Predictors of Health Service Use Over the Palliative Care Trajectory

Abstract

Background:

Health system restructuring coupled with the preference of patients to be cared for at home has altered the setting for the provision of palliative care. Accordingly, there has been emphasis on the provision of home-based palliative care by multidisciplinary teams of health care providers. Evidence suggests that these teams are better able to identify and deal with the needs of patients and their family members. Currently there is a lack of literature examining the predictors of palliative care service use for various professional service categories.

Objective:

The purpose of this study was to examine the predictors of the propensity and intensity of five main health service categories in the last three months of life for home-based palliative care patients.

Design:

This was a prospective cohort study. The predictors of service use were assessed using a two-part model, which treats the decision to use a service (propensity) and the amount of service use (intensity) as two distinct processes. Propensity was modeled using a logistic regression and intensity was modeled using ordinary least squares regression.

Results:

The results indicate that each service category emerged with a different set of predictor variables. Common predictors of health service use across service categories were patient age and functional status. The results suggest that a consistent set of predictors across service categories does not exist, and thus the determinants of access to each service category are unique.

Conclusion:

These findings will help case managers, health administrators, and policy decision makers better allocate human resources to palliative patients.

Introduction

In Canada there has been a shift in the setting where palliative health care resources are delivered, with more emphasis being placed on the home.¹ This shift is partially attributable to health system restructuring, such as, hospital bed closures and consolidations, limits to health expenditure growth, and the diffusion of sophisticated health technologies to the home setting.² Patient and caregiver preferences for care provided in the home rather than in an institutional setting has also influenced this reallocation of resources, and accordingly, health service utilization.^3,4

Increasing financial resources to the community has resulted in a larger number of patients receiving palliative care services, as well as an increase in the participation and collaboration of community-based physicians and nurses in the care of the terminally ill.¹ Home-based palliative care programs seek to provide services through community care teams consisting of physicians, nurses, personal support workers, and case managers.⁵ Evidence suggests that such teams are better able to identify and deal with the needs of patients and their family members compared to conventional care.⁶

In Ontario, palliative home-based services are paid for by the publicly funded health care system, which covers professional services, diagnostic and laboratory tests, and prescription medications. Those services not covered by the health care system are paid out of pocket by patients. Home-based palliative care teams have been developed, such as the Temmy Latner Centre for Palliative Care (TLCPC) in Toronto, to provide specialized care. The TLCPC consists of palliative physicians who provide symptom management and practical and emotional support in the community 24 hours a day, 7 days per week. Regionally based, Community Care Access Centres (CCACs) provide a range of specialized services to help support patients at home (ie., nursing, speech language therapy, physiotherapy, case management, and homemaking). Together, the TLCPC and the CCACs collaborate to provide team-based multidisciplinary care in the community. Referrals to the TLCPC are made by physicians in the hospital system, oncologists, or the CCAC.

Currently, there is a paucity of literature examining the predictors of palliative health care service use for various service categories. A paper was published⁷ on the predictors of the total cost of home-based palliative care, which included both public and private costs. The study objective addressed in this current paper was to determine the predictors of the propensity and intensity of health service use for five main service categories (home-based physician visits, nurse visits, personal support visits, ambulatory physician visits, and other ambulatory and home-based visits) in the last three months of life for palliative care patients receiving care through a home-based palliative care program. Knowledge of the predictors for the separate service categories may be used to improve overall efficiency and equity in the provision of end-of-life care. This information may aid case managers and decision makers in health planning and resource allocation decisions.

Methods

This was a prospective cohort study designed to capture health care service utilization among palliative care patients. Family caregivers were recruited from the TLCPC at Mount Sinai Hospital in Toronto, Ontario. A TLCPC patient receptionist identified and telephoned potential participants who met the eligibility criteria and asked if they would be interested in speaking with the research officer to learn about the study. The research officer verbally explained the study, and for those who indicated that they were willing to participate, a written consent form was completed. Participants were eligible if they were (1) primary caregivers (family or friends) of patients who were diagnosed with a malignant neoplasm; (2) fluent in English; and (3) 18 years of age or older. Palliative care patients classified into other diagnostic categories were excluded from this analysis, since patients with a malignant neoplasm made up 85% of the patient population and the remaining 15% were divided among a dozen other categories. The median time patients were in the study was 28 days.

