complications of gynaecologic surgery and neurologic injuries in abstracts. Keywords concerning complications (cluster A), unspecific; neurologic issues (cluster B); surgery (generic words) (cluster C); specific gynaecologic operations (cluster D); and specific gynaecologic operations for pelvic organ prolapse and urinary incontinence (cluster E) were extracted. Associations among clusters of keywords were assessed by using multiple runs of text-mining software Semantic Brand Score (SBS, https://semanticbrandscore.com/#primary). Association scores were converted into probabilities. The rate of neurologic complications in urogynaecologic surgery was calculated ("a priori" probability) by applying Bayes' theorem. Textual estimates of neurological injuries in urogynaecologic surgery are 0.035554 (95% confidence intervals 0.019607−0.0515001; no quantitative data were found). To test if the probability calculated on textual information was the same as quantitative data reports ("a posteriori" probability), the rate of neurologic complication of all gynaecologic surgery was calculated using a meta-analytics approach and was compared with the textual analysis value. Results: The rate of neurologic complications in gynaecologic surgery after meta-analytic data synthesis has been 0.016489 (95% confidence intervals 0.012163−0.022320), which is equal to the textual estimate (0.016889, 95% confidence intervals 0.019607−0.051501). Therefore, 0.035554 is a reliable likelihood to observe a neurologic complication in urogynaecologic surgery. Conclusion: Iatrogenic nerve injuries in urogynaecologic surgery are higher than whole gynaecologic surgery. Text-mining software SBS and probability conversion can provide reliable answers from overall scholars' opinions on unsolved clinical questions when better evidence is lacking.