Do social factors affect where patients die: an analysis of 10 years of cancer deaths in England

Transcription

1 Journal of Public Health Medicine Vol. 21, No. 1, pp Printed in Great Britain Do social factors affect where patients die: an analysis of 10 years of cancer deaths in England Irene J. Higginson, Brian Jarman, Paul Astin and Susan Dolan Abstract Background This study investigated whether indices of social deprivation were related to the proportion of cancer patients who died at home. Methods Data were derived from death registrations for all cancer deaths in England. Two indices of deprivation (Underprivileged Area Score (UPA), or Jarman, and Townsend scores) were calculated for each electoral ward; 1991 Census data were used. The scores use combinations of variables, including the percentage in overcrowded homes, the percentage of elderly people living alone, the percentage of one-parent families, etc. A high score indicates more deprivation. The main outcome measures were the proportion (in five and ten year averages) of cancer deaths which occurred at home, calculated for every electoral ward (with populations usually ranging from 5000 to ). Spearman rho was used to test for correlations between the proportion of cancer deaths at home and deprivation score. Electoral wards were categorized by deprivation score into three groups of equal size and analysed over 10 years. Multivariate analysis was used to determine the relative association of different patient based and electoral ward variables with cancer death at home. p < 0:05 (two-tailed) was taken as significant. Results There were over 1.3 million death registrations from cancer in the 10 years. The proportion who died at home was 0.27, in hospital 0.47, and other setting There were wide variations ( ) in the proportion of people who died at home in different electoral wards. Small inverse correlations were found between the percentage who died at home and the UPA (¹0.35; p < 0:001) and Townsend (¹0.26; p < 0:001) scores. The correlations were greatest in North Thames (¹0.63, UPA) and smallest in West Midlands ( 0.20, UPA). The proportion of home deaths for the different bands of deprivation were: 0.30 (low deprivation), 0.27 (middle deprivation) and 0.24 (high deprivation). Plotting the trends over 10 years suggests no change in this relationship. Multiple regression analysis predicted several ward and patient characteristics and accounted for 30 per cent of the variation. Increased age (patient variable), Jarman score and ethnic minorities (both ward variables) were associated with fewer patients dying at home. Being male and having cancer of the digestive organs were associated with home death. Conclusion There are wide variations in the percentage of cancer deaths at home in different electoral wards. Social factors are inversely correlated with home cancer death, and may explain part of this variation. Home care in deprived areas may be especially difficult to achieve. Keywords: palliative care, terminal care, home, primary care, social class, deprivation, cancer Introduction Predictions for the year 2020 show an ageing population with world-wide more people who die from chronic rather than acute diseases. 1 Thus health care will increasingly be concerned with achieving the best possible quality of life for patients and for their families, and with providing palliative care. Part of this includes meeting patients and families wishes of place of care and of death. 2 However, patients and families preferences are often not met: far more patients wish to be cared for and to die at home than is currently achieved. Dunlop et al., 3 Townsend et al. 4 and Ashby et al. 5 found that between 50 and 70 per cent of cancer patients would prefer to be cared for at home for as long as possible and to die at home. Had circumstances been favourable, Townsend et al. found that 41 out of 59 (67 per cent) of patients would have wished to die at home. In a longitudinal study of patients in the care of a hospice home care team, Hinton found even towards the end of their lives 54 per cent of patients and 45 per cent of relatives wished home to be the place of care. 6 In many countries there has been an increasing trend towards the hospitalization of death. In England, studies based on random samples of deaths which had an identified terminal period, showed that the proportions of patients in England who Department of Palliative Care and Policy, King s College School of Medicine & Dentistry and St Christopher s Hospice, Lawrie Park Road, Sydenham, London SE26 6DZ. Irene J. Higginson, Professor and Head of Department of Palliative Care and Policy Department of Primary Health Care and General Practice, Imperial College School of Medicine, Norfolk Place, London W2 1PG. Brian Jarman, Professor of Primary Health Care and Head of Department Paul Astin, Research Fellow Susan Dolan, Research Analyst Programmer Address correspondence to Professor Higginson. Faculty of Public Health Medicine 1999

