Survey
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
PhD Thesis Factors of importance for organising palliative care at home Register-based population studies of end-stage cancer patients Birgit Aabom Research Unit for General Practice, Institute of Public Health Faculty of Health Sciences 2006 This thesis is based on three papers (I-III) I. Population-based study of death of patients with cancer: implications for GPs. Br J Gen Pract 2005; 55: 684-9. II. Defining cancer patients as being in the terminal phase: who receives a formal diagnosis and what are the effects? J Clin Oncol 2005; 23: 7411-6. III. Does persistent involvement by the GP improve palliative care at home for end-stage cancer patients? Palliative Medicine 2006; 20; 507-12. 2 1 PREFACE This PhD study was carried out during my appointment as a research fellow at the Research Unit for General Practice in Odense, University of Southern Denmark, in the period of 2002-2006. I sincerely thank my chief supervisor Jakob Kragstrup, MD, Professor and Head of the Research Unit for General Practice for his encouragement and guidance that made this thesis possible. I also want to thank my other three supervisors, Henrik Støvring, PhD, Hindrik Vondeling, PhD, and Leiv S. Bakketeig, Professor, for their inspiring supervision and support throughout the process. Henrik Støvring especially for, from the very start, enthusiastically supporting the idea of using the Danish registers to analyse factors of importance for end-stage cancer patients and his creative thinking, not only in statistical issues. Special thanks to secretary Lise Stark for her competent proofreading and bright comments to the manuscripts. I would like to thank Hans Henrik Storm and Gerda Engholm from the Department of Cancer Prevention and Documentation, Danish Cancer Society, for inspiring input in the planning phase and access to data from the Danish Cancer Register. Also thank you to Palle Pedersen, MD, from the Municipality of Odense and to Michael Bruus, Department of Health Sector Planning in the County of Funen, for access to data. My hours in front of the computer linking register data would have been a weary task without the first-class social life with the other PhD students, with local GPs working in the field and with the administrative staff at the Research Unit. I want to thank all for support, for provocative questions and inspiring hours during formal and informal breaks in the coffee room. Last but not least, I wish to thank my husband Thorben and Mette, Sofie and Bodil for their love and daily support throughout the process. Odense, March 2006 Birgit Åbom 3 2 TABLE OF CONTENTS Page 1. PREFACE 3 2. TABLE OF CONTENTS 4 3. ABBREVIATIONS 7 4. DEFINITIONS 8 5. THIS THESIS AT A GLANCE 9 6. INTRODUCTION 10 6.1. Place of death of cancer patients 10 6.1.1. Place of death in western countries – short historical view 10 6.1.2. Is dying at home the same as a “good death”? 11 6.2. Organisation of palliative services in Denmark 12 6.2.1. National Health System 12 6.2.2. Palliative care at basic and specialised level 12 6.2.3. Palliative care at basic level 13 6.2.3.1 Defining cancer patients as terminal? 13 6.2.3.2. The terminal declaration (TD) 14 6.2.3.3. GP in end-of-life cancer care 15 6.3. Research in palliative care 16 6.3.1. Historical view and methodological challenges 16 6.3.2. Mortality follow-back design 17 7. AIMS 19 8. STUDIES 20 4 8.1. General methodological issues 20 8.1.1. Setting 20 8.1.2. Observation period (Time scale) 20 8.1.3. Health Care Data 21 8.1.3.1. Danish Civil Registration System 21 8.1.3.2. The Danish Cancer Register 21 8.1.3.3. The Danish National Hospital Register 22 8.1.3.4. The Danish Register of Causes of Death 22 8.1.3.5. The Danish National Health Service Register 22 8.1.3.6. Registration of community nurse activity and TD 23 8.2. Study I: Population-based study of death of patients with 23 cancer: implications for GPs 8.3. 8.2.1. Study I: Patients 23 8.2.2. Study I: Data analysis 24 8.2.3. Study I: Results 26 Study II: Defining cancer patients as being in the terminal phase: 31 Who receives a formal diagnosis and what are the effects? 8.3.1. Study II: Patients 31 8.3.2. Study II: Data analysis 32 8.3.3. Study II: Results 32 8.4. Study III: Does persistent involvement by the GP improve palliative care 35 at home for end-stage cancer patients? 8.4.1. Study III: Patients 36 8.4.2. Study III: Data analysis 36 8.4.3. Study III: Results 37 5 9. DISCUSSION 41 9.1. Principal findings 41 9.2. Methodological considerations 9.2.1. Choice of study design 41 9.2.2. Study populations 43 9.2.3. Quality of data 43 9.2.4. Confounding-interaction-effect modification 46 9.2.5. Place of death as outcome measure – pros et cons 49 9.3. Discussion of the results in relation to other studies 49 10. CONCLUSIONS 52 11. PERSPECTIVES 53 11.1. Health care 53 11.2 53 Future research SUMMARIES 55 12.1. Dansk resume [Danish summary] 55 12.2. English summary 58 13. REFERENCES 61 14. APPENDICES 73 12. 14.1. Paper I -III I. Population-based study of death of patients with cancer: implications for GPs. Br J Gen Pract 2005; 55: 684-9. II. Defining cancer patients as being in the terminal phase: who receives a formal diagnosis and what are the effects? J Clin Oncol 2005; 23: 7411-6. III. Does persistent involvement by the GP improve palliative care at home for end stage cancer patients? Palliative Medicine 2006; 20: 507-12. 6 3 ABBREVIATIONS CI Confidence interval CPR Central person registration EORTC European Organisation for Research and Treatment in Cancer GP General practitioner LPR Landspatientregister [The Danish National Hospital Register] NHS National Health Service OR Odds ratio QoL Quality of Life RCT Randomised controlled trial TD Terminal declaration UK United Kingdom US United States of America WHO World Health Organisation 7 4 DEFINITION The WHO definition of palliative care: Palliative care is an approach which improves the quality of life of patients and their families facing life-threatening illness, through the prevention and relief of suffering by means of early identification and impeccable assessment and treatment of pain and other problems, physical, psychosocial and spiritual. Palliative care: Provides relief from pain and other distressing symptoms; Affirms life and regards dying as a normal process; Intends neither to hasten nor postpone death; Integrates the psychological and spiritual aspects of patient care; Offers a support system to help patients live as actively as possible until death; Offers a support system to help the family cope during the patient’s illness and in their own bereavement; Uses a team approach to address the needs of patients and their families, including bereavement counselling, if indicated; Will enhance quality of life, and may also positively influence the course of illness, It is applicable early in the course of illness, in conjunction with other therapies that are intended to prolong life, such as chemotherapy or radiation therapy, and includes those investigations needed to better understand and manage distressing clinical complications. www.who.int/hiv/topics/palliative/care/en/ (17/02/2006) 8 5 THIS THESIS AT A GLANCE 5.1. What is already known on this subject? A majority of seriously ill cancer patients has a preference for dying at home. However, only a minority of patients actually die at home in most Western countries Often physicians do not define cancer patients as being terminal or their prognostic estimates tend to be optimistic Cancer patients emphasise the importance of the relationship with their GP in endof-life cancer care but the influence of the GP is not fully understood 5.2. What does this study add? Home visits conducted by GPs were strongly associated with the possibility of patients with end-stage cancer dying at home Being explicit in the shift from curative to palliative focus in end-of-life cancer care was associated with fewer admissions and more patients dying at home Women and the elderly were less likely to receive an explicit terminal diagnosis Provided that correct temporal relations are used, the mortality follow-back design can be a suitable and an ethical research method to highlight and monitor end-oflife cancer care 9 6 INTRODUCTION 6.1. Place of death of cancer patients Dying at home and remaining at home as long as possible is of great importance to a majority of seriously ill cancer patients and their families.1-8 However, in most western countries only a minority of patients actually die at home.9-21 One of the first prospective studies to present this antagonism was Townsend et al in BMJ in 199022: Of patients dying in hospital, 69% had preferred to die elsewhere. 6.1.1 Place of death in Western countries – short historical view Until the first decades of the 19th century, place of death was usually the patient’s home. Dying was part of life. Modernism, especially from the start of the 20th century, changed the attitudes towards death. Modernism represented a tendency to spare society from the hideousness of death struggle and from death in everyday life. Death and grief ceremonies (e.g. condolences and wearing black clothing) diminished or disappeared. This trend increased between 1930 and 1950. Dying moved from home surroundings to hospitals (institutional death) where the patients often died alone. This tendency increased due to the development of modern medical science and the possibilities of conquering the dying process, e.g. antibiotics and development of the ventilator during the polio epidemic in 1951-52.23-25 A shift in society’s view on death and dying initiated in 1967, where Dr. Cicely Saunders26 opened St. Christopher's Hospice in South London, England, as the first academic hospice. Hospice is a concept of caring derived from medieval times, symbolising a place where travellers, pilgrims and the sick, wounded or dying could find rest and comfort. The contemporary hospice offers comprehensive programmes of care to patients and families facing a life-threatening 10 illness. In 1969, Dr. Elizabeth Kubler-Ross published the book: “On death and dying” based on interviews with more than 200 terminally ill patients.27 In the years to follow, her theories of the dying process were pioneering. However, still in the seventies and eighties the proportion of hospital deaths increased17. In 1949 49%, in 1958 61%, in 1977 70%, and in 1990 80% of cancer patients died in hospital in the US.28 In the last part of the nineties, interest in death, end-of-life and palliative care has developed considerably.29-33 The World Health Organisation (WHO) formulated in 1986 a definition of palliative care in the wake of the HIV epidemic, with an update in 200234 (See definitions). Further, palliative care as a concept was refined by an international working group convened by the European School of Oncology in 2004.35 6.1.2. Is dying at home the same as a “good death”? Despite the growing interest in palliative care, dying alone in pain and fear of death is still the situation for many end-stage cancer patients throughout the world.36 To achieve a “good death” 37 various physical, structural, emotional and spiritual issues are important, i.e. receiving adequate pain and symptom management,38 adequate informal care resources being available,8 possibilities of preparing for death,39,40 so as to achieving a sense of completion, a sense of control and avoiding inappropriate prolongation of dying.41 Dying at home is doubtless only one component of a “good death”. However, dying at home might implicitly include some of the above-mentioned issues. Dying at home might, if the necessarily help is accessible, favours that sense of control and patient autonomy that strengthens a “good” death. 11 6.2. Organisation of palliative services in Denmark 6.2.1. National Health Services Care for end-stage cancer patients is part of the responsibility of the National Health Service (NHS).32 The Danish NHS is divided into three administrative levels, where the responsibility for the managing and financing of the hospital sector lies with 15 administrative units (counties), whereas community nurse services and nursing homes are managed by the 271 municipalities. In 2007, a politically approved Local Government Reform42 will take effect, centralising the 15 counties into 5 regions and reducing the number of municipalities from 271 till 98. Before and after the Local Government Reform, the GPs will continue to be self-employed, tax-funded in a mixed remuneration system with capitation fee and fee for service. GPs have the role of gatekeepers for secondary care and nearly the total Danish population (97%) is listed with a GP. GPs have daily office hours from 08:00 to 16:00. In all counties GPs are organised in out-ofhours co-operatives providing GP services to patients between 16:00 and 08:00.43 Specialised palliative care services have been developing since the first Danish Hospice opened in 1992.44 In 2005, a total of 8 institutions with 101 palliative care specialist beds and 16 palliative teams were available for end-of-life care.45 6.2.1. Palliative care at basic and specialised level The Danish National Board of Health (http://www.sst.dk; 17/02/2006) operates (as WHO) with palliative care at basic and expert level: “Palliative care at basic level implies the care provided in hospital departments and at home. The care in hospital and nursing home is conducted by the usual staff and at home by community nurses and the GPs, while specialist palliative care is targeted patients with complex symptoms and is 12 conducted by health care services with palliative care as their core specialty” (www.sst.dk/Planlaegning_og_behandling/Planer_Indsatser/Palliativ_indsats ;17/02/2006) 6.2.3. Palliative care at basic level In most cases, end-stage cancer patients are discharged from hospital for terminal care in the primary health care sector. The basic level includes the GP, the community nurse and informal carers. Problems in planning and conducting quality end-of-life cancer care at home can appear if the shift in focus from active, potentially curable anti-cancer