DAta MIning to Evaluate Novasure Treatment

Overview

Abnormal Uterine Bleeding (AUB) is a common health problem that affects approximately 30% of women of reproductive age and can have several underlying causes. It significantly affects quality of life, use of medical resources and health costs. Endometrial ablation is a commonly used minimally invasive surgical procedure for the treatment of AUB that destroys endometrial tissue. This procedure is an alternative treatment to hysterectomy because it is less invasive and has a shorter recovery period. Several ablation techniques are available to remove endometrial tissue, including bipolar radiofrequency (NovaSure treatment). While patient satisfaction with this form of endometrial ablation for the treatment of AUB is high, approximately 10-20% of women undergoing endometrial ablation require additional invasive surgery, primarily because of persistent blood loss or pain. There is therefore a need to identify and evaluate factors that can improve women's outcomes, or that can be building blocks for prognostic models that can be used to influence clinical practice. In this 10-year single-center retrospective cohort study, we aim to apply data mining and machine learning techniques to uncover hidden relationships/patterns between variables, and identify factors and patients at increased risk for Novasure treatment failure. With multiple time variables, this is not possible with a simple statistical analysis. Discovering these patterns and risk factors could help improve medical care, patient counseling and patient satisfaction.

Ten-year retrospective single-center cohort study in which we will: 1. retrospectively follow patients who underwent endometrial ablation using Novasure between 2009-2019 at the Máxima Medical Center and observe the incidence of outcomes and features. 2. assess which patient and procedure features are associated with failure of NovaSure treatment, defined as occurrence of any subsequent invasive procedure related to AUB within 3 years, of women who underwent this treatment in Maxima MC between 2008 and 2018. 3. assess which patient and procedure features are associated with a successful Novasure treatment, defined as freedom of subsequent invasive procedure within 3 years related to AUB, of women who underwent this treatment in Maxima MC between 2008 and 2018. 4. develop predictive and prognostic models that assess the probability of treatment success or failure based on these patient and procedure variables. Procedure In this observational study, a clinical data collector will be used to review women (with automatically pseudonymised data) who had endometrial ablation with the Novasure. An overview will be given of patient and procedure features that may have played a role in the failure of the treatment. To be able to investigate this, all women who underwent the intervention at the Máxima Medical Center between 2008-2018 are included in the study. They received the usual care and did not have to follow any additional procedures. All patients will be included in the descriptive analysis. All descriptive analyses, including total population size and follow-up time, will be given. Continuous variables are presented as mean with standard deviation, or median with interquartile range, depending on the distribution. Categorical variables are presented as number with percentage. Binary data are presented as frequency and percentage. Time between Novasure and reintervention will be given as mean (SD) and time-to-event (Kaplan-Meier). Moreover the relative risk for reintervention is calculated. In addition to classical statistical analyses, data mining (linear regression) and machine learning techniques are applied. Supervised classification models are learned to compute the importance factors of independent variables with respect to the dependent variable. In this research only retrospective file research on pseudonymised data is carried out with outcome measures that are part of the standard quality control. All data is obtained from the Clinical data Collector. The collector extracts and pseudonymizes data from the electronic patient record. Patients are not subjected to additional actions or are not imposed to any rules of conduct for this research. There are also no questionnaires sent to patients that affect the psychological integrity or that are perceived as stressful by the test subjects.