Caregivers of patients were interviewed over the telephone every two weeks from the time of study enrollment until death, between July 2005 and September 2007. Caregivers of patients were interviewed rather than patients because it would become too difficult for patients to participate in interviews as their condition deteriorated, and interviews with patients close to death would not have been possible. Caregivers were asked to recall the patient's health service use over the previous two weeks. A two-week recall period was used, since it is short enough to minimize recall bias while not overburdening the family with frequent interviews.

The utilization of five main health service categories, home-based physician visits, nurse visits, personal support visits, ambulatory physician visits, and other ambulatory and home-based visits, in the last three months of life was assessed. These service categories were chosen because they represent the services most commonly utilized by the study sample. The “other ambulatory and home-based visits” category included visits and tests that were not captured by the other categories (e.g., case coordinator, blood work, physiotherapist). Each visit was assigned a monetary value in 2008 Canadian dollars and inflated to 2011 costs using the consumer price index. Further details on the costing methodology are published elsewhere.^7,8

The selection of predictor variables was guided by Andersen and Newman's Behavioral Model of Health Service Utilization⁹ and was comprised of predisposing factors (patient and caregiver age, patient and caregiver sex, and socioeconomic status); enabling factors (coresidency); and need-based factors (the patient's functional status).

Data sources

The Ambulatory and Home Care Record (AHCR) (©Coyte and Guerriere) was used to prospectively collect data on health service use.¹⁰ The AHCR captures four categories of costs: (1) publicly financed health system costs (health professional costs, diagnostic tests, etc.; (2) out-of-pocket costs (medications, visits not publicly funded, travel expenses, etc.); (3) time costs (time devoted by family caregivers associated with providing care; and (4) third party private insurance costs. This study focused on the costs incurred by the publicly financed health care system, namely category (1) listed above. Assessment of the reliability and validity of the AHCR demonstrated that there was agreement between participants' reports and administrative data (kappa ranging from 0.41 to 1.00).¹⁰ The AHCR has been used in numerous other studies, for a wide range of patient populations and clinical settings^7,11–17

The Palliative Performance Scale (PPS) was used to collect data on functional and cognitive status of the patient during each caregiver interview.¹⁸ The PPS assesses the physical functioning in ambulation, activity, and evidence of disease, ability for self-care, intake (i.e., person's eating habits), and level of consciousness. The possible values range from 100% (full ambulation and healthy) to 0% (death), decreasing in 10% decrements. Reliability testing of the PPS-reported intra-class correlation coefficients range from 0.93 to 0.96.^18,19

Information on age, sex, coresidency, and socioeconomic characteristics was obtained from the TLCPC database using the patient's unique health care number. Information on caregiver age and sex was obtained from the caregiver interviews. Postal codes of patients were used as a proxy for socioeconomic status. Postal codes were linked to Canadian census file data, and socioeconomic status was obtained through the computation of a modified Carstairs deprivation score.²⁰ This deprivation score is an unweighted combination of three standardized variables that together yield a summary statistic (Z score) for each enumeration area within Toronto. The three standardized variables used were percentage of blue-collar workers among men (social class), percentage of unemployment rate among men, and percentage of those living in households that were below the low-income cutoff. The deprivation score was calculated by taking the sum of the three standardized variables and dividing by three to take the average. Higher deprivation scores were reflective of more deprived geographic regions.

Data analysis

Data were analyzed using statistical software SAS (SAS version 9.1.3 for Windows; SAS Institute Inc., Cary, NC). The predictors of service use were assessed using a two-part model,²¹ treating as two distinct processes the binary decision (yes/no) to use a health care service (propensity) and the decision regarding the level of service use (intensity). Propensity is defined as the probability that a patient will receive at least one visit; and intensity refers to the amount (measured in dollars) of service received, conditional on being a service user. Propensity (i.e., use yes/no) was modeled using a logistic regression. Intensity (i.e., cost) was modeled using an ordinary least-squares regression (OLS). Since the costs for each service category were skewed, cost data were log transformed using log to base 10. The coefficients were retransformed back to their original scale in order to draw useful conclusions about the original variables; therefore, parameter estimates and confidence intervals were exponentiated.²²

