Is chronic low back pain related to sleep disturbances? A self-organizing maps for unsupervised machine learning-based patient phenotyping.
AI interpretation is pending for this paper.
Open original publication →What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
BACKGROUND CONTEXT: Chronic low back pain (CLBP) is a multifactorial condition and a leading cause of disability worldwide. Among the various contributors to symptom persistence, sleep quality has emerged as a critical yet underexplored factor. Evidence suggests a bidirectional relationship between poor sleep and pain intensity, disability, and psychological distress in individuals with CLBP. PURPOSE: To explore the relationship between sleep quality and CLBP using machine learning techniques, specifically the Self-Organizing Map (SOM) algorithm, while accounting for demographic, clinical, psychosocial, and occupational factors. STUDY DESIGN/SETTING: This was a cross-sectional study conducted at Campus Bio-Medico Hospital Foundation in Rome, Italy, between July 2024 and January 2025. PATIENT SAMPLE: The study included 279 adult working patients (aged 18-65) with a clinical diagnosis of CLBP due to degenerative causes, all of whom were candidates for conservative treatment. Patients with cancer, trauma, spinal deformities, or infections were excluded. OUTCOME MEASURES: Pain intensity was assessed using the Visual Analog Scale (VAS), while sleep quality was measured with the Pittsburgh Sleep Quality Index (PSQI), a validated tool capturing various dimensions of sleep disturbance. Disability related to low back pain was evaluated using the Oswestry Disability Index (ODI). Depressive symptoms were identified using the Patient Health Questionnaire-2 (PHQ-2). Functional outcomes included assessment of work capacity through the Work Ability Index (WAI) and activity limitations via the Nordic Musculoskeletal Questionnaire, focusing on how often back pain interfered with usual daily activities. METHODS: Validated questionnaires and anthropometric data were collected. SOM, an unsupervised machine learning algorithm, was applied to detect clusters based on multidimensional patient data. Subsequent k-means clustering was used to define patient subgroups. Statistical comparisons between clusters were performed using Kruskal-Wallis and chi-squared tests. Multiple imputation addressed missing data. RESULTS: A total of 4 distinct patient clusters were identified. Cluster 1 included physically active individuals with moderate pain and no depression. Cluster 2 consisted of highly educated women with sedentary jobs and minimal disability. Cluster 3 showed the worst outcomes: high pain intensity, severe disability, depression, and the poorest sleep quality. Cluster 4, composed entirely of men, showed high work ability, good sleep, low pain, and no depressive symptoms. Poor sleep quality correlated strongly with depressive symptoms, high pain, disability, and reduced work capacity. These associations persisted even when sleep was excluded from clustering inputs, confirming its predictive linkage. CONCLUSIONS: Sleep quality is a key determinant in the clinical profile of CLBP patients. Poor sleep is closely associated with worse pain perception, psychological distress, and functional impairment. SOM analysis revealed hidden patterns not captured by traditional methods. Integrating sleep evaluation into multidisciplinary care models may improve outcomes for CLBP patients. Longitudinal studies are needed to clarify the directionality of these effects.