The association between physical behaviour and chronic low back and neck pain: Acquiring data from the Norwegian HUNT4 Study

In the 4th survey of the Nord-Trøndelag Health Study (HUNT4), we collected 7-days physical behaviour data in 33 000 adults. Two accelerometers attached to the skin on the right thigh and low back were used to collect the data. By utilising state-of-the-art machine-learning techniques, we have developed a data model that classifies time with sitting, standing, walking, running, cycling and lying down with an overall accuracy of about 95%. Along with the recordings of physical behaviour, a range of health parameters are collected in HUNT4, including pain in low back and neck. The participants also filled in a simple diary to indicate start/end of work. In the current project, we have focused on quality assurance and cleaning of the data (e.g., some participants removed one or both sensors before the end of the 7-day recording, some sensors had malfunction etc.). Moreover, we have processed the data with our most up-to-date model to classify time with different physical behaviours and extracted a limited number of other health data relevant for prognostic models related to neck and low back pain. These data will be used to expand the ongoing EU-funded Back-UP project. The overall aim of Back-UP is to create a prognostic model to support more effective and efficient management of neck and low back pain. Data from HUNT4 will enable us to improve the prognostic model and thereby strengthen our contribution to the project. Furthermore, the data on objective measurements of physical behaviour in HUNT4 are unique, giving us a competitive advantage that increases the probability of attracting further EU funding.

The data from HUNT4 has strengthened our contribution to the Back-UP project. Moreover, we are currently working on a new Horizon 2020 proposal where the unique data from HUNT4 play a major role.

The funding applied for in the current proposal will be used to acquire data on about 10 000 participants in the ongoing 4th survey of the Nord-Trøndelag Health Study (HUNT4). The data will be used to expand the ongoing EU-funded Back-UP project. In HUNT4, we are collecting 7-days physical behaviour data with two accelerometers attached to the skin on the right thigh and low back. By utilising state-of-the-art machine-learning techniques, we have developed a data model that classify time with sitting, standing, walking, running, cycling and lying down with an overall accuracy of about 95%. Along with the recordings of physical behaviour, a range of other relevant health parameters are collected in HUNT4, including pain in low back and neck. The HUNT4 data collection started in September 2017 and runs until February 2019. Based on the current participation rate, we will obtain objective measurements of physical behaviour from about 33 000 people. Thus, these data provide and a unique opportunity to study the association between physical behaviour and neck and/or low back pain. To establish the HUNT4 dataset, it is necessary to 1) perform a quality check and clean the data (e.g., some participants removed one or both sensors before the end of the 7-day recording, some sensor may have malfunctions etc.), 2) process the data with our most up-to-date model to classify time with different physical behaviours, and 3) extract a limited number of other health data relevant for the prognostic model. We foresee that 3-4 person-months are required to create the HUNT4 dataset. The dataset will have significant impact on the quality and potential impact of the prognostic models developed in Back-UP. Moreover, the uniqueness of the physical behaviour data in HUNT4, and the potential for ground-breaking research based on these data, will strengthen our possibilities for attracting further EU funding.