Health and Social Care in the Community (2015) 23(6), 605–618

doi: 10.1111/hsc.12170

Societal costs of home and hospital end-of-life care for palliative care patients in Ontario, Canada Mo Yu

MSc,

Denise N. Guerriere

PhD

and Peter C. Coyte

PhD

Institute of Health Policy, Management and Evaluation, University of Toronto, Toronto, Ontario, Canada Accepted for publication 15 October 2014

Correspondence Peter C. Coyte Institute of Health Policy, Management and Evaluation University of Toronto 155 College Street Toronto, Ontario, Canada M5T 3M6 E-mail: [email protected]

What is known about this topic

• •

Home-based palliative care improves patients’ quality of life and satisfaction at the end-of-life. Home death is often associated with lower financial burden (or costs) for the public sector.

What this paper adds

•

• •

Although patients in home-based palliative care had lower hospitalisation costs, they incurred much higher unpaid caregiver time costs and home care service costs than hospital-based patients. From a societal perspective, homebased palliative care may be just as expensive as hospital-based palliative care. Families and caregivers were shouldering more than half of the financial burden in home-based palliative care.

Abstract In Canada, health system restructuring has led to a greater focus on home-based palliative care as an alternative to institutionalised palliative care. However, little is known about the effect of this change on end-oflife care costs and the extent to which the financial burden of care has shifted from the acute care public sector to families. The purpose of this study was to assess the societal costs of end-of-life care associated with two places of death (hospital and home) using a prospective cohort design in a home-based palliative care programme. Societal cost includes all costs incurred during the course of palliative care irrespective of payer (e.g. health system, out-of-pocket, informal care-giving costs, etc.). Primary caregivers of terminal cancer patients were recruited from the Temmy Latner Centre for Palliative Care in Toronto, Canada. Demographic, service utilisation, care-giving time, health and functional status, and death data were collected by telephone interviews with primary caregivers over the course of patients’ palliative trajectory. Logistic regression was conducted to model an individual’s propensity for home death. Total societal costs of end-of-life care and component costs were compared between home and hospital death using propensity score stratification. Costs were presented in 2012 Canadian dollars ($1.00CDN = $1.00USD). The estimated total societal cost of end-of-life care was $34,197.73 per patient over the entire palliative trajectory (4 months on average). Results showed no significant difference (P > 0.05) in total societal costs between home and hospital death patients. Higher hospitalisation costs for hospital death patients were replaced by higher unpaid caregiver time and outpatient service costs for home death patients. Thus, from a societal cost perspective, alternative sites of death, while not associated with a significant change in total societal cost of end-of-life care, resulted in changes in the distribution of costs borne by different stakeholders. Keywords: end-of-life care, home care, palliative care, place of death, societal cost

Introduction In Canada, health system restructuring has changed palliative care provision and resulted in a greater focus on home-based palliative care as an alternative to institutionalised palliative care. Programmes were established to help patients stay in the comfort of their own homes during the final stage of life, as it improves patients’ quality of life, increases satisfac© 2014 John Wiley & Sons Ltd

tion and alleviates burden on the healthcare system (Hughes et al. 2000, Jordhøy et al. 2000, Romanow 2002, Carlsson & Rollison 2003, Teno et al. 2004, Peters & Sellick 2006, Brumley et al. 2007). However, home-based palliative care may place enormous burden on families, as caring for terminally ill patients is extremely demanding physically, mentally and financially (Grunfeld et al. 2004, Grov et al. 2005, Dumont et al. 2010, Guerriere et al. 2010, Williams et al. 2014). 605

M. Yu et al.

Currently, economic evidence for shifting palliative care away from acute care institutions and into the home is lacking in the literature. Little is known about the effect of this change on end-of-life care costs and the extent to which the financial burden of care has shifted from the acute care public sector to families. While existing studies have reported that home-based palliative care may be less expensive than institutionalised palliative care (Greer et al. 1986, Menec et al. 2004, Hollander 2009, Walker et al. 2011), a full conclusion cannot be drawn due to programme heterogeneity across studies and some concerns with both methodological rigour and validity (Garc
ıaP
erez et al. 2009). Furthermore, most of the existing studies only examined public health system costs or hospital-related costs (Greer et al. 1986, Menec et al. 2004, Hollander 2009, Walker et al. 2011). Excluding family-related costs and other cost components in palliative care, studies may underestimate the amount and type of resources consumed. According to a Canadian study, unpaid caregiver time costs constituted up to 70% of total end-of-life care costs within established home-based palliative care programmes (Guerriere et al. 2010). Thus, studies which excluded private and family-borne costs may not offer a comprehensive assessment of the actual cost of end-of-life care. Moreover, patients who receive home care may be intrinsically different from those who are in hospitals. For example, patients with more debilitating illnesses, less social support or less financial resources may be more likely to end up in a hospital, and eventually die in that setting. Thus, these two groups of patients may not be directly comparable, and the costs and outcomes cannot be conclusively attributed to either the site of care or to patient-specific characteristics. This is an example of treatment assignment and self-selection bias, often inherent in clinical studies (Higgins & Green 2011). Previous studies have not addressed this issue and thus may be subject to biases. Due to rapid population ageing (Canadian Institute of Health Information (CIHI) 2011), the number of seniors living with chronic conditions and terminal illnesses in the community continues to increase. Consequently, the financial burden on the healthcare system and the demand for quality end-of-life care will rise substantially. It is thus imperative that we identify cost-effective and sustainable models of palliative care provision, to offer high quality and accessible end-of-life care to those in need. Understanding the comprehensive cost profiles of providing end-oflife care in different settings will inform programme planning and health policy decision-making. 606

The purposes of the current study were: (i) to comprehensively assess the societal costs of end-oflife care associated with two places of death (hospital and home) within the context of a home-based palliative care programme in Toronto, while adjusting for potential treatment assignment and self-selection biases; and (ii) to examine the differences in cost components between patients who died at home and who died in a hospital.

Methods Participants Primary caregivers of terminal cancer patients were recruited from the Temmy Latner Centre for Palliative Care (TLCPC) at Mount Sinai Hospital in Toronto, Canada. The TLCPC offers community and teambased multidisciplinary palliative care to patients at home 24 hours a day, 7 days a week. TLCPC provides quality home care as an alternative to hospital care at the end-of-life. However, acute hospital care services are available when needed. TLCPC is a part of usual care and the services offered are covered by the provincial health insurance plan. Enrolment into the programme requires physician referral. The patient population in TLCPC is representative of the Toronto population. All new admissions into TLCPC were reviewed by a TLCPC programme clerk to determine eligibility for the study. Inclusion criteria included: (i) primary caregivers of patients diagnosed with a malignant neoplasm (terminal cancer) who had been admitted to the programme within the last 7 days; (ii) caregivers 18 years of age or older; and (iii) caregivers fluent in English. There were no explicit exclusion criteria. Eligible caregivers were approached by a programme clerk and asked if they would be interested in participating in the study. Names and contact information of caregivers who expressed interest were forwarded to the researcher, who then contacted the caregivers to provide additional study information and facilitate the consent and data collection process. Written consent was mailed to caregivers, who would sign and mail it back to the researcher. A total of 986 charts were reviewed, 528 were determined to be eligible and 215 participated in the study. Of those, 29 patients who did not have either ‘home’ or ‘hospital’ recorded as the place of death were excluded as these were the only places of death considered in this study. Consequently, the analysis was based on 186 participants. Patients were followed prospectively from programme enrolment until death and primary © 2014 John Wiley & Sons Ltd