An interview-based analysis was conducted, and as such, generalized estimating equations (GEE) were used to account for within-group correlation. Within a GEE model, the outcome produces both a matrix of the coefficients and a matrix with the inverse of the variance. Since the data are correlated, the variances are multiplied against a working matrix of correlation coefficients to correct for correlation within subjects. Multicollinearity was assessed using the variance inflation factor (VIF); variables with VIF >4 were removed from the model.²³

Backward stepwise regression was conducted for each service category in order to determine a final model of predictor variables. Coefficients with p≥0.1 were removed from the model. This study received approval from both the University of Toronto and Mount Sinai Hospital ethics review boards.

Results

Over the data collection period, 359 caregivers of patients were identified by the TLCPC as being eligible for participation. Of those who were reached by the receptionist, 169 (63.8%) agreed to receive information by mail. Consent to participate was given by 151 (89.3%) of those who received information; however, eight patients were deemed ineligible because of hospitalization before the first interview and six subjects dropped out before the first interview because the caregiver lacked time, was under stress, or for other reasons. The resulting study sample of 137 patients and caregivers was reviewed for the analysis, and a further 28 participants were excluded because their date of death was not known (n=7) or because they did not have interviews three months prior to death (n=21). Consequently, results are presented for 109 participants, representing 64.5% of all subjects originally contacted by the receptionist. These participants yielded a total of 305 interviews. Caregivers participated in an average of three interviews.

The characteristics of the study sample are presented in Table 1. The mean age of the patients was 71.1years and more than half were women (53%). The majority of patients lived with someone (88%). The mean age of the caregiver was 56.7 years and the majority were women (66%). The caregivers were either spouses (52%) or children (39%), and a small percentage were either friends or neighbors (3.7%). Most of the caregivers were either employed (33%) or retired (32%).

Table 1.

Characteristics of Patients and Caregivers

Variable	Number of individuals (%)
Sample size	109
Patient sex, male	51 (46.8%)
Patient age, mean (median)	71.1 (72.0)
Patient age
≤61	27 (24.8%)
62–72	28 (25.7%)
73–82	27 (24.8%)
≥83	27 (24.8%)
Coresidency, lives alone	12 (11.5%)
Socioeconomic status (Deprivation Score)
≤50	27 (25.0%)
51–64	27 (25.0%)
65–1.06	27 (25.0%)
≥1.07	27 (25.0%)
Unknown	1
Caregiver sex
Male	37 (34.3%)
Caregiver age, mean (median)	56.7 (58.0)
Caregiver age
≤46	27 (25.2%)
47–57	26 (24.3%)
58–67	27 (25.2%)
≥68	27 (25.2%)
Unknown	2
Retired	35 (32.1%)
Relationship to patient
Spouse	57 (52.3%)
Child	42 (38.6%)
Friend/neighbor	4 (3.7%)
Unknown	6 (5.5%)
Employment status
On leave	17 (15.6%)
Employed	36 (33.0%)
Unemployed	19 (17.4%)
Student	2 (1.8%)
Retired	35 (32.1%)
Cancer diagnosis
Lung/respiratory	28 (25.7%)
Digestive/gastrointestinal	28 (25.7%)
Gynecologic/genitourinary	23 (21.1%)
Breast	10 (9.2%)
Hematologic/blood	6 (5.5%)
Skin	4 (3.7%)
Brain	4 (3.7%)
Head and neck	3 (2.7%)
Unknown	3 (2.7%)

Table 2 presents the number of caregiver interviews in which a patient did and did not use a particular service. The proportion of interviews in which a service was used is reported for each service category. Home-based nurse visits were the most used service category, with reports of this service in 92.1% of the interviews. In contrast, only 15.4% of the interviews reported an ambulatory home-based physician visit.

Table 2.

Number of Reports of Service Use by Service Category (n=305)

	Service use, # interviews
Service category	No	Yes	Proportion of service use, %
Home-based nurse visits	24	281	92.1
Home-based PSW visits	176	129	42.2
Home-based physician visits	130	175	57.3
Ambulatory physician visits	258	47	15.4
Other ambulatory and home-based visits	133	172	56.4

PSW, personal support worker.

