Background: Whether non-communicable diseases (NCDs) are diseases of poverty or affluence in low-and-middle income countries has been vigorously debated. Most analyses of NCDs have used self-reported data, which is biased by differential access to healthcare services between groups of different socioeconomic status (SES). We sought to compare self-reported diagnoses versus standardised measures of NCD prevalence across SES groups in India.

Methods: We calculated age-adjusted prevalence rates of common NCDs from the Study on Global Ageing and Adult Health, a nationally representative cross-sectional survey. We compared self-reported diagnoses to standardized measures of disease for five NCDs. We calculated wealth-related and education-related disparities in NCD prevalence by calculating concentration index (C), which ranges from -1 to +1 (concentration of disease among lower and higher SES groups, respectively).

Findings: NCD prevalence was higher (range 5.2 to 19.1%) for standardised measures than self-reported diagnoses (range 3.1 to 9.4%). Several NCDs were particularly concentrated among higher SES groups according to self-reported diagnoses (Csrd) but were concentrated either among lower SES groups or showed no strong socioeconomic gradient using standardized measures (Csm): age-standardised wealth-related C: angina Csrd 0.02 vs. Csm -0.17; asthma and lung diseases Csrd -0.05 vs. Csm -0.04 (age-standardised education-related Csrd 0.04 vs. Csm -0.05); vision problems Csrd 0.07 vs. Csm -0.05; depression Csrd 0.07 vs. Csm -0.13. Indicating similar trends of standardized measures detecting more cases among low SES, concentration of hypertension declined among higher SES (Csrd 0.19 vs. Csm 0.03).

Conclusions: The socio-economic patterning of NCD prevalence differs markedly when assessed by standardized criteria versus self-reported diagnoses. NCDs in India are not necessarily diseases of affluence but also of poverty, indicating likely under-diagnosis and under-reporting of diseases among the poor. Standardized measures should be used, wherever feasible, to estimate the true prevalence of NCDs.