Societal costs of home and hospital end-of-life care

caregivers were interviewed by telephone every 2 weeks over the study period. During interviews, caregivers were asked to provide information on patients’ and caregivers’ demographics, socioeconomic status, health service utilisation (both inside and outside the home), travel, provision of unpaid care-giving time, perceived caregiver burden and patients’ functional status. Data collection took place from August 2010 to September 2012. This study was approved by University of Toronto and Mount Sinai Hospital Research Ethics Boards. Measures Patient and caregiver demographic information was collected during the first interview or by extraction from TLCPC databases and included age, sex, employment status, education, marital status, postal code, patient living arrangement (living alone or living with someone else), caregivers’ relationship with patient and patients’ preferred place of death (at the time of programme enrolment). Date and location of death were also recorded in the TLCPC database. The Ambulatory Home Care Record (AHCR) was used to collect information on health service utilisation during each interview. The AHCR is a tool designed to measure end-of-life resource utilisation and costs from a societal perspective (Guerriere & Coyte 2011). It has been validated in the Canadian palliative care setting (Guerriere et al. 2006) and used extensively in the literature (Stevens et al. 2006, Guerriere et al. 2008, 2010, 2014, McGillion et al. 2008). The AHCR captures both public and private resource consumption. Public consumption includes all health services or products provided inside and outside the home that are paid by the government, such as doctor appointments, tests, hospitalisations and emergency room (ER) visits. Private consumption includes any health services or related costs paid for by the family out-of-pocket or by third party insurance, such as medications, travel expenses, personal support and respite services. The AHCR also captures the unpaid time devoted to care-giving by the caregivers. Prior assessment of the psychometric properties of the AHCR found moderate to high level of agreement between participants’ responses and administrative data (kappa = 0.41–1.00) (Guerriere et al. 2006). The Carstairs deprivation score measures regional deprivation and was used as a proxy for socioeconomic status (Carstairs & Morris 1991). It was calculated using Statistics Canada Census files linked to patients’ postal codes and based on each region’s stan© 2014 John Wiley & Sons Ltd

dardised measures of (i) percentage of men as blue-collar workers; (ii) unemployment rate for men; and (iii) percentage of households below the low-income cutoff. A higher deprivation score suggests that a region is more deprived (lower socioeconomic status). The Caregiver Burden Scale in End-of-Life Care is an 18-item scale used to measure self-perceived burden of care in caregivers of terminally ill patients during each interview (Dumont et al. 2008). The questionnaire was validated and has been commonly used in the Canadian context (Dumont et al. 2008). The Palliative Performance Scale (PPS) is a selfreported measure of patients’ functional status and was administered during each interview. It is a scale from 0 to 100, with incremental changes of 10. A score of 100 equals perfect health and 0 equals death (Anderson et al. 1996). Prior assessment of the psychometric properties of the PPS demonstrated intraclass correlation coefficients ranging from 0.93 to 0.96 (Downing et al. 2007, Ho et al. 2008). The Charlson comorbidity score (Charlson et al. 1987, 2008) was used to measure the level of comorbidity and was calculated based on comorbid conditions documented in patients’ medical records using ICD-9 codes. Data on comorbid conditions and diagnoses were extracted by chart review of patients’ medical records. Estimation of cost The costing perspective in this study was societal so that all costs irrespective of payer were included. Costs were aggregated over three different timeframes of analysis (6, 2 and 1 month prior to death) and presented in 2012 Canadian dollars ($1.00CDN = $1.00USD) (Bank of Canada 2012). Costs were calculated as six separate cost components, based on methods used in previous work with the AHCR (Guerriere et al. 2008, 2010, Guerriere & Coyte 2011):

•

• • •

Outpatient service costs: costs paid by the provincial government for all health and social care services delivered in the home or on an outpatient basis, such as physician or nurse home visits, personal support workers, tests, government’s share of medication costs, home palliative care services, etc. Out-of-pocket costs: costs paid by the patient/family for travel, supplies, medications, etc. Third party costs: costs paid by private insurance for additional services not covered by the provincial health insurance plan, such as private nursing care. Unpaid caregiver time costs: costs of time loss from employment, leisure and household work due to care-giving responsibilities. 607

M. Yu et al.

• •

Hospitalisation costs: costs paid for by the provincial government for all inpatient services, medications and/or supplies used in the hospital during patients’ admission to the hospital. Emergency room visit costs: costs paid for by the provincial government for treating patients who present to the ER.

All resource utilisation (including publicly and privately funded services as well as unpaid care-giving hours) was recorded using the AHCR during each interview. This served as the basis of cost estimation in this study. Outpatient service costs, out-of-pocket costs and third party costs were calculated as the product of utilisation (recorded via the AHCR) multiplied by a unit price. Unit prices are listed in the Ontario Health Insurance Plan schedule of benefits and fees for physician and laboratory services, Ontario Drug Benefit fee schedule, and the fee schedule for Community Care Access Centre-funded services (home-based public services). Costs were adjusted to 2012 values using the Consumer Price Index (CPI). Detailed valuation methods have been described in previous studies by the research team (Guerriere & Coyte 2011). Unpaid caregiver time was considered in three components (time lost from employment, leisure and household work) and each was valued separately. If the caregiver was employed, the time lost from employment was assigned a monetary value using the Human Capital Approach (Rice & MacKenzie 1989, Torgerson et al. 1994). This approach applies current average wages by age and sex categories to lost labour market time. Caregiver time lost from leisure and household work was valued as a replacement cost using the hourly wage of a homemaker. Total unpaid caregiver time costs were determined by adding time lost from employment and leisure/ household work. If the caregiver indicated that they provided 24-hour care, 8 hours were deducted per day from the total number of hours to account for sleep time. Hospitalisation costs were estimated by the average cost of an episode of hospitalisation in Ontario weighted by a patient-specific adjustment factor (the resource intensity weight, RIW; Canadian Institute of Health Information (CIHI) 2013). The RIW is a weighting factor that adjusts inpatient costs by a patient-specific level of resource use associated with their underlying diagnosis and procedures performed. The RIW was computed based on the patient’s primary cancer diagnosis, age and comorbid conditions, using CIHI’s Case Mix Group methodology (Canadian Institute of Health Information (CIHI) 608

