Assessing Acceptance Level of a Hybrid Clinical Decision Support Systems

Area of Science:

• Medical Informatics
• Clinical Decision Support Systems
• Diabetes Mellitus Research

Background:

• Clinical Decision Support Systems (CDSS) are increasingly used in healthcare.
• Type 2 Diabetes Mellitus (T2DM) risk prediction requires efficient and reliable tools.
• Understanding physician acceptance of CDSS is crucial for effective implementation.

Purpose of the Study:

• To evaluate user acceptance of a CDSS for T2DM risk prediction.
• To assess the influence of data-driven and rule-based models on physician efficiency and acceptance.
• To determine the impact of risk scales and model explanations on CDSS usability.

Main Methods:

• A user acceptance study involving physicians evaluating CDSS outputs.
• Three CDSS output settings were tested: data-driven (DD) model, DD with FINDRISK scale, and DD with scale and explanation.
• Lankton's model was used to assess user acceptance based on agreement, understanding, and perceived usefulness.

Main Results:

• Physicians tended to blindly trust CDSS outputs when risk scales were absent.
• The presence of risk scales and explanations significantly impacted physician agreement and understanding.
• Interpretability of the CDSS output was identified as a key factor in user acceptance.

Conclusions:

• Interpretability is critical for physician acceptance of CDSS in T2DM risk prediction.
• Providing risk scales and clear explanations enhances trust and utility of CDSS.
• Future CDSS development should prioritize transparency and explainability for better clinical integration.