48
Journal of Pain and Symptom Management
Vol. 50 No. 1 July 2015
Original Article
The Association Between Home Palliative Care Services and Quality of End-of-Life Care Indicators in the Province of Quebec Bruno Gagnon, MD, MSc, Lyne Nadeau, MSc, Susan Scott, MSc, Serge Dumont, PhD, Neil MacDonald, MD, CM, Mich ele Aubin, MD, PhD, FCMF, CCMF, and Nancy Mayo, PT, PhD Department of Family Medicine and Emergency Medicine (B.G., M.A.); School of Social Work (S.D.); and Cancer Research Center (B.G., S.D.), Laval University, Que bec City; Division of Clinical Epidemiology (B.G., L.N., S.S., N.May.), McGill University Health Centre, Montreal; and Department of Oncology (N.Mac.) and School of Physical and Occupational Therapy (N.May.), McGill University, Montreal, Que bec, Canada
Abstract Context. In Canada, governments have increased spending on home care to promote better end-of-life care. In the province of Quebec, Canada, home palliative care (PC) services (HPCS) are provided by Public Local Community-Based Health Care Service providers (Centres Locaux de Services Communautaires [CLSC]) with universal coverage. Accordingly, there should be no regional variations of these services and their effect on quality of end-of-life PC (QEoLPC) indicators. Objectives. To test if all the CLSCs provided the same level of HPCS to cancer patients in the province of Qu ebec, Canada, and the association between level of HPCS and QEoLPC indicators. Methods. Characteristics of 52,316 decedents with cancer were extracted from administrative databases between 2003 and 2006. Two gender-specific ‘‘adjusted performance of CLSCs in delivering HPCS’’ models were created using gender-specific hierarchical regression adjusted for patient and CLSC neighborhood characteristics. Using the same approach, the strength of the association between the adjusted performance of CLSCs in delivering HPCS and the QEoLPC indicators was estimated. Results. Overall, 27,255 (52.1%) decedents had at least one HPCS. Significant variations in the adjusted performance of CLSC in delivering HPCS were found. Higher performance led to a lower proportion of men having more than one emergency room visit during the last month of life (risk ratio [RR] 0.924; 95% CI 0.867e0.985), and for women, a higher proportion dying at home (RR 2.255; 95% CI 1.703e2.984) and spending less time in hospital (RR 0.765; 95% CI 0.692e0.845). Conclusion. Provision of HPCS remained limited in Qu ebec, but when present, they were associated with improved QEoLPC indicators. J Pain Symptom Manage 2015;50:48e58. Ó 2015 American Academy of Hospice and Palliative Medicine. Published by Elsevier Inc. All rights reserved. Key Words Palliative medicine, administrative databases, home services, nursing, oncology
Introduction For many patients with advanced cancer, home is the preferred place of care and place of death.1e3 There is a growing interest in improving quality of care at end of life. Various indicators of aggressive care at end of life have been proposed and studied using administrative databases.4e7 Some of these
Address correspondence to: Bruno Gagnon, MD, MSc, Centre de recherche du CHU de QuebecdL’H^ otel-Dieu de Qu ebec, 9, rue McMahon, Quebec City, Quebec G1R 2J6, Canada. E-mail: [email protected] Ó 2015 American Academy of Hospice and Palliative Medicine. Published by Elsevier Inc. All rights reserved.
indicators include chemotherapy in the last two weeks of life, number of emergency room (ER) visits and hospital or intensive care unit (ICU) admissions in the last month of life, and late enrollment in hospice programs. One study, by Barbera et al,8 using the Ontario Cancer Registry and administrative databases, found that in a cohort of 21,323 people who died in
Accepted for publication: November 21, 2013.
