Sharp increases are forecast in the burden of chronic diseases in low and middle income countries (LMIC), as a result of demographic ageing, and the health transition [1]. In 2005, the WHO announced an ambitious goal to reduce chronic disease deaths by 2% per year [2], considered achievable through population prevention strategies targeting dietary salt and tobacco use [3]. The control of chronic diseases requires functioning, affordable and equitable primary healthcare since it is through access to these services that those at high risk can be identified, advised and treated [4]. Unfortunately, healthcare systems and services in LMIC are often unsuited to the needs of their ageing populations [5]. Structural reforms introduced fees for service, necessitating substantial out-of-pocket payments. Those with chronic diseases needing continuous treatment over long periods, and older people with little or no personal income are particularly affected. A paradigm shift has been called for, from preoccupation with simple curative interventions to chronic disease management, long-term support and care. Given the frailty of many older people there is also a need for outreach, assessing and managing patients in their own homes [4, 6].

The notion of equity in health care is predicated on principles of social justice underpinned by the view that access to health care should be based upon 'need' while financing should be determined according to 'ability to pay' [7]. The right to healthcare is an essential component of the universal right to health, enshrined in Article 25 (1) of the Universal Declaration of Human Rights (1948) [8], and underlined in Article 1 of the UN Convention on the Rights of Persons with Disabilities (2006) [9]. Paragraphs 74-77 of the Madrid International Plan of Action on Ageing (2002) call for the elimination of social and economic inequalities in access to healthcare and the development of healthcare and long-term care to meet the needs of older persons [10]. However, there have been very few studies of healthcare utilization among older people in LMIC, and surprisingly little focus on equity. In a large cross-sectional survey in India of those aged 60 years and over, socioeconomic status was inversely associated with ill health but positively associated with hospitalisation in the previous year in both sexes, after controlling for health status, age and social support [11]. In surveys in 10 countries in Latin America and the Caribbean [12]; health service utilisation increased with household expenditure among those aged 65 years and over for all countries other than Argentina and Chile. We were unable to find any information on the impact of health insurance coverage on older people in countries with low and middle incomes, but in the USA Health and Retirement Survey, women aged 55 to 64 years of age were two to seven times more likely to use health services for which they had insurance cover [13]. Much research has focussed upon the effects of gender on healthcare utilization among older people. Older women tend to use primary care and community health services more than men [14, 15]. Gendered differences in illness constructions, help-seeking behaviour, and accessibility of services have been suggested as mediating factors [14, 15]. Some investigators have discerned a 'health disadvantage' for older women, but this depends on the health outcome studied. Men are more likely to use emergency services, and as likely if not more likely to be hospitalised [11, 14], and have a shorter life expectancy and a higher mortality than women in late life [16].

We carried out catchment area surveys of representative samples of older people in six Latin American countries (Cuba, Dominican Republic, Puerto Rico, Venezuela, Mexico and Peru), India, China and Nigeria. In all we completed 17,944 interviews with a high response rate in most sites. Our eight urban and five rural catchment area sites encompassed wide variation in age distributions, prevailing economic circumstances, levels of education, and pension and health insurance coverage. Our data adds considerably to current understanding of patterns of service utilisation among older people in LMIC, with respect to the range of services assessed (using a standard protocol in all sites), the detailed assessment of health status, and the comprehensive assessment of predisposing and enabling variables, including the impact of health insurance coverage. There are potential limitations, particularly with respect to the self-reported data on health service use. Utilization of healthcare (in practice) is not the same as access to healthcare (in principle), which may be the more relevant construct with respect to issues of equity [7]. Information systems in most LMIC are inadequate for routinely recorded data to be used. Studies have shown that self-reported health service use may be biased by a slight tendency towards underreporting, and that measures are most accurate for recent episodes of medical care [27]. We used data on recent use (the three months preceding the interview). Misclassification is likely to have been random with respect to correlates of healthcare utilisation, biasing any associations towards the null. This may not have been the case for dementia since those with the condition might have been less likely to recall service use, and the family informant that was relied upon under such circumstances may not have always been aware. Finally, while the common research protocols and demonstrable cross-cultural validity of our methods allow direct comparisons to be made between samples from the different countries surveyed, our findings should be generalized with caution and only to populations similar to those that we focused upon. Certain of the local health systems could not be considered to be typical of rural and urban settings in the country concerned, particularly in India, where Voluntary Health Services in urban Chennai and the Christian Medical College in rural Vellore operate as charitable non-governmental service providers running comprehensive primary care, inpatient and outpatient hospital services free of charge to those who cannot afford to subscribe or pay.