2013). The Case Mix Group is a grouping methodology that aggregates patients by the similarity of their underlying diagnoses and by their pattern of inpatient health service utilisation. The cost of each individual patient’s episode of hospitalisation in our study was estimated by multiplying the average cost of hospitalisation in the Ontario by the patient-specific RIW. This is a way of obtaining relatively accurate estimates of patient-level costs in the absence of actual cost data. The average hospitalisation cost in Ontario was $5558.35 (adjusted to 2012 values using the CPI; Wodchis et al. 2011). Emergency room visits were valued at $188.44 per visit (adjusted to 2012 values using the CPI; Wodchis et al. 2011). Analysis timeframes Previous studies have suggested that there may be temporal trends in palliative patients’ resource use. Hollander (2009) and Jung et al. (2012) found that health service utilisation and costs of end-of-life care tended to increase as patients approached death. Thus, to take into consideration temporal effects in end-of-life resource use, the analysis was conducted using three reference timeframes: 6, 2 and 1 month prior to death. Each analysis timeframe included only the data which were collected within that timeframe; conversely, data which fell outside that timeframe would be excluded from that analysis. In other words, the 1-month analysis only looked at data which were collected during the final month of life, while the 6-month analysis looked at data collected during the period up to 6 months before death. The purpose of using different analysis timeframes was to see whether the results would change if we used only the final month’s data, or the last 2 months, or up to 6 months of data. Time prior to death was calculated from the date of the interview to the date of death. Analysis: adjustment for treatment assignment and self-selection bias In a non-randomised environment, the outcome (place of death) may be biased by differences in characteristics that influence the caregivers’, patients’ or physicians’ decision about admission to hospital or home care. Thus, home death patients and hospital death patients may be significantly different from the onset and not directly comparable. Certain patient, caregiver or provider characteristics may predispose the patients to be more likely to have a hospital admission or to be cared for at home (e.g. living alone, functional status). Thus, two types of biases © 2014 John Wiley & Sons Ltd

Societal costs of home and hospital end-of-life care

may be present in this study: (i) treatment assignment bias in which caregivers or physicians may assign the patient to home or hospital care due to specific patient, family and provider characteristics and (ii) the patient may self-select into home or hospital care due to personal characteristics. To adjust for potential treatment assignment and self-selection biases when comparing the costs of care between patients who died at home and who died in a hospital, the propensity score stratification method was used (Rosenbaum & Rubin 2007). This form of stratification is a method of defining groups of comparable patients in a non-randomised environment that allows for valid comparisons within a stratum of similar patients. The propensity score or individual patient’s likelihood of home death, conditional on observed covariates, was calculated using a logistic regression model. The covariates used were as follows: number of days patients spent at home, timeframe-specific baseline caregiver burden, timeframe-specific baseline PPS score, caregiver age, caregiver sex, patient comorbidity score, patient age, patient sex, preferred place of death, patient living arrangement and the Carstairs deprivation score. These covariates were selected based on the following criteria: previous findings by the study team (Masucci et al. 2010, Guerriere et al. 2014); univariate analysis results; correlations among the variables and other findings from the predictors of place of death literature (Cantwell et al. 2000, Jordhøy et al. 2000, Gallo et al. 2001, Brazil et al. 2002, Burge et al. 2003, Carlsson & Rollison 2003, Gyllenhammar et al. 2003, Fukui et al. 2004, Gomes & Higginson 2006, Aoun et al. 2007, Brink & Smith 2008). The propensity score summarised the data in our model into a single number: the patient’s predicted probability of dying at home conditional on observed covariates. Based on the estimated propensity scores, patients were stratified into quartiles. Stratification based on the propensity score created strata of similar patients in which the home and the hospital death groups were balanced in the observed covariates included in the model. It effectively eliminated the majority of the bias due to measured confounders, created groups of comparable patients and led to more reliable estimates of the treatment effect (Rosenbaum & Rubin 2007). The societal costs of end-of-life care were compared between patients who died at home and who died in hospital, globally and also within propensity score quartiles. P-values were computed using t-tests to determine statistical significance (significance level = 0.05). Variations in cost components were also examined. The comparison was performed for three © 2014 John Wiley & Sons Ltd

analysis timeframes (6, 2 and 1 month prior to death), using STATA version 12.

Results Demographic characteristics Table 1 reports the demographic information for the sample. The average age of the patients and their caregivers at the time of admission to the palliative care programme was 73 and 58 years respectively. More than half (54.84) of the patients were female; 65.05% of caregivers were female. In 45.16% of the cases, the caregiver was the child of the patient and 44.09% were the patient’s spouse. The majority of the caregivers were married (75.81%). Approximately, 41.94% of caregivers were employed and 31.35% were retired. Most of the patients lived with someone (81.72%), while 18.28% lived alone. Hospitalisation, unpaid care-giving and place of death Of the entire sample, 56.45% patients (n = 105) died at home and 43.55% died in the hospital (n = 81). When we compared the actual place of death with patients’ preferred place of death, we found that the majority of patients (66.7%) died in their preferred place of death. Prior to death, the average patient in the cohort spent about 4.7 days in the hospital during the palliative trajectory, which on average was 4 months. Home death patients spent on average 2.3 days in the hospital, while hospital death patients spent 7.9 days in the hospital. The average caregiver spent about 5829 hours for unpaid care-giving over the palliative trajectory. Caregivers of home death patients spent more than twice as much time for unpaid caregiving as caregivers of hospital death patients, 7760 and 3326 hours respectively. Patient characteristics by place of death Table 2 reports patient characteristics by location of death. We found that home and hospital death patients were significantly different from each other for several variables. Home death patients had significantly (P < 0.05) lower baseline caregiver burden, higher baseline PPS and higher deprivation score than patients who died in hospitals. Furthermore, patients who lived with others and who had a preference for home death were also significantly (P < 0.05) more likely to die at home. This result is consistent with 609

M. Yu et al.

Table 1 Demographic characteristics of study sample (n = 186)

Patient characteristics Place of death Home Hospital Age 27–69 70–96 Sex Male Marital status Married Education High school or less Post-secondary or higher Living arrangement Lives alone Lives with others Baseline PPS (n = 185) 13–30 (low functioning) 31–50 (moderate functioning) Baseline comorbidity score 6–7 8–9 10–12 Baseline deprivation score Quartile 1 (≤0.38) Quartile 2 (0.43–0.63) Quartile 3 (0.64–0.99) Quartile 4 (1.00–2.07) Preferred place of death Home Hospital Hospice Missing Caregiver characteristics Age ≤45 46–69 ≥70 Sex Male Marital status Married Education High school or less Post-secondary or higher Employment Unemployed/retired/on leave Part-time employed Full-time employed Relationship with patient Caregiver is spouse Caregiver is others Baseline caregiver burden (n = 174) 16–30 31–59