treatments to palliative and terminal treatment in the primary sector is blurred or hampered by organisational, relational or personal barriers.46,47 6.2.3.1. Defining cancer patients as terminal? Deciding that a cancer course has come to a terminal phase is an ongoing challenge for physicians.48-51 Despite development of prognostic models and instruments for predicting survival in hospitalised late-stage cancer patients52,53 most physicians do either not define patients as being terminal, or their prognostic estimates tend to be optimistic.54-57 This might affect patients’ appropriate and timely referral to specialist palliative care services58,59 or can lead to unplanned hospitalisation due to poorly co-ordinated or otherwise inadequate supportive care services being available at home.46 It may also lead to problems such as poor symptom control, i.e. restricted morphine prescription60-62 or “carer fatigue” due to insufficient support to informal carer.63-67 Defining a patient as terminal might provide opportunities to plan the terminal phase68 which could in the end result in better quality of end-of-life for patient and for carer.41 Entering the terminal phase of a cancer disease may be clinically recognised by indistinct patterns such as lack of response to treatment, increased disease progression, the onset of anorexia, 13 or the loss of will to live. Cancer type may be of importance for terminal definition since some types have a well-known poor prognosis, for example, lung cancer69 and pancreatic cancer.70 This might facilitate definition of the terminal status while other types of cancers or an inaccurate diagnosis might tend to hold back a clear terminal diagnosis. In the elderly a cancer disease might be less obvious, less well-diagnosed due to co-morbidity. This could hamper or delay a clear terminal diagnosis.71 6.2.3.2. The terminal declaration (TD) The Danish “terminal declaration” (TD) 24,72 makes the important and difficult shift from curative to palliative care explicit (as contrasted with implicit and not expressed). A TD can be issued to patients by a physician (general practitioner or hospital physician), if the physician finds the patient to be terminal (not likely to benefit from curable treatment and not likely to live for more than six month). The intent of the Danish Social Services Law 73-75 is to support the patient and the family either at home or in a hospice. The TD entitles the patient to full reimbursement of drug expenses, reimbursement of expenses for equipment used in home care. Furthermore, an informal carer may be entitled to full-time or half-time paid leave.76-78 Informed consent from the patient is needed, but the patient does not have to sign the declaration. Before March 2000, all requests for TD had to be applied for at the municipality. In 2000, TD was separated in two; reimbursement for drug expenses had to be applied for at the Danish Medicines Agency79, an agency under the Ministry of the Interior and Health, while medical aids and paid leave still had to be applied for at the municipality. However, the key content of TD was the same as before 2000. 14 6.2.3.3. GP in end-of-life cancer care The influence of the GP in end-stage cancer care and on place of death of cancer patients is not well researched,80 but some elements of the GP role have been studied in small, selected populations or commented in editorials. Physician home visit was found pivotal in improving patient outcomes in a study of 298 patients81. In bereavement interviews of informal caregivers, the association of dying at home was significant, when the family physician made a home visit. 2,82 In a study of 73 patients with cancer, support by a family physician was associated with home death.83 In a retrospective interview of 48 carers, accessibility of the GP was valued important65,84 and cancer patients have emphasised the importance of an individual relationship with their doctor.85 In other papers, e.g. editorials, case stories, authors consider some of the possible functions of the GP in end-of-life care besides control of symptoms, i.e. a willingness to talk about the process of dying40,86,87 or highlight the GP skills to champion the philosophy of terminal care,29,88-91 but also insufficiency of the GP,76,92-94 frustration of the GP due to a poor opinion of the quality of the care provided46,95,96 and the need and initiatives for improving the necessary GP qualifications.47,97-101 However, very few researchers have analysed the influence of the GP in end-of life cancer care in population-based studies. In a retrospective study of 9714 Canadian cancer patients an association between family physician continuity of care measured by a Modified, Modified Continuity Index (MMCI) and location of death was found102 which indicates an association between GP values, i.e. continuity of care and location of death. Through the studies in this thesis, we want to explore the hypothesis derived from former studies and editorials that the GP seems to have various fundamental values in end-oflife cancer care. We decided to incorporate a careful time-analysis of the health services 15 provided for all cancer patients in a well-defined region and time-frame. Futhermore, we aimed to analyse the effect of the GP in the subpopulation receiving TD (home death planned) and in the subpopulation not receiving TD. 6.3. Research in palliative care 6.3.1. Historic view and methodological challenges A few case stories, e.g. an article in The Lancet in 1952, ”opened a door on some very dark scenes” about end-of-life cancer care in the UK103 but very few studies on end-of-life cancer care had been carried out at that time. Since the mid 1980s, medical research in palliative medicine and end-of-life cancer care has increased considerably.104 One of the main challenges has been to find outcome measures and measuring instruments which cover the complex physical, emotional and spiritual needs of end-of-life cancer patients.105 The EORTC Quality of Life Group has recently presented a quality-of-life instrument aimed at palliative care.106 Anyhow, in palliative care research, the persistent problem is frailty of the patients. Results from RCTs, questionnaires and interview studies are often biased by dropouts either due to deterioration of the condition or death of the patient.4,45,107-113 Place of death can, on the other hand, easily be measured and made operational. Well aware that place of death is an indirect measure of quality of end-of-life cancer care, place of death is, for the time being, probably the most valuable outcome measure in population-based studies.114 16 6.3.2. Mortality follow-back design “Twenty years ago, the challenge was to engage healthcare workers in the care of the dying. Ten years ago, the challenge was to engage healthcare organisations in quality improvement efforts on end-of-life care. Today, the challenge is to develop systematic and comprehensive information on the quality of end-of-life care at the population level.” Editorial BMJ 2003; 327: 173-4. Despite the rapidly growing research in palliative medicine in the last years, population-based data are still lacking. Most research by far has been carried out among patients already attached to specialist palliative care services or hospices, home hospices and other specialist programmes for end-of-life care. Research groups in Canada2,115 and in Boston, US116 have in recent years been working to identify and evaluate potential indicators for the quality of end-of-life care, focusing on processes that can be measured using data already collected electronically e.g. data in registers, claims, and other administrative data. The group in Boston is currently validating the measures against prospectively collected assessments of quality of care made by patients and their family members (http://www.dana-farber.org/pat 13/02/2006). The approach of working back from dateof-death to study care received by patients over a period of time prior to their death has been recommended in the US in 2003 by the National Cancer Policy Board.115-120 However, the value of such studies with a retrospective design has recently been the subject of contention. 121 The major problem raised was the absence of a clearly defined origin of the terminal phase. Was the patient actually dying (~in a clinically recognised terminal phase) or did the patient die in potentially curable phase (for instance during surgery)? One could argue that if there were a clear date of origin of the terminal phase it would be possible by this method to mimic an ordinary cohort study with respect to analysis and interpretation. From this perspective the Danish National Health Service offers a unique opportunity to conduct valid mortality follow-back studies.122,123 Not only are extensive data available for all cancer patients – including their use of health care services – but 17 the rather large group who receives a terminal declaration (TD) may actually be viewed as a cohort of terminal cancer patients with a reasonably well-defined origin of their terminal phase. 18 7. AIMS The purpose of the thesis was to identify factors of importance to meet the desires of a majority of end-stage cancer patients in the western countries, i.e. to stay at home as long as possible and to die at home. Based on a mortality follow-back design we aimed: 1) To explore factors associated with place of death in an unselected population of patients dying from cancer and, in particular, to explore whether GP and community nurse services were associated with place of death adjusted for patient-related factors. 2) To find population-based patient-related factors of importance for the formal terminal diagnosis and to analyse the implications of such a diagnosis on some important and measurable parameters of quality of end-of-life care in cancer. 3) To analyse the effect of GP home visits on granting a terminal declaration (TD) and on place of death. 19 8. THE STUDIES Section 8 of the thesis first describes general methodological issues. The following sections give a short version of the 3 studies published as separate papers (Papers I-III). The first study describes the study population and analyses factors of importance for place of death of patients with end-stage cancer. The second study analyses the important shift from curative to palliative focus of care and its implications for planning end-of-life cancer care, while the third study, in depth, analyses GP home visit as a marker of persistent involvement in end-of-life cancer care. 8.1. General methodological issues 8.1.1. Setting The setting of the three population-based studies was the island and the County of Funen in 1996-98 with a population of approx. 470 000. Data in the studies represent the latest validated data at the time of writing. The study population comprised all patients in the county who died of cancer in the 3-year period. 8.1.2. Observation period (Time scale) We obtained healthcare information on deceased cancer patients in their last 3 months of life. The decision to start data collection at 3 months prior to date of death (left censoring) was an estimate based on literature studies with reference to duration of the terminal phase.29,54,57,124-131 20 8.1.3. Health care data We obtained health care information from the Danish Civil Registration System,132 the Danish Cancer Register,133 the Danish National Hospital Register,134 the Danish Register of Causes of Death,135 the National Health Service Register136 and the registration of community nurse activity and the terminal declaration (TD) in the municipalities. The personal identification number assigned to every citizen in Denmark at birth or immigration permits accurate, individual-based linkage of information between registers. 8.1.3.1. Danish Civil Registration System (CPR) The Danish Civil Registration System132 was set up in 1968 on the basis of the previously manually compiled municipal registers. The civil register contains information about each citizen, including the personal identification number, former (historical) addresses, names, civil status and citizenship. The road system register in CPR contains current information on all roads in the nation. This information was used to find nursing home addresses. The Danish Ministry of the Interior and Health's Central Office of Civil Registration (the CPR-Office) is in charge of the Danish Civil Registration System and functions as the main provider of basic personal information to public authorities and the private sector. 8.1.3.2. The Danish Cancer Register Since 1943, the Danish Cancer Register134 has collected information on all individuals in Denmark with cancer.133,137 The registration is based on notification forms completed by hospital departments and GPs whenever a case of cancer is diagnosed or found at autopsy, and whenever there are changes to an initial diagnosis. By annual linkages to the death certificate files the cause of death is noted. Unclear or contradictory information, either within a 21 notification form or between forms, leads to queries in approximately 10% of the notifications received. Comprehensive evaluation has shown that the register is 95% to 99% complete and valid.138-140 8.1.3.3. The Danish National Hospital Register The Danish National Hospital Register (Landspatientregistret, LPR) keeps records of nearly all (99%) the somatic hospitalisations in Denmark since 1977.141 The register receives new data from all Danish hospitals once a month. This computerised central register contains, for each discharge, information on the personal identification number of the patient, dates of admission/discharge, up to 20 diagnoses, and every surgical procedure performed during the respective admission, as well as all out-patient contacts by date and diagnosis. The register is frequently used for research. Ad hoc validation of the administrative information has shown agreement in 98.5% of the cases. 