OUP user menu

A tale of two cities: Factors affecting place of cancer death in London and New York

Sandra L. Decker, Irene J. Higginson
DOI: http://dx.doi.org/10.1093/eurpub/ckl243 285-290 First published online: 26 October 2006

Abstract

Background: Most American and English cancer patients prefer to die at home. Factors associated with greater likelihood of dying at home have been contradictory in many studies and no studies have compared the effects of factors in different countries. The objective of this paper is to compare the factors affecting place of cancer death in two major cities, New York and London. Methods: We use data on all individuals aged ≥40 dying of cancer in London (59 604) and New York City (51 668) in the years 1995 through 1998. The probability of death at home is examined in each city as a function of gender, age group (40–55, 56–64, 65–74, 75+), year, type of cancer, and area socioeconomic status, using multiple logistic regression. Results: Although the probability of death at home is the same in the two cities (∼1 in 5), being female lowers the odds of death at home by ∼7% in London, and raises it by ∼22% in New York. Older age is associated with increased odds of dying at home in New York but decreased odds of dying at home in London. Being in the lowest tercile of socioeconomic status (relative to the highest) lowers the odds of death at home by 22% in London and 39% in New York. Conclusion: Site of death varies significantly by patient and area characteristics in both cities, an understanding, which should be taken account of in future planning of end-of-life care.

  • place of cancer death

Introduction

Policies surrounding end-of-life care are likely to become increasingly important as the population ages, and a key issue in end-of-life care is where people die.1,2 There is considerable evidence that a large majority of both American and English cancer patients prefer to die at home.35 A systematic review of 18 studies of preferences found that between 49 and 100% of respondents wanted to be at home.5 Preferences are individual and shaped by many things; home may be preferred because it is associated with feeling connected to the community, familiarity, comfort and the presence of family and loved ones.6,7 However, inpatient care may be preferred because of concerns regarding practical needs and the quality of care at home, feeling a burden on the family, needs for equipment, treatment or symptom management.6,7 Also, some patients have mixed views about the presence of professional carers within the home. Although professional carers provide support, these ‘strangers’ may be regarded as intrusive and compromising the ideal of ‘home’.6 Previous experience with home care (positive or negative) may change a preference, particularly if service provision has broken down in the past, or there have been problems with coverage at night or on weekends.5 Home care is a preference for many people, including those from different cultures,8 although a qualitative study in the UK found that some older people do recognize that institutional death may be preferable in certain circumstances.6

Most previous research on actual place of death has focused on deaths due to cancer, a cause-of-death often associated with above average chances of dying in the hospital.912 Using national vital statistics data, Flory et al.9 found that ∼38% of Americans dying of cancer died at home in 1998, a large increase of 15% from the 1980 fraction, although this analysis excluded 31% of deaths because of inconsistencies in data coding. Using similar data, the percentage of cancer patients who died at home in England rose between 1985 and 1994 from only 1% to 27%.10 In addition to differences across countries, there are considerable differences in place of death within countries. In the USA, for example, it has been found that the proportion of African Americans dying in the hospital is significantly higher than for whites.2,9,13 In Italy, Costantini et al.14 found wide geographic variations. In England, the proportion dying in the hospital was higher for those living in areas with lower socioeconomic status (SES) compared with those in areas with higher SES.15 While some factors, such as haematological malignancy, have consistently been associated with hospital death,12 other factors, such as age, gender, SES and health care system factors vary in their effects between studies.2,3,9,1113,1619 The factors affecting place of death have never been compared across cities or countries, to explore which factors are important in different contexts. In this study, therefore, we compare factors affecting place of death in two major cities, London and New York.

Methods

The populations of London (England) and New York (USA) are ∼7.2 and 8.0 million respectively. Both cities have urban core centres, a mix of high and low income populations and 13 (London) and 19 (New York) teaching hospitals in their inner core. Population and family structures are similar. The proportion of people over 85 years in the inner core of London is 1.5% (1998) and in New York 1.7% (2000).20 The goal of this study is to analyse differences in the place of death in these two cities that have similar demographic attributes but very different health care systems. Data on place of death and patients' characteristics were obtained for all cancer deaths occurring in England and Wales in the years 1995 through 1998 from the UK Office of National Statistics.10 Deaths among patients aged ≥40 who were residents of the London Health Region were selected. Data on place of death and patients' characteristics were also obtained for all deaths in New York City in the years 1995 through 1998 from the New York City Department of Health and Mental Hygiene's Bureau of Vital Statistics. Of the deaths occurring in New York City ∼94% were among New York City residents. Most of the remaining deaths occurred among residents of New York State and were included in the analysis. Deaths due to cancer and among patients at least 40 years old were selected.

