The aim of this study was to describe the clinical course of patients with chronic, non-specific neck pain undergoing a public health covered, exercise-based rehabilitation program and to identify predictors of poor outcome. A prospective cohort study was carried out on patients with non-specific neck pain (6 months or longer), referred by their general practitioner to a 6-session program, including education and individually tailored exercise. The primary outcome measure for the course of neck pain was the Northwick neck pain questionnaire (NPQ) administered on baseline, discharge, and 1 year from discharge. Poor outcome was defined as NPQ score improving <30% (minimal clinically important difference-MCID-NPQ). The potential predictors included demographics, general health and psychological factors, neck pain history, and the clinical features described by NPQ. From January 2008 to June 2009, 212 patients were consecutively assessed for eligibility: 178 were enrolled and 162 completed follow-up (mean age = 65.3; 75% women). Baseline NPQ average score (40.7 + 17.1) improved by MCID on discharge (26.1 + 16.3) and at 1 year (28.5 + 17.3%). The poor outcome was reported by 45% patients on discharge and by 56% at follow-up. Pain-related medication intake independently predicted poor short- (OR 4.24; 95% CI 1.83-9.84; p = 0.001) and long-term (OR 2.69; 95% CI 1.19-6.06; p = 0.017) outcome, and catastrophizing (OR 2.91; 95% CI 1.31-6.48; p = 0.009) predicted poor outcome at 1 year. Our cohort of patients with chronic neck pain undergoing an exercise-based rehabilitation program reported improvement by or beyond MICD-NPQ in 55% cases on discharge and in 44% cases at 1 year. Poor outcome was predicted by pain-related medication intake in the short and long term, and by catastrophizing in the long term.