Predictors of the propensity of service use

Table 3 presents the results of the final logistic model using backward stepwise regression. Patient age was a significant predictor of the propensity of service use for home-based physician visits, home-based personal support worker (PSW) visits, and other ambulatory and home-based visits. For home-based physician visits, those patients who were 62–72 years of age had a 72% lower likelihood of having at least one home-based physician visit compared to those in the youngest age group (OR, odds ratio 0.28). For other ambulatory and home-based visits, those patients in the older age groups (73–82 and ≥83 years) were 67% and 76% less likely, respectively, to use this service compared to those in the youngest age group (OR 0.33, OR 0.24). Patient age was also a predictor of receiving a home-based PSW visit, although the opposite relationship was observed; patients aged 62–72 had almost a fourfold greater likelihood of using this service compared to those in the youngest age group.

Table 3.

Predictors of the Propensity of Service Use (Logistic Regression) ^*

	Home-based visits
Variable	Physician OR (95% CI)	PSW OR (95% CI)	Ambulatory physician OR (95% CI)	Other ambulatory and home-based OR (95% CI)
Patient age
34–61 (Quartile 1)	1.00	1.00	1.00	1.00
62–72 (Quartile 2)	0.28^** (0.13, 0.60)	3.77^** (1.28, 11.13)	0.43 (0.15, 1.19)	0.45 (0.17, 1.21)
73–82 (Quartile 3)	0.58 (0.25, 1.37)	2.42 (0.89,6.56)	0.29 (0.08, 1.04)	0.33^** (0.13, 0.85)
≥83 (Quartile 4)	0.67 (0.30, 1.53)	1.92 (0.69,5.33)	0.28^** (0.10, 0.77)	0.24^** (0.10, 0.59)
PPS level	0.97^* (0.95, 0.98)	0.97^** (0.95, 0.99)	1.02 (1.00, 1.05)	N/A
Lives alone (baseline lives with others)	N/A	0.32 (0.10, 1.08)	N/A	N/A

Using an elimination criterion of p>0.1, there were no variables left in the final model for nurse visits.

significant at p≤0.01.

CI, confidence interval; OR, odds ratio; PPS, Palliative Performance Scale; PSW, personal support worker.

As expected, patients with higher functional status were less likely to receive home-based visits. For each one-unit increase in PPS level, there was a decrease in the likelihood of receiving a home-based physician visit, PSW visits, and nurse visits. Male caregivers experienced more than a fourfold greater likelihood of using a home-based nurse visit compared to female caregivers.

Predictors of the intensity of service use

Table 4 presents the results of the final OLS regression model. Patient age was a predictor of the intensity of service use for home-based nurse and physician visits. For home-based nurse visits, patients aged 62–72 and 73–82 had a 35% and 43%, respectively, lower intensity of nurse visits compared to those in the youngest age group. For home-based physician visits, those in the two oldest groups had a 27% and 31%, respectively, lower intensity of physician visits compared to those in the youngest age group.

Table 4.

Predictors of the Intensity of Service Use (Ordinary Least-Squares Regression)

	Home-based visits
Variable	Physician exp (β) (p-value)	Nurse exp (β) (p-value)	PSW exp (β) (p-value)
Patient age
34–61 (Quartile 1)	1.00	1.00	N/A
62–72 (Quartile 2)	0.75 (0.06)	0.65^** (0.02)	N/A
73–82 (Quartile 3)	0.73^** (0.02)	0.57^* (0.00)	N/A
≥83 (Quartile 4)	0.69^* (0.01)	0.92 (0.66)	N/A
Caregiver age
19–46 (Quartile 1)	1.00	1.00	N/A
47–57 (Quartile 2)	0.83 (0.19)	N/A	N/A
58–67 (Quartile 3)	0.74^** (0.03)	N/A	N/A
≥68 (Quartile 4)	0.68^* (0.01)	N/A	N/A
Socioeconomic status
0–0.50 (Quartile 1)	N/A	1.00	N/A
0.51–0.64 (Quartile 2)	N/A	0.98 (0.92)	N/A
0.65–1.06 (Quartile 3)	N/A	1.07 (0.71)	N/A
≥1.07 (Quartile 4)	N/A	1.74^* (0.00)	N/A
Male (baseline female patient)	N/A	N/A	N/A
Male caregiver (baseline female caregiver)	N/A	0.62^* (0.00)	N/A
PPS level	0.99^** (0.04)	0.99^* (0.01)	0.99^* (0.00)
Lives alone (baseline lives with others)	0.77 (0.11)	0.69 (0.07)	N/A

significant at p-value≤0.01.

significant at p-value≤0.05.

exp, exponentiated; PPS, Palliative Performance Scale.

