SAMSS data have shown that health inequalities are stable, if not increasing, in SA with health disparities for lower educated and lower income groups, measured by those unable to save, appearing to be increasing in most of the health variables considered. Although it is not an aim of this manuscript to speculate about possible causes, which are quite often difficult to define [22], it is conceivable that most of these increases in inequality can be related to interventions not targeted to specific groups, or not specifically designed to be capable of reducing the gap [23,24,25]. Interventions targeted to the general population [23], such as, e.g., building cycling tracks (now covering most of the South Australian urban area), could be of benefit for those already active (typically higher educated, with higher income) and have no impact on those more vulnerable that are unable to buy a bike. This eventually produces an increase in the gap in the level of sedentary activity between classes. A call for more action and better understanding for more effectively targeting of the interventions is warranted. Staying with the biking example, this would mean health promotion activities involving more vulnerable communities. This could include, for example, offering free bikes, activities aimed to engage people in small bike tours, offering bikes to cycle to school and creating the conditions where this could easily happen.
It is evident, from the analysis presented in this study, the substantial role played by non-communicable disease and behavioural risk factor surveillance systems, in showing the evolutionary aspects of health disparities. Certainly, for these analyses the availability of a ‘real’ surveillance system [26, 27] rather than a few scattered surveys is fundamental. In our opinion, yearly or even with less frequently repeated surveys (the WHO suggest ‘at least every five years’) [28] provide little information when studying and showing trends. We believe that these analyses have only scratched the surface: much information can still to be obtained from surveillance data, particularly in understanding the mechanisms [29] which create health inequalities and, as we have seen, increase these inequalities. Specific analyses for sub-groups, defined on the basis of socio-demographic variables, but also geographically, can provide further information [30,31,32]. Data from surveillance systems, highlighted in this study, could be even more useful when linked with data from other sources (e.g. Census data), to study other potentially influencing social determinants such as social and cultural capital [33] or urban settings [34]. An even more important role in the future could have surveillance showing ‘what works’ in reducing health inequalities (when targeted interventions are implemented), given the potential use of these systems for evaluation purposes [4, 35, 36].
In this first paper we purposely limited the analyses so as to show simple time trends. Research is needed on these data, to better understand interaction of different social determinants of health, and the possible underlying mechanisms which creates and reinforces gaps. Certainly, for instance, the fact that over the years the number of those falling into more deprived groups has decreased in SA, due to the selective effect of social mobility, and could have left individuals in the lower strata individuals with characteristics that (directly or indirectly) induce worse health attitudes and behaviours.
However, using simple analyses to show how much the health inequality gap remains relevant, also creates several limitations. Some of these are related to the available data, and some are associate with the analysis conducted. A first weakness is the limiting of the risk factors assessed to two (BMI and fruit consumption) and three specific chronic conditions (diabetes, mental health and psychological distress). In addition, only two socio-economic related variables were used. Although the use of ‘ability to save’ as an indicator has been shown in Australia to be a valid indicator of financial security [37], it is acknowledged that other more reliable questions could have been used such as income. However, over the 10 year period, income earnings have increased for the whole of Australia which made it difficult to have comparable income groupings across the years. This study uses self-reported surveys which can potentially be subjected to bias due to socially desirable responses leading to possible over- or under-estimation of behaviors or health conditions, such as having a mental health condition or overweight and obesity due to incorrect reporting of height and weight [38]. However, these biases are of little importance if the aim of SAMSS is to study changes in the behaviour or health condition over time, assuming that the level of bias is constant over time.
The use of listed telephone numbers as the sampling frame can be considered a weakness of this study due to an increasing number of mobile-only households with the majority of these types of telephone numbers not being listed [39]. However, studies have shown that using this sample frame is still a viable source and reliable estimates can be produced when applying more effective weighting techniques, such as raked weighting [10], to overcome the sampling bias as well as non-response bias.
A further weakness is the lack of power in terms of data on Aboriginal status. In Australia, recent policy actions have focused on improving the health of Aboriginal populations with the Prime Minister of Australia in 2008 signing a statement committed to developing a long-term plan of action to end health inequalities between indigenous and non-indigenous populations [32]. Although SAMSS collects this information, the limited sample size does not permit analysis by Aboriginal status. It is also acknowledged that some of the increases in prevalence of mental health problems reported in this analysis could be the result of better diagnosing, which has been supported by additional funding from the Federal government in recent years. This could also impact our analyses by social class, leading to an underestimation of inequalities, since individuals who are more educated are also more likely to seek health care services and receive a diagnosis. The acknowledgment of public health campaigns in increasing the fruit consumption has also not been fully explored although earlier work with this surveillance system has shown promising results [35].
