Online prediction tool for female pelvic floor dysfunction: development and validation.

INTRODUCTION AND HYPOTHESIS: The aim was to develop and validate (internally and externally) a prediction model for the presence and diagnosis of pelvic floor dysfunction (PFD) in women, including pelvic organ prolapse, stress urinary incontinence and/or overactive bladder via a patient-completed online tool.

METHODS

Using a retrospective cohort of women aged >18 years, from multiple tertiary gynaecology units within Queensland, Australia (2014-2018), the prediction model was developed via penalized logistic regression with internal and external validation utilizing multiple clinical predictors (42 questions from the Australian Pelvic Floor Questionnaire and demographics: age, body mass index, parity and mode of delivery). The main outcome measures were the accuracy of the model in predicting a diagnosis of pelvic floor dysfunction and its specific conditions of prolapse and incontinence.

RESULTS

A total of 3,501 women were utilized for model development and internal validation and 449 for external validation. On internal validation the model correctly identified those with PFD with 97% sensitivity, 74% specificity and a concordance index (C-index) of 0.96. Predictions of pelvic organ prolapse were also accurate, with 86% sensitivity, 83% specificity, C-index 0.83, as was stress urinary incontinence, 84% sensitivity, 87% specificity, C-index 0.87, and overactive bladder, 76% sensitivity, 77% specificity, C-index 0.77. External validation confirmed the model's accuracy with a similar C-index in all parameters.

CONCLUSIONS

This model provides an accurate online tool to differentiate between those with and without PFD and diagnoses of common pelvic floor disorders. It serves as a valuable self-assessment for women and primary care providers.