610

Number

%

105 81

56.45 43.55

71 115

38.17 61.83

84

45.16

112

60.22

85 101

45.70 54.30

34 152

18.28 81.72

84 101

45.41 54.59

147 35 4

79.03 18.82 2.15

49 47 45 45

26.34 25.27 24.19 24.19

130 40 1 15

69.89 21.51 0.54 8.06

32 122 32

17.20 65.59 17.20

65

34.95

141

75.81

40 115

21.51 61.83

97 18 60

52.15 9.68 32.26

82 104

44.09 55.91

124 50

71.26 28.74

The caregiver burden scale is a score from 16 to 64, 16 being lowest burden and 64 being the highest (Dumont et al. 2008). The palliative performance score (PPS) is scaled from 0 to 100, where 0 means death and 100 means perfect health (Anderson et al. 1996). The Carstairs deprivation score was used to measure the level of deprivation and was used as a proxy for socioeconomic status (Carstairs & Morris 1991). A higher deprivation score suggests that a region is more deprived (lower socioeconomic status). The Charlson comorbidity score (Charlson et al. 1987, 2008) was used to measure comorbidity. Minimum score is 0 and maximum score is 39.

the larger place of death study by the research team (Guerriere et al. 2014). Other patient characteristics were not significantly different between home and hospital death patient groups. This evidence provided further justification for the use of the propensity score stratification method to correct for potential assignment and self-selection biases. Overall costs of care The average total societal cost of end-of-life care for all patients (both home and hospital death) was $34,197.73 over the entire palliative trajectory (4 months on average). Public health system costs comprised 46.72% of total costs; private costs comprised 53.28%. Table 3 provides a detailed breakdown of the total cost by public and private sources. Looking at costs of end-of-life care by location of death, the average total cost was $37,699.37 for patients who died at home and $29,658.57 for patients who died in hospitals (a difference of $8040.80). Logistic regression results Table 4 shows the logistic regression results for three analysis timeframes. In general, higher baseline PPS, living with others and a preference for home death led to a significant increase in the probability of home death (P < 0.05). Higher baseline caregiver burden led to a lower probability of a home death (P < 0.05). The logistic regression results reported in Table 4 yield the determinants of the probability of a home death against the alternative of a hospital death for each of the analysis timeframes. The regression coefficients tell us the effect of each covariate on the probability of the binary outcome. The propensity score is the predicted probability of the dependent variable (the probability of a home death) conditional on values of the independent variables used in the model we specified.

© 2014 John Wiley & Sons Ltd

Societal costs of home and hospital end-of-life care

Table 2 Patient characteristics by place of death at baseline Home death mean Continuous variables Days home Baseline burden Baseline PPS Patient age Caregiver age Comorbidity score Deprivation score

Binary variables Patient sex Male Female Caregiver sex Male Female Preferred POD Home Other Living arrangement Alone With others

73.35 25.98 32.57 73.15 58.54 6.81 0.76 Home death (n)

Hospital death mean

60.65 29.50 28.69 72.38 56.59 6.86 0.62 Hospital death (n)

P-value for between-group difference

0.285 0.003 0.001 0.677 0.304 0.730 0.020

42 63

42 39

0.107

41 64

24 57

0.182

92 13

38 43

0.000

13 92

21 60

0.018

PPS, Palliative Performance Scale; POD, place of death.

A propensity score for each patient was calculated from the logistic regression model. Patients were stratified into quartiles based on their propensity scores. Costs of end-of-life care were compared between home and hospital death patients, both globally and within propensity score quartiles. Cost comparisons pre-stratification (unadjusted) Table 5 shows the total cost of end-of-life care and cost by component for patients who died at home and who died in hospitals using different analysis timeframes (6, 2 and 1 month prior to death). Figure 1 provides a graphical representation of the costs and magnitude of each cost component relative to the total cost. While total societal cost of end-of-life care for home death patients was higher than that for hospital death patients (except for 1 month prior to death), the difference was not statistically significant (P > 0.05) for any timeframe. Hospital death patients had significantly (P < 0.05) higher hospitalisation costs and higher emergency department costs, which was expected as they tended to remain in the hospital for a longer period of time prior to death. On the other hand, home death patients had significantly (P < 0.05) higher outpatient service costs and unpaid © 2014 John Wiley & Sons Ltd

Table 3 Distribution of total costs over the entire palliative trajectory across cost components

Publicly financed costs Outpatient service costs Public – Medications Public – Supplies Public – tests Public – appointments inside the home Public – appointments outside the home Public – hospice Hospitalisation costs Hospitalisation ER visit costs ER visits Total public costs Privately financed costs Out-of-pocket costs Out-of-pocket – medications Out-of-pocket – supplies Out-of-pocket – appointments inside the home Out-of-pocket – travel costs Unpaid caregiver time costs Time lost from household work Time lost from leisure Time lost from employment Third party costs Third party – medications Third party – supplies Third party – appointments inside the home Third party – hospice Total private costs Total cost

$

% of total cost

$619.87 $4451.31 $418.79 $5491.93

1.81 13.02 1.22 16.06

$327.49

0.96

$2.64

0.01

$4558.61

13.33

$105.36 $15,976.00

0.31 46.72

$103.81 $207.17 $1429.82

0.30 0.61 4.18

$277.71

0.81

$2085.17 $13,728.42 $84.06 $241.79 $7.34 $54.1 $2.34 $18,221.73 $34,197.73

6.10 40.14 0.25 0.71 0.02 0.16 0.01 53.28

caregiver time costs, as a significant amount of homecare services and unpaid caregiver time were required if the patient was being cared for at home. These findings indicate that although patients who died in the home had lower acute care costs (hospitalisation and ER cost) than patients who died in hospitals, they had much higher costs at home in terms of unpaid caregiver time and outpatient services. Thus, the overall cost was not significantly different between home and hospital death patients. Out-ofpocket costs and third party costs were not found to be significantly different between home and hospital death patients for any timeframe. Cost comparisons post-stratification (adjusted by propensity score) Table 6 reports the total cost of end-of-life care and cost by component for patients who died at home 611

M. Yu et al.

Table 4 Multivariate logistic regression results for 6, 2 and 1 month prior to death (dependent variable = place of death; 1 = home, 0 = hospital) Analysis timeframes Six months prior to death (N = 178); R2 = 0.2418 Β (SE) Number of days spent at home Baseline caregiver burden Baseline PPS Caregiver age Caregiver sex (male = 1) Charlson comorbidity score Patient age Patient sex (male = 1) Preferred place of death (home = 1) Living arrangement (Lives with others = 1) Carstairs deprivation score Constant

0.00 0.07 0.07 0.00 0.04 0.12 0.00 0.25 1.92 1.08 0.66 1.77

P (0.00) (0.03) (0.03) (0.02) (0.42) (0.19) (0.02) (0.42) (0.43) (0.51) (0.45) (1.95)

0.534 0.004 0.011 0.779 0.932 0.521 0.838 0.550 0.000 0.035 0.142 0.365

Two months prior to death (N = 163); R2 = 0.2635

One month prior to death (N = 138); R2 = 0.3050

Β (SE)

Β (SE)

0.02 0.06 0.09 0.01 0.24 0.01 0.01 0.08 2.07 1.33 0.51 3.69

P (0.01) (0.02) (0.03) (0.02) (0.50) (0.19) (0.02) (0.49) (0.47) (0.57) (0.47) (2.10)

