The Isabel Differential Diagnostic Tool: Results of evaluation on a set of real-life clinical case scenarios
P Ramanarayan, A Tomlinson, J Britto

Background: Isabel is a decision-support system on the Internet that incorporates the use of a novel differential diagnostic tool (IDDT), powered by textual pattern recognising software searching standard paediatric textbooks. In response to entering clinical features from a patient, it produces a list of differential diagnoses for the doctor to consider.

Aims: To assess the accuracy of the IDDT in a variety of clinical scenarios drawn from real-life patients.

Materials and methods: Data was collected from October-December 2000 on an unselected group of children presenting to the emergency departments in four hospitals ( 2 teaching and 2 district general). This included the age group, presenting clinical features, results of initial investigations, the examining doctors' working diagnosis and the final diagnosis as recorded in the discharge summary. Presenting clinical features were entered into the IDDT by one investigator not involved in the data collection and the 15 differential diagnoses generated by the IDDT were recorded. The proportion of case in which the final diagnosis were present in the IDDT list was calculated (binary measure of diagnostic quality).

Results: A total of 114 cases were analysed from a total of 144 forms (the rest ineligible due to incomplete data collection). A total of 55 unique diagnoses were represented in the dataset. In 77% of cases, the final diagnosis was present in the IDDT list.

Conclusions: The IDDT showed a clinically reasonable degree of accuracy in generating differential diagnoses and the final diagnosis in a range of real-life clinical presentations. Further studies are underway to measure the clinical impact of the IDDT in the hands of real users.