Abstract
The implications of an ageing population are frequently used as the justification for all kinds of policy decisions about health care funding. Though it is clear that the costs of care are much greater in older age groups, when forecasting future expenditure over long time periods, the relationship is not so simple. The additional years of life may be relatively healthy with lower care costs, but with some high costs concentrated towards the end of life, hence the importance of the idea that costs are shaped more by time to death than age per se. 1 However, teasing out the separate effects of age and time to death requires some novel approaches and powerful data sets – ones that are capable of looking at large numbers of people over long time periods. The easiest approach is through the exploitation of administrative data and the use of linkage techniques to piece together individual people's histories.
The paper by Moorin and colleagues 2 exploits just such data. They looked at a group of people who died with defined causes of death, and then used linked data to study their Medicare claims for the five years before death. The authors were able to repeat this analysis for three eras from 1990 to 2004. These data were able to show differences in the relationship between service use, age and time to death. Moreover, they were able to explore these relationships for different health care services. Such studies are possible through the exploitation of data linkage techniques applied to increasingly available administrative data generated by health and care systems. 3 They exploited claims data from the Australian Medicare Benefit Scheme. Yet many other countries have similar data sets that have been collected fairly consistently over many years and use common patient identifiers. This means it becomes possible to trace an individual's use of care over many years.
Data linkage in health care is not especially new. In England, we have benefitted from the work of the Oxford Record Linkage Studies 4 which started looking at hospital readmissions in the 1960s despite the limited information technology of that time. The rise of computerized information systems and the internet has meant that the use of linked data is commonplace in our daily lives whether it is the choices offered by online book sellers or supermarkets exploiting data from loyalty cards. Within the health sector, most interactions with care services involves some form of electronic record, for booking appointments, billing or logging details of treatment provided. We are getting better at finding ways to exploit these data and in ways that protect an individual's identity through the use of anonymized identifiers. Exploiting operational data avoids the need for costly new data collection and allows analysis of a full range of service users, not just a pre-specified subset or client group. Many countries, including the UK, are making progress in developing the skills to undertake data linkage 5 for both research and in more routine analysis such as the use of predictive risk tools. In Scotland, the long history of work on record linkage has led to a national framework for data linkage, 6 whilst in Wales the Secure Anonymised Information Linkage (SAIL) initiative was established to link the widest possible range of anonymized, person-based data. 7 Yet there is still significant untapped potential in using linked data sets and areas where the approach can be improved and extended.
Moorin and colleagues’ analysis included an important twist by using data sets that looked not only at individuals over time, but also across sectors of care. They noted how the relationship between care use and age was quite different when looking at primary care as opposed to specialist services or diagnostic and therapeutic services. Such linkage of data sets across different services is needed to test whether changes in one sector have implications for another. Too often we focus on one sector, yet the wider view across different forms of care is important, especially for the management of complex long term health problems and in care at the end of life. It is often the case that the people who consume the most resources will have multiple health problems and be in contact with a wide range of interacting, and in some cases interdependent care providers. One area where this can be seen in England is at the interface between health care (provided by the National Health Service) and social care delivered by independent providers and funded either by local government or by people paying for themselves. The majority of people receiving social care will also be receiving health care and there are studies to show how these services interact so that receipt of one service reduces the need for the other.8,9 It appears that the level of use of social care can affect the level of hospital care and vice versa. This is as important as ever in a time of significant financial constraints on public services.
Similarly, improving care at the end of life is an agreed national priority for health and social care in England yet the strategy has been hampered by a lack of information about the patterns and quality of care delivered. 10 Studies of cost at the end of life have shown the patterns seen by Moorin et al, namely that older age groups tend to use fewer hospital resources at the end of life than younger people. This could be a reflection of differences in underlying disease coupled with patient and professional preferences for treatment options. But the reduction in hospital costs in older age groups is offset by greater use of social care, particularly nursing home resources. 11 Moreover analyses from England using linked health and social care data suggest that the inverse relationship between local authority funded social care inputs and NHS hospital costs in the last 12 months of life exists in all age groups. 12
Extracting and linking pseudonymous data from across care settings could be used more in a wider range of studies. Moreover, this approach can be extended beyond the research environment to inform practitioners directly, for example, by providing continuous monitoring of activity following service change and innovation. For policy makers, these analytical approaches should form essential tools to assess the success of policy initiatives across all care services.