Chronic obstructive pulmonary disease (COPD) is characterised by high morbidity and mortality. It remains unknown which aspect of lung function carries the most prognostic information and if simple spirometry is sufficient. Survival was assessed in COPD outpatients whose data had been added prospectively to a clinical audit database from the point of first full lung function testing including spirometry, lung volumes, gas transfer and arterial blood gases. Variables univariately associated with survival were entered into a multivariate Cox proportional hazard model. 604 patients were included (mean ± SD age 61.9 ± 9.7 years; forced expiratory volume in 1 s 37 ± 18.1% predicted; 62.9% males); 229 (37.9%) died during a median follow-up of 83 months. Median survival was 91.9 (95% CI 80.8-103) months with survival rates at 3 and 5 years 0.83 and 0.66, respectively. Carbon monoxide transfer factor % pred quartiles (best quartile (>51%): HR 0.33, 95% CI 0.172-0.639; and second quartile (51-37.3%): HR 0.52, 95% CI 0.322-0.825; versus lowest quartile (<27.9%)), age (HR 1.04, 95% CI 1.02-1.06) and arterial oxygen partial pressure (HR 0.85, 95% CI 0.77-0.94) were the only parameters independently associated with mortality. Measurement of gas transfer provides additional prognostic information compared to spirometry in patients under hospital follow-up and could be considered routinely.