0885-3924/$ - see front matter http://dx.doi.org/10.1016/j.jpainsymman.2014.12.012
Vol. 50 No. 1 July 2015
Quality of HPCS
2001, 27% had at least one ER visit, 5% an ICU admission, and 16% chemotherapy treatments in the last two weeks of life. In Ontario, Canada, a cohort study of 9018 decedents looking at the association between home palliative care (PC) services (HPCS) and quality of end-of-life PC (QEoLPC) found that earlier referral to intensive home services were associated with reduced risk of having a hospitalization or an ER visit in the last two weeks before death.9 However, this study only looked at patients referred to home PC. The above-mentioned study by Barbera et al8 found that HPCS were associated with a lower proportion of ER visits, ICU admissions, and chemotherapy treatment in the last two weeks of life. These studies, however, did not look at local variations of HPCS across the province and their potential effects on systemwide QEoLPC indicators.10 From a health policy point of view, it is paramount to identify factors associated with provision of HPCS at the local and regional geographic levels. This will improve our understanding of the complexity of the health care system in view of guiding decisions on policies on structural changes and allocation of funding. In the province of Qu ebec, all citizens receive HPCS by Local Public Communityebased Comprehensive Healthcare Service providers named ‘‘Centres Locaux de Services Communautaires’’ (CLSC)11 within a welldefined small geographical area, which in turn are within Regional Health Care Districts (RHCDs).11,12 Hypothetically, such universal coverage should be associated with the same level of HPCS and equal quality of QEoLPC indicators across the province. The aims of this study were to estimate if all CLSCs provided the same level of HPCS to cancer patients in the province of Qu ebec, Canada, between 2003 and 2006, and to determine the association between level of HPCS and QEoLPC indicators.
Methods Data Sources We accessed four data sources from Qu ebec, Canada, namely: 1) the Death Registry (Institut de la Statistique du Qu ebec); 2) the Hospital Discharge database (Maintenance et Exploitation des Donnees pour l’Etude de la Client ele Hospitali ere [MedEcho]); 3) the Physician Fee-for-Service Billings database (R egie de l’Assurance Maladie du Qu ebec [RAMQ]); and 4) the Home Services database (Banque des donn ees des services offerts par les CLSC). These databases are linked through a unique identity number (Numero d’assurance maladie du Qu ebec [NAM]). The MedEcho database, among other information, provided each person’s gender and date of birth. It
49
also provided a summary of all hospital admissions, including the principal medical conditions justifying each admission, all comorbidities, topography codes for cancer diagnoses, and sites of metastasis. Dates of admission and discharge, type of bed to which the person was admitted (acute, chronic, or palliative), discharge status (alive or dead), and orientation at discharge were also included in the summary. The RAMQ database provided all physicians’ fee for services and type of visits, including a special code for home visits and dates. The CLSC database included all services (PC and others) provided by each professional (physicians, nurses, and so on), and the date and place where the services were provided. We requested all MedEcho, Death Registry, RAMQ, and CLSC data available for all the decedents included in the population (Appendix). To maintain confidentiality, all the different databases were linked by the Infocentre, and the data were transmitted to our research center with a scrambled unique subject identifier number. The 2001 census data from Statistics Canada,13 which contains sociodemographic characteristics for all postal code areas, was used to characterize the neighborhood of each CLSC. Information needed for this research project was obtained as enumerated in the approved list by the ‘‘Commission d’acces a l’information.’’ Ethical approval was obtained from the McGill University Health Center Research Ethics Board.
Defining the Population As Quebec, Canada, does not maintain a systematic provincial-wide tumor registry, we requested the list of people diagnosed with cancer during the five years preceding the study period (2003e2006) from MedEcho and obtained the death certificates of those who died. To be included in the study, a person must have had cancer as the cause of death on his/her death certificate, or have had metastatic cancer according to MedEcho data, and a cause of death closely related to cancer determined from an algorithm that we previously validated.14
Measures Definition of HPCS. In Quebec, Canada, the population is entitled to complete health care coverage for physicians’ services and HPCS. The HPCS are usually provided through the CLSC of a patient’s neighborhood by either a specialized PC team including physicians, nurses, physical therapists, occupational therapists, social workers, psychologists, and personal physical care support or by some of these health care providers working on an ad hoc basis. The procedure to establish HPCS begins with a request to the CLSC from the patient, his family, or any health
50
Gagnon et al.