2 DO SOCIAL FACTORS AFFECT WHERE PATIENTS DIE? 23 died in institutions increased between 1969 and 1987 from 46 per cent to 50 per cent (hospitals) and 5 per cent to 18 per cent (hospices and other institutions). Meanwhile the proportion who died at home decreased from 42 per cent to 24 per cent. 7 There was also an increase in the proportions living in institutions and being admitted to hospitals in the last year of life. Currently in England and Wales, 26 per cent of cancer deaths occur at home. 8,9 In the United States, Mann et al. found that, between 1980 and 1990, 22 per cent of 468 gynaecological cancer patients died at home. Examination of the variable year of death demonstrated that the likelihood of death in the hospital generally increased from 1980 to 1990 despite aggressive efforts by caregivers to facilitate and encourage death at home. 10 The factors associated with increased likelihood of admission have been reported as: breast cancer, being of minority ethnic origin, lower social class or income, increasing age, patients living alone or with unfit relatives, carers becoming over-burdened and being unable to continue, poor symptom control, and longer dying trajectories All these indicate a social component, although this has not been quantified. Indices of social deprivation have been developed to indicate general practitioner (GP) workload and to assist in the local allocations of health and social resources. 17,18 These are based on weighting factors in the 1991 Census data, such as the proportion of people in social class IV and V and the proportion of overcrowded households. They can be used to estimate deprivation and/or need for health care in small geographical areas (electoral wards) which have populations usually ranging from 3000 to people. The indices are correlated with admission rates, mortality rates and morbidity data. One health district in London found eight-fold variations in the proportion of home cancer deaths between electoral wards. 19 This was inversely correlated with deprivation (Spearman rho ¼¹0:65). However, it is not known whether this finding exists more widely. This study calculated the proportion of cancer patients who died at home in each electoral ward in England, and determined whether this was related to social deprivation. Methods Data Data on place of death and patients characteristics were derived from death registrations, which are collected for all of England by the Office for National Statistics (ONS; previously Office of Population Censuses and Surveys). The death registration data were purchased for the last 10 years ( inclusive) for all cancer deaths in England. The 1991 Census was used to calculate variables which have been shown to be associated with economic and social deprivation at electoral ward level. Values for the following variables were calculated: economic activity (unemployed by age-group), no car, low social class or unskilled persons, overcrowding, households lacking or sharing amenities, nonpermanent accommodation, housing tenure (not owneroccupier), persons permanently sick by age-group, dependants in households, persons in one-parent households, children under five years, lone elderly persons, one-year migrants, ethnic minorities (new Commonwealth and Pakistani) by age-group. Composite deprivation indices Underprivileged Area Score (UPA or Jarman score), Townsend and Carstairs scores were calculated using the standard formula for each electoral ward. 17,18 Main outcome measure The main outcome measure was the proportion of home cancer deaths in each electoral ward between 1985 and Analysis The relationship between the proportions of cancer deaths at home in each ward and both the UPA and Townsend scores was examined using Pearson correlation coefficients. These were calculated for the whole data set and then separately by Regional Health Authority (1996 boundaries). We examined the 10 year trends in the above separately for: (1) those areas which have high and low underprivileged area scores, and (2) urban and rural areas. For (1), the wards were grouped into three equal-sized (by number of wards) bands: top third, middle third and lowest third UPA scores. For (2), wards were grouped into urban (metropolitan and city), mixed and rural using the definitions and variables from the 1991 Census. To explore the relationship between home cancer death and the possible explanatory or predictor variables multivariate analysis was performed. Four approaches to multivariate analysis were undertaken, in an attempt to develop the best predictive model and to determine which variables were independently most strongly associated with home death. These were: multiple regression, factor and logistic regression analysis and multi-level modelling. The analysis was performed on data for the 10 years and the last five years to assess whether any significant change occurred. Results During the 10 year period there were over 1.3 million death registrations from cancer, with slightly more men than women. The numbers were constant over the years. Three-quarters of these were aged over 65 years. The proportion who died in an NHS hospital fell gradually from 0.58 (1985) to 0.47 (1994), whereas the proportion who died in private hospitals or communal establishments increased from 0.12 (1985) to 0.26 (1994). The percentage who died at home fell initially from 0.27 (1985) to (1992) but appears to have been increasing since then and in 1994 was

