A novel method of interface design--user driven, evidence based experimental design--was developed which approximates the usual clinical way of maturing science and technology in the healthcare domain. The method is user-driven and the clinician remains in control of gathering and evaluating evidence of relevance to the project--as well as specifying the details of the user interface. Information not obtainable from the literature was gained experimentally and used to achieve a deeper understanding of the problem before the design phase. The design was subsequently validated experimentally by ordinary users with no connection to the software or design team. After applying this method to the problem of gaining a satisfactory clinical overview of a single patient's record, we recommend that clinical IT interfaces have clinical logic, sufficient complexity, and are well structured. Developers should use computer power to support "building blocks" such as anatomical problem lists and summaries of history, status and treatment, personal notes, and should support clinical browsing using text and graphics.