Are estimates of socioeconomic inequalities in chronic disease artefactually narrowed by self-reported measures of prevalence in low-income and middle-income countries? Findings from the WHO-SAGE survey.

Background: The use of self-reported measures of chronic disease may substantially underestimate prevalence in low-income and middle-income country settings, especially in groups with lower socioeconomic status (SES). We sought to determine whether socioeconomic inequalities in the prevalence of non-communicable chronic diseases (NCDs) differ if estimated by using symptom-based or criterion-based measures compared with self-reported physician diagnoses.

Methods: Using population-representative data sets of the WHO Study of Global Ageing and Adult Health (SAGE), 2007-2010 (n=42 464), we calculated wealth-related and education-related concentration indices of self-reported diagnoses and symptom-based measures of angina, hypertension, asthma/chronic lung disease, visual impairment and depression in three 'low-income and lower middle-income countries'-China, Ghana and India-and three 'upper-middle-income countries'-Mexico, Russia and South Africa.

Results: SES gradients in NCD prevalence tended to be positive for self-reported diagnoses compared with symptom-based/criterion-based measures. In China, Ghana and India, SES gradients were positive for hypertension, angina, visual impairment and depression when using self-reported diagnoses, but were attenuated or became negative when using symptom-based/criterion-based measures. In Mexico, Russia and South Africa, this distinction was not observed consistently. For example, concentration index of self-reported versus symptom-based angina were: in China: 0.07 vs. -0.11, Ghana: 0.04 vs. -0.21, India: 0.02 vs. -0.16, Mexico: 0.19 vs. -0.22, Russia: -0.01 vs. -0.02 and South Africa: 0.37 vs. 0.02.

Conclusions: Socioeconomic inequalities in NCD prevalence tend to be artefactually positive when using self-report compared with symptom-based or criterion-based diagnostic criteria, with greater bias occurring in low-income countries. Using standardised, symptom-based measures would provide more valid estimates of NCD inequalities.