The relation between musculoskeletal pain and sickness absence was tested in an adult county population. Maximal explained variance in absence from work due to chronic musculoskeletal pain (sickness absence) was tested in a model in which subjective health was expected to mediate the associations between such pain and dysphoria, respectively, and work efficacy. In turn, work efficacy was expected to mediate the link between subjective health and sickness absence. All the residents in the County of Nord-Trøndelag, Norway, aged 20 and older, were invited to take part in a public health survey during 1995-97 (HUNT-2), and 66,140 (71.2%) participated. Prevalence of musculoskeletal pain, dysphoria, subjective health and work efficacy were assessed, as well as sickness absence last year due to musculoskeletal pain. The model test was performed by use of the LISREL procedure based upon data from 30,158 employees reporting chronic musculoskeletal pain last year. The measurement model fitted the data well: χ2 = 9075, df = 52, p < .0004, Critical N = 1041, RMSEA = 0.038, CFI = 0.99, SRMR = 0.020. The structural model fitted the data equally well, and the best prediction of sickness absence was obtained with lower back pain, upper and lower extremity pain, as well as dysphoria as the primary variables affecting subjective health that, in turn, was the convergent predictor of work efficacy that, finally, best explained the variance in sickness absence (56%). The data supported an indirect sequence of complaint-health-efficacy (CHE-model) as the best predictor of sickness absence due to musculoskeletal pain.