Physical Activity Levels of Patients with Chronic Low Back Pain and Central Sensitization: Insights from A Machine Learning Method

Research output: Contribution to conferenceAbstractAcademic


Chronic low back pain (CLBP) is the leading cause of high levels of disability. Central sensitization (CS) is often present in CLBP and is defined as the increased responsiveness to noxious and non-noxious stimuli. It is reported that physical activity (PA) can reduce the risk of disability, and is often recommended in the management of CLBP. However, the evidence of the relation between PA levels and CLBP is inconsistent and even less is known about association with CS.
This study aimed to investigate PA patterns in patients with CLBP and low and high CS using an unsupervised machine learning method.
Based on the Central Sensitization Inventory, 42 patients were divided into: low CS (N=23; CLBP-) and high CS (N=19; CLBP+). With an accelerometer sensor attached at the right hip, PA during daily life was recorded. For each patient, 4 days of data were used for analyses. Accelerometer data were corrected for gravity and from the x, y, z axes, the vector magnitude was calculated. For each group, a Hidden semi Markov Model (HSMM) was made to examine PA patterns based on their vector magnitude. To determine the number of hidden states (PA patterns), Bayesian Information Criterion (BIC) was used. Differences in state duration time, state occupation time in day, state transitions outcomes were assessed with independent t-tests, and state composition with Jensen–Shannon divergence (JSD).
Indicated by BIC scores, 5 hidden states HSMMs were selected. JSD showed no difference between two HSMMs’ hidden state compositions. These states were defined as: long-time sedentary (e.g., sleeping), sedentary (e.g., desk work), light activities (e.g., vacuuming), light locomotion (e.g., slow walking), moderate-vigorous activities (e.g., fast walking, running). Differences in overall PA levels between CLBP- and CLBP+ were non-significant (p>0.05). Differences in state transitions showed that between light and moderate activity, CLBP- more frequently have a sedentary state.
This study observed that CLBP- and CLBP+ have no differences in overall PA levels, but have differences in the distribution of PA patterns over the course of performing light and moderate activity. CLBP- rested more frequently between light and moderate activities while CLBP+ maintained light and moderate activities longer.

Original languageEnglish
Publication statusPublished - 23-Jun-2021
EventInternational Conference on Ambulatory Monitoring of Physical Activity and Movement (ICAMPAM) - online
Duration: 23-Jun-202124-Jun-2021


ConferenceInternational Conference on Ambulatory Monitoring of Physical Activity and Movement (ICAMPAM)

Cite this