AI-driven Optimisation of Quality of Recovery (QoR) in Remote Patient Monitoring

Researchers developed QoR-compact, a five-question alternative to the 15-item Quality of Recovery survey for remote patient monitoring, achieving statistically comparable predictive accuracy (AUC-ROC 0.968) while reducing patient burden by two-thirds. The streamlined tool addresses low compliance rates in daily post-surgical assessments while maintaining clinical reliability for predicting recovery outcomes.

Remote patient monitoring systems face a critical usability challenge: while comprehensive health surveys provide rich diagnostic data, patients often abandon them due to completion burden. This research tackles that tension directly by engineering a minimal viable instrument from the QoR-15 survey, a gold-standard recovery metric originally designed for occasional use that now faces daily administration in post-operative contexts. The five-item QoR-compact selection—spanning physical comfort, psychological well-being, pain severity, and emotional state—preserves predictive power while reducing friction.

The study's methodology demonstrates disciplined AI-driven optimization: exhaustive evaluation of 3,003 possible five-question combinations against real deployment data where baseline compliance was only 55% over 30 days. The resulting tool matches full-form performance in predicting readmission events and recovery severity, suggesting that information density matters less than targeted feature selection in clinical prediction tasks. This aligns with broader healthcare AI trends emphasizing practical implementation over theoretical maximization.

The clinical implications are substantial. Patient adherence directly impacts monitoring efficacy; a simpler daily input could materially improve data collection rates and outcome tracking. For healthcare technology developers and remote monitoring platforms, QoR-compact offers a template for optimizing survey instruments without sacrificing predictive validity. The authors appropriately note this remains preliminary work requiring prospective validation on larger cohorts and external datasets before clinical deployment.

The research exemplifies how constraint-driven engineering—setting a specific reduction target and systematically evaluating alternatives—can produce clinically useful tools. Future validation will determine whether the improved completion rates hypothetically enabled by reduced burden translate to actionable advantages in real-world deployments.

• →QoR-compact achieves 0.968 AUC-ROC performance with five items versus 15, matching full-instrument predictive accuracy for post-operative recovery monitoring.
• →Patient compliance in remote monitoring was only 55% with the full survey, suggesting significant burden reduction could improve data collection rates.
• →The five selected items span physical and psychological recovery axes, including rest quality, comfort, well-being, pain, and anxiety measurements.
• →External validation on larger cohorts is required before clinical implementation despite promising preliminary results.
• →This work demonstrates AI-driven feature optimization can enhance usability in healthcare without sacrificing predictive performance.