Hypothesis / aims of study
Recurrence after native tissue repair for pelvic organ prolapse (POP) surgery remains a major concern, particularly after anterior colporrhaphy. Prediction models have been developed to estimate recurrence risk, yet their performance compared to clinical judgment is unclear. This study compares two validated prediction models against (uro)gynaecologists' estimates of recurrence risk. In addition, we aimed to investigate how (uro)gynaecologists communicate recurrence risks to patients, and whether providing individualized risk information could influence their counseling approach.
Study design, materials and methods
A vignette-based study was conducted using real patient data from three previous cohort studies (n=152). Each vignette included key predictors of recurrence. Clinicians (members of the Dutch Pelvic Floor Working Group) estimated the probability of anatomical and composite POP recurrence on a scale from 0 to 100%, which was then compared to model-derived estimates. Discrimination was assessed using the area under the receiver operating characteristic curve (AUC). Calibration-in-the-large and calibration plots assessed systematic over- or underestimation. Furthermore, clinicians completed a risk counseling survey.
Results
Sixty-seven clinicians completed the survey and vignettes. For anatomical recurrence, AUCs were 66.2% (95% CI 58.4–74.0) for the prediction model versus 59.3% (95% CI 51.2–67.4) for clinicians (p=0.181). For composite recurrence, AUCs were 69.3% (95% CI 60.8–77.9) for the model versus 50.1% (95% CI 40.6–59.7) for clinicians (p<0.001).
For the anatomical outcome, calibration-in-the-large was poor for clinicians and good for the prediction model as the mean predicted probability for clinicians was 27.6% (±16.6%) and the mean predicted probability for the prediction model was 52.1% (±16.1%), where the observed actual anatomical outcome was 50.5%.
For the composite outcome the same pattern was observed with a poor calibration-in-the-large for clinicians.
The mean predicted probability was 19.3% (±12.5%), whereas the mean predicted probability for the prediction model was 29.3% (±13.3%) and the observed actual composite outcome was 25.8%.
Most clinicians (76.1%) reported that recurrence risk influenced management decisions, and 97.0% indicated that individualized risk estimates would affect their counseling approach.
Interpretation of results
Prediction models demonstrated superior discrimination (only for composite recurrence) and calibration over clinical judgment in estimating POP recurrence. Clinicians, while discussing recurrence routinely, systematically underestimated recurrence risk with poor calibration-in-the-large. These findings support incorporating individualized prediction tools into clinical counseling to improve shared decision-making and risk stratification.