Clinician-in-the-Loop Smart Home System to Detect Urinary Tract Infection Flare-Ups via Uncertainty-Aware Decision Support

TL;DR

This work addresses UTI flare-up detection in older adults with chronic conditions by marrying ambient-smart-home sensing with a clinician-in-the-loop framework. The core innovation is the Conformal-Calibrated Interval (CCI), a distribution-free uncertainty quantification method that yields prediction intervals with finite-sample coverage guarantees and an abstention option to avoid uncertain decisions, enhancing clinical trust. Evaluations on 117 labeled days from eight CASAS homes show that CCI improves recall and reduces abstention while producing tighter intervals than naive uncertainty baselines, without sacrificing accuracy or precision. A nurse survey corroborates the practical value of interval outputs for clinical decision-making, supporting the potential for deployment in remote health monitoring and proactive UTI management in aging populations.

Abstract

Urinary tract infection (UTI) flare-ups pose a significant health risk for older adults with chronic conditions. These infections often go unnoticed until they become severe, making early detection through innovative smart home technologies crucial. Traditional machine learning (ML) approaches relying on simple binary classification for UTI detection offer limited utility to nurses and practitioners as they lack insight into prediction uncertainty, hindering informed clinical decision-making. This paper presents a clinician-in-the-loop (CIL) smart home system that leverages ambient sensor data to extract meaningful behavioral markers, train robust predictive ML models, and calibrate them to enable uncertainty-aware decision support. The system incorporates a statistically valid uncertainty quantification method called Conformal-Calibrated Interval (CCI), which quantifies uncertainty and abstains from making predictions ("I don't know") when the ML model's confidence is low. Evaluated on real-world data from eight smart homes, our method outperforms baseline methods in recall and other classification metrics while maintaining the lowest abstention proportion and interval width. A survey of 42 nurses confirms that our system's outputs are valuable for guiding clinical decision-making, underscoring their practical utility in improving informed decisions and effectively managing UTIs and other condition flare-ups in older adults.

Figures (6)

• Figure 1: Overview of the clinician-in-the-loop smart home system for UTI flare-up detection. The system analyzes daily sensor data to extract behavioral markers and predict UTI occurrence in an uncertainty-aware manner. Clinicians, integrated into the loop, use these prediction/uncertainty intervals to guide decision-making and treatment planning.
• Figure 2: Illustration of interval-based decisions for three participants. The purple bands represent predicted intervals.
• Figure 3: SHAP Feature importance for all extracted features for the base model (averaged over 20 independent runs)
• Figure 4: Confusion matrices for Logistic Regression, Neural Network, and Random Guess. Values are averaged over 20 runs.
• Figure 5: Survey responses distribution of Likert scale responses for base model (point prediction) across six dimensions: clarity in conveying UTI likelihood, confidence in using to support diagnosis, trust in the information, potential influence on decision-making, help in communicating with caregivers/patients, and consideration for use in clinical workflow. Responses for the Naive Interval show a wider spread with higher disagreement, while those for CCI are predominantly positive with strong agreement.