OBJECTIVES: To develop a simple dichotomous scale for the assessment of RLS severity to be used for facilitating therapeutic decisions. METHODS: A prospective epidemiological study was conducted in primary healthcare in France. First, patients fulfilling the RLS diagnostic criteria from the International Restless Legs Study Group were identified. Secondly, severity of identified RLS subjects was assessed using the IRLS rating scale. This scale includes ten items, each rated on a five point Lickert scale from zero to four (“non” to “very severe”) and summed in a global score from zero to 40 (0-20 : non-severe ; 21-40: severe). Using a segmentation analysis applied to the rating scale items and other measures of RLS impact, we attempted to find a small number of items able to distinguish severe from non-severe RLS patients as the rating scale would have done. RESULTS: A total of 537 patients were analyzed. Three items came up from the segmentation analysis, consolidated into “RS-3”. All three were part of the rating scale. They dealt with: overall RLS severity, overall sleep disturbance and mood disturbance. The decision rule was that an answer zero,one or two (ie. none, mild or moderate) to at least one of the three items assigned an RS-3 grade of 'non-severe'. The sensitivity and specificity of RS-3 were 82% and 95% respectively. The positive and negative predictive values (PPV and NPV) were 92% and 88%. Sensitivity analyses with thresholds of 15, 25 and 30 were performed without improving sensitivity and specificity. The reproducibility of the RS-3 was assessed in another RLS population (731 subjects from the INSTANT Study) with sensitivity / specificity of 88% and 91%; and PPV and NPV of 73% and 97%, respectively. CONCLUSIONS: The metrology of the RS-3 scale appears to be strong enough to be used in larger sample.