8.1.3.4. The Danish Register of Causes of Death Since 1875, the Danish National Board of Health has published the causes of death based upon the death certificates. From 1970 the register has been fully computerised.142,135 Declining autopsy rates (about 12% of deaths) threaten the quality of the data; however, in our studies we used the cross-checked, well-validated causes of death from the Danish Cancer Register and only used place of death from the Register of Causes of Death.143 8.1.3.5. The Danish National Health Service Register Date and type of contacts with GP were collected from the National Health Service Register. The Danish National Health Service Register is a data system available for counties and 22 municipalities to manage the National Health Insurance covering primary health care providers. The counties use the register for administrative purposes, especially for the settling of accounts with providers. The register contains data on all citizens, providers, and health care services reimbursed by the health authorities.136 8.1.3.6. The registration of community nurse activity and the terminal declaration (TD) Information on community nurse services at home as well as information on the TD was extracted manually from administrative data files in each local municipal administration in the county. Existence of some degree of misclassification in the registration system in the municipal administration must be considered. Misclassification due to incomplete or erroneous registration in the municipal administration was, however, likely to be non-differential with respect to the studied variables and tends to be reduced by routinely conducted external auditing. The next pages summarise the methods and results from the three studies. The pooled discussion of the results in the papers is presented in section 9.3. More detail of the separate studies can be obtained in Appendices 1-3. 8.2. Study I: Population-based study of death of patients with cancer: implications for GPs 8.2.1. Study I: Patients After initial description of place of death in the total cohort (n = 4,386) we excluded the population living in nursing homes at the beginning of the observation period (3 months prior to death) and patients moving to nursing homes during the last 3 months of life (Fig 1). 23 Residence 3 months prior to death Home Home 4208 4208 Nursing Nursing Home Home 178 178 116 Residence at time of death Home Home 4092 4092 Nursing Nursing Home Home 294 294 438 34 Unknown Unknown place place of of death death 34 34 11 % % 1221 Dying Dying at at Home Home 1221 1221 28 28 % % 35 2399 259 Dying Dying in in Hospital Hospital 2434 2434 55 55 % % Dying Dying in in Nursing Home Nursing Home 697 697 16 16 % % Fig 1 Place of death and change of residence during last three months of life for cancer patients (n=4386). Thus the cohort used in the analysis were those with residence at home (n=4,092). The design was a case-control design where cases were cancer patients dying in hospital. Cases comprised 2,399 patients dying in hospital and 438 patients dying in nursing homes plus 34 with unknown place of death, in total 2,871 patients. Controls were 1,221 patients dying at home. The patients dying in nursing homes in the analysis cohort were those in respite care, who could reasonably be considered to die an institutional death (cases). Because of automatic deletion of records more than 5 years old, the analysis was limited to deaths which occurred between 27 March 1997 and 31 December 1998. This subgroup was, however, compatible with the total study group. 8.2.2. Study I: Data analysis We estimated the odds ratio of dying in hospital by logistic regression models and calculated 95% confidence intervals (CIs). In the models we adjusted for potential confounding factors: 24 type of cancer, duration of disease, severity of disease by number of ambulatory contacts, gender, age, marital status, number of children, geographic area and citizenship. Primary cause of death was categorised into six categories: four separate categories for the four most frequent cancers (lung, breast, colon-rectum, prostate), one category for the haematological cancers (the course of which is rapid and treatment intense), and one final category containing other cancers. Age was categorised as 18-39, 40-65, 65+, and finally being married and having children were dichotomised. We calculated survival time by subtracting date of diagnosis from date of death and subsequently dichotomised the result (less than 1 month, more than 1 month). GP activities were measured by number and type of contacts (surgery consultation, telephone consultation, home visit and out-of-hour services in general practice co-operatives). Community nurse activity was measured in a sub-analysis in the largest municipality (40% of total population), and patients were dichotomised: with and without nurse contact. Time at home was calculated as days not hospitalised from diagnosis to death, left censored at 90 days, i.e., days possible to receive home visits. The home visit rate was calculated by the number of home visits divided by time at home in the last 90 days before death (~risk time). A multivariate logistic model was fitted using the rate of home visits along with previously identified predictors. Further, a matched case-control study was conducted. Among cases (hospital death) we found the day of admittance for the ultimate admission and matched this day among randomly chosen controls (home death). The matching process allowed identification of 423 randomly selected cases with a corresponding control in the analysed cohort. By this, we compared the ultimate period at home for cases (hospital death) with matching controls (home death). We then adjusted for the previous explored covariates. Finally, we described the population not receiving home visits. All analyses were conducted using Stata 8 (StataCorp, College Station, TX, USA). 25 8.2.3. Study I: Results During the three-year study period, 4,386 adult patients (at least 18 years of age) died with cancer as primary cause of death. Figure 1 shows the place of residence, the place of death and migration patterns of cancer patients in the 3 months prior to death. In the total cohort only 28% died at home, 55% died in hospitals, 16% in nursing homes and for one percent place of death was unknown. Some 9% (438) were temporarily admitted to nursing home where they died. Table 1 shows characteristics of cases (death in hospital or nursing home) and controls (death at home). 26 Table 1. Characteristics of cases and controls in Study I Institutional death (cases) (N=2871) Sex, number (%) Male Female Age at death, number(%) 18-39 40-65 >65 Marital status, number (%) Married Widowed, divorced or single Children, number (%) 0 1 or more Primary cancer; number (%) Lung Colon-rectum Breast Prostate Haematological Others Time of diagnosis to death: ≥ 1 month < 1 month Geographic area: Urban Rural Nationality: Danish Other Home death (controls) (N=1221) 1490 1381 (52) (48) 655 566 (54) (46) 35 717 2119 (1) (25) (74) 12 349 860 (1) (29) (70) 1882 989 (66) (34) 916 305 (75) (25) 1353 1518 (47) (53) 505 716 (41) (59) 609 273 205 175 109 1500 (21) (10) (7) (6) (4) (52) 277 133 117 90 20 584 (23) (11) (10) (7) (1) (48) 2253 618 (78) (22) 1117 104 (91) (9) 2209 662 (77) (23) 956 265 (78) (22) 2850 21 (99) (1) 1208 13 (99) (1) In the univariate analyses we found a higher proportion of women than men dying at home in the 40-65 year age group. Being married and having children made it less likely to die in hospital, while we found no association with geographic locations. On the other hand, we found 27 general practitioner home visit and visits by community nurse to be strongly inversely associated with hospital death and we found an association between patients dying in hospital and short survival time as well as patients with haematological cancer. We examined the association of these variables in an adjusted multiple logistic regression model (Table 2). 28 Table 2. Associations with hospital death General practitioner: No home visits Home visits Community nurse: No contact Contact Sex: Male Female Age at death(years): 18-39 40-65 >65 Marital status: Single, divorced, widowed Married No. of children: 0 1 or more Primary cancer: Lung Colon/rectum Breast Prostate Haematological Others Time of diagnosis to death: ≥ 1 month < 1 month Geographic area: Urban Rural Crude Odds ratio(95 % CI ) Adjusted Odds ratio (95 % CI ) 1.00 0.09 † (0.06 to 0.13) 1.00 0.08† (0.06 to 0.12) 1.00 0.32† (0.24 to 0.42) 1.00 0.36† (0.26 to 0.48) 1.00 0.96 (0.81 to 1.15) 1.00 0.92 (0.74 to 1.14) 0.98 (0.45 to 2.17) 0.86 (0.70 to 1.04) 1.00 0.61 (0.24 to 1.52) 0.70† (0.56 to 0.90) 1.00 1.00 0.65† (0.53 to 0.78) 1.00 0.68† (0.56 to 0.85) 1.00 0.83 (0.69 to0.99) 1.00 0.86 (0.70 to1.06) 0.87 (0.68 to1.09) 0.78 (0.58 to 1.02) 0.66 (0.49 to 0.92) 0.87 (0.62 to1.23) 1.43 (0.81 to2.56) 1.00 0.99 (0.77 to1.25) 0.79 (0.58 to 1.09) 0.73 (0.51 to 1.05) 1.04 (0.71 to 1.54) 1.35 (0.74 to 2.50) 1.00 1.00 2.94† (2.17 to 3.85) 1.00 2.27† (1.69 to 3.13) 1.08 (0.87 to 1.32) 1.00 1.09 (0.87 to 1.35) 1.00 († p<0.05) The strongest association was the general practitioner paying at least one home visit during the last 3 months, but also community nurse visits had an impact. We then wanted to find out whether GP home visit was solely a consequence of an acute need of care. To explore this, we 29 made a separate analysis excluding all general practitioner services in the last 14 days. However, the difference between the two groups declined but was still significant: 0.52 (0.43 to 0.63). To find out whether the association persisted with another cut-off point, we analysed the last period at home before last admission (for cases) in a matched case-control study, matched with date of last admittance and adjusted for the previous explored covariates. In this analysis, we retrieved the significant association: 0.22 (0.15 to 0.41) between GP home visit and place of death. Finally, the home visit rate was a strong predictor of home death (Table 3), i.e. the odds of hospital death consistently decreased with increasing rate of home visits by general practitioner. In other words, by comparing patients with equal time at home, those with added GP home visits will be more likely to die at home. Table 3. Association of rate of GP home visits with hospital death Rate of home visits (per month) Crude Odds ratio (95%CI) Adjusted Odds ratio (95%CI) 0 0.1 - 0.5 0.6 - 1 1-2 2-4 >4 1.00 0.15 (0.10 to 0.23) 0.12 (0.08 to 0.18) 0.08 (0.05 to 0.11) 0.06 (0.04 to 0.09) 0.05 (0.03 to 0.06) 1.00 0.14 (0.09 to 0.21) 0.11 (0.07 to 0.17) 0.07 (0.05 to 1.00) 0.06 (0.04 to 0.08) 0.05 (0.03 to 0.08) 30 8.3. Study II: Defining cancer patients as being in the terminal phase: who receives a formal diagnosis and what are the effects? 8.3.1. Study II: Patients The patient cohort in this study was 3,445 cancer patients from 18 municipalities ( Figure 2). Due to incomplete or missing information we had to eliminate 13 small municipalities with a total of 941 cancer patients. However, those 941 patients were comparable to the rest of the cohort with regard to age, gender, marital status, number of children and types of cancer. Fig 2 Cohort of patients with cancer as primary cause of death County of Funen 31 municipalities 471,989 inhabitants Deceased patients 1996-98 N=16,580 Other causes of death N=12,194 Cancer as primary cause of death N=4,386 13 municipalities N=941 18 municipalities N=3,445 Terminal diagnosis N=1,156 34 % No terminal diagnosis N=2,389 66 % 31 8.3.2. Study II: Data analysis We calculated the duration of the terminal declaration (TD). We could collect the exact date of registration, also before the observation period of 3 months prior to date of death. We first used the TD as a binary outcome variable, and hence estimated the odds ratio of TD by logistic regression models and calculated 95% confidence intervals (CIs). In the models we adjusted for potential confounding factors. Age could be a confounder because elderly people may delay seeking medical advice if suspicion of cancer and analogous to this we assume they may also delay seeking help in a terminal phase. Gender must also be considered a confounder because of different health service utilisation by men and women, and type of cancer because different types of cancer have distinctly different natural courses of disease. Being married and having children are also possible confounders, because informal carers are able to ask for assistance and support in the health care system. As expected, we found a dependency between the odds of TD and week number before death, which we allowed for by including a separate constant for each week on which the risk of TD could depend. In the second analysis, we estimated the rate of admission to hospital for patients with a cancer diagnosis (a maximum of 12 weeks and accounting only for the time period after diagnosis of cancer) by using TD as a binary covariate. Finally, we analysed the risk of hospital death by logistic regression. 8.3.3. Study II: Results For 34% of the patients in the total cohort (n = 3,445) the physician signed a terminal declaration (TD) (Fig 2.). However, when survival time, i.e. time from cancer diagnosis until death, was less than one month, only 8% received a TD. The declaration was signed 52 days (median, interquartile range from 20 to105) before death with a mean of 100 days (95% CI: 91-110) (Table 4). 