Purpose: To review the literature on risk stratification after acute m
yocardial infarction in the reperfusion era and to propose an algorith
m for early and continual risk assessment. Data Sources: A MEDLINE sea
rch of the English-language literature on humans was done using the te
rms myocardial infarction, prospective studies, and prognosis. This se
arch was supplemented by narrowed searches for subheadings (such as ca
rdiogenic shock, thrombolytic therapy, and stress testing) and surveys
of references cited in review articles and book chapters. Study Selec
tion: Literature on prognosis and myocardial infarction published from
1981 to 1996 was considered. From the literature on stress testing me
thods, studies that enrolled patients before 1980, enrolled patients f
or indications other than myocardial infarction, tested patients more
than 6 weeks after infarction, were missing outcome data, or had inade
quate follow-up were excluded. Data Extraction: Because too few random
ized trials were available to allow the cross-comparison of risk strat
ification methods, the available observational data were synthesized a
nd supplemented with clinical judgments to produce recommendations. Da
ta Synthesis: Risk stratification must begin when acute myocardial inf
arction is diagnosed. High-risk patients (such as those with cardiogen
ic shock) and candidates for reperfusion therapy must be identified qu
ickly if ideal emergency care is to be given. At specific points durin
g hospitalization, specialized tests may be useful if they add increme
ntal information to the results of clinical evaluations. High-risk pat
ients who have complications after infarction or significant left vent
ricular dysfunction probably benefit from early angiography; patients
without these conditions are at low risk for recurrent events and shou
ld have noninvasive stress testing for further risk stratification. Co
nclusions: Physicians should continually reappraise risk throughout ho
spitalization to optimize both patient outcomes and cost containment.