A Machine Learning-Based Approach for Predicting Patient Punctuality in Ambulatory Care Centers

Area of Science:

• Healthcare Operations Research
• Applied Machine Learning
• Patient Flow Management

Background:

• Late patient arrivals disrupt ambulatory care clinic schedules.
• Consequences include increased wait times, reduced care quality, and physician burnout.
• Rescheduling late patients can delay access to necessary medical care.

Purpose of the Study:

• To develop and compare machine learning (ML) models for predicting individual patient risk of late arrival.
• To identify critical predictors associated with patient lateness.
• To evaluate the utility of ML as a decision support tool for clinic management.

Main Methods:

• Utilized electronic medical record data from two ambulatory care facilities.
• Extracted and derived predictor variables relevant to patient punctuality.
• Compared four distinct ML algorithms: logistic regression, random forests, gradient boosting machine, and artificial neural networks.

Main Results:

• Machine learning models demonstrated accuracy in predicting patient lateness.
• No single ML model excelled across all metrics (predictive performance, training time, interpretability).
• Consistent critical predictors identified by all models were prior history of late arrivals, patient age, and afternoon appointment scheduling.

Conclusions:

• ML-based prediction of patient lateness is feasible and accurate.
• A tailored ML approach may be necessary, considering trade-offs between model performance and interpretability.
• Integrating ML tools into appointment systems can proactively manage and mitigate patient tardiness, improving clinic efficiency.