A valid method for classifying chronic pain patients into more homogenous groups could be useful for treatment planning, that is, which treatment is effective for which patient, and as a marker when evaluating treatment outcome. One instrument that has been used to derive subgroups of patients is the Multidimensional Pain Inventory (MPI). The primary aim of this study was to evaluate a classification method based on the Swedish version of the MPI, the MPI-S, to predict sick leave among chronic neck and back pain patients for a period of 7 years after vocational rehabilitation. As hypothesized, dysfunctional patients (DYS), according to the MPI-S, showed a higher amount of sickness absence and disability pension expressed in days than adaptive copers (AC) during the 7-years follow-up period, even when adjusting for sickness absence prior to rehabilitation (355.8days, 95% confidence interval, 71.7; 639.9). Forty percent of DYS patients and 26.7% of AC patients received disability pension during the follow-up period. However, this difference was not statistically significant. Further analyses showed that the difference between patient groups was most pronounced among patients with more than 60days of sickness absence prior to rehabilitation. Cost-effectiveness calculations indicated that the DYS patients showed an increase in production loss compared to AC patients. The present study yields support for the prognostic value of this subgroup classification method concerning long-term outcome on sick leave following this type of vocational rehabilitation.