This study examined whether staff working within a cancer centre treating patients with gastrointestinal malignancy routinely identified individuals from outpatients for referral to a dietitian. A nutrition screening tool is employed only for in-patient admissions. Height, current and usual weight were recorded prospectively in all patients referred for consideration of treatment. First appointment with the dietitian, first hospital admission, demographic and clinical details were obtained from hospital records. Time from first appointment to referral to a dietitian was examined. Between September 2002 and March 2004, 920 patients were included. Five hundred and seventeen patients had lost weight, of whom 223 patients had lost between 5% and 10% and 294 patients had lost more than 10% of their pre-morbid weight. Three hundred and twenty-seven patients (36%) were referred to dietitians. Twenty eight (9%) of referrals were made by staff in outpatients. Two hundred and ninety-nine were referred during or after an inpatient admission but only 39% of these occurred within the first seven days following admission. One third of patients with more than 10% weight loss were not referred for dietary assessment, even following admission. The likelihood of referral was significantly associated with the degree of weight loss (univariate analysis hazard ratio (HR) 1.75, 95% Confidence Interval (CI) 1.4-2.19, multivariate HR 1.65, 95% CI 1.22-2.23) and was independent of factors such as performance status and clinical setting. Few patients were identified early in their treatment for referral to a dietitian. Since most chemotherapy is now given on an outpatient basis, patients are unlikely to be referred if they do not require admission. This study suggests that an out-patient dietetic screening tool is urgently required. Such screening is likely to result in considerable improvements to the clinical care of cancer patients with weight loss.