A decision support system for the management of oral hypoglycaemic therapy in type II diabetes was evaluated. The ruleset contained therein forms the basis of a prototype computer programme, but in order to assess the robustness of the individual rules, it was decided it was necessary to use a paper-based form of the ruleset. A nurse with no previous experience of managing type II diabetes was trained to use the system and then undertook the exclusive management of half of all new type II diabetics, from a district population of 300,000, over a 16-month period. General practices within this area were divided into two groups, study and control, matching for size, geographical area and standards of existing diabetes care. Patients (n = 102) from the study group practices were then assigned to her care. Those patients (n = 116) in the control group of practices were treated according to their normal procedures. The decision support system for oral hypoglycaemic therapy was based on the following criteria: the current type of treatment (six levels); current glycaemic control (HbA1 and FBS)-whether improving, steady or worsening; and weight-%IBW, whether rising, steady or falling. Each of these parameters was carefully defined on the basis of established practice and clinical experience. Patients after initial education were seen at their usual clinic by the nurse only, on a monthly basis, until satisfactory glycaemic control was established and thereafter reviewed 3 monthly. She was also responsible for ensuring the organisation of Diabetes Annual Review procedures. The medical records of the control group patients were examined at the end of the study and data on glycaemic control and Annual Reviews extracted. In the study group 98% patients achieved HbA1 levels within the normal range and all patients had full annual reviews performed. The control practices achieved much poorer degrees of metabolic control (P < 0.01) and completed fewer annual reviews. The study group did not demonstrate a significantly increased frequency of clinical hypoglycaemia consequent upon better blood sugar control. No exceptions to the ruleset, as initially defined, were detected. In conclusion, this decision support system was successful at achieving standards of diabetes control and care equal to or better than conventional structures of diabetes care. Implementation of such a system, on a simple computer platform, could greatly assist and possibly improve diabetes management in general practice.