D S Bell1. 1. Department of Medicine, UCLA School of Medicine, USA.
Abstract
BACKGROUND: Many patients who need coronary angiography fail to get it and they have decreased survival as a result. This study demonstrates the use of decision analysis to predict the survival value of an alerting system for necessary angiography. METHODS: Data on the use of angiography and survival after myocardial infarction (MI) were taken from a published cohort study. The expected value of information (EVI) was calculated for alerts that angiography is necessary. Maximal EVI was estimated by assuming that alert advice is always followed. Sensitivity analysis relaxed that assumption. Hypothetical data were generated to demonstrate EVI analysis for narrower subcohorts. RESULTS: A maximally effective alerting system would increase survival in this cohort by 2.2% over 1-4 years after MI. The system would therefore need to be applied to 46 people to prevent one death. Its effectiveness would decrease linearly with decreasing adherence to its advice. Given sufficiently detailed outcome and prevalence data, EVI analysis could also predict the survival value of the system's individual data elements. CONCLUSIONS: An alerting system that ensures necessary angiography post-MI should have a survival value comparable to the value of t-PA over streptokinase. EVI analysis provides a framework for predicting the overall effectiveness of information systems and for understanding the contribution of individual features to a system's effectiveness.
BACKGROUND: Many patients who need coronary angiography fail to get it and they have decreased survival as a result. This study demonstrates the use of decision analysis to predict the survival value of an alerting system for necessary angiography. METHODS: Data on the use of angiography and survival after myocardial infarction (MI) were taken from a published cohort study. The expected value of information (EVI) was calculated for alerts that angiography is necessary. Maximal EVI was estimated by assuming that alert advice is always followed. Sensitivity analysis relaxed that assumption. Hypothetical data were generated to demonstrate EVI analysis for narrower subcohorts. RESULTS: A maximally effective alerting system would increase survival in this cohort by 2.2% over 1-4 years after MI. The system would therefore need to be applied to 46 people to prevent one death. Its effectiveness would decrease linearly with decreasing adherence to its advice. Given sufficiently detailed outcome and prevalence data, EVI analysis could also predict the survival value of the system's individual data elements. CONCLUSIONS: An alerting system that ensures necessary angiography post-MI should have a survival value comparable to the value of t-PA over streptokinase. EVI analysis provides a framework for predicting the overall effectiveness of information systems and for understanding the contribution of individual features to a system's effectiveness.
Authors: J V Selby; B H Fireman; R J Lundstrom; B E Swain; A F Truman; C C Wong; E S Froelicher; H V Barron; M A Hlatky Journal: N Engl J Med Date: 1996-12-19 Impact factor: 91.245
Authors: D B Mark; M A Hlatky; R M Califf; C D Naylor; K L Lee; P W Armstrong; G Barbash; H White; M L Simoons; C L Nelson Journal: N Engl J Med Date: 1995-05-25 Impact factor: 91.245