32 Table 4. Characteristics of the study population in Study II. No Terminal diagnosis N % Sex: Female Male Age at death: 18-29 30-39 40-49 50-59 60-69 70-79 >=80 Married: Yes No Children: 0 ≥1 Primary cancer: Colon-rectum Haematol. Breast Other Prostate Lung Survival time: < 1 month ≥ 1 month Total Terminal diagnosis Duration in days (Median, interquartile range) N % 1,103 1,186 66 67 563 593 34 33 56 (21,109) 47 (19,102) 3 15 69 203 422 790 787 37 48 48 50 55 70 81 5 16 74 200 339 334 188 63 52 52 50 45 30 19 61 (11,90) 96 (35,145) 60 (24,110) 67 (24,140) 47 (21,92) 52 (19,109) 40 (15,90) 1,470 819 64 72 836 320 36 28 52 (19,102) 54 (21,107) 1,221 1,068 74 59 422 734 26 41 50 (19,100) 54 (20,107) 230 95 175 1,203 156 430 68 89 65 68 68 59 109 12 96 565 75 299 32 11 35 32 32 40 60 (22,99) 32 (13,63) 66 (23,155) 45 (18,101) 97 (41,219) 49 (21,90) 552 1,737 2,289 92 61 66 49 1,107 1.156 8 39 34 15 (8;36) 55 (21,107) 52 (20,105) We found the probability of a TD to be significantly lower in age groups 70+ years, also after eliminating patients living in nursing homes (n = 225). Female gender compared to male gender and haematological cancers compared to “other” cancers showed significantly lower 33 probability of a TD while being married vs. single (widowed, divorced or not married), having cancer of the prostate or the lung increased the probability of a TD (Table 5). Table 5 The odds ratio of receiving a TD (terminal diagnosis) Variable Odds Ratio Sex: Female Male Age at death: 18-29 30-39 40-49 50-59 60-69 70-79 ≥ 80 Married: Yes No Children: 0 ≥1 Primary cancer: Colon-rectum Haematological Breast Other Prostate Lung Survival time ≤ 1 month > 1 month 95% CI 0.81 1 0.69 to 2.42 1.46 1.10 1 0.72 0.44 0.24 0.57 to 10.23 0.71 to 3.01 0.77 to 1.60 0.57 to 0.34 to 0.18 to 0.92 0.56 0.32 1.20 1 1.01 to 1.44 0.85 to 1.25 1.20 0.20 1.08 1 2.14 1.33 0.92 to 0.09 to 0.79 to 1.57 0.41 1.46 1.55 to 1.10 to 2.96 1.60 0.10 1 0.07 to 0.15 1 1.03 0,96 Looking at the patients (n = 619) with short survival time ≤ 1 month, we found the same age distribution as in the total cohort. We found, however, significantly fewer cancers with a known extended course, i.e. prostate and breast cancers. Looking at the effects of TD, we found that the admission rate increased less with TD (Fig 3). 34 Fig.3. Effect of terminal diagnosis on admissions per week. Admission per week % 80 60 40 20 0 No terminal diagnosis Terminal diagnosis -13 -11 -9 -7 -5 -3 -1 Week number Interestingly, however, we found significantly longer admission time with TD (35.3 days vs. 30.2 days). In our cohort 56% (n = 1,935) died in hospital, 28% (n = 965) at home, 16% (n = 518) in nursing home and 1% had unknown place of death. Hospital death was found in 65% (n = 1,487) of cancer patients without TD, while only 39% (n = 448) of cancer patients with TD died in hospital. With TD, the adjusted OR of hospital death was 0.25 (95% CI: 0.21-0.29). 8.4. Study III: Does persistent involvement by the GP improve palliative care at home for end stage cancer patients: Mortality follow-back design In the first study, we found after careful considerations concerning alternative explanations that GPs making home visits resulted in more cancer patients dying at home. In the second study, we found 35 that the explicit terminal diagnosis in the form of a TD resulted in fewer hospital admissions and more patients dying at home. We then wondered whether the effect of GP home visit was because of the TD (the TD implies that the patient, the informal carer, community nurse and GP are aiming for care at home) or whether the GP home visit has an effect on its own (Fig. 4) Fig 4. Possible effects of GP home visit in end-of-life cancer care indicated by arrows Cancer patient Effect of GP visit? Explicit terminal diagnosis (TD) Effect of GP visit? Place of death Effect of GP visit? 8.4.1. Study III: Patients The cohort used for analysis comprised those cancer patients who were living at home three months prior to death ( n=2,025). We excluded patients living in nursing homes at the beginning of the observational period as well as those moving to nursing home during the 3-month period prior to death. 8.4.2. Study III: Data analysis In the first analysis we used the terminal declaration (TD) as an intermediate outcome and GP home visit in the last week before TD as exposure. We used a matched case-control study 36 where cases were diagnosed cancer patients receiving a TD, and matched controls were diagnosed cancer patients with same interval to death without having received a TD (yet). The first matched pair analysis included 510 pairs. In all pairs, we eliminated the pair if either the case or the control were in-hospital patient in the week before a TD. Then we repeated the analysis with GP home visit in the last 4 weeks before a TD as exposure. The second analysis included 652 matched pairs. By conditional logistic regression for matched case-control groups we found the crude and adjusted odds ratio (OR) of home visit in the week immediately preceding TD, and in the extended period of 4 weeks before TD. Finally, in order to examine the unobserved confounding of individual frailty, we analysed the sub-populations of patients (40 pairs) receiving a TD very short time (1 week) before dying. In the second analysis we separated the data into two strata: patients with and without TD. Main outcome was hospital death. We calculated the risk time, i.e. time at home (not hospitalised and left censored at 90 days before death). In the TD stratum we found all GP home visits after the date of TD. In the no TD stratum, we found all GP home visits in the risk time of the last 30 days. Then, in both strata, we used the categorised (6 categories) home visit rate as explanatory variable. Multivariate logistic models were fitted using the home visit rate as predictor along with the previously identified predictors and hospital death as our main outcome variable. We calculated the 95% confidence intervals (CI) for all estimates. 8.4.3. Study III: Results A total of 2,025 patients died of cancer, 56% died in hospital, and 38% received a terminal declaration (TD) (Table 6). The median age at death was 74 years (interquartile range: 64 to81). 37 Table 6. Characteristics of 2025 terminal cancer patients stratified with respect to having received a terminal declaration ( + TD) or not (-TD). + TD N =774 (38 %) N - TD N =1251 (62 %) Hospital death % N Hospital death % Sex Female Male 375 (48 %) 399 (52 %) 35 43 616 (49 %) 635 (51 %) 65 68 Age at death(years) 18-29 30-39 40-49 50-59 60-69 70-79 80- 4 (1%) 12 (2%) 51 (7%) 127 (16%) 218 (28%) 222 (29%) 140 (18%) 0 50 47 50 42 38 21 3 (0%) 6 (0%) 34 (3%) 115 (9%) 223 (18%) 441 (35%) 429 (34%) 67 100 82 79 76 72 51 Married Yes No 553 (38 %) 221 (62 %) 37 40 780 (29%) 471 (71%) 68 64 Children 0 ≥1 285 (37 %) 489 (63 %) 31 43 645 (52 %) 606 (48 %) 58 76 Primary cancer Lung Colon-rectum Breast Prostate Haematological Others 193 (25 %) 71 (9 %) 61 (8 %) 55 (7 %) 7 (1 %) 387 (50 %) 46 34 34 31 43 38 223 (18 %) 131(10 %) 89 ( 7 %) 85 ( 7 %) 54 ( 4 %) 669 ( 53 %) 74 66 49 49 74 68 38 We found a strong association between TD and GP home visit in the week preceding the declaration (OR= 11.5, 95% CI: 6.0-24.6). Increasing the period before the TD to 4 weeks, we found a weaker but still strong association to GP home visits (OR: 6.4 (4.5-9.2)). The association was increased when we adjusted for patient characteristics, i.e. gender, age at death, marital status, children and type of cancer. The association also existed in a subsample of patients (n = 80), who received the TD one week or less before death (Table 7). Table 7. Chance of getting a terminal declaration (TD) as a function of preceding GP home visit analysed in matched case control design (652 pairs) by conditional logistic regression. Cases received a TD and controls were selected among those who had not received a TD. Period for TD 3 months 1 week Exposure period for home visit Discordant pairs Cases visited / control unvisited Case unvisited / control visited 1 week 115 10 4 weeks 218 34 1 week 19 3 Crude OR (95% CI) Adjusted OR (95% CI) 11.5 (6.0-24.6) 6.4 (4.5-9.2) 16.8 (8.2-34.4) 9.7 (6.4-14.6) 6.3 (1.9-21.4) 39 To see whether we could find an effect of GP home visit despite TD, we stratified on TD in the next series of analyses. In both strata we found GP home visit inversely associated with hospital death, in TD stratum (OR: 0.18 (0.11-0.29)) and in no-TD stratum (OR: 0.08 (0.05-0.13). We also found dose-response relationship in both strata (Table 8). Table 8. Risk of dying in hospital in relation to GP home visits (+/- GP) among patients with and without a terminal declaration (+/- TD) + TD GP home visits per week N - TD OR (95%CI) Hospital death % N OR (95% CI) Hospital death % + GP visit (0/1variable) 774 0.18 (0.11-0.29) 39 1251 0.08 (0.05-0.13) 66 0 0.1-0.19 0.2-0.39 0.4-0.59 0.6-1.49 ≥1.5 160 153 116 155 155 31 1 0.28 (0.17-0.47) 0.11 (0.06-0.20) 0.08 (0.04-0.13) 0.09 (0.06-0.16) 0.07 (0.03-0.18) 78 77 48 30 26 12 591 178 211 142 109 20 1 0.14 (0.09-0.20) 0.14 (0.09-0.21) 0.07 (0.05-0.11) 0.09 (0.06-0.15) 0.10 (0.04-0.27) 92 89 51 34 43 40 40 9. DISCUSSION The discussion is divided into three sections: Principal findings, methodological considerations i.e. study design, study populations, data quality, confounding - interaction- effect modification and choice of outcome measures. Finally, the results are discussed in relation to other studies. 9.1. Principal findings The present studies from primary care show that GPs play a central role in management of endstage cancer patients. The studies also highlight the importance of making the switch from curative to palliative focus explicit in end-of-life cancer care. 9.2. Methodological considerations 9.2.1. Choice of study design Choice of study design will always be a balance between the most robust method and feasibility of the method in the actual research area. In palliative care research there are important methodological problems: One of the main barriers is the frailty of the patients involved. Randomised, controlled trials (RCTs) generally considered to have a high level of evidence often fail to report robust significant results in end-of-life studies. Attrition due to deterioration of the condition, drop-outs and small sample size104,107 are well-known problems in palliative care research which can hamper the conducting of RCTs. Besides, prediction of a patient death cannot be precise,54 i.e., entry to a prospective study will introduce selection bias. Further, a randomised trial in our studies would, apart from the mentioned problem of drop-outs, include a “no home visit” arm or a “no terminal diagnosis” arm. That would restrain patients in the terminal phase from receiving a GP home visit or the benefits from a terminal declaration which obvious would be unethical. 41 Methodological alternatives to randomised trials are prospective or retrospective observational studies. Due to the above mentioned problem of prediction of death and absence of a clearly defined starting point of terminal phase, prospective cohort studies are very difficult to conduct, whilst the retrospective cohort designs, i.e. mortality follow-back studies might be more convenient for analysing the last months of life. In mortality follow-back studies all healthcare services can be examined for a defined period of time prior to death.116-120 The major problem with this type of studies raised recently in JAMA is the absence of a clearly defined origin of the terminal phase.121 However, with a clearly defined origin of terminal phase it is possible to mimic an ordinary cohort study with respect to analysis and interpretation. From this perspective, the Danish healthcare system offers unique opportunities to conduct valid mortality follow-back studies. Not only are extensive data available for all cancer patients – including their date of use of healthcare services – but the rather large group who receives a terminal declaration (TD) may actually be viewed as a cohort of terminal cancer patients with reasonably well-defined origin of their terminal phase. An important problem induced by the retrospective data collection is the leftcensoring, i.e., restricting the analysis to the last three months. We had to introduce left-censoring (date of death minus 3 months) to get good data quality but as a consequence we cannot allow for any events prior to the three months before death in our study. We did, however, find similar results for those with a diagnosis less than three months before death and those with a diagnosis prior to the three months before death. 42 9.2.2 Study populations In the studies we have minimised selection bias by analysing all cancer deaths in a well-defined region and time period. We have included all patients dying with cancer as primary cause of death and with residence in the County of Funen in 1996-98. We have thus also included patients who in the last weeks and days of their lives were transferred to and dying in a hospital out of the county (0.5%). In some of the analyses, we had to do with subpopulations. As a consequence of routine removal of records more than 5 years old, the analysis of services provided in general practice had to be limited to deaths which occurred between 27 March 1997 and 31 December 1998. This subgroup was, however, compatible with the total study group in relation to gender, age, distribution of types of cancer, marital status and children. Further, in the analysis of community nurse services and TD, we had as a result of incomplete or missing information to eliminate 13 small municipalities with a total of 941 cancer patients. However, those 941 patients were not different from the rest of the cohort with regard to age, gender, marital status, children and types of cancer. 9.2.3. Quality of data We used one disease register, i.e. the Danish Cancer Register133,137 and four administrative registers. By means of the personal identification number in the Civil Registration System we could conduct accurate, individual-based linkage of information between the 5 registries in the studies. The disease register, the Danish Cancer Register, has since 1943 collected information on all individuals in Denmark with cancer. The registration is based on notification forms that are completed by hospital departments and general practitioners whenever a case of cancer is diagnosed 43 or found at autopsy, and whenever there are changes in an initial diagnosis. By annual linkages to the death certificate files the cause of death is noted. Unclear or contradictory information, either within a notification form or between forms, leads to queries in approximately 10% of the notifications received. Comprehensive evaluation has shown that the register is 95% to 99% complete and valid.139,144,145 The Danish National Discharge Register keeps records of nearly all (99%) the somatic hospitalisations in Denmark since 1977. The administrative data in the register, i.e. admission date, discharge dates and type of admission have been compared with the medical records of the source hospital, finding an agreement of 98.5%.141 The National Health Insurance has information on number of services given by all GPs. The information is used for reimbursement for services provided in general practice. In Denmark, most primary care is free of charge. General practitioners are self-employed but funded through taxes in a mixed remuneration system with capitation fee and fee for service. Some 97% of the population is listed with a GP. Due to the remuneration component of the registration the data are likely to be complete and they are checked for errors by the county administration. Community nurse activity and information on TD were extracted from administrative data files in the main municipality, Odense (approx. 