professional. After the evaluation of the patient’s needs, resources are dispatched. Medical support is provided by family physicians who usually do not have specific training in PC.15 Each CLSC is responsible for organizing HPCS delivery and managing community associations for a delimited geographic sector.16 The dates of the home PC visits by physicians and nurses to each decedent during the last three months of life were extracted from the CLSC database using the PC code and from the RAMQ database using the home services code. For the purpose of this study, any home PC visit by physicians and nurses defined HPCS, while recognizing that other health care professionals offer HPCS. Definition of Factors Associated with Home PC Visits. Patient demographics were extracted from the RAMQ database, and cause of death was extracted from the death certificate. Information from the MedEcho database for the preceding five years before death was used to extract comorbidities that could be associated with home health care services (congestive heart failure, cerebrovascular accident, peripheral vascular disease, dementia, chronic obstructive pulmonary disease, hemiparaplegia, and liver disease). Time at home for the last three months of life was defined as the number of days not spent in a hospital, community-based PC setting, or ER. People needed to be at home for at least seven days during this period to be included in the analysis. The number of admissions to the ICU and the number of blood product transfusions for the last three months of life were extracted from the MedEcho database, and for the same period, the number of chemotherapy treatments was extracted from the RAMQ database. Because we did not have the socioeconomic data of each person, a weighted average of the neighborhood characteristics of each CLSC was determined by linking the information available from Statistics Canada using the postal codes of their geographic sector. The social and material status for each neighborhood was estimated using the social and material deprivation indices developed and validated by Pampalon et al.17 In certain regions of Qu ebec, nonprofit organizations provide HPCS. To adjust for their association with HPCS delivery, a variable to identify the CLSCs covered by these organizations was created. Definition of QEoLPC Indicators. We used some of the QEoLPC indicators as defined by the ‘‘Institut National de Sant e Publique du Qu ebec’’18 and suggested in the 4e7 literature : place of death (home and ER), having at least one visit to the ER during the last 28 days excluding the one associated with the last admission leading to death, and having more than one hospitalization or
Vol. 50 No. 1 July 2015
spending more than 14 days in the hospital during the last month of life. People were considered to have died at home if the place of death on the death certificate was a private address. These specific QEoLPC indicators were selected because we hypothesized that they would be more responsive to HPCS.
Statistical Analysis As a high proportion of patients were not receiving any HPCS, this outcome was dichotomized into receiving any or not. To account for sex-specific cancer types, data from men and women were analyzed separately. The main exposure variable was the proportion of people receiving HPCS from their respective CLSC, which is called ‘‘performance of each CLSC in delivering HPCS’’ in this article. As such, multivariate gender-specific hierarchical risk ratio (RR) regression models (Fig. 1, Model A) were carried out to account for the cluster effect within each CLSC of all the covariates under study on the delivery of HPCS (lower box of Model A). The resultant of each model is an adjusted performance of each CLSC in delivering HPCS (expressed by the b estimate). Because time spent in the community was likely to influence the referral pattern for HPCS, proxy variables hypothesized to influence this pattern of referral were included in the model, such as time spent at home, ICU admissions, and chemotherapy treatment. In view of the complexity of the model, it was necessary to be parsimonious in the selection of variables to be included in the multiple gender-specific hierarchical RR regression models of the strength association between the adjusted performance of each CLSC in delivering HPCS and QEoLPC indicators. Therefore, a series of multiple gender-specific hierarchical RR regression models adjusting only for age and time spent at home were carried out for each covariate to select the ones with clinical relevance (with an RR value #0.90 or $1.10). Continuous covariates were categorized by regrouping their quartile values with similar RR (
