Is Internet assistance with acute paediatric diagnosis feasible? A pilot evaluation of the 'Isabel' diagnostic tool.
P Ramnarayan , RC Tasker , A Reeve , G Wilson , S Nadel , J Britto

Background: Medical error leads to nearly 40,000 deaths in the UK each year. Lack of easy access to relevant point-of0care information may contribute to diagnostic error by emergency medical staff.

Aims: To evaluate the performance of an Internet-based diagnostic tool (Isabel Diagnostic Tool, IsDT; Charity Site, www.isabel.org.uk) in generating differential diagnoses for a given set of clinical features.

Methods: Junior medical staff from a paediatric Infectious Disease unit submitted forms with a set of clinical features and the final diagnosis/es of patients. These clinical features were entered into the IsDT and the output was compared to expected diagnosis/es. The IsDT output was derived from software (Autonomy) which understands, categorises and searches for information by context, rather than word. The software had been primed with Nelson's Textbook of Paediatrics (15th edition). The presence of the correct diagnosis (or at least 50% of the differential provided) in the first 10 choices of the IsDT output was considered accurate.

Results: Out of a total of 81 forms used for analysis, 46 (56.8% suggested a single diagnosis as the appropriate answer and 35 (43.2%) listed a differential as possible answers. The IsDT provided accurate results in 61 (75.3%) out of the 81 forms, and in 40 (87%) out of the 46 forms that asked for a single diagnosis.
Conclusions: The IsDT provided clinically relevant and reasonable differential diagnoses, with an accuracy of 75-87%. We believe that extending the use of such a tool to a near-patients setting would greatly enhance the emergency paediatrician's diagnostic skills and serve as a real time educational aid.