Related Experiment Videos
Stakeholder perspectives on machine learning models predicting diabetic foot ulcers and amputations in diabetes care:
Iris Ten Klooster1, Saskia M Kelders1,2, Hanneke Kip1,3
1Centre for eHealth and Wellbeing Research, Department of Psychology, Health and Technology, University of Twente, Enschede, Netherlands.
Background:
Predictive machine learning models can support timely interventions for diabetes management. However, there is limited insight into stakeholder perspectives on their use in healthcare, which is important for alignment with expectations and in turn fostering adoption.
Objective:
This study aimed to identify stakeholder perspectives on the (1) scenarios of use, and (2) the values and attributes of two machine learning models predicting diabetic foot ulcers and amputations in diabetes care.
Materials And Methods:
Five diabetes patients, two healthcare workers and four experts involved in system integration aspects participated in semi-structured interviews. Three scenario components were presented to help participants articulate their needs and preferences. Attributes were inductively identified from interview data and directly linked to scenarios of use, with values analyzed based on the identified attributes.
Results:
We identified four scenarios of use namely (1) supporting healthcare workers' decision making, (2) supporting patient empowerment, (3) improving appointment planning based on risk, and (4) providing early warnings concerning acute risks. In addition, thirteen values were identified: supporting patient awareness, hybrid approach, detecting acute risks, collaboration between healthcare workers at different levels, unobtrusiveness, integration into existing systems, explainability, translating patient data from EHRS into clinically relevant insights, risk-based stratification of care, continuous development, interoperability, privacy, and regulatory compliance.
Discussion:
The insights of this study can guide the creation of a system that integrate machine learning models for the prediction of diabetic foot ulcers and amputations.