“Ambulatory physician visits” and “Other ambulatory and home-based visits” categories are not shown because they were N/A.

Caregivers in the older age groups had a 26% and 32% greater intensity of home-based physician visits. Caregiver sex was a predictor of the intensity of home-based nurse visits. Male caregivers had a 38% lower intensity of home-based nurse visits compared to females.

Socioeconomic status emerged as a predictor of the intensity of nurse visits; patients in the highest deprivation level (≥1.07) had a 74% greater intensity of nurse visits compared to those in the lowest deprivation group (0–0.50). Lastly, PPS level was a significant predictor of the intensity of home-based physician, nurse and PSW visits. For each incremental one-unit increase in PPS level, there was a 1% decrease in the intensity of visits.

Discussion

We believe this is the first study to examine the predictors of the propensity and intensity of health service use for five main palliative care service categories. The multivariate results for the propensity of service use indicated that patient age, caregiver sex, and PPS level predicted the use of particular service categories. In the multivariate model, patient age, caregiver age, socioeconomic status, caregiver sex, and PPS level predicted the intensity of service use of service categories.

Our study found that PPS level predicted both the propensity and intensity of service use for home-based physician, nurse and PSW visits. Patients with lower functional status were more likely to have a home-based visit and higher intensity of service use. This finding was expected, because as a patient's health declines, he or she may need increased support in the home with activities of daily living, and may have increased medical/medication needs. This is consistent with the findings of an Italian study that reported a negative correlation between a patient's Karnofsky Performance Status score (modified PPS) and average weekly home-based care costs.²⁴ However, these results are not directly comparable to ours, as the Italian study did not assess utilization by separate service categories.

The use of home-based nurse visits differed by caregiver gender. Patients with male caregivers were more likely to have at least one home-based nurse visit compared to those with female caregivers. This contrasts with the findings of a Canadian study that did not report gender differences in the likelihood of using home-based nurse visits.²⁵ This may be a result of population differences, as their study investigated those who were terminally ill and not specifically those diagnosed with a malignant neoplasm; and they did not examine services delivered through an established palliative care program.

Although our study revealed that patients with male caregivers were less likely to receive a nurse visit, we also observed a gender difference with respect to intensity. Patients of male caregivers had a lower intensity of nurse visits compared to female caregivers. It is possible that female caregivers lobbied more for nurse services compared to male caregivers. There may have been other factors that we were not able to observe, such as the physical status of the caregivers.

In our study, patient age predicted the propensity of service use for home-based physician visits, PSW visits, and other ambulatory and home-based visits. For home-based physician visits, patients 62–72 years of age were less likely to use at least one physician visit than those in the youngest age group. A Canadian study found that patients aged 65 and older were more likely to have a home-based physician visit than those in the youngest age group.²⁶ However, their study examined home-based physician visits in the last six months of life and consisted of patients who were registered in a palliative care program and those who were not, with the majority being the latter.

Patient age also predicted the intensity of service use for home-based physician visits and nurse visits, with a decrease in intensity as age increased. Studies that have assessed the predictors of receiving home-based services overall and not by service category report that patients in older age groups had significantly lower costs compared to younger patients.^27–30

In the intensity model, patients with caregivers in the older age groups of 58–67 and ≥68 experienced a lower intensity of home-based physician visits compared to caregivers in the youngest age group. The reason for this finding may be that younger caregivers are more effective at lobbying for care compared to older caregivers; and they may seek additional care due to life circumstances such as laborforce participation and childrearing.

