Article Text

Download PDFPDF

Seven-year hospital and nursing home care use according to age and proximity to death: variations by cause of death and socio-demographic position
  1. Pekka Martikainen1,
  2. Michael Murphy2,
  3. Niina Metsä-Simola1,
  4. Unto Häkkinen3,
  5. Heta Moustgaard1
  1. 1Population Research Unit, Department of Social Research, University of Helsinki, Helsinki, Finland
  2. 2Department of Social Policy, London School of Economics, London, UK
  3. 3Centre for Health and Social Economics, National Institute for Health and Welfare, Helsinki, Finland
  1. Correspondence to Professor Pekka Martikainen, Population Research Unit, Department of Sociology, University of Helsinki, P.O. Box 18, FIN-00014 Helsinki, Finland; pekka.martikainen{at}


Background Provision of hospital and long-term care services for the growing number of older people is a major policy concern. The authors estimate hospital and nursing home care use by age and proximity to death for selected causes and by gender, education and marital status.

Methods A 40% random sample of the Finnish population aged 65+ years alive at the end of 1997 was followed to death in 1998–2002. Use of hospital and nursing home care was assessed up to 7 years prior to death for those who died and prior to the end of 2002 for survivors.

Results In the 7-year period, before death total average care days were 294 (95% CI 286 to 301) for men and 430 (95% CI 423 to 438) for women. For surviving men and women, the corresponding figures were 89 (95% CI 86 to 92) and 136 (95% CI 130 to 141) days. Use of hospital and particularly nursing home care increased rapidly with age, while proximity to death was more important for hospital care. The married used less care than the non-married. Care use of those dying from dementia was approximately twice that for all causes combined and was substantial for an extended period before death.

Conclusions The effects of age are more substantial for nursing home than for hospital care use, and both are larger the older the age at death. Care use will be considerably higher among the non-married. Increasing longevity coupled with a rising trend of dementia is likely to mean a major shift towards higher nursing home care use in the future.

  • Hospital
  • nursing home
  • end of life
  • cause of death
  • social factors
  • social epidemiology
  • mortality
  • ageing
  • registers
  • marital status
  • demography
  • health expectancy
  • morbidity
  • longitudinal studies

Statistics from


One of the major concerns relating to population ageing is the provision and financing of care services for the growing number of older people and especially the ‘old–old’. A major part of the previous evidence relates to hospital care and shows that the average cost and use of hospital care services rises substantially with age. However, it is argued that the real reason for this is not to do with age per se, but with the fact that at older ages, more people are close to death. A typical finding in the USA and elsewhere is that about one-third of a person's total lifetime use of hospital care services takes place in the last 2 years of life.1 ,2 The implications of whether healthcare needs are affected more by proximity to death than by age are substantial. If healthcare use is heavily determined by experiences shortly before death, then expected costs of population ageing are likely to be less than if care use is mainly determined by old age. These cost savings may be brought about in two ways. First, pushing out the age at death in populations with increasing life expectancy reduces the number of deaths occurring in a given year and second the healthcare costs in the last years of life may be lower for people who die at older rather than at younger ages.3 ,4 Less is known about the relationship of nursing home care costs with age and proximity to death. However, existing evidence suggests that although the use and costs of nursing home rises with age, proximity to death is much less important for nursing home than for acute hospital care.5–11

Care use is of course determined not only by age, and proximity to death, but also by other socio-demographic factors. Gender, education and marital status are of central importance because they are all well-established determinants of health, hospital care and nursing home care use. Educational differences in health and mortality are one of the most consistent findings in epidemiology.12 Education is likely to influence care use both indirectly, through its effect on health, and directly, through care seeking behaviour or ability to purchase or better access health services. Being single, widowed or divorced is also associated with poorer health and higher costs of medical care than being married.13 It is well recognised that those living with a partner are less likely to enter and stay in nursing home care than those living alone, at least partially because of greater availability of informal care.14 Finally, different underlying health problems and causes of death may be differently associated with functioning and end-of-life care.15–17 In particular, strong need for hospital and nursing home care during an extended period before death could be assumed for conditions associated with severe often long-term functional decline—in particular dementia.16 ,18

To our knowledge, none of the previous studies on the effect of age and proximity to death on care needs have been able to simultaneously incorporate information on both hospital and nursing home care, socio-demographic factors and cause of death in general population samples. Furthermore, most analyses are limited to relatively short periods, typically 1–2 years, before death, and will thus underestimate and possibly bias estimates of longer term care use trajectories.19 Using large registration-based data sources, the current study extends previous work with in-depth analyses of the effects of age and proximity to death on hospital and nursing home care for selected groups of causes of death in a 7-year period. We compare care use patterns in the period before death for the deceased at a particular age to those who survived to the same age. The results are presented separately by gender, education and marital status as these factors may be associated with care needs.