0.136 0.013 0.002 0.634 0.639 0.967 0.766 0.864 0.000 0.020 0.270 0.078

0.11 0.04 0.08 0.03 0.60 0.08 0.01 0.28 2.15 1.55 0.08 4.11

P (0.04) (0.02) (0.05) (0.02) (0.58) (0.21) (0.02) (0.61) (0.55) (0.62) (0.48) (2.51)

0.010 0.044 0.082 0.148 0.304 0.717 0.677 0.649 0.000 0.013 0.866 0.101

B is the logistic regression coefficient and (SE) is the robust standard error in the logistic regression analysis. PPS, palliative performance score.

and who died in hospitals, stratified by propensity score quartiles, for three analysis timeframes (6, 2 and 1 month prior to death). After propensity score stratification, the findings were essentially the same as the pre-stratification results. Total societal cost of end-of-life care for home and hospital death patients was not significantly different (P > 0.05) in any propensity score quartiles. In most cases, similar to the pre-stratification results, home death patients had higher outpatient service costs and unpaid caregiver time costs, while hospital death patients had higher hospitalisation costs and ER costs (but the difference was not significant in most cases). Stratum-specific treatment effects were also pooled and weighted by propensity scores to estimate an average treatment effect (ATE) for home versus hospital death. Table 7 reports the ATEs estimated for each analysis timeframe. For all three, the average treatment effect was not significant (P > 0.05). This is consistent with the un-pooled results showing that different places of death did not result in a significant change in total cost of end-of-life care.

Discussion Summary of findings The estimated average societal cost of end-of-life care for all patients in the cohort was $34,197.73 (over 612

entire palliative trajectory). Unpaid caregiver time costs accounted for 46% ($15,897.65) of total costs; outpatient/home service costs 33% ($11,312.03) and hospitalisations 13% ($4558.61). Looking at different places of death, we found that the total societal cost was $37,699.37 for patients who died at home and $29,658.57 for patients who died in hospitals. Two other studies have estimated the societal costs of end-of-life care in the Canadian context. Guerriere et al. (2010) found that for a home-based palliative care programme, mean total monthly costs were $24,549 per patient; 71% ($17,465) was attributed to unpaid caregiver time costs and 2.6% ($630) to hospitalisations. Dumont et al. (2010) examined societal costs of end-of-life care in community and home-based programmes and found that during the final month of life, the mean total cost was $7765 per patient; 58.8% ($4565) was for inpatient care; 29.4% ($2285) for outpatient care and 11.8% ($915) for unpaid care-giving. The differences in cost estimates between our study and the previous studies are likely due to different proximities to death examined and different methods used to estimate end-of-life costs. For example, we used the Case Mix Group methodology to estimate inpatient cost weighted by each patient’s specific level of resource use. This has not been done in previous studies. Furthermore, this study has also gone to great lengths to capture all costs associated with end-oflife care. The costing exercise in our study may be © 2014 John Wiley & Sons Ltd

Societal costs of home and hospital end-of-life care

Table 5 Pre-stratification results: total costs of end-of-life care and costs by component for home and hospital death patients (6, 2 and 1 month prior to death) Six months prior to death (n = 178)

Costs

Home death mean

% of total cost

Hospital death mean

% of total cost

P-value for betweengroup difference

Pooled mean cost

% of total cost

Total cost Outpatient services cost Hospitalisation cost ER visits cost Out-of-pocket cost Unpaid caregiver time cost Third party cost

$31,910.54 $12,103.93 $2111.84 $59.12 $1998.19 $15,324.88 $312.59

37.93 6.62 0.19 6.26 48.02 0.98

$29,116.62 $7837.75 $7733.86 $156.20 $1800.42 $11,416.84 $171.55

26.92 26.56 0.54 6.18 39.21 0.59

0.467 0.015 0.000 0.002 0.727 0.050 0.258

$30,717.63 $10,282.41 $4512.25 $100.57 $1913.75 $13,656.28 $252.37

33.47 14.69 0.33 6.23 44.46 0.82

22.33 37.26 0.59 6.03 33.36 0.44

0.525 0.000 0.000 0.001 0.584 0.001 0.307

$20,351.21 $6568.00 $3777.38 $63.58 $1297.16 $8515.40 $129.69

32.27 18.56 0.31 6.37 41.84 0.64

17.27 46.85 0.46 6.00 28.90 0.53

0.457 0.000 0.000 0.000 0.837 0.001 0.998

$12,828.57 $4037.89 $2772.07 $31.41 $838.98 $5077.33 $70.89

31.48 21.61 0.24 6.54 39.58 0.55

Two months prior to death (n = 163) Total cost Outpatient services cost Hospitalisation cost ER visits cost Out-of-pocket cost Unpaid caregiver time cost Third party cost

$20,867.13 $8025.84 $1445.72 $28.84 $1374.64 $9832.97 $159.11

38.46 6.93 0.14 6.59 47.12 0.76

$19,573.38 $4370.02 $7292.80 $115.96 $1180.36 $6528.89 $85.34

One month prior to death (n = 138) Total cost Outpatient services cost Hospitalisation cost ER visits cost Out-of-pocket cost Unpaid caregiver time cost Third party cost

$12,483.59 $4981.37 $827.80 $14.82 $856.61 $5732.04 $70.94

39.90 6.63 0.12 6.86 45.92 0.57

$13,455.17 $2324.23 $6303.52 $61.53 $806.96 $3888.15 $70.79

Societal cost of end-of-life care Outpatient services

Out-of-pocket

Third party

Caregiver time

Hospitalisation

ER visits

40,000 $37,699.67 35,000 $31,910.54 $29,658.57 $29,116.62

30,000

25,000 $20,867.13 $19,573.38

20,000

15,000 $12,483.59

$13,455.17

10,000

5000

0 Over all home

Over all hospital

6 months home

6 months hospital

2 months home

2 months hospital

1 month home

1 month hospital

Figure 1 Total cost of end-of-life broken down by cost component for each analysis timeframe.

much more comprehensive and detailed than previous studies, thus resulting in our estimate of some cost components being higher than other studies. © 2014 John Wiley & Sons Ltd

Using propensity score stratification methods, this study also demonstrated that there were no significant differences in the societal costs of end-of-life care 613

M. Yu et al.

Table 6 Post-stratification results: total costs of end-of-life care and costs by component for home and hospital death patients (6, 2 and 1 month prior to death) stratified by propensity score quartile Six months prior to death (n = 178)

Variable Quartile 1 Total cost Outpatient services cost Hospitalisation cost ER visits cost Out-of-pocket cost Unpaid caregiver time cost Third party cost Quartile 2 Total cost Outpatient services cost Hospitalisation cost ER visits cost Out-of-pocket cost Unpaid caregiver time cost Third party cost Quartile 3 Total cost Outpatient services cost Hospitalisation cost ER visits cost Out-of-pocket cost Unpaid caregiver time cost Third party cost Quartile 4 Total cost Outpatient services cost Hospitalisation cost ER visits cost Out-of-pocket cost Unpaid caregiver time cost Third party cost

614

Two months prior to death (n = 163)