Patients in the highest deprivation group (i.e., low socioeconomic status) had higher intensity of home-based nurse visits. Those with high socioeconomic status may have reported lower intensity of home-based nurse visits as caregivers may have been able to take time off work to provide care and may have the financial ability to supplement care with privately paid care. Our study results are corroborated by a study conducted in the United Kingdom that found those receiving home-based nurse visits were more likely to be in lower SES classes; however in their study, this relationship did not exist in the multivariate regression analysis, and they did not look at service intensity.³² In contrast, two studies reported that those who were more economically deprived utilized fewer home-based services.^26,31 However, one of these studies³¹ focused on the utilization of all home-based services as an aggregate and the other²⁶ focused on only home-based physician visits.

Limitations

Our study has a few notable limitations. There were few statistically significant predictors of the propensity and intensity of service use. There may be other characteristics not included in our model that predicted health service utilization, for example, symptom severity.

This study included patients receiving care through a regional palliative care program, and as such, the findings may not necessarily generalize to individuals receiving services from other types of palliative programs. However, the population serviced by the palliative program is diverse in clinical, demographic, and ethnic background, which may help improve generalizability.

Lastly, since some of the data were acquired from telephone interviews with caregivers, there was a potential for bias in recall and based on social desirability. However, because the psychometric properties of the study instruments have met acceptable standards, we believe this bias is minimal.

Conclusions

There is a paucity of literature examining the predictors of use for individual service categories for palliative care patients. This study sought to fill a critical gap in the literature by examining the predictors of service use of five main health service categories used at the end of life. Our findings demonstrate that a consistent set of predictor variables across service categories does not exist, and thus each service category should be assessed separately. These findings will aid case managers and decision makers in the allocation of human resources across service categories. Understanding the predictors of service use can enable more patients to remain and die at home if they so choose, and can ensure that the amount of care received is tailored towards patient characteristics.

To ensure accessibility, palliative care programs could target their efforts to patients who may have more difficulty accessing services, such as those in older age groups. Upon referral to the home-based palliative care program, a screening tool can be developed to determine whether the patient may need more specialized services based on predefined characteristics (e.g., patient age, caregiver age, PPS level). Since it was found in all models that PPS level predicted both the likelihood and amount of home-based services use, CCAC case managers can play close attention to these scores in order to determine the amount of services allocated to each patient. Further research should be conducted to improve our understanding of the predictors of palliative care service use and the characteristics of patients who use these services.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

References

Bruera

, Sweeney

. Palliative care models: International perspective. J Palliat Med, 2002; 5,2:319–327.

Stajduhar

. Examining the perspectives of family members involved in the delivery of palliative care at home. J Palliat Care, 2003; 19,1:27–35.

Higginson

, Sen-Gupta

. Place of care in advanced cancer: A qualitative systematic literature review of patient preferences. J Palliat Med, 2000; 3,3:287–300.

Romanow

. Building on Values: The Future of Health Care in Canada–Final Report. Saskatoon: Commission on the Future of Health Care in Canada, 2002. http://publications.gc.ca/collections/Collection/cp32-85-2002E.pdf

Carstairs

. Still Not There: Quality End-of-Life Care: A Progress Report. Ottawa: Senate of Canada, 2005. http://www.sharoncarstairs.ca/stillNotThereJune2005.pdf

Hearn

, Higginson

. Do specialist palliative care teams improve outcomes for cancer patients? A systematic literature review. Palliat Med, 1998; 12,5:317–332.

Guerriere

, Zagorski

, Fassbender

, Masucci

, Librach

, Coyte

. Cost variations in ambulatory and home-based palliative care. Palliat Med, 2010; 24,5:523–532.

Masucci

. Health Care Service Provision Over the Palliative Care Trajectory. MSc Thesis. University of Toronto. Ann Arbor: ProQuest/UMI, 2011(Publication No. 916718702.)

Andersen

, Newman

. Societal and individual determinants of medical care utilization in the United States. Milbank Mem Fund Q Health Soc, 1973; 51,1:95–124.

10.

Guerriere

, Ungar

, Corey

, Croxford

, Tranmer

, Tullis

et al. Evaluation of the ambulatory and home care record: Agreement between self-reports and administrative data. Int J Technol Assess Health Care, 2006; 22,2:203–210.

11.

Guerriere

, Tullis

, Ungar

, Tranmer

, Corey

, Gaskin

et al. Economic burden of ambulatory and home-based care for adults with cystic fibrosis. Treat Respir Med, 2006; 5,5:351–359.