Our main outcome measure is the probability that an individual cancer death occurring in New York or London between 1995 and 1998 takes place at home In each city, we examine this probability as a function of gender, age group (40–55, 56–64, 65–74, 75+), SES, year, and type of cancer, using multiple logistic regression. Type of cancer was divided into 10 categories according to the primary cause of death, which was coded using the International Classification of Diseases, 9th Revision (ICD-9) in both cities during the 4 years of data analysed.

Mortality data from London were merged with information on area SES from the 1991 UK Census by electoral ward of residence. Electoral wards are small geographic areas, usually ranging from 3000 to 11 000 people. Mortality data from New York was matched to measures of area SES by zip code from the 2000 USA Census. The median size of a zip code in New York City is ∼60 000 individuals, though zip codes boundaries within the city generally mark distinct neighbourhood divisions. Although data on place of death in London and New York were merged with census data on local area SES in each city, available measures of area socioeconomic status from the USA and UK censuses are not comparable. For London, the 1991 census was used to calculate an index of ‘income deprivation’ by electoral ward. This index is a commonly used measure of socioeconomic status in England, and is based on local census information on the fraction of individuals in various income support programs.21 Higher numbers indicate higher levels of deprivation. The mean area index of income deprivation for the London data is 24.2, and the range is 1.9 (Westminster)–61.9 (Tower Hamlets). For New York, the 2000 USA Census was used to calculate the proportion of the population under the Federal poverty line by zip code. In the New York data, the mean percent under the poverty line is 17.4%, and the range is 5–50% (Upper East Side of Manhattan–South Bronx). Terciles of socioeconomic status were calculated within each city based on all deaths in the sample.

Altogether, 1317 observations were dropped in London due to missing information (mostly on place of death) and 5324 in New York (mostly due to missing zip codes of residence). The final sample size for analysis during the 4-year period is 59 604 cancer deaths among individuals at least 40 years old in London, and 51 668 in New York City.

Results

The demographic and clinical characteristics of cancer deaths in the two cities are similar. Approximately half of the deaths are among females (49.2% in London and 51.7% in New York), and 40.8% are among those aged 75+ years in each city (mean age is 70.4 in London and 72.2 in New York). The distribution of primary cancers follows normal patterns, with cancers of the lung and intra-thoracic, upper gastrointestinal (GI), and colon and rectum being the most common (24.4, 16.8 and 10.2%, respectively of cancer deaths in London; 23.4, 16.5 and 12.1%, respectively in New York). Nearly 40% of deaths in New York were among non-whites. Ethnicity or race is not coded in the London data, although 12.5% of deaths in the London sample were among individuals who were born outside of the UK.

Table 1 presents the mean proportion of cancer deaths at home by patient and area characteristics. In both cities ∼1 in 5 deaths were at home (20.0% in London and 19.9% in New York). Although not reported in the table, there were similar proportions of deaths in nursing and residential homes in each city (5.5% in London and 6.3% in New York). In New York, the balance of deaths was recorded as inpatient deaths (73.7%). In London, only 56% of deaths were in hospitals, while 18.5% occurred in inpatient hospices.

View this table:
Table 1

Proportion of cancer deaths at home by selected patient and area characteristicsa

LondonNew YorkDifference (London − New York)P-valueb
All20.019.90.10.935
Male20.918.52.40.000
Female19.021.3−2.30.000
White (USA)/non-immigrant (UK)20.023.6−3.60.000
Non-White (USA)/immigrant (UK)19.414.25.20.000
Age group
    Ages 40–5523.715.97.80.000
    Ages 56–6423.119.04.10.000
    Ages 65–7421.619.52.10.000
    Ages 75+17.222.0−4.80.000
SES tercile
    Highest21.424.8−3.40.000
    Middle20.120.8−0.70.117
    Lowest18.314.14.20.000
Year
    199521.318.72.60.000
    199620.019.40.60.184
    199719.520.5−1.00.042
    199819.121.3−2.20.000
Type of cancer
    Head and neck20.717.33.40.027
    Lung and intrathoracic22.920.42.50.000
    Upper gastrointentinal (GI)23.920.13.80.000
    Colon and rectal22.623.0−0.40.616
    Female breast18.520.1−1.60.039
    Female genital18.520.8−2.30.025
    Male genital17.420.2−2.80.005
    Kidney and bladder17.721.7−4.00.000
    Lymphatic and haematopoietic tissue11.613.6−2.00.004
    Other23.715.97.80.000
  • a: All cancer deaths (59 604 in London, and 51 668 in New York City) among those aged 40+ in the years 1995–1998

  • b: Wald test

Although the overall fraction of deaths at home is not significantly different between the two cities, the probability of at home death is increasing over time in New York and falling in London. Compared with London, deaths among women and older individuals were more likely to be at home in New York. In contrast, those with lower socioeconomic status are less likely to die at home in New York compared with London.

