Today an evidence-based medicine has given number of medical practice clinical guidelines and protocols. Clinical guidelines systematically assist practitioners with providing appropriate health care for specific clinical circumstances. However, a significant number of guidelines and protocols are lacking in quality. Indeed, ambiguity and incompleteness are more likely anomalies in medical practices. From last few years, many researchers have tried to address the problem of protocol improvement in clinical guidelines, but results are not sufficient since they believe on informal processes and notations. Our objective is to find anomalies and to improve the quality of medical protocols using well known formal techniques, such as Event B. In this report, we use the Event B modeling language to capture guidelines for their validation. We have established a classification of possible properties to be verified in a guideline. Our approach is illustrated with a guideline which published by the National Guideline Clearing House (NGC) and AHA/ACC Society. Our main contribution is to evaluate real-life medical protocols using refinement based formal method like Event B for improving quality of the protocols. Refinement based formalisation is very easy to handle any complex medical protocols. For this evaluation we have selected a real-life reference protocol (ECG Interpretation) which covers a wide variety of protocol characteristics related to the several heart diseases. We formalise the given reference protocol, verify a set of interesting properties of the protocol and finally determine anomalies. Our main results are: to formalise an ECG interpretation protocol for diagnosing the ECG signal in optimal way; to discover an hierarchical structure for the ECG interpretation efficiently using incremental refinement approach; a set of properties which should be satisfied by medical protocol; verification proofs for the protocol and properties according to the medical experts; and perspectives of the potentials of this approach. Finally, we have shown the feasibility of our approach for analysing medical protocols.