Journal of Pain and Symptom Management
Vol. 50 No. 1 July 2015
Original Article
The Association Between Home Palliative Care Services and Quality of End-of-Life Care Indicators in the Province of Quebec Bruno Gagnon, MD, MSc, Lyne Nadeau, MSc, Susan Scott, MSc, Serge Dumont, PhD, Neil MacDonald, MD, CM, Mich ele Aubin, MD, PhD, FCMF, CCMF, and Nancy Mayo, PT, PhD Department of Family Medicine and Emergency Medicine (B.G., M.A.); School of Social Work (S.D.); and Cancer Research Center (B.G., S.D.), Laval University, Que bec City; Division of Clinical Epidemiology (B.G., L.N., S.S., N.May.), McGill University Health Centre, Montreal; and Department of Oncology (N.Mac.) and School of Physical and Occupational Therapy (N.May.), McGill University, Montreal, Que bec, Canada
Abstract Context. In Canada, governments have increased spending on home care to promote better end-of-life care. In the province of Quebec, Canada, home palliative care (PC) services (HPCS) are provided by Public Local Community-Based Health Care Service providers (Centres Locaux de Services Communautaires [CLSC]) with universal coverage. Accordingly, there should be no regional variations of these services and their effect on quality of end-of-life PC (QEoLPC) indicators. Objectives. To test if all the CLSCs provided the same level of HPCS to cancer patients in the province of Qu ebec, Canada, and the association between level of HPCS and QEoLPC indicators. Methods. Characteristics of 52,316 decedents with cancer were extracted from administrative databases between 2003 and 2006. Two gender-specific ‘‘adjusted performance of CLSCs in delivering HPCS’’ models were created using gender-specific hierarchical regression adjusted for patient and CLSC neighborhood characteristics. Using the same approach, the strength of the association between the adjusted performance of CLSCs in delivering HPCS and the QEoLPC indicators was estimated. Results. Overall, 27,255 (52.1%) decedents had at least one HPCS. Significant variations in the adjusted performance of CLSC in delivering HPCS were found. Higher performance led to a lower proportion of men having more than one emergency room visit during the last month of life (risk ratio [RR] 0.924; 95% CI 0.867e0.985), and for women, a higher proportion dying at home (RR 2.255; 95% CI 1.703e2.984) and spending less time in hospital (RR 0.765; 95% CI 0.692e0.845). Conclusion. Provision of HPCS remained limited in Qu ebec, but when present, they were associated with improved QEoLPC indicators. J Pain Symptom Manage 2015;50:48e58. Ó 2015 American Academy of Hospice and Palliative Medicine. Published by Elsevier Inc. All rights reserved. Key Words Palliative medicine, administrative databases, home services, nursing, oncology
Introduction For many patients with advanced cancer, home is the preferred place of care and place of death.1e3 There is a growing interest in improving quality of care at end of life. Various indicators of aggressive care at end of life have been proposed and studied using administrative databases.4e7 Some of these
Address correspondence to: Bruno Gagnon, MD, MSc, Centre de recherche du CHU de QuebecdL’H^ otel-Dieu de Qu ebec, 9, rue McMahon, Quebec City, Quebec G1R 2J6, Canada. E-mail: [email protected] Ó 2015 American Academy of Hospice and Palliative Medicine. Published by Elsevier Inc. All rights reserved.
indicators include chemotherapy in the last two weeks of life, number of emergency room (ER) visits and hospital or intensive care unit (ICU) admissions in the last month of life, and late enrollment in hospice programs. One study, by Barbera et al,8 using the Ontario Cancer Registry and administrative databases, found that in a cohort of 21,323 people who died in
Accepted for publication: November 21, 2013.