189 000), and manually from the other municipalities in the county. The munipalities receive 50% repayment from the county administration if a patient with a TD receives care at home by a community nurse or receives medication aid. Because of this remuneration practice (the same situation as with the general practitioners) there will undoubtedly be a tendency to correct registration of services provided. Existence of misclassification in all the registries must, however, be considered carefully. (Misclassification is the erroneous classification of an individual into a category other than the one to which he/she should be assigned). Misclassification due to erroneous 44 registration in the municipal administration is likely to be non-differential with respect to the studied variables and will in any case tend to be reduced by the routinely conducted external auditing. Misclassification in the other registers is likely to be negligible due to the intense validation process and non-differential with respect to the data used in the study. A more important question is whether a misclassification exists with respect to classification of patients as terminal or not: We cannot be certain whether the patients under study actually are in a clinically terminal phase, or whether they died in a potentially curative phase during for example surgery or intensive chemotherapy treatment. If we have included patients who were not in the terminal phase, we could not expect the healthcare system to have mobilised for end-of-life cancer care. This would tend to weaken the association between GP home visit and dying out of hospital. However, in the third study, this type of misclassification is taken into consideration by using the TD: A patient provided with a TD consolidates that a physician has diagnosed the patient terminal (with a life prognosis of less than 6 months), i.e. there is a clear cut-of point (date) beyond which the patient has a terminal diagnosis. Misclassification might also occur when using the written declaration as a proxy for the terminal diagnosis. Because the TD carries with it benefits around reimbursement for expenses (medicine and equipment for home care) and especially paid leave for informal caregivers, it may reflect more than simply the prognosis of the patients. For example, caregivers for the elderly (who may be old-age pensioners themselves) are less likely to benefit from paid leave from work. Although the declaration would have other advantages for such patients (and patients without a caregiver) they might be less likely to receive a TD. The TD is, therefore, most reliable as a proxy for a terminal diagnosis in patients with a positive TD. For patients without a TD we cannot be sure that the physician did not make up his mind (i.e. diagnosing the patient as terminal). The physician may have made an implicit diagnosis. The administrative 45 work of signing the declaration or considering patients without need of financial support might have held back the physician from signing a TD. Or, the physician might weigh up the advantages against the disadvantages of confronting the patient and the family with a terminal diagnosis. Studies have shed light on existence of this bias. Collusion in doctor-patient communication might have acted as a mechanism of delayed TD,146 and GPs have been less willing to diagnose patients as terminal compared with other doctors.147 From the patient’s point-of-view, patient coping strategies constrained an information-seeking behaviour148 Especially patients, but also informal carers, wish to maintain hope.149 Consequently, professionals may well have treated some of the patients without a TD as terminal patients. This type of misclassification may lead to some degree of underestimation of the effects of the terminal diagnosis. In another scenario, if the patient showed no sign of an approaching death we would not expect the physician to sign a TD.121 However, when analysing patient characteristics for receiving a TD, we analysed at the point of death and it must therefore be assumed that terminal signs existed for most patients.In our analysis of the effects of terminal diagnosis, we broke down the total follow-up period into shorter intervals of time (weeks) and by this demonstrated an increasing effect as death approaches. 9.2.4. Confounding – interaction – effect modification If we want to draw inference of our studies, it is essential to reflect alternative explanations like confounding, interaction and effect modification. In our studies, by using the dates in the registers we could conduct careful time analysis which fulfilled the absolute most important criterion of a causal relationship: that the exposures of interest had taken place before the outcomes (temporal relationship). In all our studies, the inter-relations of dates of admission and discharge from hospital, dates of GP home visits and 46 date of TD were analysed. Further, time at risk, i.e. time possible (not hospitalised) to receive GP home visit, was analysed. The analytical strategy employed was the use of GP home visit rates (the number of home visits relative to time spent at home) as the primary exposure in the analysis of place of death. It is worth mentioning that using home visit rates as exposure implicitly accounts for varying lengths of patients time at risk. Our analysis identified the rate of home visits as a predictor of place of death and showed a strong dose-response relationship with respect to place of death. In the studies we have adjusted for a series of observed confounders, i.e. age, gender, type of cancer and proxies for informal carers (marital status and children). However, those observed confounders only to some extent portrayed real life. In the three studies, some potentially important, incompletely observed confounders have been discussed: 1) Inequality in disease status and need of care in the comparison groups, 2) Patient factors as patients’ preference and awareness of an approaching death. Re 1) Inequality in disease status and need of care: In Study I, to make the groups more equal we eliminated all services the last 2 weeks of life in the adjusted regression and in the matched case-control study, matching on the last period before the last admission (of cases) and found, that the association persisted. In Study III, the matching was done with respect to time to death (considering this the most relevant time scale due to severity of the disease) and combined with adjusting for observed patient characteristics and in the last part stratified on TD status. Concerned that this might not capture all relevant information of the patient’s disease status and need of care, a sub-analysis using only those matched pairs, where the TD was received less than one week before death (assuming that near death the patients were more alike in the comparison groups) was performed. As expected this removed a significant portion of the previously identified effect, which supported a suspicion of confounding by indication. But 47 although diminished, the effect of GP home visits on the chance of receiving a TD remained very strong and statistically significant, despite the small sample size available for this analysis. The most important analytic strategy was the second analysis in Study III, i.e., stratification on TD status. We found a strong, dose-response relationship in the TD group which we assume must be more equal. Re 2) Patient factors as patients’ preference, committed spouse and awareness of an approaching death: In case of the patient’s capacity to influence the GP home visit whatsoever it was the patient with very strong desire of a home death and/ or strong social network, i.e. a dedicated and committed spouse (who understands the health system making it more likely that the GP was called in appropriately). The results are interesting, since the established dose-response relationship would seem to confirm the usefulness of GP visits for implementing such a wish. GP home visit is probably neither sufficient nor necessary for home death but there might be a important relation. Exposure (GP home visit) occurred before outcome (temporal relationship). A strong relationship (strength of association) was found and as the dose of exposure (GP home visit’s) increased, the odds of home death also increased (dose-response relationship). As cancer patients approach death one must assume that the need for care increases (the rule of plausibility fulfilled). We have discussed alternative explanations (confounding) and the consistency with other knowledge. However, GP home visit is undoubtedly not a single causal factor for cancer patients’ place of death. The unobserved confounders mentioned: Patients’ preference for place of death and informal carer’s support in home care might act as causes to prevent/ not prevent end-of-life hospital admission, i.e. interaction (synergistic or antagonistic). Effect modifiers in 48 this study will be factors that modify the effect of GP home visit and GP persistent involvement, for example impractical housing conditions. 9.2.5. Using place of death as outcome measure - pros et cons Place of death has been used as a suitable outcome measure in palliative care research for many years and in many countries.10,11,13,15,18,150 Although we consider home death as advantageous, not all cancer patients can or ought to die at home. Patients living alone, patients with no informal carer or carers with no interest or no capability to cope with end-of-life home care are presumably better cared for in hospital, hospice or nursing home. Further, although cancer patients stated that home is the preferred place for dying,2,3,4,5 in some cases this preference fades slightly as death approaches. Reasons have been found to be tacit protection of the informal carer, not to be a burden for the family, terminal delirium, intractable pain or other unbearable symptoms resembling risk of bleeding or fear of death. However, home is the predominantly preferable place of death, and dying at home represents some of the key values in “quality of end-of-life cancer care”. Place of death as outcome measure is therefore useful. 9.3 Discussion of the results in relation to other studies Our findings that GP home visit and persistent involvement by the GP are associated with place of death confirm our hypothesis that the GP seems to have various fundamental values in the quality of end-of-life cancer care. Our results are, however, the first to show in an unselected population, that GP home visit in the last weeks of end-stage cancer patients’ lives not only has an important impact on elements of quality of end-of-life care, i.e. staying at home and dying at home, but that this impact is largely independent of whether or not the patient received a formal terminal diagnosis. Further, it is interesting that GP home visits appear to influence, first, the chance of receiving a TD, which in 49 itself lowers the risk of dying in hospital, and, second, the risk of dying in hospital. This double finding seems to suggest that persistent involvement by the GP has beneficial consequences for endstage cancer patients on multiple levels. Undoubtedly, a complicated network of factors affect place of death of cancer patients. Homecare, home visits, patients’ preference, living with relatives, extended family support and caregivers’ preference were the most important factors in a recent review in BMJ.160 These findings do not contradict our results. We confirmed previous studies, which have demonstrated that patients with haematological cancer are more likely to die in hospital.12,13,161,162 In Study II, a smaller number of female and elderly cancer patients received the formal terminal diagnosis and TD was associated with fewer hospital admissions and increased possibilities of dying at home. To our knowledge, this is one of the first studies in an unselected population of cancer patients to highlight the very difficult transition from curative to palliative focus in end-of life cancer care. Hospice referral or referral to a specialised palliative home care programme (which could be used as a marker of that switch) often implies more intensive palliative treatment. In North America the average patient stays less than 3 weeks in a hospice before dying. 