Participants and measurement

Dates of death from 1998 to 2002 were recorded for a 40% random sample of the Finnish population aged 65+ and alive at the end of 1997. Statistics Finland provided data on socio-demographic characteristics in 1997 and dates and causes of death in 1998–2002, and the National Institute for Health and Welfare provided data on entry to and exit from nursing home care and hospital care in 1995–2002. Statistics Finland linked these data using unique personal identification codes available in the different data sources (permission TK 53-576-04). The data comprised 301 263 persons of whom 73 451 died during follow-up in 1998–2002. Non-linkage, loss to follow-up due to emigration or other reasons was negligible at these ages (0.07% of the study cohort).

Days in nursing home care included days spent in nursing homes, service homes with 24-hour assistance, and rehabilitation care. Hospital care days included overnight stays and day surgery in hospitals and health centres. Information on care use was based on institutional discharge registers and end of the year patient censuses. Using data on entry and exit dates, we identified 1 909 746 care episodes in either nursing home or hospital care.

Marital status and education at the end of year 1997 were used throughout the analysis. Marital status was categorised as: (1) married (2) non-married. The two educational categories were based on the highest educational qualification: (1) higher or intermediate and (2) basic education. Age at death or end of follow-up was used as a continuous variable in single years.

We categorised deaths according to the underlying cause of death recorded on the death certificate using the International Statistical Classification of Diseases, 10th Revision. The following categories were identified: ischaemic heart diseases (I20–I25), cerebrovascular diseases (I60–I69), other diseases of the circulatory system (I00–I15, I26–I28, I70–I99), dementia and Alzheimer's disease (F01, F03, G30, R54), malignant neoplasms (C00–C97), diseases of the respiratory system (J00–J64, J66–J99), accidents and violence (V01–X44, X46–Y89), and all other causes.

Statistical methods

We used growth curve models to analyse the number of days spent in nursing home and hospital care each year preceding death or end of follow-up. To achieve this, we assessed the care status for each participant for each single day (when alive) during the study period as (1) at home, (2) in nursing home care or (3) in hospital care. If a person had overlapping days of nursing home care and hospital care (altogether 1.4% of all care days), we gave priority to hospital care. We then aggregated these care data for successive one-year periods before the day of death for those who died in 1998–2002 and before the end of follow-up (31 December 2002) for the survivors.

Age-adjusted growth curve models were set up to compare care use patterns among those deceased at a particular age in any given year 1998–2002 to those who survived to the same age at the end of the follow-up; for example, we compared those who died at the age of 75 with survivors aged 75 years at the end of 2002. Age, gender, marital status and education constituted the fixed effects part of the model, while only the coefficients of year before death or end of follow-up were allowed to vary randomly between subjects.20 STATA was used for all the calculations.21


Use of care increases rapidly with age particularly in the case of nursing home care, and in all age groups, women use both types of care more than men (table 1). Use of care at a given age is also much higher for the deceased than the survivors. In relative terms, the total 7-year excess in use of hospital and nursing home care is much higher at younger ages, being between four and sixfold among those aged 70–79 years but under twofold among those aged older than 90 years. However, those dying at older ages used more care prior to death than did those dying young.

Table 1

Age-adjusted average number of hospital, nursing home and total care days in the previous 7 years among the deceased and the survivors at the end of follow-up by age and sex

Among the deceased, use of care rises rapidly towards the end of life (figure 1). Hospital care days in the last year of life are twofold compared with the preceding year. However, nursing home care days rise more smoothly over a longer period of time before death and peak at a lower level. Compared with the survivors, total care use is much higher among the deceased, being about four to fivefold in the last year of life but about threefold in the third year before death.

Figure 1

Age-adjusted average number of hospital, nursing home and total care days in the previous 7 years among the deceased and the survivors at the end of follow-up by sex.

Age-adjusted educational differences in hospital care use are small in the 7-year period before death. For nursing home care, the deceased with basic education spent about 40 days more in care than for those with higher education (table 2). Differentials in total care by marital status are much more substantial, are largely due differentials in nursing home care, and are larger among men than among women. Married men use nursing home care on average 130 days less than the non-married during the 7-year period before death. For women, the corresponding difference in nursing home care was 100 days. Among both men and women who survive to the end of follow-up marital status differences are in the same direction as among the deceased but smaller. The excess care use of the deceased was larger among the non-married than among the married. For married men, the cumulative 7-year excess in total care is 162 days, and for the non-married, it is 243 days. For women, the corresponding differences are 233 and 305.

