Scoping Literature Review on the Mathematical and Computation Models Currently Available to Predict Causes of Abnormal Uterine Bleeding
Abstract
Introduction: Abnormal uterine bleeding (AUB) affects 3-30% of women of reproductive age. Per the International Federation of Gynecology and Obstetrics system, etiologies of AUB can be classified as structural (PALM - polyp, adenomyosis, leiomyoma, and malignancy/hyperplasia) or non-structural (COEIN - coagulopathy, ovulatory dysfunction, endometrial, iatrogenic, and not yet classified). The diagnostic process frequently involves numerous office visits and invasive imaging and procedures such as transvaginal ultrasounds and endometrial biopsies to rule out various etiologies in a stepwise fashion. The purpose of this review was to determine which mathematical and computational models were currently available to diagnose abnormal uterine bleeding and guide initial treatment.
Main Body: Research articles from five databases were screened based on inclusion criteria of "abnormal uterine bleeding" and "model" and exclusion criteria of "postpartum hemorrhage" and "pregnancy." Then articles were further screened based on whether or not they generated a predictive model for AUB. The full text of these articles was retrieved and evaluated for inclusion in the review. Pertinent information was charted using Excel. The full text of nineteen articles were reviewed. None of them evaluated for all possible causes of AUB at the same time. Two studies have developed diagnostic models that distinguish between endometrial polyps, endometrial myoma, endometrial hyperplasia without atypia, and endometrial carcinoma. Seventeen articles generated predictive models for AUB due to endometrial malignancy/hyperplasia (AUB-M). The remaining two articles focused on AUB-O and AUB-L.
Conclusion: While there are currently no comprehensive diagnostic models for AUB, the work done by Wynants et al and Shang et al were the first models to be able to differentiate between multiple etiologies of AUB. Their methods provide a strong foundation for future studies to build upon and expand the diagnostic scope of predictive models. Based on the articles in this review, next steps have been suggested to create a more comprehensive predictive model.