10.3389/fpsyg.2012.00132

Vollständiger Abstract

Worum geht es in dieser Arbeit?

<h4>Background</h4>Patients in inpatient rehabilitation spend significant time of their day in sedentary activities, limiting recovery potential. Current monitoring relies on manual, intermittent nursing observations that are labor-intensive and produce fragmented data. Thermography-based artificial intelligence (AI) monitoring offers a privacy-preserving, continuous alternative; however, its impact on rehabilitation engagement when combined with patient-facing feedback has not been rigorously evaluated. Conventional pre-post and parallel-group designs are poorly suited for ward-level digital health interventions because of contamination risks, ethical constraints on withholding technology, and failure to exploit dense within-patient time-series data.<h4>Objective</h4>This study aims to evaluate the effectiveness of the PreSAGE thermography-based AI activity monitoring and feedback dashboard system (the DREAMT intervention) in enhancing patient physical activity levels, rehabilitation outcomes, and nursing workflow, and to explore the experiences and contextual factors influencing uptake.<h4>Methods</h4>This is an explanatory sequential mixed-methods study embedding a randomized multiple baseline design across participants (RMBD) for the quantitative component and semi-structured interviews for the qualitative component. Conducted over 24 months at Tan Tock Seng Hospital, Singapore, the study proceeds through three phases: (1) AI algorithm development via thermographic data collection and manual labelling (<i>n</i> = 50); (2) algorithm validation and user-centered dashboard co-design; and (3) RMBD evaluation (<i>n</i> = 50), in which each patient undergoes continuous monitoring from admission and the feedback dashboard is activated at a randomly assigned day (3-7 post-admission), with each patient serving as their own control. The primary outcome is daily total active time. Secondary outcomes include exercise frequency, Functional Independence Measure scores, length of stay, nursing workload, and dashboard utilization. Quantitative data will be analyzed using piecewise multilevel growth models. Qualitative interviews with 10-15 patients, 2-3 ward sisters, and 4-6 clinicians will be analyzed using reflexive thematic analysis, with integration via a joint display framework.<h4>Conclusions</h4>This protocol outlines a methodologically rigorous evaluation of a novel privacy-preserving AI monitoring and feedback system. The RMBD utilizes continuous within-patient data to yield stronger causal inference than conventional designs while remaining feasible in a single-site setting. If effective, DREAMT could shift rehabilitation monitoring from episodic, subjective observation to continuous, personalized, data-driven care.

Abstract: PubMed · Datensatz