Objective To build a prediction model of X-ray-positive, based on clinical symptoms,signs and hemogram. Methods A total of 1080 suspicious CAP cases from outpatients of our hospital are studied. General appearance, respiratory rate, cough severity, cough duration, fever spike, fever duration, rales, rhonchi, wheezes, white blood cell count, neutrophil percentage, lymphocyte percentage, monocyte percentage, oxyphile leukocytes percentage, C-reaction protein and chest X ray representation of the cases are analysed by logistic regression. Results 11 variables enter, composed of respiratory rate, cough severity, fever duration, rales, rhonchi, wheezes, neutrophil percentage, lymphocyte percentage, monocyte percentage, oxyphile leukocytes percentage, C-reaction protein. Their regression coefficients are , respectively, 1.477、2.147、0.623、1.568、-0.870、1.422、-0.627、-0.691、-0.621、-0.936、1.122，with the constant of 52.103. General appearance, cough duration, fever spike, white cell count are rejected. Conclusion Pneumologues may need revaluate some variables that were important according to experience to make an impersonal diagnosis of CAP. Aided by the prediction model, they could obtain the probability of a positive result if a patient is sent for a chest X-ray and therefore most patients may avoid X-ray exposure.