0885-3924/$ - see front matter http://dx.doi.org/10.1016/j.jpainsymman.2014.12.012
Vol. 50 No. 1 July 2015
Quality of HPCS
2001, 27% had at least one ER visit, 5% an ICU admission, and 16% chemotherapy treatments in the last two weeks of life. In Ontario, Canada, a cohort study of 9018 decedents looking at the association between home palliative care (PC) services (HPCS) and quality of end-of-life PC (QEoLPC) found that earlier referral to intensive home services were associated with reduced risk of having a hospitalization or an ER visit in the last two weeks before death.9 However, this study only looked at patients referred to home PC. The above-mentioned study by Barbera et al8 found that HPCS were associated with a lower proportion of ER visits, ICU admissions, and chemotherapy treatment in the last two weeks of life. These studies, however, did not look at local variations of HPCS across the province and their potential effects on systemwide QEoLPC indicators.10 From a health policy point of view, it is paramount to identify factors associated with provision of HPCS at the local and regional geographic levels. This will improve our understanding of the complexity of the health care system in view of guiding decisions on policies on structural changes and allocation of funding. In the province of Qu ebec, all citizens receive HPCS by Local Public Communityebased Comprehensive Healthcare Service providers named ‘‘Centres Locaux de Services Communautaires’’ (CLSC)11 within a welldefined small geographical area, which in turn are within Regional Health Care Districts (RHCDs).11,12 Hypothetically, such universal coverage should be associated with the same level of HPCS and equal quality of QEoLPC indicators across the province. The aims of this study were to estimate if all CLSCs provided the same level of HPCS to cancer patients in the province of Qu ebec, Canada, between 2003 and 2006, and to determine the association between level of HPCS and QEoLPC indicators.
Methods Data Sources We accessed four data sources from Qu ebec, Canada, namely: 1) the Death Registry (Institut de la Statistique du Qu ebec); 2) the Hospital Discharge database (Maintenance et Exploitation des Donnees pour l’Etude de la Client ele Hospitali ere [MedEcho]); 3) the Physician Fee-for-Service Billings database (R egie de l’Assurance Maladie du Qu ebec [RAMQ]); and 4) the Home Services database (Banque des donn ees des services offerts par les CLSC). These databases are linked through a unique identity number (Numero d’assurance maladie du Qu ebec [NAM]). The MedEcho database, among other information, provided each person’s gender and date of birth. It
49
also provided a summary of all hospital admissions, including the principal medical conditions justifying each admission, all comorbidities, topography codes for cancer diagnoses, and sites of metastasis. Dates of admission and discharge, type of bed to which the person was admitted (acute, chronic, or palliative), discharge status (alive or dead), and orientation at discharge were also included in the summary. The RAMQ database provided all physicians’ fee for services and type of visits, including a special code for home visits and dates. The CLSC database included all services (PC and others) provided by each professional (physicians, nurses, and so on), and the date and place where the services were provided. We requested all MedEcho, Death Registry, RAMQ, and CLSC data available for all the decedents included in the population (Appendix). To maintain confidentiality, all the different databases were linked by the Infocentre, and the data were transmitted to our research center with a scrambled unique subject identifier number. The 2001 census data from Statistics Canada,13 which contains sociodemographic characteristics for all postal code areas, was used to characterize the neighborhood of each CLSC. Information needed for this research project was obtained as enumerated in the approved list by the ‘‘Commission d’acces a l’information.’’ Ethical approval was obtained from the McGill University Health Center Research Ethics Board.
Defining the Population As Quebec, Canada, does not maintain a systematic provincial-wide tumor registry, we requested the list of people diagnosed with cancer during the five years preceding the study period (2003e2006) from MedEcho and obtained the death certificates of those who died. To be included in the study, a person must have had cancer as the cause of death on his/her death certificate, or have had metastatic cancer according to MedEcho data, and a cause of death closely related to cancer determined from an algorithm that we previously validated.14
Measures Definition of HPCS. In Quebec, Canada, the population is entitled to complete health care coverage for physicians’ services and HPCS. The HPCS are usually provided through the CLSC of a patient’s neighborhood by either a specialized PC team including physicians, nurses, physical therapists, occupational therapists, social workers, psychologists, and personal physical care support or by some of these health care providers working on an ad hoc basis. The procedure to establish HPCS begins with a request to the CLSC from the patient, his family, or any health
50
Gagnon et al.