Table 2

Age-adjusted average number of hospital, nursing home and total care days in the previous 7 years among the deceased and the survivors at the end of follow-up by education, marital status and sex

Care use varies strongly by cause of death (table 3). On average, those recorded as dying of dementia use about twice the number of hospital and nursing home care days as do those dying from other causes. Although among women, dementia accounted for only 12% of all deaths, these older women use about a quarter of all hospital and nursing home care days among the deceased women. Those dying from dementia are particularly heavy users of nursing home care with a cumulated total of 553 days in the last 7 years of life for women and 397 for men.

Table 3

Distribution of deaths and age-adjusted average and total (in 1000s) number of hospital and nursing home care days in the previous 7 years by cause of death and sex

Although care use for dementia is largest in the last year of life, high use is observed in all 7 years before death (figure 2, selected causes). Also, cerebrovascular deaths are associated with extended care use profiles. Conversely, deaths from malignant neoplasms are associated with clearly increased hospital care days only towards the last year of life.

Figure 2

Age-adjusted average number of hospital, nursing home and total care days in the previous 7 years before death by cause of death and sex.


Main findings and comparison to other studies

We have studied the end-of-life use of hospital and nursing home care in terms of average care days. Most previous studies have assessed care use over a relatively short period before death, usually 1–2 years. However, we show that higher use of care before death is not limited to this period. The average total use of care is over 100 days more among women in the 12-month period before death than among surviving women and 175 days in the last 2 years. The average cumulative difference in the last 7 years of life is about 300 days. Our results thus clearly show that the excess care use before death is not only limited to the last two years of life but cumulates for a longer time.22 Our results show that care use is higher the older the age at death, and this is true for both nursing home and hospital care. Exact comparisons between our results and those from previous studies are difficult for various data collection and definitional reasons as well as for differences in study aims. However, the pattern of our descriptive results are consistent with previous analyses in the USA and elsewhere with regard to, for example, hospital care and nursing home care,6 ,8 ,9 end-of-life cause-specific care trajectories.15

We observed large and consistent effects of being married on lower total care use that were mostly due to nursing home care (results corroborated with comparison of those living with a partner and those living alone). These differentials are likely to be somewhat underestimated as marital status is measured only once at the beginning of the follow-up. In particular, a part of those classified as married in the current analyses will have become widowed during the follow-up and thus are actually on a higher care use trajectory. This bias is conservative in nature and will result in underestimation of differentials between the married and non-married older people. In spite of this potential underestimation of differentials, we observed substantially larger proximity of death care use among the non-married. Married are less likely to enter nursing homes and remain in care.14 ,23–29 In addition to varying patterns of ill health between the marital status groups, this effect can possibly best be understood by availability of informal care—both emotional support and task help—for the married within the household allowing married people to postpone or avoid nursing home care entry. Furthermore, among those already in nursing home care not having access to informal care in the household of potential return may make it more difficult to return to the community. Education was relatively weakly associated with hospital and nursing home care use, a result consistent with a previous finding showing relatively small care cost differentials by income.10

We found those dying from dementia to be heavy users of nursing home care with a cumulated use of 553 days in the last 7 years of life for women and 397 for men. The high level of nursing home care use is in line with the severe functional consequences of dementia. Further analyses confirmed that among those dying of dementia, the married had substantially reduced nursing home care use; married men used about half and women about 70% of the care used by the non-married (data not shown). This suggests that the availability of informal care strongly affects the use of long-term care. Also, those dying of cerebrovascular and respiratory diseases were heavy care users, but average use of care was only about 50% of that for those dying of dementia. Cancer deaths were characterised by low use of care that only increased in the last year of life.

Lunney et al 15 suggest and empirically assess different functional decline patterns in the last 12 months of life; for example, sudden death being associated with no prior decline of functioning, organ failure with declining but fluctuating functioning and cancer with relatively rapid and sustained decline at the very end of life. Many of the patterns of hospital and nursing home care use observed in this study can be understood within this framework and have important implications for provision of healthcare. By showing very different care use trajectories by cause of death, our results are consistent with the idea that different types of health and nursing home care resources and clinical approaches are needed for different underlying health conditions. In this context, the particular advantage of this data set is that it allows for the assessment of care in both hospitals and nursing homes and for long periods before death.