3 24 JOURNAL OF PUBLIC HEALTH MEDICINE Table 1 Correlation of 1991 UPA and Townsend ward scores and number (with proportion given in parentheses) of those dying of cancer at home ( ) presented by RHA r for Region r for UPA Townsend Anglia & Oxford ¹0.231 ¹ (0.29) North Thames ¹0.625 ¹ (0.22) North West ¹0.302 ¹ (0.27) Northern & Yorkshire ¹0.243 ¹ (0.28) South & West ¹0.281 ¹ (0.22) South Thames ¹0.370 ¹ (0.27) Trent ¹0.261 ¹ (0.28) West Midlands ¹0.201 ¹ (0.26) Overall ¹0.343 ¹0.266 r, correlation coefficient. All correlations, p < 0:001. At ward level the number of cancer deaths in the 10 years varied from one to 1102 (mean 156, SD 112) and the deaths at home from zero to 292 (mean 41, SD 32). There were wide variations between wards in the proportions of cancer deaths at home. Excluding those wards with fewer than 30 cancer deaths in the 10 years, the proportion of home cancer deaths ranged from 0.05 to For the 10 years data combined, modest negative correlation coefficients were found between the percentage who died at home and the UPA and Townsend index (¹0.345 and ¹0.266, respectively; p < 0:001). Findings for other composite measures were similar to those for the Townsend index. The correlations were highest in the two Thames Health Authorities and lowest in the West Midlands (see Table 1 and Figs 1 and 2). When wards were grouped into low, medium and high UPA bands there was a marked difference between the three bands, which remains constant over the years: the high band (most deprived) shows the proportion of deaths at home at 0.24 and the low band (least deprived) at The 95 per cent confidence intervals demonstrate that these differences are highly significant (see Fig. 3). No such differences were found in the percentage of cancer deaths at home when wards were categorized into the three groups, largely urban, mixed and largely rural, beyond the existing findings of lower proportions of home cancer deaths in wards with high deprivation and in the Thames Regions, especially in the central London districts. In multivariate analysis the results of multiple regression, logistic regression, factor analysis and multi-level modelling were broadly similar. Multiple regression analysis gave the best predictive model. The R 2 statistic was 0.30, indicating that about 30 per cent of the variance is accounted for in the model. In this (Table 2) the most important variables were: age, Jarman score, head of household in social class IV or V, ethnic minorities, high-dependence households, cancer of digestive organs, over-65s. Increased age (patient variable), Jarman score and ethnic minorities (both ward variables) were associated with fewer patients dying at home. Being male and having cancer of the digestive organs were associated with home death. Once the effects of age and Jarman score were accounted for, small further effects of increased likelihood of a home death Figure 1 North Thames; proportion of cancer deaths at home in each electoral ward versus UPA score,

4 DO SOCIAL FACTORS AFFECT WHERE PATIENTS DIE? 25 Figure 2 West Midlands; proportion of cancer deaths at home in each electoral ward versus UPA score, were found if people lived in a ward with a higher percentage of: head of household in social class IV or V, high-dependence households, people over 65 years. There was no difference in these results when data for the last five years were analysed independently. Discussion Both the UPA and Townsend index were negatively related to the proportion of cancer patients dying at home. Thus, the higher the deprivation the smaller, in general, was the proportion dying at home. There may be several explanations for this finding. Higher deprivation reflects lower income and fewer resources. This may result in less ability to provide adequate care at home; for example, families with more resources may be able to afford home nursing or support. However, the UPA index was originally developed to reflect primary care workload and pressure on the services of GPs. An alternative explanation would be that in areas with high levels of primary care workload, few patients can be cared for at home. There may also be differences between areas because of Figure 3 Trends in deaths at home from cancer by deprivation band (one-third of population, sorted by Jarman UPA score).

