The aim of the study is to determine the diagnostic value of (combinations of) signs, symptoms and simple laboratory test results for colorectal cancer in patients with rectal bleeding presenting in general practice. Initial complaints and findings were compared with the final diagnoses based on clinical follow-up after at least 1 year. Patients studied were those presenting overt rectal bleeding to the general practitioner (83 GPs in the South of the Netherlands). Outcome measures are sensitivity, specificity, predictive values, odds ratios and a prediction model derived from multiple logistic regression analysis. Age, change in bowel habit and blood mixed with or on stool show a statistically significant independent value in the discrimination between patients with a low and those with a high probability of colorectal cancer. Many other variables did not show predictive value. The prediction model has a sensitivitiy of 100% and a specificity of 90%. Although the number of patients with colorectal cancer is small (n = 9) it was possible to identify three characteristics which can be helpful in the prediction of presence or absence of colorectal cancer in general practice. Application of the model presented might prevent 90% of 'unnecessary' invasive diagnostic procedures for patients with rectal bleeding who do not have colorectal cancer (true negative). Testing the performance of the model in other general practice populations is recommended.