Graded back pain revisited – Do latent variable models change our understanding of severe back pain in the general population?

Abstract 

Back pain severity has extensively been targeted in clinical and epidemiologic studies. However, despite the importance of a valid pain severity grading its adequate conceptualization in the general population has received comparatively little attention. The potentially misleading influence of measurement error remains unclear. Latent variable models allow for a versatile assessment of disease severity and will be applied to propose a model-based grading of back pain. This cross-sectional postal survey was carried out in Germany between 2003 and 2004 to address back pain severity in the general adult population. 8756 subjects, aged 18–75years, provided data on measures of pain intensity and disability. Latent class analysis and confirmatory factor analysis were used to assess and compare categorical and dimensional representations of back pain severity. The results show that beyond differences in their location on a severity continuum, the subjects did not report markedly different pain intensity/disability profiles. Our analyses disconfirmed the presence of a sizeable high pain intensity, low disability subgroup. A comparison of the different latent variable models yielded a usable classification into five severity subtypes. This classification showed statistically significant and clinically important associations to health-related variables. Our results confirm the high burden of back pain in the general population but suggest a different categorization of those with severe back pain. This entails consequences on how to best target this important health problem from a public health perspective.

Keywords: Back pain, Pain severity, Pain intensity, Disability, Graded back pain, Population based, Cross-sectional, Confirmatory factor analysis, Latent class analysis, General population