Clinicians' ethical considerations regarding the use of prediction models in hand and wrist care: a qualitative study.

OBJECTIVE

To explore clinicians' ethical considerations regarding the use of prediction models in clinical decision-making in hand and wrist care.

DESIGN

We conducted semi-structured interviews and analysed them using hermeneutic-interpretive phenomenology. An example prediction model was used to predict whether a patient will achieve the improvement they aim for in a domain relevant to them, given their baseline score.

SETTING

Data were collected in the Netherlands from two specialized clinics for hand and wrist care and two academic hospitals.

PARTICIPANTS

We purposefully selected seven hand surgeons and five hand therapists.

INTERVENTIONS: Not applicable MAIN OUTCOME MEASURE: Semi-structured interviews

RESULTS

We identified four main themes. First, clinicians acknowledge the patient's voice in shared decision-making, but ultimately, they decide. Second, clinicians want to maintain their autonomy in treatment decisions because they have a strong clinical intuition and feel that a prediction model will never be as good as their clinical judgment. Third, deviating from predictions is sometimes needed due to equity issues, the right to receive care, and the principle of striving for the least invasive treatment. Fourth, clinicians require sufficient information to build confidence in using the prediction model effectively.

CONCLUSIONS

Clinicians consider prediction models ethically acceptable when they support, rather than replace, clinical judgment. Their main concerns are preserving professional autonomy and responsibility, respecting patient needs, ensuring equitable access to care, and being able to deviate from predictions when necessary. As prediction models improve, increased accuracy and transparency may enhance trust but may not resolve broader ethical issues, including patient preferences, fairness, responsibility, and accountability. Addressing these concerns is an important step toward enabling prediction models to support meaningful patient involvement, which could contribute to shared decision-making and improve quality of life.