One month prior to death (n = 138)

Hospital death mean

P-value for betweengroup difference

Home death mean

Hospital death mean

P-value for betweengroup difference

Home death mean

Hospital death mean

P-value for betweengroup difference

$46,760.71 $11,962.25

$31,215.06 $7142.65

0.124 0.121

$12,983.39 $3273.29

$22,672.48 $3762.11

0.169 0.747

$7897.28 $0.21

$13,024.79 $1937.59

0.210 0.930

$743.62

$10,623.85

0.123

$2378.66

$11,124.81

0.110

$1015.74

$6865.28

0.067

$94.22 $3008.53

$204.14 $2143.07

0.404 0.593

$53.84 $1287.79

$170.20 $1118.59

0.323 0.824

$0.00 $607.64

$65.23 $486.39

0.095 0.768

$29,566.90

$10,923.28

0.001

$5895.24

$6378.04

0.811

$4171.46

$3570.92

0.591

$1385.18

$178.08

0.002

$94.57

$118.74

0.885

$56.83

$99.38

0.823

$26,978.36 $8939.72

$25,256.43 $6911.28

0.813 0.424

$19,383.52 $6450.05

$14,906.17 $4221.54

0.216 0.188

$12,742.94 $4636.03

$12,531.11 $2072.20

0.939 0.050

$3399.35

$5983.04

0.369

$2424.71

$3355.25

0.600

$2025.46

$4980.73

0.152

$78.52 $1877.83

$83.75 $1399.51

0.928 0.671

$30.15 $1161.84

$101.47 $1495.28

0.222 0.625

$28.27 $918.74

$47.11 $1557.94

0.499 0.208

$12,353.57

$10,655.04

0.623

$9134.17

$5729.48

0.077

$5132.85

$3870.07

0.146

$329.38

$223.81

0.731

$182.60

$3.14

0.162

$1.60

$3.07

0.607

$23,095.35 $7439.76

$30,944.34 $12,059.37

0.282 0.061

$17,559.85 $5994.77

$17,060.22 $5231.04

0.890 0.595

$9927.65 $3884.62

$14,341.18 $7063.08

0.216 0.071

$1890.72

$3068.09

0.328

$1341.24

$3160.21

0.124

$310.79

$3715.05

0.022

$48.63 $1374.76

$154.18 $1175.94

0.098 0.759

$34.90 $1185.32

$51.39 $1394.05

0.561 0.777

$13.00 $782.92

$62.81 $210.07

0.144 0.484

$12,238.74

$14,276.22

0.678

$8976.55

$7164.26

0.420

$4888.85

$3290.17

0.244

$102.74

$210.55

0.401

$27.07

$59.27

0.489

$47.48

$0.00

0.428

$42,224.39 $18,609.80

$36,152.60 $10,432.53

0.649 0.288

$26,825.29 $11,531.55

$17,833.48 $6520.38

0.117 0.197

$16,113.87 $6716.40

$17,648.49 $3291.91

0.675 0.053

$1811.74

$7106.27

0.003

$881.21

$4248.75

0.035

$644.90

$7427.94

0.000

$53.84 $2530.30

$134.60 $2984.76

0.176 0.841

$23.55 $1243.33

$0.00 $234.73

0.516 0.197

$13.46 $675.52

$47.11 $316.17

0.401 0.473

$18,874.97

$15,465.39

0.578

$12,844.56

$6803.91

0.070

$7900.59

$6535.36

0.548

$343.74

$29.06

0.453

$301.09

$25.72

0.445

$163.00

$30.00

0.707

Home death mean

© 2014 John Wiley & Sons Ltd

Societal costs of home and hospital end-of-life care

Table 7 Re-weighting by propensity score: average treatment effect (ATE). Outcome = total cost; treatment = place of death (1 = home; 0 = hospital) Total cost

ATE coefficient

1 month 2 months 6 months

6282.36 1350.22 394.82

SE

P

5088.38 3826.79 7213.75

0.217 0.724 0.956

SE, standard error.

between home and hospital deaths for palliative care recipients. The higher cost of hospitalisations for patients who died in hospital was replaced by higher unpaid caregiver time cost and higher outpatient service cost for patients who died at home. Thus, alternative places of death (home versus hospital) were not associated with a significant change in total societal cost of end-of-life care. This finding was confirmed by both stratum-specific treatment effects and pooled average treatment effect. However, this is contrary to most studies in the literature, which have typically found dying in acute care hospitals to be much more expensive than dying in the home setting (Greer et al. 1986, Menec et al. 2004, Hollander 2009, Walker et al. 2011). The main reason for this discrepancy was that our study adopted a societal cost perspective, while previous studies measured only direct health system cost or hospital cost. By taking a societal perspective, we were able to capture unpaid caregiver time costs, outof-pocket costs borne by families and third party private insurance costs that were not included in other studies. These private costs constituted more than half of the total cost of care. Consequently, the societal cost estimates from our study may be considered as a more comprehensive and accurate reflection of the actual cost of end-of-life care. Lastly, the propensity score adjusted and unadjusted analyses led to the same results. This suggested that while potential treatment assignment and selection biases may occur, in the context of this study when controlling for such potential biases using the observed covariates, we were not able to identify such biases in the data. Limitations This study has the following limitations. First, the sample size was relatively limited. However, small sample studies are especially common in palliative care research, and the model we specified was able to incorporate most of the key variables needed for the analysis. © 2014 John Wiley & Sons Ltd

Second, data were self-reported during telephone interviews and may be susceptible to recall bias. To address this issue, the 2-week interview interval was selected. It was short enough to minimise recall bias, but not so short that it overburdened caregivers. Furthermore, self-reported data may also be subject to social desirability bias (Norman & Streiner 1986). To address this issue, it was made clear to the participants in advance that all data would be kept strictly confidential and that the interviewer was not affiliated with the TLCPC programme. Third, the accuracy of the methods used to derive the cost data may be debated. CIHI’s Case Mix Group methodology and the Human Capital Approach, in the absence of available cost data, were used to derive hospitalisation costs and to assign monetary value for time lost from employment. Although there are disadvantages and assumptions associated with both methods, they have been validated and used widely in the literature (Rice & MacKenzie 1989, Pink & Bolley 1994, Torgerson et al. 1994). Fourth, the TLCPC programme is based in a large tertiary hospital in a metropolitan centre. Although the patient population in the TLCPC programme is representative of the Toronto population in terms of socio-demographic distribution, the generalisability of the results to other regions (especially rural and remote areas) needs to be further investigated. Furthermore, the cohort was exclusively terminal cancer patients due to the majority (85%) of TLCPC clients being cancer patients. Resource use and care demands for patients with other illnesses may differ. Fifth, as patients approached death, it often became difficult for caregivers to participate in interviews and loss to follow-up increased with proximity to death. A total of 138 caregivers were interviewed during the final month of life compared to 163 during the last 2 months of life. We have tried to minimise the impact of attrition by conducting the analysis using three different timeframes in the palliative care trajectory. The results of the 1-month analysis were the same as the results of the 2-month and 6-month analysis. Although this does not solve the attrition bias, it does build confidence in our findings despite attrition. Finally, evidence has shown that home death improves psychological outcomes for caregivers, as emergency admissions and aggressive treatments are largely avoided in home death (Wright et al. 2008, Detering et al. 2010). In this study, we also found self-reported caregiver burden to be significantly (P < 0.05) lower for the home death group than the hospital death group, when measured at baseline and during the last interview before death. This is an 615