12.

Guerriere

, Choiniere

, Dion

, Peng

, Stafford-Coyte

, Zagorski

et al.

The Canadian STOP-PAIN Project: Part 2, What is the cost of pain for patients on waitlists of multidisciplinary pain treatment facilities?

Can J Anaesth, 2010; 57,6:549–558.

13.

Guerriere

, Tranmer

, Ungar

, Manoharan

, Coyte

. Valuing care recipient and family caregiver time: A comparison of methods. Int J Technol Assess Health Care, 2008; 24,1:52–59.

14.

Guerriere

, Wong

, Croxford

, Leong

, McKeever

, Coyte

. Costs and determinants of privately financed home-based health care in Ontario, Canada. Health Soc Care Community, 2008; 16,2:126–136.

15.

McGillion

, Croxford

, Watt-Watson

, Lefort

, Stevens

, Coyte

. Cost of illness for chronic stable angina patients enrolled in a self-management education trial. Can J Cardiol, 2008; 24,10:759–764.

16.

Stevens

, Guerriere

, McKeever

, Croxford

, Miller

, Watson-MacDonell

et al. Economics of home vs. hospital breastfeeding support for newborns. J Adv Nurs, 2006; 53,2:233–243.

17.

Masucci

, Guerriere

, Cheng

, Coyte

. Determinants of place of death for recipients of home-based palliative care. J Palliat Care, 2010; 26,4:279–286.

18.

Downing

, Lau

, Lesperance

, Karlson

, Shaw

, Kuziemsky

et al. Meta-analysis of survival prediction with Palliative Performance Scale. J Palliat Care, 2007; 23,4:245–252.

19.

, Lau

, Downing

, Lesperance

. A reliability and validity study of the Palliative Performance Scale. BMC Palliat Care, 2008; 7:10.

20.

Carstairs

, Morris

. Deprivation, Health in Scotland. Aberdeen: Abderdeen University Press, 1991.

21.

Puhani

. The Heckman correction for sample selection and its critique. J Econ Surveys, 2000; 14,1:53–68.

22.

Diehr

, Yanez

, Ash

, Hornbrook

, Lin

. Methods for analyzing health care utilization and costs. Annu Rev Public Health, 1999; 20:125–144.

23.

O'Brien

. A caution regarding rules of thumb for variance inflation factors. Quality and Quantity, 2007; 41,5.

24.

Maltoni

, Travaglini

, Santi

, Nanni

, Scarpi

, Benvenuti

et al. Evaluation of the cost of home care for terminally ill cancer patients. Support Care Cancer, 1997; 5,5:396–401.

25.

Brazil

, Thabane

, Foster

, Bedard

. Gender differences among Canadian spousal caregivers at the end of life. Health Soc Care Community, 2009; 17,2:159–166.

26.

Burge

, Lawson

, Johnston

. Home visits by family physicians during the end-of-life: Does patient income or residence play a role? BMC Palliat Care, 2005274,1:1.

27.

Chan

, Jacobs

, Yeo

, Lai

, Hazlett

, Mok

et al. The cost of palliative care for hepatocellular carcinoma in Hong Kong. Pharmacoeconomics, 2001; 19,9:947–953.

28.

Mukamel

, Bajorska

, Temkin-Greener

. Health care services utilization at the end of life in a managed care program integrating acute and long-term care. Med Care, 2002; 40,12:1136–1148.

29.

Shnoor

, Szlaifer

, Aoberman

, Bentur

. The cost of home hospice care for terminal patients in Israel. Am J Hosp Palliat Care, 2007; 24,4:284–290.

30.

Tamir

, Singer

, Shvartzman

. Taking care of terminally-ill patients at home: The economic perspective revisited. Palliat Med, 2007; 21,6:537–541.

31.

Jakobsson

, Bergh

, Ohlen

, Oden

, Gaston-Johansson

. Utilization of health-care services at the end-of-life. Health Policy, 2007; 82,3:276–287.

32.

Grande

, Farquhar

, Barclay

, Todd

. The influence of patient and carer age in access to palliative care services. Age Ageing, 2006; 35,3:267–273.