163 A diagnosis of “terminal stage” can often be made earlier and the median period of survival after receiving a TD was in our study more than double as long (52 days). It is important to emphasize that the TD does not change the patients overall rights in the Danish healthcare system and the patient does, for example, not relinquish any acute care benefits as a consequence. In contrast, referral to a hospice in US may change the healthcare access for patient and family in the US Medicare program and this may have an influence on the timing of hospice admission.119 This influence of the healthcare system has recently been discussed in an editorial and an article based on our study.164,165 Many patients are not in need of intensive terminal care when they receive the TD. 50 But what is very important, the TD gears the healthcare system towards palliative care which might ensure cancer patients’ sufficient symptom control and sufficient support to informal carer.64,67 In our study we found that fewer female and fewer elderly cancer patients received TD, even after adjusting for important confounders. The gender difference could be explained by the fact that women are still the primary caregivers in the family and hence are more proactive getting the services needed for their spouse. That elderly patients were less likely to get terminal declaration may partially be explained by the hypothesis that elderly cancer patients are more reluctant to contact healthcare services. A relationship is observed between increasing age and advanced stage of the disease at time of diagnosis.166 From a healthcare service point-of-view, elderly patients (65+) may, however, also be less likely to receive cancer-related treatment116 and physicians tend to under-treat pain in older patients.167 The elderly were also found to suffer delays in accessing terminal benefits168 and hospice referral71 which corresponds well with our results. As expected, short survival time, i.e. time from cancer diagnosis until death, was found to reduce the proportion receiving a terminal declaration. Aggressive progression of the disease provides a reduced amount of time for the physician to detect a beginning terminal phase. 51 10. CONCLUSIONS 1. Dying at home rather than at hospital was associated with GP visit and, to a lesser extent, community nurse visits. Clinical and socio-demographic characteristics of cancer patients only showed weak associations to home death. 2. Women and the elderly were less likely to receive a formal terminal diagnosis (TD). The formal terminal diagnosis reduced hospital admissions and increased the possibilities of dying at home. 3. Persistent involvement by the GP is associated with increased possibilities of dying at home. Provided that the time dimension is considered, the mortality follow-back design can be a suitable and an ethical research method to highlight and monitor end-of-life cancer care. 52 11. 11.1. PERSPECTIVES Healthcare For the healthcare providers, our studies show the importance of making the terminal diagnosis explicit. It is not surprising that an explicit terminal diagnosis facilitates planning of palliative care and thus has an effect on end-of-life care. However, only one third of the patients in our study received a terminal declaration. The late recognition of an approaching death might not allow healthcare professionals time to discuss feasibility of care and death at home. What is more important is that our results shed further light on the probably insufficient services to some subgroups of end-stage cancer patients. Further studies and initiatives should focus on these groups to facilitate the necessary access to palliative care services. The future challenge might be a more differentiated triage of patients with palliative care needs. It is remarkable in the present studies that even among patients receiving a TD (where the vast majority might be expected to desire death at home) the importance of involvement by the GP is very strong, suggesting that future research should focus on the organisation of the daily GP workload in order to support the GP in incorporating end-of-life cancer care in daily practice. Such organisational initiatives could include financial support in terms of improved pay to GPs conducting home visits or guaranteeing access to sufficient supervision and facilities in daily GP work. 11.2. Future research Our quantitative studies have generated research questions that demand other research methods to clarify the influence of the GP in end-of-life cancer care. A future study with in-depth analysis of the GP in the context of home-based palliative care is planned. In organisational 53 research, studies may identify barriers for GP home visits in end-stage cancer care and studies may focus on the organisation of the daily GP workload in order to support the GP in incorporating end-of-life cancer care in daily practice. The results based on registers cannot elucidate mechanisms in doctor-patient communication or mechanisms in patient or doctor inner life. However, provided that the time dimension is analysed, the mortality follow-back design can be a suitable and an ethical research method to highlight problems and monitor quality in end-of-life cancer care. 54 12. SUMMARIES 12.1. Dansk resume [Danish summary] Ph.d. afhandlingen består af en oversigt og tre originalarbejder. Afhandlingen udgår fra Forskningsenheden for Almen Praksis i Odense, Syddansk Universitet. Baggrund At tilbringe så megen tid som muligt i eget hjem og om muligt at dø hjemme er af stor betydning for terminalt syge kræftpatienter og deres familier. Imidlertid dør betydeligt færre end ønsket i eget hjem i de fleste vestlige lande. Det er uklart, hvilke faktorer der har indflydelse på kræftpatienters dødssted. Der er specielt mangel på befolkningsanalyser af primærsektorens indflydelse, dvs. af praktiserende lægers og hjemmesygeplejens rolle. Læger får ofte enten ikke defineret kræftpatienter som værende i terminalstadie, eller deres prognostiske estimater kan være for optimistiske. Dette kan antages at have betydning for forberedelse til eventuel hjemmedød, rettidig sufficient smertebehandling, aflastning af pårørende og for, at patienter i tide kan blive henvist til palliativ specialistbistand. Formål Afhandlingens formål var at identificere betydende faktorer for kræftpatienters dødssted. Der er lagt særlig vægt på praktiserende lægers involvering ved sygebesøg og på effekten af terminal erklæring (TD). Metode Designet var et retrospektivt befolkningsstudie af sundhedsydelser i de sidste tre måneder af livet til patienter i Fyns Amt hos hvem kræft var primær dødsårsag i 1996-98. Patienterne blev via CPRregistret identificeret i Cancerregistret og information om sundhedsydelser blev koblet med oplysninger fra Landspatientregistret, Sygesikringsregistret og registreringer af terminal erklæring (TD) og hjemmesygepleje i de enkelte kommuner. 55 Resultater I den 3-årige studieperiode, døde 4,386 voksne patienter (18 år eller over) med kræft som primær dødsårsag. I alt 28% døde hjemme, 55% på hospital, 16% på plejehjem (ni procent under aflastningsophold) og 1% havde ukendt dødssted. Vi fandt en stærk sammenhæng mellem hjemmedød og den praktiserende læges sygebesøg samt hjemmeplejens besøg. At være gift og have børn på dødstidspunktet var associeret med hjemmedød, mens vi ikke fandt forskel mellem land og by. Kort overlevelsestid fra diagnose til død og hæmatologisk kræftsygdom var associeret med død på hospital. Terminalerklæring blev udstedt til 34% af kræftpatienterne gennemsnitligt 52 dage før død. Når overlevelsestiden fra diagnose til død var under en måned, blev der kun udstedt terminalerklæring til 8%. Udstedelse af terminalerklæring var hyppigere til gifte og patienter med visse kræftsygdomme (prostata og lunge). Endvidere var udstedelsen omvendt associeret med alder over 70 år, kvindekøn og hæmatologisk kræftsygdom. Vi fandt signifikant nedsat indlæggelsesrate efter udstedelse af terminalerklæring og risikoen for hospitalsdød var kun 25%. Vi fandt stærk association mellem udstedelse af terminalerklæring og praktiserende lægers sygebesøg i ugen inden, men også 4 uger før. Desuden var der en kraftig dosis-respons sammenhæng mellem praktiserende læges sygebesøg og dødssted både hos patienter med og uden terminalerklæring. En del af sammenhængen mellem sygebesøg og dødssted kan naturligvis skyldes, at patienter, der dør hjemme, er mere tilgængelige for sygebesøg end patienter, som er indlagt. Vi valgte imidlertid at anvende en sygebesøgsrate (sygebesøg/risikotid i hjemmet) som variabel i de logistiske regressioner. Sammenhængen fandtes endvidere i en række delanalyser: 1) delstudie, som ikke inddrog de sidste 14 levedage, hvor det må antages, at der i højere grad kan opstå et ikke-planlagt akut behov for lægehjælp (ikke nødvendigvis et udtryk for, at lægen har valgt et aktivt engagement) 56 2) matchet case-controlstudie, hvor observationsperioden var tiden forud for sidste indlæggelse før død 3) matchet case-controlstudie, hvor patienterne var sammenlignelige med hensyn til terminal erklæring Studierne taler for, at kræftpatienters mulighed for at få en terminal erklæring og for at dø i eget hjem øges, når den praktiserende læge er involveret med sygebesøg (uanset patientens alder, køn, kræfttype, samboende ægtefælle eller hvorvidt, der er børn). Konklusion og perspektiver Resultaterne tyder på, at det er væsentligt for den palliative indsats, at den praktiserende læge er involveret og at den terminale diagnose gøres eksplicit ved en terminalerklæring. Disse fund kræver bekræftelse fra studier, hvor patientens behov og præferencer inddrages mere dybtgående, end det er muligt med registerdata. Under nøje hensynstagen til den rette tidsmæssige sammenhæng kan det retrospektive design af leverede sundhedsydelser være en brugbar og etisk metode til at undersøge og monitorere den palliative indsats. 57 12.2. English summary This PhD thesis is based on an overview and three original papers. The studies were carried out at the Research Unit for General Practice, University of Southern Denmark, Odense. Background Dying at home and remaining at home as long as possible are of great importance to a majority of seriously ill cancer patients and their families. However, only a minority of patients actually die at home in most Western countries. Factors influencing place of death of cancer patients are not fully understood. There is, especially, a lack of assessment of the influence of nonspecialised community services, i.e. GP and community nurse services in an unselected population of cancer patients. Physicians do either not define cancer patients as being terminal or their prognostic estimates tend to be optimistic. This might affect preparation for home death and patients’ appropriate and timely referral to specialist palliative care services and can lead to unintended acute hospitalisation. Objectives The purpose of the thesis was to find factors of importance for place of death in unselected populations of patients dying from cancer. Special emphasis was on the effect of GP home visit and on the effect of a formal terminal diagnosis. Methods The design was a mortality follow-back study of health services the last three month of life applied to all patients with cancer as primary cause of death in the study period 1996-98 in the County of Funen. Patients were identified in the Danish Cancer Register, and the health services provided were collected from the Danish National Hospital Register, the Danish National Health Service 58 Register and the registration of community nurse activity and formal terminal declaration (TD) in the municipal administrations. Results In the 3-year study period, 4,386 adult patients (aged at least 18 years of age) died with cancer considered the primary cause of death. In the total cohort 28% died at home, 55% died in hospitals, 16% in nursing homes (nine percent on a temporary stay) and for one percent place of death was unknown. Being married and having children made it less likely to die in hospital. GP home visit so as visits by community nurse were found to be strongly inversely associated with hospital death while no association was found with geographical location. Short survival time and haematological cancer were associated with hospital death. For 34% of the patients a terminal declaration (TD) was issued. However, when survival time, i.e. time from cancer diagnosis until death, was less than one month, only 8% received a TD. The declaration was issued 52 days before death. The TD was associated with being married, having cancer of the prostate or the lung and was inversely associated with 70+ years, female gender and haematological cancer. Looking at the effects of TD, with TD the admission rate increased less and the risk of hospital death was 25%. A strong association between issue of TD and GP home visit in the week before TD, but also 4 weeks before the TD, was found. Furthermore a strong dose-response relationship was found between GP home visit and non-hospital death both among patients with TD and patients without TD. Some of the association between GP home visit and place of death could of course relate to patients dying at home being more accessible to GP home visit than patients admitted. However, we used a home visit rate i.e. home visit / risk time at home as explanatory variable in the logistic regressions. 59 Futhermore, the association was found in several subanalysis: 1) The association persisted eliminating all services the last 2 weeks in the adjusted regression (in the last 2 weeks a no-planned acute need for physician assistance is more likely which is not necessarily a marker for GP involvement.) 2) The association persisted in a matched case-control study, matching according to the time period before the last admission 3) The association persisted in the subpopulation of patients receiving a TD. The studies indicate that cancer patients’ possibility of getting a TD and of dying at home increases when the GP involves himself, e.g. home visits ( irrespective of the patient’s age, gender, cancer type, marital status or whether or not the have children. Conclusion and perspectives In summary, this population-based study points to two important factors in palliative care in the community, i.e. the importance of persistent involvement by the GP in end-of-life cancer care and the effect of making the terminal diagnosis explicit . These studies need to be confirmed by more profound studies of patients’ desires and preferences in end-of-life cancer care than possible with register-based data. Provisional perspectives could be to ensure, that GP’s are involved in organisation of palliative care and if possible, a terminal declaration is made explicit. Provided that correct temporal relations are used, the mortality follow-back design can be a suitable and an ethical research method to highlight and monitor end-of-life cancer care. 60 13. REFERENCES 1. Kmietowicz Z. Report calls for more care for people wanting to die at home. BMJ 2004;329 (7460): 248. 