Resting-State Heart Rate Variability and Symptom-Level Conditional Associations in Hospitalized Patients with Musculoskeletal Pain: An Exploratory Cross-Sectional Network Analysis

Abstract

Background Pain, depressive and anxiety symptoms, sleep disturbance, and autonomic regulation may coexist in hospitalized patients with musculoskeletal pain. This study integrated item-level symptoms with resting-state photoplethysmography-derived heart rate variability (HRV) to characterize their regularized association structure. Methods This exploratory cross-sectional study retained 172 adults hospitalized with musculoskeletal pain-related conditions. Pain-duration information was available for all participants, and 169 complete cases entered the primary 58-node network. Regularized partial rank associations were estimated from Spearman correlations using EBICglasso (γ = 0.50) without additional thresholding. Sensitivity analyses examined participants with pain lasting at least 3 months and the influence of overlapping insomnia items. Results Median pain duration was 4.00 months (IQR, 0.57–24.33; range, 0.50–365.00), and 91 of 172 participants (52.9%) had pain lasting at least 3 months. The primary network retained 332 of 1653 possible edges. Prominent within- and cross-domain connections involved pain intensity and mood, gastrointestinal, cognitive, somatic, and insomnia symptoms. HRV indices formed a comparatively distinct physiological cluster; only 24 of 441 possible HRV–symptom edges were retained, all with absolute weights below 0.029. Strength and expected-influence estimates showed low bootstrap stability (CS coefficients = 0.000 and 0.207, respectively). In the ≥3-month subgroup, correlations with the primary network were 0.868 for edge weights, 0.552 for strength, 0.729 for expected influence, and 0.925 for bridge expected influence. Network structure and bridge patterns were partly preserved, but centrality rankings were more duration-sensitive. After overlapping insomnia items were removed, the global structure remained similar, but pain-related sleep interference fell from fourth to 27th in bridge expected influence. Conclusion Item-level network analysis identified closely connected pain and psychological symptom patterns, whereas short-term PPG-derived HRV showed sparse and weak direct network connectivity with clinical symptoms. Sleep-related bridging was partly attributable to overlapping insomnia content. These exploratory findings support multidimensional inpatient assessment but require validation in larger longitudinal studies.