In the Netherlands a prediction model has been developed to detect incident heart failure among symptomatic patients in general practice using routine health care data. The authors present a prognostic model that should help predict the development of heart failure within 10 years in community people aged 35 years or over with shortness of breath and/or oedema. They used longitudinal routine care data for establishing heart failure. However, there are over 30 causes of shortness of breath, and both patient and general practitioner primarily want to know the cause of it (the diagnosis), because only then adequate treatment can be given. Thus, a cross-sectional diagnostic model is needed for these patients, not a prognostic one. Moreover, routine care data are insufficient for establishing heart failure because many patients will be misclassified as 'no heart failure'. Finally, there are over 1300 prediction models for cardiovascular disease and validation of an existing model should be preferred above developing a new one.