5 26 JOURNAL OF PUBLIC HEALTH MEDICINE Table 2 Results of multiple regression analysis; significant ward level variables (given in descending order of importance) and patient characteristics in order of importance along with their contribution to the R 2 statistic Variables R 2 contribution Cumulative R 2 Coefficient Age UPA score Head of household in social class IV or V Male Ethnic minorities High-dependence households Cancer of digestive organs Over-65s differences in the preferences of local people or in local services, such as community, nursing, specialist palliative care or 24 hour care. There was a wide spread of deprivation scores in the two Thames regions, which may be, in part, responsible for the higher correlations in these two regions. Our findings are different from those found by Sims et al. in Doncaster, 20 where cancer patients with a lower social class died at home in similar proportions compared with other groups. In their study, those of lower social class were least likely to die in a hospice. However, their study was based on a relatively small socially deprived area. Our data analysed underprivilege within electoral wards, which might reflect area effects, rather than individuals social class. We attempted to analyse individuals social class in our data set, but found that social class data were missing in over 50 per cent of the individuals, mainly among those aged over 75. This analysis is limited in three respects. First, although place of death is reported here, place of care is equally or more important. However, data are not available to allow us to study this across large populations. Furthermore, place of death is important because many more patients wish to die at home than actually achieve this. Last-minute admissions can be inappropriate and distressing. Power and Kearney found that of 100 admissions to a hospice, 40 died within a week of admission, 16 within 48 hours and 11 within 24 hours. They reported that distress may be caused to patient, family and staff when death occurs shortly after admission. 21 Boyd analysed the records of 47 patients (9.8 per cent of all admissions during a six month period) who died within 48 hours of admission to a London hospice. Respiratory symptoms were commonest. Although some patients were reported to have delayed admission for as long as possible, choosing to come in when they were close to death, in other cases the family and hospice staff regretted admission. 22 Studying the trends of place of death can help to understand service needs, although the findings must be interpreted with caution, given the lack of information on place of care. Second, as a result of basing the study on routine statistics, we were unable to explore relationships in depth and to include variables such as family circumstances, coping strategies or services available. This would require a study of a different kind, probably collecting prospective data. Our study has provided information which could be explored and tested in such studies. Finally, small area statistics, such as those produced by examining data by electoral ward, often involve small numbers where fluctuations can appear to be significant when they are not. In some small areas we found enormous variations in the cancer deaths at home (from zero to 0.80). However, although some of these are due to small numbers, in most instances they were not. Home care is often reported to be more cost-effective than hospital care. 23,24 In this study we found that social deprivation explained a significant component of the variation in home cancer deaths. The 6 per cent difference in the percentage who die at home between the least and most deprived thirds is a larger effect than any new service has been able to demonstrate within a whole community. 25 Thus deprivation not only contributes to shortening life, 26 but also makes home death, if desired, significantly less likely. Social class and cancer site are associated, therefore cancer site may be an important confounding variable. However, in a separate analysis we found that people who died from lung cancer were significantly more likely to die at home than those who died from other cancers. 27 Lung cancer is more common in lower social classes at all ages Thus the higher number of deaths from lung cancer in people of lower social class would tend to increase the proportion of home deaths the opposite to our findings. Beyond the finding of an association with deprivation scores, there remains marked variation which is unexplained. Multiple regression analysis identified the independent effects of several ward and patient characteristics, and managed to account for about 30 per cent of the variation, which is reasonably good. Age and UPA score were the most important predictors. Some individual variables in the UPA index were also important. However, variables contained in the UPA index will often be close to the UPA composite variable

