Safety reasoning in medical decision support

In protocol based medical care, protocols and guidelines are employed to standardise some aspect of treatment: to regularise the management of a disease or to enforce a particular protocol during a clinical trial of a new therapeutic agent. Some guidelines do not need to be followed rigorously, whereas adherence to a clinical trial protocol is essential if statistical analysis of trial results is to be scientifically valid or if the safety of patients is to be ensured. The use of computers in the application of clinical guidelines and protocols is becoming more widespread and improved compliance and more complete data capture for subsequent analysis of clinical trial results have been reported. In particular, decision support systems for protocol based care in oncology have been studied for more than 20 years. The safety aspects of such systems were studied in an earlier project and some of the resulting generic safety principles were encoded in OaSiS, a prototype knowledge based system supporting protocol based care in cancer management. A small number of computer aids for designing new clinical trials, therapy plans and associated protocol documents have been implemented. Notable examples are DaT and OPAL. We summarise work on the empirical derivation, formalisation and implementation of safety reasoning for cancer management and describe the reuse of the safety knowledge obtained to generate specific safety clauses semi automatically during clinical trial design and protocol document assembly.