Predicting Anxiety in Routine Palliative Care Using Bayesian-Inspired Association Rule Mining.

Area of Science:

• Computational biology
• Medical informatics
• Data mining

Background:

• Anxiety is a common concern in palliative care.
• Routinely collected data offers a rich source for understanding patient conditions.
• Existing methods for anxiety classification in palliative care may lack accuracy or explainability.

Purpose of the Study:

• To develop and evaluate a novel knowledge extraction method for classifying anxiety in palliative care patients.
• To leverage Bayesian-inspired association rule mining on heterogeneous, routinely collected data.
• To compare the performance of the proposed method against the state-of-the-art.

Main Methods:

• Employed Bayesian-inspired association rule mining with lift and local support criteria for rule selection.
• Classified anxiety by assessing evidence supporting or rejecting anxiety for each patient.
• Evaluated predictive accuracy using the area under the receiver operating characteristic curve (AUC).

Main Results:

• Achieved a high predictive accuracy with an AUC of 0.89, significantly outperforming the previous state-of-the-art (AUC = 0.72).
• Extracted 55 atomic rules linking variables such as pain, nausea, and medications to anxiety.
• Identified discrepancies between data-driven rules and expert opinions, particularly regarding patient sex, suggesting potential knowledge gaps.

Conclusions:

• The proposed method provides an explainable and effective approach for predicting anxiety in palliative care using routinely collected data.
• The extracted association rules offer valuable insights into anxiety predictors and highlight areas for further investigation in palliative care.
• This Bayesian-inspired approach enhances the understanding of anxiety in palliative settings and demonstrates the utility of data mining in clinical decision support.