PT - JOURNAL ARTICLE AU - Bradford, Daniel AU - Allik, Mirjam AU - McMahon, Alex AU - Brown, Denise TI - OP108 Potential endogeneity bias when using composite measures of multiple deprivation in health inequalities research AID - 10.1136/jech-2023-SSMabstracts.107 DP - 2023 Aug 01 TA - Journal of Epidemiology and Community Health PG - A53--A53 VI - 77 IP - Suppl 1 4099 - http://jech.bmj.com/content/77/Suppl_1/A53.1.short 4100 - http://jech.bmj.com/content/77/Suppl_1/A53.1.full SO - J Epidemiol Community Health2023 Aug 01; 77 AB - Background Composite measures of multiple deprivation are composed of multiple domains e.g., income, education, health. There is a concern that using deprivation-dependent health outcomes as indicators in the health domain of these measures to predict health inequalities will lead to biased results due to endogeneity. The objective of this study was to assess the effects of this potential endogeneity bias.Methods The Scottish Index of Multiple Deprivation (SIMD) was used as an exemplar composite measure of multiple deprivation as it includes health-related indicator variables in the calculation of area deprivation scores. Area deprivation scores were calculated at the Data Zone level which divides Scotland into approximately 7,000 geographic units. Data Zones were assigned to deprivation deciles calculated independently using the original SIMD, SIMD with the health domain excluded, and the income domain in isolation. Health inequalities were assessed using population-wide mortality data for two time periods (2010—2012 and 2017—2019) and self-rated health data from the 2011 Scottish census as dependent variables. Health inequalities were assessed separately for males and females using age-standardised rates, as well as both the slope and relative indices of inequality.Results There were no significant differences in slope or relative indices of inequality when comparing 912 pairs of values calculated using SIMD with either the health-excluded version or the income domain in isolation. Similarly, only two of 160 pairwise comparisons of age-standardised rates between SIMD and the two alternative methods were significantly different: mortality in females in decile 9 was 943 (95% CI: 904, 983) using SIMD versus 854 (95%CI: 816, 893) using the income domain; self-rated bad health in males in decile 6 was 45.8 (95%CI: 45.0, 46.7) using SIMD versus 47.6 (95%CI: 46.8, 48.6) using the income domain. Mortality and health outcomes were consistently poorer in more deprived areas. This pattern of inequalities held for all three methods used to calculate deprivation deciles.Conclusion Endogeneity due to the inclusion of health indicators in composite measures of multiple deprivation did not have any substantial effect on estimates of age-standardised mortality and health outcome rates or the measurement of health inequalities. Further work is needed to confirm this result is generalisable to composite measures of multiple deprivation other than the Scottish Index of Multiple of Deprivation and to other health outcomes. Researchers should continue to be thoughtful in their selection of deprivation measure but should be reassured that endogeneity bias of this type will not substantially affect results.