2. Brazil K, Howell D, Bedard M, Krueger P, Heidebrecht C. Preferences for place of care and place of death among informal caregivers of the terminally ill. Palliat Med 2005; 19: 492-9. 3. Tiernan E, O`Connor M, O`Siorain L, Kearney M. A prospective study of preferred versus actual place of death among patients referred to a palliative care home-care service. Ir Med J 2002; 95(8), 232-5. 4. Higginson I, Sen-Gupta GJA. Place of care in advanced cancer: A qualitative systematic literature review of patient preferences. J Palliat Med 2000; 3(3): 287-300. 5. Strang P. Most terminal patients want to die at home [De fleste terminalt sjuka vill dø hemma][Swedish] Lakartidningen 2002; 99(8): 742-6. 6. Lee A PW. Preferred place of death - a local study of cancer patients and their relatives. Singapore Med J 1998; 39(10): 447-50. 7. Wilson DM. End-of-life care preferences of Canadian senior citizens with caregiving experience. J Adv Nurs 2000; 31(6): 1416-21. 8. Thomas C, Morris SM, Clark D. Place of death: preferences among cancer patients and their carers. Soc Sci Med 2004; 58(12): 2431-44. 9. Weitzen S, Teno JM, Fennell M, Mor V. Factors associated with site of death. A national study of where people die. Med Care 2003; 41(2): 323-35. 10. Higginson I, 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(5): 353-63. 11. Hunt RW, Fazekas BS, Luke CG, Roder DM. Where patients with cancer die in South Australia, 1990-1999: a population-based review. Med JAust 2001; 175: 526-9. 12. Bruera E, Russell N, Sweeney C, Fisch M, Palmer L. Place of death and its predictors for local patients registered at a comprehensive cancer center. J Clin Oncol 2002; 20: 2127-33. 13. Burge F, Lawson B, Johnston G. Trends in the place of death of cancer patients, 1992–1997. CMAJ 2003; 168(3): 265-70. 14. Karlsen S, Addington-Hall JM. How do cancer patients who die at home differ from those who die elsewhere? Palliat Med 1998; 12(4): 279-86. 15. Axelsson A, Christensen SB. Place of death correlated to sociodemografic factors - A study of 203 patients dying of cancer in a rural Swedish country in 1996. Palliat Med 1996; 10(4): 329-35. 61 16. Wilson DM, Smith SL, Anderson MC, Northcott HC, Fainsinger RL, Stingl MJ, Truman CD. Twentieth-century social and health-care influences on location of death in Canada. Can J Nurs Res 2002; 34(3):141-61. 17. Heyland DK, Lavery JV, Tranmer JE, Shortt SE, Tayler SJ. Dying in Canada. Is it an institutionalized, tecnologically supported experience? J Palliat Care 2000; 16 Suppl: S10-6. 18. McCusker J. Where cancer patients die: an epidemiologic study. Public Health Rep 1983; 98(2), 170-6. 19. Huang J, Boyd C, Tyldesley S, Zhang-Salomons J, Groome PA, Mackillop WJ. Time spent in hospital in the last sex month of life in patients who died of cancer in Ontario. J Clin Oncol 2002; 20(6): 1584-92. 20. Angus DC, Barnato AE, Linde-Zwirble WT, Weissfeld LA, Watson S, Rickert T et al. Use of intensive care at the end of life in the United States: An epidemiologic study. Crit Care Med 2004; 32(3): 638-43. 21. Wilson DM, Northcott HC, Truman CD, Smith SL, Anderson MC, Fainsinger RL, Stingl MJ. Location of death in Canada. Eval Health Prof 2001; 24(4):385-403. 22. Townsend J, Frank AO, Fermont D, Dyer S, Karran O, Walgrove A. Terminal cancer care and patients preference for place of death: a prospective study. BMJ 1990; 301(6749), 4157. 23. Aries P. Western attitudes toward death: From middle ages to the present. London: The John Hopkins University Press, 1974. 24. Floto I. An expedition in the history of death. [Strejftog i dødens historie][Danish]. Historisk Tidsskrift 2006; 105: 164-78. 25. Brock S, Samsoe A. (ed) As we are waiting for death. Death in ethical and philosophic light. [Mens vi venter på døden. Døden i etisk og kulturfilosofisk belysning][Danish]. Aarhus: Philosophia, 2005. 26. Richmond C. Dame Cicely Saunders. BMJ 2005; 331(7510): 238. 27. Kubler-Ross E. On Death and Dying, 1969. [Døden og den døende][Danish] København: Nordisk Forlag, 1995. 28. Nuland SB. How we die: Reflections of life's final chapter. New York: Alfred A Knopf, 1994. 29. Thorpe G. Enabling more dying people to remain at home. BMJ 1993; 307(6909): 915-8. 30. Kaasa S. From hospice movement to palliative medicine[Fra hospicebevegelse til palliativ medisin][Norwegian]. Tidsskr Nor Laegeforen 1995; 115(1), 58-60. 62 31. Kaasa S. Palliative care [Palliativ behandling og pleie.Nordisk lærebok] Oslo: Ad Notam Gyldendal, 1998. 32. Danish Regions [Amtsrådsforeningen], The Ministry of Health [Sundhedsministeriet], Local Government Denmark[Kommunernes Landsforening]. ”Help with living until ones dies” [”Hjælp til at leve til man dør”][Danish]. 2001. 33. Bruera E, Neumann CM, Gagnon B, Brenneis C, Kneisler P, Selmser P, Hanson J. Edmonton regional palliative care program: impact on patterns of terminal cancer care. CMAJ 1999; 161(3): 290-3. 34. World Health Organization. Pain relief and palliative care. In: National cancer control programmes. Policies and managerial guidelines. 2nd ed. Geneva: World Health Organization, 2002; 83-91. 35. Ahmedzai SH, Costa A, Blengini C, Bosch A, Sanz-Ortiz J, Ventafridda V, Verhagen SC; international working group convened by the European School of Oncology. A new international framework for palliative care. Eur J Cancer 2004; 40: 2192-200. 36. Benson LR. End-of-life cancer care: Progress still needed. J Natl Cancer Inst 2001; 93(18):1368-9. 37. Emanuel EJ, Emanuel LL. The promise of a good death. Lancet 1998; 351:21-9. 38. Singer PA, Martin DK, Kelner M. Quality End-of-Life Care. Patients Perspective. JAMA 2001; 281(2):163-8. 39. Aspinal F, Hughes R, Dunckley M, Addington-Hall J. What is important to measure in the last months and weeks of life? A modified nominal group study. Int J Nurs Stud. 2005 Aug 13; [Epub ahead of print] 40. Steinhauser KE, Christakis N, Clipp EC, McNeilly M, Grambow S, Parker J et al. Preparing for the end of life: Preferences of patients, families, physicians, and other care providers. J Pain Sympt Manage 2001; 22(3):727-37. 41. Steinhauser KE, Chistakis NA, Clipp EC, MvNeilly M, McIntyre L, Tulsky JA. Factors considered important at the end of life by patients, family, physicians, and other care providers. JAMA 2000; 284: 2476-82. 42. Ministry of the Interior and Health. The Local Governement Reform. http://www.im.dk/publikationer /government_reform_in%20brief/index.htm., 13-02-2006. 43. Christiansen T. Organization and financing of the Danish health care system. Health Policy 2002; 59:107-18. 44. Hee I, Henriksen H, Riis J. Sankt Lukas' Home Hospice - palliative team in County of Copenhagen [Sankt Lukas Hjemmehospice - palliativt team i Københavns Amt] [Danish]. Hellerup: Diakonissehuset Sankt Lukas Stiftelsen, 1998 63 45. Goldschmidt D. Evaluation of palliative home care: views of patients, carers, general practitioners and district nurses. PhDThesis. Copenhagen:University of Copenhagen, 2006. 46. Åbom BM, Obling NJ, Rasmussen H, Kragstrup J. Unplanned emergency admission of dying patients. Causes elucidated by focus group interviews with general practitioners [Utilsigtet akut indlæggelse af døende][Danish] Ugeskr Laeger 2000; 162(43), 5768-71. 47. Groot M, Vernooij_Dassen MJFJ, Crul BJP, Grol RPTM. General practitioners (GPs) and palliative care: perceived tasks and barriers in daily practice. Palliat Med 2005; 19(2):11118. 48. Lamont EB, Christakis NA. Complexities in prognostication in advanced cancer: "to help them live their lives the way they want to". JAMA 2003; 290(1):98-104. 49. Gordon EJ, Daugherty CK. 'Hitting you over the head': Oncologists' disclosure of prognosis to advanced cancer patients. Bioethics 2003; 17(2):142-68. 50. Cappuccini F, Petty W, Cain J. Palliative care: a critical component of care for women. Current Obstetrics and Gynaecology 2003; 13: 166-72. 51. Lamont EB, Christakis NA. Some elements of prognosis in terminal cancer. Oncology (Huntingt) 1999; 13(8):1165-70. 52. Glare PA, Eychmuller S, McMahon P. Diagnostic accuracy of the palliative prognostic score in hospitalized patients with advanced cancer. J Clin Oncol 2004; 22 (23):4823-8. 53. Hoang T, Xu R, Schiller JH, Bonomi P, Johnson DH. Clinical model to predict survival in chemonaive patients with advanced non-small-cell lung cancer treated with third-generation chemotherapy regimens based on eastern cooperative oncology group data. J Clin Oncol 2005; 23(1): 175-83. 54. Glare P, Virik K, Jones M, Hudson M, Eychmuller S, Simes J, Christakis N. A systematic review of physicians' survival predictions in terminally ill cancer patients. BMJ 2003; 327(7408): 195. 55. Christakis NA, Lamont EB. Extent and determinants of error in doctors' prognoses in terminally ill patients: prospective cohort study. BMJ 2004; 320(7233): 469-72. 56. Chow E, Harth T, Hruby G, Finkelstein J, Wu J, Danjoux C. How accurate are physicians' clinical predictions of survival and the available prognostic tools in estimating survival times in terminally ill cancer patients? A systematic review. Clin Oncol (R Coll Radiol) 2001; 13(3): 209-18. 57. Vigano A DM, Bruera E, Suarez-Almazor ME. The relative accuracy of the clinical estimation of the duration of life for patients with end of life cancer. Cancer 1999; 86(1), 170-6. 64 58. Brinckner L, Scannell K, Marquet S, Ackerson L. Barriers to hospice care and referrals: survey of physicians' knowledge, attitudes, and perceptions in a health maintance organization. J Palliat Med 2004; 7(3): 411-8. 59. Ogle K, Mavis B,Wang T. Hospice and primary care physicians: attitudes, knowledge, and barriers. Am J Hosp Palliat Care 2003; 20(1):41-51. 60. Zenz M, Willweber -Strumpf A. Opiophobia and cancer pain in Europe. Lancet 1993; 341(1075): 1076. 61. Bercovitch M, Adunsky A. Patterns and high-dose morphine use in a home-care hospice service. Should we be afraid of it? Cancer 2004; 101: 1473-7. 62. Carr DB, Goudas LC, Balk EM, Bloch R, Ioannidis JPA, Lau J. Evidence report on the treatment of pain in cancer patients. J Natl Cancer Inst Monogr 2004; 32: 23-31. 63. Grunfeld E, Coyle D, Whelan T, Clinch J, Reyno L, Earle C, Willan A, Viola R, Coristine M, Janz T, Glossop R. Family caregiver burden: results of a longitudinal study of breast cancer patients and their principal caregivers. CMAJ 2004; 170(12): 1795-1801. 64. Visser G, Klinkenberg M, Broese van Groenou MI, Willems DL, Knipscheer CPM, Deeg DJH. The end of life: informal care for dying older people and its relationship to place of death. Palliative Medicine 2004; 18:468-77. 65. Grande G, Farquhar M, Barclay S, Todd C. Valued aspects of primary palliative care: content analysis of bereaved carers' descriptions. Br J Gen Pract 2004; 54(507):772-8. 66. Simon C KT. Informal carers--the role of general practitioners and district nurses. Br J Gen Pract 2001; 51(469):655-7. 67. Given B, Wyatt G, Given C, Sherwood P, Gift A, DeVoss D, Rahbar M. Burden and depression among caregivers of patients with cancer at the end of life. Oncol Nurs Forum 2004;31(6):1105-17. 68. Saunders Y, Ross JR, Riley J. Planning for a good death: responding to unexpected events. BMJ 2004; 327(7408):204-7. 69. Earle CC. Outcomes research in lung cancer. J Natl Cancer Inst Monogr 2004; 33:56-77. 70. de Braud F, Cascinu S, Gatta G. Cancer of pancreas. Crit Rev Oncol Hematol. 2004; 50: 147-55. 71. McCarthy EP, Burns RB, Davis RB, Phillips RS. Barriers to hospice care among older patients dying with lung and colorectal cancer. J Clin Oncol 2003; 21:728-35. 72. Danish Medicines Agency. Reimbursement for the terminally ill [Terminal tilskud]. http://www.laegemiddelstyrelsen.dk/1024/visLSArtikel.asp?artikelID=872, 13-2-2006. 65 73. Law on Social Services.[Lov om social service] www.retsinfo.dk/ INDEX. 9-2-2006. 74. Michelsen N, Jensen BT, Nielsen CW, Schroll M. Clinical Social Medicine. [Klinisk Social Medicin][Danish] København: Munksgaard, 2000. 75. Ministry of Social Affairs.[Socialministeriet] Government notice of Law of Social Service [Bekendtgørelse af lov om social service]LBK nr 844 of 24/09/2001, chapter 20. Care of dying [Pasning af døende], 2001. 76. Timm H. Evaluation of terminal leave [Evaluering af plejevederlagsordningen][Danish] København: UCSF & Socialministeriet, 1999. 77. Mikkelsen EM. Terminal leave.[Terminal plejeorlov][Danish] Master of Public Health Aarhus Universitet, 1999. 78. Terminal leave.[Terminal plejeorlov] http://www.social.dk/kontante_ydelser/doeende/plejevederlag.html, 2006. 79. Danish Medicines Agency.[Lægemiddelstyrelsen] isUKLSForside.asp?artikelID=728. 2006. 80. Gysels M, Higginson I. Improving supportive and palliative care for adults with cancer. Research evidence. National Institute for Clinical Excellence 2004: 1-384. 81. Cherin DA, Enguidanos SM, Jamison P. Physicians as medical center "extenders" in end-of life care: Physician home visits as the lynch pin in creating an end-of-life care system. Home Health Care Serv Q. 2004; 23(2): 41-53. 82. Brazil K, Bedard M, Willison K. Factors associated with home death for individuals who receive home support services; a retrospective cohort study. BMC Palliat Care 2002; 1(1): 2. 83. Cantwell P, Turco S, Brenneis, Hanson J, Neumann CM, Bruera E. Predictors of home death in palliative care cancer patients. J Palliat Care 2000; 16(1): 23-8. 84. Borgsteede SD, Graafland-Riedstra C, Deliens L, Francke AL, van Eijk JT, Willems D. Good end-of-life care according to patients and their GPs. Br J Gen Pract 2006; 56(522): 206. 85. Wright EB, Holcombe C, Salmon P. Doctors' communication of trust, care, and respect in breast cancer: qualitative study. BMJ 2004; 328 (7444): 864-7. 86. Lee E. Learning from death. Br J Gen Pract 2002; 52: 267-8. 87. Rabow MW, Hauser JM, Adams J. Supporting family caregivers at the end of life: "They don't know what they don't know". JAMA 2004; 291(4): 483-91. 88. Gwinn RB. Family physicians should be experts in palliative care. Am Fam Physician 2000; 61(3), 636-42. http://www.dkma.dk/ 1024 / 66 89. Farber SJ, Egnew TD, Herman-Bertsch JL. Issues in end-of-life care. Family practice faculty perceptions. J Fam Pract 1999; 49, 525-30. 