New statistical techniques to model biomarkers for social science research
New research from the Institute for Social and Economic Research using data from Understanding Society, the UK Household Longitudinal Study, proposes new statistical techniques to model biomarker data for analysing socio-economic inequalities in health. The new research, co-authored by Dr Apostolos Davillas, a health economist based at ISER, with Professor Andrew Jones from the University of York, has been published in Health Economics.
Recent developments in social surveys include the integration of biological samples such as the blood-based biomarkers. Unlike self-reported assessments of health, biomarkers are more objective heath measures, provide information on pre-disease conditions and insights on the biological links between socio-economic status and health. As a result, over and above the medical literature, biomarkers became popular to growing numbers of social science and economics studies that explore socio-economic inequalities in health.
Despite these advantages, biomarker data impose a number of statistical challenges. Typically, biomarkers are distributed asymmetrically reflecting the fact that a large part of the population experience good health, while there are parts of the population with ill health or severe illnesses. However, most of the existing studies aiming to explore the role of socio-economic position on biomarkers employ statistical techniques that ignore the asymmetric nature of biomarkers distribution. This raises doubts about the robustness of such studies based on these conventional statistical techniques.
The new study seeks to propose new statistical techniques for modelling biomarkers and highlights their importance and potential of socio-economic inequalities in health research. The study uses Understanding Society data and a set of blood-based biomarkers, such as inflammatory, diabetes and cholesterol biomarkers. The paper introduces and compares the performance of a set of flexible parametric statistical techniques for modelling biomarkers. Of particular interest, it has been shown that the conventional estimation techniques used by most of the previous studies to model biomarkers may lead to important biases for analysis of the socio-economic inequalities in health.
Dr Apostolos Davillas said, “An important recent development in research based on large-scale social surveys is the integration of biological samples, in addition to traditional self-reported health assessments. However, biomarkers impose statistical challenges since they are often characterised by asymmetric distributions with heavy tails. The new estimation techniques that are proposed and compared in our study are particularly useful for public health researchers, social science researchers and policy makers interested in modelling biomarkers for exploring socio-economic inequalities in health. Eliminating prediction bias at the tails of the biomarkers distribution is of policy interest because of the elevated health risks and associated health-care costs.”
Download the paper Parametric models for biomarkers based on flexible size distributions