M. Yu et al.

important consideration that should be given economic value as it will indirectly impact the societal cost of end-of-life care. Unfortunately, the study was not designed to capture caregiver’s psychological outcomes. It would, however, be an important point to consider for future research. Policy implications From a policy perspective, this study demonstrated that home-based palliative care may not be so costsaving if we were to consider the costs of care from a societal perspective. In fact, dying at home may be just as expensive as dying in the hospital, due to the high demand for home care services and the high costs incurred by private sources (such as patient families and third party insurance). In particular, caring for a dying loved one at home places an enormous burden on the families and caregivers. In our data, the families and caregivers were shouldering more than half of the total cost of home-based palliative care. Fifty-two per cent (52%) of the total societal cost of end-of-life care was due to unpaid caregiver time and out-of-pocket costs borne by families. Unpaid caregiver time was by far the largest contributor to total end-of-life cost in home-based palliative care (46%; $15,897.65), far exceeding outpatient services (33%; $11,312.03) or hospitalisation costs (13%; $4558.61). Furthermore, caregivers of home death patients spent more than twice as much time for unpaid care-giving as caregivers of hospital death patients, 7760 and 3326 hours respectively. Despite the presence of an array of fiscal policies that yield tax benefits or subsidies to caregivers (e.g. the Employment Insurance benefits for caregivers and the Caregiver tax credit), these subsidies tend to be small and rarely cover the full cost of providing unpaid care as reported in our study. As a result, by encouraging home-based palliative care without providing the necessary support or reimbursement to caregivers at home, we may have shifted the burden of care from the health system to families. This raises important questions both ethically and in terms of system sustainability. This study encourages us to re-think and revisit where this shift from hospital-based to home-based palliative care has been the most beneficial and where there is room for improvement. Undoubtedly, being able to remain in the comfort of one’s home during the final stage of life has many benefits for the patient and the family. However, this also places enormous burden on families and caregivers both psychologically and financially. To ensure the sustainability and viability of home-based palliative care, we need to 616

find better ways to relieve caregivers of the burden placed on them by the high demands of home care and to appropriately compensate them for the work that they do.

Acknowledgements This study was funded by the Canadian Institutes of Health Research (Grant #: PHE-101530). We thank the palliative physicians, nurses and frontline healthcare workers at the TLCPC who have made this programme possible and have been an inspiration to all. We thank Dr Audrey Laporte and Dr Marcus Hollander for their advice on this paper. We also thank Robin McLernon and Brandon Zagorski for the critical role they played in this project.

Conflict of interest disclosure The authors of this paper have no conflict of interests in this study.

References Anderson F., Downing G.M., Hill J., Casorso L. & Lerch N. (1996) Palliative performance scale (PPS): a new tool. Journal of Palliative Care 12 (1), 5–11. Aoun S., Kristjanson L.J., Currow D., Skett K., Oldham L. & Yates P. (2007) Terminally-ill people living alone without a caregiver: an Australian national scoping study of palliative care needs. Palliative Medicine 21 (1), 29–34. Bank of Canada (2012) Financial markets department: year average of exchange rates. Available at: http://www.bankofcanada.ca/stats/assets/pdf/nraa-2012.pdf (accessed on 1/2/2012). Brazil K., Bedard M. & Willison K. (2002) Factors associated with home death for individuals who receive home support services: a retrospective cohort study. BMC Palliative Care 1 (1), 2. Brink P. & Smith T.F. (2008) Determinants of home death in palliative home care: using the interRAI Palliative Care to assess end-of-life care. American Journal of Hospice and Palliative Medicine 25 (4), 263–270. Brumley R., Enguidanos S. & Jamison P. et al. (2007) Increased satisfaction with care and lower costs: results of a randomized trial of in-home palliative care. Journal of the American Geriatrics Society 55 (7), 993–1000. Burge F., Lawson B. & Johnston G. (2003) Trends in the place of death of cancer patients, 1992-1997. Canadian Medical Association Journal 168 (3), 265–270. Canadian Institute of Health Information (CIHI) (2011) Health care in Canada 2011: a focus on seniors and aging. Available at: https://secure.cihi.ca/free_products/HCIC_2011_seniors_report_en.pdf (accessed on 1/2/2012). Canadian Institute of Health Information (CIHI) (2013) Patient cost estimator methodological notes and glossary. Available at: http://www.cihi.ca/CIHI-ext-portal/pdf/internet/PCE_methnotes_en (accessed on 1/2/2012). Cantwell P., Turco S., Brenneis C., Hanson J., Neumann C.M. & Bruera E. (2000) Predictors of home death in palli© 2014 John Wiley & Sons Ltd

Societal costs of home and hospital end-of-life care

ative care cancer patients. Journal of Palliative Care 16 (1), 23–28. Carlsson M.E. & Rollison B. (2003) A comparison of patients dying at home and patients dying at a hospice: sociodemographic factors and caregivers’ experiences. Palliative and Supportive Care 1, 33–39. Carstairs V. & Morris R. (1991) Deprivation and Health in Scotland. Aberdeen University Press, Aberdeen. Charlson M.E., Pompei P., Ales K.L. & MacKenzie C.R. (1987) A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. Journal of Chronic Diseases 40 (5), 373–383. Charlson M.E., Charlson R.E., Peterson J.C., Marinopoulos S.S., Briggs W.M. & Hollenberg J.P. (2008) The Charlson comorbidity index is adapted to predict costs of chronic disease in primary care patients. Journal of Clinical Epidemiology 61 (12), 1234–1240. Detering K.M., Hancock A.D., Reade M.C. & Silvester W. (2010) The impact of advance care planning on end-of-life care in elderly patients: randomised controlled trial. British Medical Journal 340, c1345. Downing M., Lau F. & Lesperance M. et al. (2007) Metaanalysis of survival prediction with palliative performance scale. Journal of Palliative Care 23 (4), 245–252. Dumont S., Fillion L., Gagnon P. & Bernier N. (2008) A new tool to assess family caregivers’ burden during end-of-life care. Journal of Palliative Care 24 (3), 151–161. Dumont S., Jacobs P., Turcotte V., Anderson D. & Harel F. (2010) The trajectory of palliative care costs over the last 5 months of life: a Canadian longitudinal study. Palliative Medicine 24 (6), 630–640. Fukui S., Fukui N. & Kawagoe H. (2004) Predictors of place of death for Japanese patients with advanced-stage malignant disease in home care settings: a nationwide survey. Cancer 101 (2), 421–429. Gallo W.T., Baker M.J. & Bradley E.H. (2001) Factors associated with home versus institutional death among cancer patients in Connecticut. Journal of the American Geriatrics Society 49 (6), 771–777. Garc
ıa-P
erez L., Linertov
a R., Mart
ın-Olivera R., SerranoAguilar P. & Ben
ıtez-Rosario M.A. (2009) A systematic review of specialised palliative care for terminal patients: which model is better? Palliative Medicine 23 (1), 17–22. Gomes B. & Higginson I.J. (2006) Factors influencing death at home in terminally ill patients with cancer: systematic review. British Medical Journal 332 (7540), 515–521. Greer D.S., Mor V., Morris J.N., Sherwood S., Kidder D. & Birnbaum H. (1986) An alternative in terminal care: results of the National Hospice Study. Journal of Chronic Diseases 39 (1), 9–26. Grov E.K., Dahl A.A., Moum T. & Foss a S.D. (2005) Anxiety, depression, and quality of life in caregivers of patients with cancer in late palliative phase. Annals of Oncology 16 (7), 1185–1191. Grunfeld E., Coyle D. & Whelan T. et al. (2004) Family caregiver burden: results of a longitudinal study of breast cancer patients and their principal caregivers. Canadian Medical Association Journal 170 (12), 1795–1801. Guerriere D.N. & Coyte P.C. (2011) The ambulatory and home care record: a methodological framework for economic analyses in end-of-life care. Journal of Aging Research 2011, 374237. Guerriere D.N., Ungar W.J., Corey M., Croxford R., Tranmer J.E., Tullis E. & Coyte P.C. (2006) Evaluation of the © 2014 John Wiley & Sons Ltd