90. Østergaard I. Terminal care in the primary sector. [Terminal pleje i primærsektoren][Danish] Medicinsk årbog. Copenhagen, Munksgaard, 1999: 31-41. 91. Humphrey A. Home visits and palliative care. N Z Med J 2000; 113: 64-5. 92. Henriksen H, Davidsen-Nielsen M. The doctor and the dying patient.[Lægen og den døende patient.][Danish] Maanedsskr Prakt Laegegern 2000; 78, 923-31. 93. Osse BH, Vernooij_Dassen MJFJ, Schade E, de Vree B, van den Muijsenbergh METC, Grol RPTM. Problems to discuss with cancer patients in palliative care: a comprehensive approach. Patient Educ Couns. 2002; 47(3): 195-204. 94. Vejlsgaard T, Addington-Hall J. Attitudes of Danish doctors and nurses to palliative and terminal care. Palliat Med 2005; 19:119-27. 95. Porta M, Busquet X, Jariod M. Attitudes and views of physicians and nurses towards cancer patients dying at home. Palliat Med 1997; 11(2): 116-26. 96. Farber NJ, Urban SY Collier VU, Metzger M, Weiner J, Boyer EG. Frequency and percived competance in providing palliative care to terminally ill patients: A survey of primary care physicians. J Pain Symptom Manage. 2004; 28: 364-72. 97. Kenyon Z. Palliative care in general practice. BMJ 1995; 311(7010): 888-9. 98. Schuit KW, Bender W, Meijler WJ, Otter R, Meyboom-Dejong B, Sleijfer DT. Learning effects of a workshop in palliative cancer care for general practitioners. J Cancer Educ. 1999; 14(1), 18-22. 99. Schuit KW. Palliative care in general practice. Research and education. (Thesis). Groningen: 1999. 100. Brenneis C, Bruera E. The interaction between family physicians and palliative care consultants in the delivery of palliative care: clinical and educational issues. J Palliat Care 1998; 14(3): 58-61. 101. Hall P, Hupe D, Scott J. Palliative care education for community-based family physicians: the development of a program, the evaluation, and its consequences. J Palliat Care 1998; 14(3): 69-74. 102. Burge F, Lawson B, Johnston G, Cummings I. Primary care continuity and location of death for those with cancer. J Palliat Med. 2003; 6(6):911-8. 103. Marie Curie Memorial. Cancer nursed at home. Lancet 1952; 1(21):1061-2. 67 104. Kaasa S, Loge JH. Quality-of-life assessment in palliative care. Lancet Oncol. 2002; 3(3):175-82. 105. Kaasa S, Loge JH. Quality of life in palliative care: principles and practice. Palliat Med 2003; 17(1):11-20. 106. Groenvold M, Petersen MA, Aaronson NK, Arraras JI, Blazeby JM, Bottomley A, Fayers PM, de Graeff A, Hammerlid E, Kaasa S, Sprangers MA, Bjorner JB, EORTC Quality of Life Group. The development of the EORTC QLQ-C15-PAL: a shortened questionnaire for cancer patients in palliative care. Eur J Cancer 2006; 42(1):55-64. 107. Higginson I, Goodwin D, Cook AM, Hood K, Edwards AGK, Douglas H-R, Norman CE. Do hospital-based palliative teams improve care for patients or families at the end of life? J Pain Symptom Manage 2002; 23(2): 96-106. 108. Grande GE, Todd CJ. Why are trials in palliative care so difficult? Palliat Med 2000; 14:6974. 109. Jordhoy MS, Kaasa S, Fayers P, Ovreness T, Underland G, Ahlner-Elmqvist M. Challenges in palliative care research; recruitment, attrition and compliance: experience from a randomized controlled trial. Palliat Med 1999; 13(4), 299-310. 110. Rinck GC, van den Bos GA, Kleijnen J, Haes de HJ, Schade E, Veenhof CH. Methodologic issues in effectiveness research on palliative cancer care: a systematic review. J Clin Oncol. 1997; 15: 1697-1707. 111. Keeley D. Rigorous assessment of palliative care revisted. Wisdom and compassion are needed when evidence is lacking. BMJ 1999; 319(7223), 1447-8. 112. Salisbury C, Bosanquet N. Assessing palliative care is difficult. BMJ 2000; 320 (7239), 942. 113. Jordhoy MS, Fayers P, Santnes T, Ahlner-Elmqvist M, Jannert M, Kaasa S. A palliative care intervention and death at home: a cluster randomised trial. Lancet 2000; 356:888-93. 114. Cardenas-Turanzas M, Grimes RM, Bruera E, Quill B, Tortolero-Luna G. Clinical, sociodemographic, and local system factors associated with a hospital death among cancer patients. Support Care Cancer 2006; 14(1):71-7. 115. Burge F, Lawson B, Johnston G. Home visits by family physicians during the end-of-life: Does patient income or residence play a role? BMC Palliat Care 2005; 4(1):1. 116. Earle CC, Neville BA, Landrum MB, Ayanian JZ, Block SD, Weeks JC. Trends in the aggressiveness of cancer care near the end of life. J Clin Oncol 2004; 22:315-21. 117. Earle CC, Park ER, Lai B, Weeks JC, Ayanian JZ, Block S. Identifying potential indicators of the quality of end-of-life cancer care from administrative data. J Clin Oncol 2003; 21: 1133-8. 68 118. Emanuel EJ, Yinong Young-Xu, Levinsky N, Gazelle G, Saynina O, Ash A. Chemotherapy use among Medicare beneficiaries at the end of life. Ann Intern Med 2003; 138 (8):639-43. 119. McCarthy EP, Burns RB, Ngo-Metzger Q, Davis RB, Phillips RS. Hospice use among medicare managed care and fee-for-service patients dying with cancer. JAMA 2003; 289: 2238-45. 120. Wennberg JE, Fisher ES, Stukel TA, Skinner JS, Sharp SM, Bronner KK. Use of hospitals, physician visits, and hospice care during last six months of life among cohorts loyal to highly respected hospitals in the United States. BMJ 2004; 328 (7440): 607. 121. Bach P, Schrag D, Begg CB. Resurrecting treatment histories of dead patients. JAMA 2004; 292: 2765-70. 122. Rasmussen S, Madsen M. Registers in health care. [Registre inden for sundhedsområdet] København: DIKE, 1997. 123. Registerdata- an unique research ressource in Denmark [Registerdata - en enestående dansk forskningsressource.] København: Dansk Grundforskningsfond, 2000. 124. McCusker J. The terminal period of cancer: definition and descriptive epidemiology. J Chronic Dis 1984; 37(5): 377-85. 125. Sessa C, Roggero E, Pampallona S, Regazzoni S, Ghielmini M, Lang M et al. The last 3 months of life of cancer patients: medical aspects and role of home-care services in Southern Switzerland. Support Care Cancer 1996; 4(3):180-5. 126. Llobera J, Esteva M, Rifa J, Benito E, Terrasa J, Rojas C, Pons O, Catalan G, Avella A. Terminal cancer. Duration and prediction of survival time. Eur J Cancer 2000; 36(16):203643. 127. Costantini M, Toscani F, Gallucci M, Brunelli C, Miccinesi G, Tampaci E et al. Terminal cancer patients and timing of referral to pallitive multicenter prospective cohort study. Italian Cooperative Research Group on Palliative Medicine. J Pain Symp Manage.1999; 18(4), 243-52. 128. Vigano A, Dorgan M, Buckingham, Bruera E, Suarez-Almazor ME. Survival prediction in terminal cancer patients: a systematic review of the medical literature. Palliat Med 2000; 14(5), 363-74. 129. Vigano A, Bruera E, Jhangri GS, Newman SC, Fields AL, Suarez-Almazor ME. Clinical survival predictors in patients with advanced cancer. Arch Intern Med 2000; 160(6), 861-8. 130. Maltoni M, Pirovano M, Nanni O, Marinari M, Indelli M, Gramazio A, Terzoli E, Luzzani M, De Marinis F, Caraceni A, Labianca R. Biological indices predictive of survival in 519 Italian terminally ill cancer patients. Italian Multicenter Study Group on Palliative Care. J Pain Symptom Manage. 1997; 13(1):1-9. 69 131. den-Daas N. Estimating length of survival in end-stage cancer: A review of the literature. J Pain Symptom Manage. 1995; 10(7):548-55. 132. Malig C. The Civil Registration System in Denmark. Bethesda, MD: International Institute for Vital Registration and Statistics 1996; Technical paper No 66. 133. Storm HH, Michelsen EV, Clemmensen I, Pihl J. The Danish Cancer Registry - history, content, quality and use. Dan Med Bull 1997; 44: 535-9. 134. National Board of Health.[Sundhedsstyrelsen] The Cancer Register [Cancerregistret] 1998. New figures from the National Board of Health [Nye tal fra sundhedsstyrelsen] 2002; 6(6):1-18. 135. Juel K, Helweg-Larsen K. The Danish register of causes of death. Dan Med Bull 1999; 46: 354-57. 136. Olivarius N, Hollnagel H, Krasnik A, Pedersen PA, Thorsen H. The Danish National Health Service Register. A tool for primary health care research. Dan Med Bull 1997; 44(4): 44953. 137. Storm HH. Health care system, cancer registration and follow-up of cancer patients in Denmark. IARC Sci Publ 1995; 132: 48-50. 138. Storm HH. Validity of Death Certificates for Cancer Patients in Denmark 1977. The Danish Cancer Register.[Validitet af danske dødsattester for cancerpatienter 1977][Danish] København: Stougaard Jensen, 1984. 139. Storm HH. Completeness of cancer registration in Denmark 1943-1966 and efficacy of record linkage procedures. Int J Epidemiol. 1988; 17: 44-9. 140. Thorsteinsson R, Sørensen M, Jensen TL, Bernhardtsen TM, Gjerris F, Carstensen H, Schmiegelow K, Raaschou-Nielsen O. Central nervous system tumours in children. An evaluation of the completeness and validity of the Cancer Registry.[Tumorer i centralnervesystemet hos børn. En evaluering af cancerregisterets komplethed og validitet] [Danish] Ugeskr Laeger 2005; 167(40): 3782-5. 141. Andersen TF, Madsen M, Jørgensen J, Mellemkjaer L, Olsen JH. The Danish National Hospital Register. Dan Med Bull 1999; 46: 263-8. 142. Juel K. Causes of death registration in Denmark [Dødsårsagsregistreringen i Danmark]. Ugeskr Laeger 1998; 160, 5019. 143. National Board of health. [Sundhedsstyrelsen] Causes of Death Register 2000. [Dødsårsagsregistret 2000][Danish]. 2002. 144. Kjaergaard J, Clemmensen IH, Storm HH. Validity and completeness of registration of surgically treated malignant gynaecological diseases in the Danish National Hospital 70 Registry. J Epidemiol Biostat 2001; 6(5):387-92. 145. Jensen AR, Overgaard J, Storm HH. Validity of breast cancer in the Danish Cancer Registry. A study based on clinical records from one county in Denmark. Eur J Cancer Prev 2002; 11(4):359-64. 146. The A, Hak T, Koeter G, van der Wal G. Collusion in doctor-patient communication about imminent death: an ethnographic study. BMJ 2000; 321(7273):1376-81. 147. Farquhar M, Grande G, Todd C, Barclay S. Defining patients as palliative: hospital doctors' versus general practitioners' perceptions. Palliat Med 2002; 16: 247-50. 148. Leydon GM, Boulton M, Moynihan C, Jones A, Mossman J, Boudioni M, McPherson K. Cancer patients' information needs and information seeking behaviour: in depth interview study. BMJ 2000; 320(7239): 909-13. 149. de Haes H, Koedoot N. Patient centered decision making in palliative cancer treatment: a world of paradoxes. Patient Educ Couns 2003; 50(1):43-9. 150. Polissar L, Severson RK, Brown NK. Factors affecting place of death in Washington State, 1968-1981. J Community Health 1987; 12(1): 40-55. 151. Grande GE, Farquhar MC, Barclay SI, Todd CJ. Valued aspects of primary palliative care: content analysis of bereaved carers' descriptions.Br J Gen Pract 2004; 54(507): 72-8. 152. Mittchell GK. How well do general practitioners deliver palliative care? A systematic review. Palliat Med 2002; 16: 457-64. 153. Walsh D, Regan J. Terminal care in the home--the general practice perspective. Ir Med J. 2001; 94(1):9-11. 154. Danish College of General Practitioners. Palliation in primary sector. [Palliation i primærsektoren][Danish], 2005. http://www.dsam.dk 155. National Institute for Clinical Excellence. Specialist palliative care services. Improving Supportive and palliative care for adults with cancer. The manual. 2004: 123-33. 156. Goldscmidt D, Groenvold M, Johnsen AT, Stromgren A, Krasnik A, Schmidt L.Cooperating with a palliative home-care team: expectations and evaluations of GPs and district nurses. Palliat Med. 2005; 19: 241-50. 157. Burt J, Barclay S, Marshall N, Shipman C, Stimson A, Young J. Continuity within primary palliative care: an audit of general practice out-of-hours co-operatives. J Public Health 2004; 26(3):275-6. 158. Mittchel GK, Cherry M, Kennedy R, Weeden K, Burridge L, Clavarino A, O’Rourke P, Del Mar C. General practitioner, specialist providers case conferences in palliative care--lessons learned from 56 case conferences. Aust Fam Physician. 2005; 34(5): 389-92. 71 159. Shipman C, Addington-Hall J, Thompson M, Pearce A, Barclay S, Cox I, Maher J, Millar D. Building bridges in palliative care: evaluating a GP Facilitator programme. Palliat Med. 2003;17(7): 621-7. 160. Gomes B, Higginson I. Factors influencing death at home in terminally ill patients with cancer: systematic review. BMJ 2006; 332(7540): 515-21. 161. Bruera E, Sweeney C, Russell N, Willey JS, Palmer JL. Place of death of Houston Area residents with cancer over a two-year period. J Pain Sympt Manage. 2003; 26(1): 637-43. 162. Gallo WT, Baker MJ, Bradley EH. Factors asssociated with home versus interstitutional death among cancer patients in Connecticut. J Am Geriatr Soc 2001; 49(771): 777. 163. Haupt BJ. Characteristics of hospice care discharges and their length of service: United States, 2000. National Center for Health Statistics. Vital Health Statistic 2003; 13(154): 136. 164. Bruera E. Improving the care of terminally ill cancer patients and their families. J Clin Oncol 2005; 23(7378): 7379. 165. Goodman A. Giving formal diagnosis of terminal found to increase probability of home death. "Oncology Times" 2006; 25: 24-5. 166. Richards MA, Westcombe AM, Love SB, Littlejohns P, Ramirez AJ. Influence of delay on survival in patients with breast cancer: a systematic review. Lancet 1999; 353(9159):111926. 167. Bernabei R, Gambassi G, Lapane K, Landi F, Gatsonis C, Dunlop R et al. Management of pain in elderly patients with cancer. SAGE Study Group. Systematic Assessment of Geriatric Drug Use via Epidemiology. JAMA 1998; 279:1877- 82. 168. Nosowska G. A delay they can ill afford: delays in obtaining Attendance Allowance for older, terminally ill cancer patients, and the role of health and social care professionals in reducing them. Health Soc Care Community 2004; 12(4):283-7. 72 14 APPENDICES 14.1. Paper I –III I. Population-based study of death of patients with cancer: implications for GPs. Br J Gen Pract 2005; 55: 684-9. II. Defining cancer patients as being in the terminal phase: who receives a formal diagnosis and what are the effects? J Clin Oncol 2005; 23: 7411-6. III. Does persistent involvement by the GP improve palliative care at home for end-stage cancer patients? Palliative Medicine 2006; 20: 507-12. 73