Predicting Survivorship Appointment Nonattendance in a Community Cancer Center: A Machine-Learning Approach

Area of Science:

• Oncology
• Health Services Research
• Data Science

Background:

• Effective cancer survivorship care is crucial for patient outcomes.
• The transition from active treatment to follow-up care includes a key consultative appointment.
• Predicting nonattendance at this appointment can improve care utilization.

Purpose of the Study:

• To develop a predictive model for nonattendance at cancer survivorship care appointments.
• To identify key patient attributes associated with appointment nonattendance.
• To aid practitioners in proactively managing follow-up care for cancer survivors.

Main Methods:

• Utilized machine learning algorithms to analyze data from 843 cancer survivors.
• Developed a predictive model using variables from electronic medical records.
• Employed k-fold cross-validation for model accuracy assessment.

Main Results:

• A parsimonious predictive model was developed.
• The model achieved a k-fold classification accuracy of 67.3%.
• The final model incorporated three significant variables.

Conclusions:

• A machine learning model can predict cancer survivors likely to miss follow-up appointments.
• Identifying at-risk patients allows for targeted interventions to improve care utilization.
• This approach supports enhanced quality of care for cancer survivors.