We found that the overall levels of use of any community healthcare services in the past three months varied very widely among sites, with little of this variation explained by compositional differences in age, sex and the prevalence of physical, mental and cognitive disorders. In almost all of the settings studied, the healthcare systems were mixed, with private providers played an important role in all countries other than Cuba and China, particularly in the Dominican Republic, Puerto Rico, Venezuela and India, where, other than in Puerto Rico, out-of-pocket payments were the rule. Out-of-pocket payments were also required for almost all government primary care and hospital outpatient consultations in China, and for a high proportion of these consultations in rural Peru and Nigeria. In China, the proportion of participants claiming to have health insurance does not provide a true picture of the extent of cover provided. In urban China, there are two employee-based health insurance schemes, one for government and the other for public and private company employees. In rural China, the government contributes to a common fund covering healthcare costs but only proportionate to the amount contributed locally. This rural Cooperative Medical System is ineffective, and the cover provided in cities is patchy; none of the schemes has kept pace with the sharp increases in charges under the new fee-for-service system [28], and the majority of healthcare costs are covered by out-of-pocket payments [29]. The extent of cover provided in other settings will have varied considerably among individuals claiming to have private health insurance. One of the limitations of our assessment of the effect of health insurance on service use is that we have individual level data only on whether or individuals have private health insurance, and not on the extent of the cover provided to the individual. Although causality is hard to demonstrate, the extent of out-of-pocket payments, coupled with the extent of pension and health insurance coverage, provide a parsimonious explanation for the observed variation in the overall levels of healthcare utilisation between sites, as well as for the type of service preferentially consulted within sites. It cannot however account for the relative unpopularity of low cost government primary care services in the Dominican Republic and in India, compared to private healthcare providers, particularly given the relatively low levels of private health insurance coverage in these sites.

Need for healthcare, as indicated by the presence of chronic health conditions, was the main determinant of use of healthcare services in all sites. However, while number of limiting physical impairments and ICD-10 depression were associated with the use of healthcare services in most sites, dementia was consistently inversely associated. This observation is consistent with the earlier impression, principally from qualitative research in India, of low levels of awareness of dementia as a medical condition, limited help-seeking, and generally unhelpful responses from healthcare services [30, 31]. It will be important in the future to explore factors associated with help-seeking among people with dementia, particularly the role of disease severity and the presence of behavioural or psychological symptoms. The effective rationing of community healthcare to those with the most evident physical or psychological morbidity was particularly apparent in those sites such as rural China and Nigeria, with generally low levels of use of healthcare services. It was not possible to determine whether demand or supply, or a combination of both, contributed to this pattern of usage.

Inequity was particularly evident in the positive association between educational level and use of healthcare services, and the strong effect of health insurance cover on use of community healthcare services at least in those sites where out-of-pocket expenses were common, and where private health insurance was an important component of healthcare financing. In this context, the heterogeneous association between household assets and use of healthcare services was puzzling. For countries such as Cuba, where healthcare is free at the point of delivery, it makes sense that educational level rather than socioeconomic position is associated with increased healthcare utilisation particularly of prevention and promotion services, and through better adherence to chronic disease management protocols. In some other settings, it may be that household assets, an index of accumulated wealth, may be less salient than regular personal income to meet out-of-pocket expenses. We found no consistent evidence for ageism in the demand for or delivery of healthcare services, although older people were significantly less likely to use healthcare services in Cuba and rural India, after adjusting for health status and other variables, with a strong trend in that direction in rural China. There was also little evidence for the exclusion of those with disabilities, at least as regards those with restricted mobility. Of course, one of the likely mechanisms for the under usage of healthcare services by people with dementia may be the impact of cognitive disability on help seeking, compounded by the lack of outreach services to meet the special needs of this group [32]. The finding that women are more likely than men to access community healthcare services is now also conclusively demonstrated for older people in LMIC. We have previously reported that the apparent excess disability among women in our surveys may be accounted for by selective underreporting by men [32]. Studies from LMIC suggest that men have a higher mortality in late-life [33, 34], and in our survey, as in others, are much more likely to be admitted acutely to hospital. Attention should therefore be directed towards encouraging timely help seeking, detection, treatment and control of chronic diseases and their risk factors among men.

The most striking finding from this international survey is the very wide variation in the proportions of older people accessing community healthcare services in the past three months between sites, even after accounting for compositional differences in age and sex and the prevalence of chronic health conditions. Aside from cultural differences in help-seeking, this seems likely to be accounted for by stark differences in the extent of out-of-pocket expenses, and the ability of older people and their families to afford them. Health service utilisation was highest in Puerto Rico, where most older people were covered by Medicare, and lowest in China, where out-of-pocket costs were almost universal; across the 13 sites, the ecological correlation between the proportion of consultations requiring out-of-pocket costs and the prevalence of health service use was large (-0.50) and of borderline statistical significance (p = 0.09). Out-of-pocket costs are linked directly to government policies on the financing and reimbursement of healthcare. Intergovernmental charters, conventions and plans stress the importance of equity in the provision of healthcare to older people, and provide a framework of rights and responsibilities which, once ratified, are binding in international law. Data from epidemiological studies can remind governments of their obligations, provide evidence of the extent to which these have, or have not been met, and hence assist stakeholders in holding governments to account. The results of our surveys suggest that health systems that rely to a significant extent upon out-of-pocket payments for financing basic medical services risk excluding three overlapping groups; the poorest older people in society, those that lack a secure regular income, and those without health insurance. The strong independent effect of education on healthcare utilisation suggests that more may need to be done to promote demand for healthcare, even in those settings, such as Cuba, where economic barriers are less of a problem. Finally, we should note that achieving equity in access to healthcare is not the same as achieving equity in health [7]. While this may be one important driver, the determinants of health inequalities are complex and operate across generations and throughout the life course.