Investigate the feasibility of identifying a well-defined treatment group and a comparable reference group in clinical register data.

There is insufficient knowledge on how to avert neck/back pain from turning chronic or to impair work ability. The Swedish Government implemented a national multimodal rehabilitation (MMR) programme in primary care intending to promote work ability, reduce sick leave and increase return to work. Since randomised control trial data for effect is lacking, it is important to evaluate existing observational data from clinical settings.

We identified all unique patients with musculoskeletal pain (MSP) diagnoses undergoing the MMR programme in primary care in the Skåne Health care Register (n = 2140) during 2010–2011. A reference cohort in primary care (n = 56 300) with similar MSP diagnoses, same ages and the same level of sick leave before baseline was identified for the same period. The reference cohort received ordinary care and treatment in primary care. The final study group consisted of 603 eligible MMR patients and 2874 eligible reference patients. Socio-economic and health-related baseline data including sick leave one year before up to two years after baseline were compared between groups.

There were significant socio-economic and health differences at baseline between the MMR and the reference patients, with the MMR group having lower income, higher morbidity and more sick leave days. Sick leave days per year decreased significantly in the MMR group (118–102 days, P < 0.001) and in the reference group (50–42 days, P < 0.001) from one year before baseline to two years after.

It was not feasible to identify a comparable reference group based on clinical register data. Despite an ambitious attempt to limit selection bias, significant baseline differences in socio-economic and health were present. In absence of randomised trials, effects of MMR cannot be sufficiently evaluated in primary care.