Their task was to consult two sets of clinical reports, each presenting the medical history of a cancer patient, and to answer ten questions about the patients’ condition. They were asked to perform this task in the context of a consultation they were about to have with a cancer patient who had been newly referred to them. Their task, then, was not to make a diagnosis or any other evaluation of the patient but to gather the important information that they would need before seeing the patient for the first time. The two patients were randomly selected from the repository of clinical records of 22,500 deceased patients FDA-approved Drug Library cell assay from the Royal Marsden Hospital in London. One (patient A) had a diagnosis of

breast cancer (breast carcinoma with bony metastases); her hospital records cover 32 consultations over four and a half years, and consist of 43 documents; the other (patient B) had a diagnosis of invasive ductal carcinoma, with records covering 8 consultations over one year and consisting of 11 documents (see Table 1 and Table 2). The records for each patient covers only the time they were treated at the Royal Marsden; patient A had received treatment elsewhere for five years prior and patient selleck chemicals llc B for one year. Although already anonymised by the hospital, the records were subject to further careful scrutiny by two experts

to remove all information that could identify the patient (e.g., occupation, consultant names, place names, etc.). Even so, all participants in our study were required to sign a non-disclosure agreement. The ten questions addressed issues that our clinical partners advised were key ones for a clinician about to see new cancer patient: • What is the presenting symptom/complaint? Each clinician was presented with a set of records for each patient. For one patient they were given the original hospital records (consisting of a collection of documents);

this mimicked the standard scenario for this website a doctor about to treat a new patient already diagnosed with cancer. For the other patient, they were given three summary records that were generated by the Report Generator: a full longitudinal summary, a summary from the perspective of clinical problems (e.g., cancer, anaemia or pain) and a summary from the perspective of curative procedures (e.g., chemotherapy, radiotherapy or surgery). Half of the subjects received the full records for Patient A and the summarised records for Patient B, and the other half received them the other way around. To avoid a biasing effect, half the subjects received the summaries before the full records, and the other half the other way around. All subjects received all questions in the same order. The clinicians read the records or summaries (in different sessions) and then answered the 10 questions. For each set, they were given 5 min for a ‘preliminary reading’ before proceeding to the questions.