Prediction models can support shared decision making. However, it remains unclear how doctors and patients will use these predicted percentages by making their choice. That is why this study investigates how outcomes of prediction models influence the clinical decision making of both patients and professionals.

Design:

Prospective survey study.

Setting:

Multicenter data collection between the 1^st of February 2019 and the 1^st of August 2020.

Patients or Participants:

Patients: Dutch-speaking females (age 18-75 years) visiting the (benign) gynaecology outpatient department.

Professionals: Dutch-speaking physicians involved in the gynaecological field.

Interventions:

The survey contained a fictional case about abnormal uterine bleeding. Endometrial ablation (EA) was suggested as treatment with a general failure percentage about 15%. Patients preference of treatment was asked (endometrial ablation vs. uterus extirpation (UE)). After using a prediction model, the failure percentage of EA was 61%. Treatment preference was asked again. A similar survey was developed for professionals.

Measurements and Main Results:

The primary outcome measure was the failure percentage of EA where both patients and doctors would no longer choose EA. A total of respectively 585 and 102 surveys of patients and were analysed. After reading the case, 86,7% (N=508) of the women and 86.3% (N=87) of the professionals preferred EA over UE. When the prediction of failure seemed to be 61% based on the fictive case, a total of 47,9% (N=243) in the patient group and 48.9% (N=43) of the professionals significantly changed their choice from EA into UE (both p-value 0.000).

The accepted average failure percentage of EA was 56.7% (range 10-100) by patients and 53,6% (range 10-100) by professionals.

Conclusion:

Patients and professionals significantly adapted their treatment preference after reading the personal failure percentage. Hereby we can conclude that it seems that a prediction model influences treatment choice. The extra information, gained by such a model, can contribute better patient counselling and optimize shared decision making.