professional. After the evaluation of the patient’s needs, resources are dispatched. Medical support is provided by family physicians who usually do not have specific training in PC.15 Each CLSC is responsible for organizing HPCS delivery and managing community associations for a delimited geographic sector.16 The dates of the home PC visits by physicians and nurses to each decedent during the last three months of life were extracted from the CLSC database using the PC code and from the RAMQ database using the home services code. For the purpose of this study, any home PC visit by physicians and nurses defined HPCS, while recognizing that other health care professionals offer HPCS. Definition of Factors Associated with Home PC Visits. Patient demographics were extracted from the RAMQ database, and cause of death was extracted from the death certificate. Information from the MedEcho database for the preceding five years before death was used to extract comorbidities that could be associated with home health care services (congestive heart failure, cerebrovascular accident, peripheral vascular disease, dementia, chronic obstructive pulmonary disease, hemiparaplegia, and liver disease). Time at home for the last three months of life was defined as the number of days not spent in a hospital, community-based PC setting, or ER. People needed to be at home for at least seven days during this period to be included in the analysis. The number of admissions to the ICU and the number of blood product transfusions for the last three months of life were extracted from the MedEcho database, and for the same period, the number of chemotherapy treatments was extracted from the RAMQ database. Because we did not have the socioeconomic data of each person, a weighted average of the neighborhood characteristics of each CLSC was determined by linking the information available from Statistics Canada using the postal codes of their geographic sector. The social and material status for each neighborhood was estimated using the social and material deprivation indices developed and validated by Pampalon et al.17 In certain regions of Qu ebec, nonprofit organizations provide HPCS. To adjust for their association with HPCS delivery, a variable to identify the CLSCs covered by these organizations was created. Definition of QEoLPC Indicators. We used some of the QEoLPC indicators as defined by the ‘‘Institut National de Sant e Publique du Qu ebec’’18 and suggested in the 4e7 literature : place of death (home and ER), having at least one visit to the ER during the last 28 days excluding the one associated with the last admission leading to death, and having more than one hospitalization or
Vol. 50 No. 1 July 2015
spending more than 14 days in the hospital during the last month of life. People were considered to have died at home if the place of death on the death certificate was a private address. These specific QEoLPC indicators were selected because we hypothesized that they would be more responsive to HPCS.
Statistical Analysis As a high proportion of patients were not receiving any HPCS, this outcome was dichotomized into receiving any or not. To account for sex-specific cancer types, data from men and women were analyzed separately. The main exposure variable was the proportion of people receiving HPCS from their respective CLSC, which is called ‘‘performance of each CLSC in delivering HPCS’’ in this article. As such, multivariate gender-specific hierarchical risk ratio (RR) regression models (Fig. 1, Model A) were carried out to account for the cluster effect within each CLSC of all the covariates under study on the delivery of HPCS (lower box of Model A). The resultant of each model is an adjusted performance of each CLSC in delivering HPCS (expressed by the b estimate). Because time spent in the community was likely to influence the referral pattern for HPCS, proxy variables hypothesized to influence this pattern of referral were included in the model, such as time spent at home, ICU admissions, and chemotherapy treatment. In view of the complexity of the model, it was necessary to be parsimonious in the selection of variables to be included in the multiple gender-specific hierarchical RR regression models of the strength association between the adjusted performance of each CLSC in delivering HPCS and QEoLPC indicators. Therefore, a series of multiple gender-specific hierarchical RR regression models adjusting only for age and time spent at home were carried out for each covariate to select the ones with clinical relevance (with an RR value #0.90 or $1.10). Continuous covariates were categorized by regrouping their quartile values with similar RR (