Table 2 presents estimates of the predictors of death at home by city, using multiple logistic regression analysis. Being female lowers the odds of death at home by ∼7% in London, and raises it in New York by ∼22%. Being non-white significantly lowers the odds of death at home in New York by a very substantial 34%. Using the point estimate at the centre of the confidence interval (CI), being an immigrant lowers the odds of death at home in London by 6%, though this relationship is not statistically significant.

View this table:
Table 2

Predictors of death at homea

LondonNew YorkP-valueb
Female0.93 (0.89–0.98)1.22 (1.16–1.28)0.000
Non-white (USA)/immigrant (UK)0.94 (0.88–1.00)0.66 (0.62–0.69)0.000
Age group (relative to 40–55 years)
    Ages 56–640.93 (0.86–1.00)1.19 (1.10–1.30)0.002
    Ages 65–740.83 (0.77–0.89)1.17 (1.08–1.26)0.000
    Ages 75+0.63 (0.58–0.67)1.26 (1.17–1.35)0.000
SES tercile (relative to highest)
    Middle0.90 (0.86–0.95)0.84 (0.80–0.88)0.006
    Lowest0.78 (0.74–0.82)0.61 (0.57–0.65)0.000
Log likelihood−29 317−25 178
  • a: The table presents odds ratios and 95% CI (in parentheses) from logit estimates of the probability of cancer death at home for those aged 40+ in the years 1995–1998. Although not reported, controls for year effects and type of cancer (10 categories) are included

  • b: The P-value reports the significance of the difference of the effect of the variable in New York City and London on the probability of death at home. This significance level is the one associated with the interaction between the variable and a city dummy in a pooled New York City/London logistic model with full city interactions

The probability of death at home is higher in London than in New York until age ∼80, when the probabilities ‘cross over’ and death at home is then more common in New York than in London (figure 1). Although not pictured, the probability of death in a nursing home is similar in the two cities until ages in the late 80s, when it does increase in London more than in New York. However, the probability of death in inpatient hospice declines substantially in London around age 80. In New York, older age is associated with an increase in both the probability of death at home and in nursing homes and a decrease in the probability of death in the hospital. In London, older age is associated with an increase in the probability of death in the hospital and nursing home, and a decrease in the probability of death at home and in inpatient hospice.

Figure 1

Proportion of cancer deaths at home, 1995–1998

Lower SES significantly lowers the probability of in home death in both cities, but significantly more so in New York than in London. Specifically, as can be seen in table 2, being in the middle tercile of SES (relative to the highest tercile) lowers the odds of death at home by 10% in London and 16% in New York. Being in the lowest tercile of SES (relative to the highest) lowers the odds of death at home by 22% in London and by 39% in New York. This finding is further pictured in figures 2 and 3. The probability of death at home by SES is noisier in London than in New York, though the probability of death at home is generally highest for those in the highest tercile of SES and lowest in the lowest tercile of SES throughout the age range. The difference in the probability of death at home in London by SES comes mostly from those at younger ages. In New York, the difference in the probability of death at home by SES is substantial throughout the age range.

Figure 2

Proportion of cancer deaths at home, NYC, 1995–1998, by SES tercile

Figure 3

Proportion of cancer deaths at home, London, 1995–1998, by SES tercile

Comments

In London and New York, one in five cancer deaths are at home, much lower than preferences for home death,3,5 including among studies in London4 and New York.22 Of the deaths in institutions, inpatient hospices played a significant role in London, accounting for 18.5% of cancer deaths, a figure that is similar to national statistics for the UK.10 The London Health Region has 389 inpatient hospice beds (54.1 per million population) in 17 inpatient hospice or palliative care units, most of which are freestanding from teaching hospitals. Six inpatient units are managed by the National Health Service and 11 by independent charities.23 A national survey in England in 2002 found that after home, inpatient hospice was the second highest preference (24%) for end-of-life care, with hospital preferred by only 11%.24 A meta-analysis showed improved pain, symptom control and psychological support in inpatient hospice compared to standard end-of-life.25 It may be worthwhile exploring whether such inpatient ‘hospice’ services would be an alternative when home is not possible in New York. A study of older patients in England also highlighted hospice as an alternative preference,26 though the chance of death in inpatient hospice declines substantially in London at age ∼80.

