Objective Self-report of physician diagnosis and ECG ‘Q’ waves are common survey measures of prior myocardial infarction (MI) prevalence. We sought to assess relative prevalence of self-reported prior MI and ECG Q-waves (ECG-MI) in populations and population subgroups with varying MI prevalence.

Design A secondary analysis of seven population-based cross-sectional surveys of prevalent MI selected from a systematic review of ischaemic heart disease epidemiology.

Setting Men and women aged 45–74 years in population-based Belgian surveys (1978–-1998, n=29 419) and US National Health and Nutrition Examination Surveys (1976–1994, n=11 107). Comparison of the US and Belgian surveys with surveys in seven other nations (United Kingdom, Russia, Lithuania, Belarus, India, Turkey and Ghana).

Main outcome measures Prevalence of prior MI measured by self-report or resting ECG Q-waves (ECG-MI; Minnesota ECG codes 1.1 and 1.2).

Results Self-reported prior MI prevalence was 1.5–2.6 times higher than ECG-MI in Belgian and US men aged 45–74 years and women 55–74 years. ECG-MI was more prevalent than self-reported MI in women <55 years old, and self-reported MI relatively low in US African–American men compared with US Caucasian men. In the overall nine-nation comparison, there was no consistent relationship between self-reported MI and ECG-MI. ECG-MI was higher relative to self-report in nations with lower prevalence of ischaemic heart disease.

Conclusions Self-reported MI and ECG-MI prevalence may only be reliable in higher ischaemic heart disease incidence groups. Self-report and ECG-MI have limited accuracy, and ECG Q-waves likely capture fewer prior MIs in the 21st century. The limitations of current survey prevalence measures of MI should be taken into account when measuring the burden of ischaemic heart disease in populations.