The Goldman algorithm revisited: Prospective evaluation of a computer-derived algorithm versus unaided physician judgment in suspected acute myocardial infarction

Background it has been nearly a decade since Goldman's computer-driven algo rithm to predict myocardial infarction was validated. Despite the potential to avoid admission of patients without acute myocardial infarction (AMI) t o the coronary care unit (CCU), the routine use of computer-generated proto cols has not been widely adopted. Methods Two hundred consecutive patients admitted to a university-affiliate community hospital with the suspected diagnosis of AMI as determined by ph ysicians without the aid of the Goldman protocol underwent a blinded prospe ctive evaluation to assess the performance of the Goldman algorithm in pred icting the presence of AMI. Over the same time period, the Goldman algorith m was applied by retrospective chart review in 762 patients with non-AMI ad mitting diagnoses. Prospective history, physical examination, and electroca rdiographic data were obtained within 24 hours of admission to the CCU by a physician blinded to each patient's clinical course. Retrospective chart r eviews were conducted for 762 patients with chest pain given with non-AMI d iagnoses. Results The diagnosis of AMI was confirmed in 68.5% (137/200) of patients w ith suspected AM I admitted to the CCU. In prospective parallel evaluations the Goldman algorithm predicted the presence of AMI in 167 (83.5%) of thes e 200 patients. All 137 confirmed patients with AMI were correctly identifi ed by the Goldman algorithm. All major in-hospital complications occurred i n the 137 patients who were diagnosed as having AMI. Of the 762 patients wi th chest pain with non-AMI diagnoses, only 27 (3.5%) sustained an AMI. The Goldman algorithm predicted the presence of AMI in 85% (23/27) of these pat ients. Adherence to the use of Goldman's algorithm in the triage of chest p ain could have prevented 16.5% of CCU admissions for AMI. Conclusions Routine adherence to the Goldman algorithm for the evaluation o f patients with acute chest pain could have decreased the number of CCU adm issions for suspected AMI by 16.5%. Because major in-hospital complications occurred only in patients with AMI, this strategy would result in signific ant cost savings to our health care system without jeopardizing patient saf ety.