The divergent relationship between factors affecting home death in the two cities is intriguing. In New York, women and those over 75 years old were more likely than others to die at home. The opposite relationships between gender and age and the probability of death at home are found in London. These differences were maintained in our multiple regression analysis and are not easily explained by symptom profiles or family structures. The cancer diagnoses were similar in the two cities. The proportions of persons over 85 years living alone in the core and first ring of the two cities are quite similar: 54 and 48% respectively in London, and 55 and 39% respectively in New York.20 New York has more acute hospital beds and physicians per 1000 than London, but it is not clear how this might affect home care.20 The correlation between greater age and increased chance of death at home in New York is supported by Levinsky et al. who found that Medicare expenditures in the last year of life are lower for older individuals on Medicare than for younger individuals on Medicare, due to less aggressive treatment of older Medicare patients (including fewer hospital admissions) compared to younger ones.27 The relationship between age and probability of death at home in New York may also reflect differences in health and social services availability for those over and under age 65, in particular the intensive home care support provided by the Medicare hospice program for some individuals who are at least 65 years old.

Numerous studies have found a relationship between lower SES and higher rates of cancer death both in England2830 and in the United States.3132 In addition to a relationship between lower SES and the probability of cancer death, our study finds a significant relationship between lower socioeconomic status and the experience of death. Patients with lower SES are more likely to die in an institutional setting in both New York City and London. This may reflect fewer resources and less ability to provide adequate care at home including nursing and other support services. The greater effect in New York may be because the range of socioeconomic status is greater in New York than in London – the scales used in the two cities are different and we divided the populations into terciles. An alternative explanation may be that the National Health Service in the UK, which is universally free at the point of delivery and provides free medication for those aged over 64 years, reduces the effects of lower SES on aspects of health care delivery. Other factors, including individual preferences, may also be important.

In New York, non-white patients were 34% less likely to die at home compared to white individuals. This finding is supported by Flory et al.'s national analysis9 and other studies.2 In London, there was no significant difference between immigrants and non-immigrants in place of death, although routine data on race or ethnicity are not collected. A small study in South London comparing matched samples of white UK born and black Caribbean patients also found few differences in home cancer death rates, although hospice deaths were less common in the black Caribbean group.33

Although the probability of death at home is about the same in London and New York, we find evidence that this probability is falling in London but rising in New York. The reasons for this difference are unclear and call for a comparative study of the US and the UK, which could compare patient experiences with the quality of home and hospital care. The greater emphasis on home care in the USA, especially through the Medicare hospice benefit, may play a role in the increasing probability of at home death in New York. Increasing managed care penetration in the USA has led to greater use of hospice in recent years, which may be one driver of falling in-hospital death rates in New York City.3435 The Medicare hospice benefit includes hands-on nursing care which is not provided by hospices or palliative care teams in the UK. The hospice benefit may also serve to enhance continuity of care in the US by providing services through one organization compared to care provided in the UK by a fragmented and complex web of providers (family doctor, district nurse, palliative care nurse, and social services). The rising probability of at home death in New York might also reflect difficulty accessing hospital care in the USA because of the Medicare Part A deductible and limits to coverage after 60 days of hospitalization in a year.

The factors we were able to study were limited to those available in national vital statistics registries. A wide range of other factors may be influencing home death rates.8,16 For example, severity of symptoms which are known to drive the need for hospitalization could not be directly assessed in this study. We were also not able to assess the relative role of family support in the two cities. Nevertheless, we find that some groups fare better than others in terms of achieving a home death in both systems. Further comparisons of end-of-life support in London and New York should help to identify which factors help patients be cared for in their preferred location, and to enable health care planners in each city learn from the successes of the other. Such research should assess the effects of funding mechanisms, health and social services, personal preferences, environmental, and other social factors.

Acknowledgments

This work was begun while Professor Higginson was the 2004 Hatch lecturer. The Harold Hatch International Lectureship in Geriatrics and Gerontology provides support for leaders in the field of aging to visit the International Longevity Center-USA and the Henry L. Schwartz Department of Geriatrics and Adult Development at the Mount Sinai School of Medicine in New York City. It is supported by the Cobble Pond Foundation and the Margaret Milliken Hatch Charitable Trust.

Key points

  • The probability of cancer death at home compared to other sites is the same in London and New York City (∼1 in 5).

  • However, site of cancer death varies significantly by patient and area characteristics within both London and New York City.

  • For example, being in the lowest tercile of socioeconomic status (relative to the highest) lowers the odds of cancer death at home by 22% in London and 39% in New York.

References

View Abstract