For decades, chronic obstructive pulmonary disease (COPD) has been considered a relentlessly progressive disease in which the deterioration of lung function is associated with an increase in symptoms, interrupted only by periods of exacerbation. However, this paradigm of COPD severity based on FEV1 has been challenged by currently available evidence. So far, three main approaches, though with contradictory aspects, have been proposed in order to address the complexity of COPD as well as to develop appropriate diagnostic, prognostic and therapeutic strategies for the disease: 1) the use of independent, clinically relevant variables, 2) the use of multidimensional indices, and 3) disease approaches based on clinical phenotypes. Multivariable systems seem superior to FEV1 in predicting prognosis and defining disease severity. However, selection of variables available from current literature must be confronted with issues of medical practice. Future evidence will be needed to reveal their effective relationship with disease long-term prognosis and to demonstrate the most adequate cutoff values to be used in clinical settings. Multidimensional scores provide a good prognostic instrument for the identification of patients with a particular degree of disease severity. Clinical phenotyping can help clinicians identify the patients who respond to specific pharmacological interventions; however, there is some controversy about the phenotypes to select and their long-term implications. Although these approaches are not perfect, they represent the first step towards patient-centered medicine for COPD. In the near-future, these different approaches should converge towards one new field to focus on the better management of COPD patients.