ambulatory and home care record: agreement between self-reports and administrative data. International Journal of Technology Assessment in Health Care 22 (2), 203–210. Guerriere D.N., Wong A.Y., Croxford R., Leong V.W., McKeever P. & Coyte P.C. (2008) Costs and determinants of privately financed home-based health care in Ontario, Canada. Health and Social Care in the Community 16 (2), 126–136. Guerriere D.N., Zagorski B., Fassbender K., Masucci L., Librach L. & Coyte P.C. (2010) Cost variations in ambulatory and home-based palliative care. Palliative Medicine 24 (5), 523–532. Guerriere D., Husain A., Marshall D. et al. (2014) Predictors of Place of Death for those in receipt of home-based palliative care services in Ontario, Canada. Journal of Palliative Care (Submitted). Gyllenhammar E., Thoren-Todoulos E., Strang P., Str€ om G., Eriksson E. & Kinch M. (2003) Predictive factors for home deaths among cancer patients in Swedish palliative home care. Supportive Care in Cancer 11 (9), 560–567. Higgins J.P.T. & Green S. (Eds) (2011) Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March 2011]. The Cochrane Collaboration, 2011. Available from www.cochrane-handbook.org. Ho F., Lau F., Downing M.G. & Lesperance M. (2008) A reliability and validity study of the palliative performance scale. BMC Palliative Care 7, 10. Norman G.R. & Streiner D.L. (2009) Costs of end-of-life care: findings from the province of Saskatchewan. Healthcare Quarterly 12 (3), 50–58. Hughes S.L., Weaver F.M. & Giobbie-Hurder A. et al. (2000) Effectiveness of team-managed home-based primary care. Journal of the American Medical Association 284 (22), 2877–2885. Jordhøy M.S., Fayers P., Saltnes T., Ahlner-Elmqvist M., Jannert M. & Kaasa S. (2000) A palliative-care intervention and death at home: a cluster randomised trial. Lancet 356 (9233), 888–893. Jung H.M., Kim J., Heo D.S. & Baek S.K. (2012) Health economics of a palliative care unit for terminal cancer patients: a retrospective cohort study. Supportive Care in Cancer 20 (1), 29–37. Masucci L., Guerriere D.N., Cheng R. & Coyte P.C. (2010) Determinants of place of death for recipients of homebased palliative care. Journal of Palliative Care 26 (4), 279–286. McGillion M.H., Croxford R., Watt-Watson J., Lefort S., Stevens B. & Coyte P. (2008) Cost of illness for chronic stable angina patients enrolled in a self-management education trial. Canadian Journal of Cardiology 24 (10), 759–764. Menec V., Lix L., Steinbach C., Ekuma O., Sirski M., Dahl M. & Soodeen R. (2004) Patterns of health care use and cost at the end of life. Available at: http://mchp-appserv.cpe. umanitoba.ca/reference/end_of_life.pdf (accessed on 1/ 2/2012). Norman G.R. & Streiner D.L. (1986) PDQ Statistics. BC Decker Inc, Toronto, Canada. Peters L. & Sellick K. (2006) Quality of life of cancer patients receiving inpatient and home-based palliative care. Journal of Advanced Nursing 53 (5), 524–533. Pink G.H. & Bolley H.B. (1994) Physicians in health care management: 3. Case mix groups and resource intensity weights: an overview for physicians. Canadian Medical Association Journal 150 (60), 889–894.

617

M. Yu et al.

Rice P. & MacKenzie E. (1989) Cost of Injury in the United States: A Report to Congress. Institute for Health and Aging, University of California; Injury Prevention Center, Johns Hopkins University, San Francisco. Romanow R.J. (2002) Commission on the future of health care in Canada. Available at: http://www.hc-sc.gc.ca/hcs-sss/ com/fed/romanow/index-eng.php (accessed on 1/2/2012). Rosenbaum P.R. & Rubin D.B. (2007) Reducing bias in observational studies using subclassification on the propensity score. Journal of the American Statistical Association 79 (387), 516–524. Stevens B., Guerriere D. & McKeever B. et al. (2006) Economics of home vs. hospital breastfeeding support for newborns. Journal of Advanced Nursing 53 (2), 233–243. Teno JM., Clarridge B.R., Casey V., Welch L.C., Wetle T., Shield R. & Mor V. (2004) Family perspectives on end-oflife care at the last place of care. Journal of the American Medical Association 291 (1), 88–93. Torgerson D.J., Donaldson C. & Reid D.M. (1994) Private versus social opportunity cost of time: valuing time in

618

the demand for health care. Health Economics 3 (3), 149–155. Walker H., Anderson M. & Farahati F. et al. (2011) Resource use and costs of end-of-life/palliative care: Ontario adult cancer patients dying during 2002 and 2003. Journal of Palliative Care 27 (2), 79–88. Williams A.M., Wang L. & Kitchen P. (2014) Differential impacts of care-giving across three caregiver groups in Canada: end-of-life care, long-term care and short-term care. Health and Social Care in the Community 22 (2), 187–196. Wodchis W.P., Bushmeneva K., Nikitovic M. & McKillop I. (2011) Guidelines on Person-Level Costing Using Administrative Databases on Ontario. Health System Performance Research Network, Toronto. Wright A.A., Zhang B., Ray A. et al. (2008) Associations between end-of-life discussions, patient mental health, medical care near death, and caregiver bereavement adjustment. Journal of the American Medical Association 300 (14), 1665–1673.

© 2014 John Wiley & Sons Ltd