Health Literacy and Sustainable Healthcare: Machine Learning Modelling of Health Literacy Disparities in Australia (Preprint)
Meng Ji
DOI: https://doi.org/10.2196/preprints.25909
2020-11-20
Abstract:BACKGROUND Health literacy is a key issue in sustainable healthcare support to reduce health inequality and disparity. In multicultural societies with large and changing migration populations, there is a pressing need to understand the disparity of health literacy among diverse, complex population segments. This study offers much-needed insights into the correlation and interaction among various underlying dimensions of health literacy among diverse populations in Australia. This is based on the 2018 Health Literacy Survey (HLS) conducted by the Australian Bureau of Statistics (ABS) with 5,790 fully responding Australian adults aged above 15. OBJECTIVE Using machine learning to identify major contributing factors (especially, specific value ranges of key health literacy domains) to health literacy disparities in Australia. METHODS Statistical machine learning models (XGBoost Tree) were used to identify and measure the disparity of health literacy between Australian populations characterised by demographic, educational and socio-economic attributes: age, sex, country of birth, main language spoken at home, labour force status, equivalised income of household (EIH), family composition of household, level of highest educational attainment, disability status, Australian states and territories, remoteness and index of relative socio-economic disadvantage (SED). RESULTS Our analysis found that among the nine domains of the 2018 Australian HLS, there were distinct patterns of disparities in health literacy among Australians. Populations which reported higher scores of self-health management ability (SHMA) (Domain 3: 3.08-3.22) were Australians aged under 35 or above 55, having Year 12 or above educational attainment, English-speaking, married with/without children, female, in the top two EIH quintiles, in the lowest two SED quintiles, having no disability or restrictive long-term health condition, and living in the states of Queensland, Victoria, Western Australia, South Australia, Northern Territory. Populations which reported lower scores of SHMA (Domain 3: 2.99-3.08) were Australians aged between 35 and 55; having Year 11 or below education; speaking languages other than English at home; living alone or single parents with dependent children, male, in the bottom three EIH quintiles, in the highest three SED quintiles; having profound or severe core activity limitation, or other disability or restrictive long-term health condition, and living in the Australian states of New South Wales, Tasmania, and the Australian Capital Territory. CONCLUSIONS Our study identified major contributing factors (especially, specific value ranges of key health literacy domains) to health literacy disparities in Australia. These include education (Year 10/11 or below), disability (profound/severe disability), household income (lowest quintiles), the relative SED index (highest quintiles), gender (male Australians), age (aged 35-55 years), main home language (other than English), geographical location (major cities, inner, outer regional, remote Australia). Higher value ranges of these variables are strongly associated with higher scores of key health literacy domains such as access to healthcare support (Domain 1), access to sufficient health information (Domain 2), ability to appraise health information (Domain 5), ability to find good health information (Domain 8) and ability to understand health information well to know how to apply the health information (Domain 9). Higher scores on these domains in turn can have real impact on the overall self-health management ability (Domain 3). CLINICALTRIAL n/a