6 DO SOCIAL FACTORS AFFECT WHERE PATIENTS DIE? 27 in terms of influence. Interestingly, the composite variable, rather than individual components, was the most strongly predictive. We were unable to develop a composite variable which explained more of the variation. Using the multiple regression models means that we can reasonably predict how many cancer patients will die at home in each ward. This may be of benefit for healthcare professionals in deciding on resource allocation. Further analysis of supply effects, such as available palliative care services, social services and GP community hospitals, is needed to investigate the variation further. Acknowledgements Funding is acknowledged from the NHS National Research and Development Programme on Cancer. We thank Karen Dunnell and Susan Kelly of ONS for their advice and support in the preparation of the data for analysis and for their comments on the full report of the findings. Detailed data providing information on individual wards or districts are available on our World Wide Web page kcsmd/palliative/top/html References 1 Murray CJ, Lopez AD. Alternative projections of mortality and disability by cause : Global Burden of Disease Study. Lancet 1997; 349(9064): WHO Expert Committee. Cancer pain relief and palliative care. World Health Organization Technical Report Series, No Geneva: WHO, Dunlop RJ, Davies RJ, Hockley JM. Preferred versus actual place of death: a hospital palliative care support team study. Palliat Med 1989; 3: Townsend J, Frank AO, Fermont D, et al. Terminal cancer care and patients preference for place of death: a prospective study. Br Med J 1990; 301: Ashby M, Wakefield M. Attitudes to some aspects of death and dying, living wills and substituted health care decisionmaking in South Australia: public opinion survey for a parliamentary select committee. Palliat Med 1993; 7(4): Hinton J. Can home care maintain an acceptable quality of life for patients with terminal cancer and their relatives? Palliat Med 1994; 8(3): Cartwright A. Changes in life and care in the year before death J Publ Hlth Med 1991; 13(2): Office of Population Censuses and Surveys. Mortality statistics general. Series DH1 No. 26. Review of the Registrar General on deaths in England and Wales, London: HMSO, Office of Population Censuses and Surveys. Population trends 74, Winter Tables 13 and 14. Deaths rates by age, sex and selected causes. London: HMSO, Mann WJ, Loesch M, Shurpin KM, Chalas E. Determinants of home versus hospital terminal care for patients with gynaecological cancer. Cancer 1993; 71(9): Clifford CA, Jolley DJ, Giles GG. Where people die in Victoria. Med J Aust 1991; 155(7): , Costantini M, Camoirano E, Madeddu L, et al. Palliative home care and place of death among cancer patients: a population-based study. Palliat Med 1993; 7(4): Hunt R, Bonett A, Roder D. Trends in the terminal care of cancer patients: South Australia, Aust N Z J Med 1993; 23(3): Buehler JA, Lee HJ. Exploration of home care resources for rural families with cancer. Cancer Nurs 1992; 15(4): Jimeno-Aranda A, Catalan R, Garcia-Ruiz M, et al. Where does the terminal patient die? Aten Primaria 1993; 11(3): Cartwright A. Social class differences in health and care in the year before death. J Epidemiol Commun Hlth 1992; 46: Jarman B. Identification of underprivileged areas. Br Med J 1983; 286: Morris R, Carstairs V. Which deprivation? A comparison of selected deprivation indices. J Publ Hlth Med 1991; 13(4): Higginson I, Webb D, Lessof L. Reducing hospital beds for patients with advanced cancer. Lancet 1994; 344: Sims A, Radford J, Doran K, Page H. Social class variation in place of cancer death. Palliat Med 1997; 11: Power D, Kearney M. Management of the final 24 hours. Irish Med J 1992; 85(3): Boyd KJ. Short term admissions to a hospice. Palliat Med 1993; 7: Zimmer JG, Groth-Juncker A, McCusker J. Effects of a physician-led home care team on terminal care. JAm Geriatr Soc 1984; 32(4): Kidder D. Hospice services and cost savings in the last weeks of life. In: Mor V, Greer DS, Kastenbaum R, eds. The hospice experiment. Baltimore: Johns Hopkins University Press, 1988: Higginson I. In: Stevens A, Raftery J. eds. Health care needs assessment: The epidemiologically based needs assessment reviews. Oxford: Radcliffe Medical Press, Soni Raleigh V, Kiri A. Life expectancy in England:

7 28 JOURNAL OF PUBLIC HEALTH MEDICINE variations and trends by gender, health authority and level of deprivation. J Epidemiol Commun Hlth 1997; 51: Higginson IJ, Astin P, Dolan S. Where do cancer patients die? Ten-year trends in the place of death of cancer patients in England. Palliat Med 1998; 12 (in press). 28 Brown J, Harding S, Bethune A, Rosato M. Incidence of health of the nation cancers by social class. Population Trends 1997; 40(7): Pollock AM, Vickers N. Breast, lung and colorectal cancer incidence and survival in South Thames Region, : the effect of social deprivation. J Publ Hlth Med 1991; 19(3): Accepted on 5 August 1998