Methodological considerations

The diagnosis of dementia as a cause of death may be a limitation of our study as diagnosis is not easy and dementia remains understated as a cause in death certificates.30 Knapp and Prince31 estimated—on the basis of estimates on dementia prevalence and excess mortality of dementia patients—that about 10% of all deaths above age 65 are attributable to dementia among men and 15% among women in the UK in 2005. These estimates are in line with data from New Zealand in 2006 and expectedly somewhat higher than what we observe in our data, 5.9% among men and 12.0% among women.32 Estimates based on proportions dying with dementia are roughly twice as high as proportions of deaths attributable to dementia.33

Longitudinal linked registration data with detailed information on nursing home and hospital care use and various socio-demographic indicators allowed us to reliably assess use of care in the years before death. The data do not suffer from missing values, participant mis-reporting or lack of power, which may be serious problems with survey-based data. Furthermore, non-linkage between various registers was negligible and loss to follow-up was mainly related to emigration (0.07% over the study period).

Implications for future demand of care

The fastest growing section of the older population in developed countries is the old–old.34 In this context, our findings on the great importance of age as a determinant of nursing home care, more so than hospital care, is of particular relevance for future care use projections. To contain future increases in care use and expenditure, health policy interventions should aim to maintain the activities of daily living of older people and thus prevent or at least delay nursing home care.10

The combination of improving mortality, reduced gender differentials in mortality and the fact that the cohorts now entering retirement age were in the prime marriage ages at the time of the marriage boom of the early 1960s imply that the proportion of older women, but not men, who are married is likely to increase for about 2 decades in most Western European countries.35 Also an increasing proportion of non-married older people will live as a co-habiting couple.36–38

Coupled with our results showing lower care use among those living with a spouse imply that any increase in the proportion of people who are partnered will tend to reduce the demand for nursing home care places. This change may partly shift the burden of care from the formal to the informal sector, particularly the family—possibly redistributing rather than reducing care. As it is likely that the proportion of partnered women but not men will grow, more demand for informal care giving may be expected from male partners, a source of informal care not traditionally relied on. Partnered older people are also more likely to have and to retain contact with their children. However, as caring for a family member with disability is often harmful for the psychological and physical health of the care giver,39 ,40 support for informal care givers is needed to avoid these adverse health consequences.

The rising trend of dementia is well established globally. Deaths due to dementia have grown rapidly in the past 2 decades and are likely to grow further.33 The increase in dementia is probably due at least partly to changes in diagnostic practices, but purely because of increasing life expectancy and age at death, the number of dementia deaths are likely to keep rising rapidly and their share of all deaths to rise, possibly doubling in the next 25 years.31 ,32 ,41 By showing high use of hospital and nursing home care among dementia patients—particularly those non-married—our results demonstrate that this is likely to mean a major shift towards higher care needs in the future.


The greater importance of age as a determinant of nursing home care use as compared with the relevance of proximity to death for hospital care is demonstrated, but both are relevant to needs. Both types of care are larger the older the age at death. The volume of nursing home care use (as measured by bed days) is greater than hospital care for the old–old. Differentials in service use by marital status are substantial, and, in particular, care use is considerably higher among the non-married in the period close to death. Increasing longevity may result in more nursing home care overall and for dementia in particular. Care use for dementia is substantial for extended periods before death and at least twice that for patients dying for any other cause of death. The results have clear implications for the patients and their caring relatives concerned about quality of life, policy makers estimating care costs and clinicians responsible for end-of-life care decisions.

What is already known on this subject

  • Hospital and nursing home care use is concentrated at the end of life.

  • Different underlying health problems are differently associated with end-of-life functioning.

What this study adds

  • Hospital and nursing home care use patterns vary strongly by cause of death and time to death with those dying of dementia using at least twice as much care for extended periods before death.

  • Old age, female gender and not being married are associated with higher care use patterns among dying patients.


Supplementary materials

Related Data


  • Funding This study was supported by the Academy of Finland (210752 and 205631) and the ESRC (RES-339-25-0002). The study sponsors had no role in the design or conduct of the study; the collection, management, analysis and interpretation of the data or the preparation, review or approval of the manuscript.

  • Competing interests None.

  • Ethics approval We are grateful to the National Research and Development Centre for Welfare and Health (STAKES) and the Social Insurance Institution and Statistics Finland (permission TK 53-576-04 and TK 53-499-05) for making the data available to us.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.