THE IMPACT OF PALLIATIVE CARE ON THE AGGRESSIVENESS OF END-OF-LIFE CARE IN PATIENTS WITH ADVANCED PANCREATIC CANCER

by

Raymond Woo-Jun Jang

A thesis submitted in conformity with the requirements for the degree of Master of Science Graduate Department of Institute of Medical Science University of Toronto

© Copyright by Raymond Woo-Jun Jang 2013

THE IMPACT OF PALLIATIVE CARE ON THE AGGRESSIVENESS OF END-OF-LIFE CARE IN PATIENTS WITH ADVANCED PANCREATIC CANCER

Raymond Woo-Jun Jang Master of Science Graduate Department of Institute of Medical Science University of Toronto 2013

Abstract Our objective was to examine the impact of palliative care (PC) on aggressive care near death for patients with advanced pancreatic cancer. Measures of aggressive care included (i) chemotherapy within 14 days of death; (ii) more than one emergency department (ED) visit; (iii) more than one hospitalization; and (iv) at least one intensive care unit (ICU) admission, all within 30 days of death. A retrospective population-based cohort study was conducted with patients diagnosed with advanced pancreatic cancer in Ontario. Multivariable logistic analyses were performed. Our final cohort consisted of 5,381 patients (median survival of 75 days). 52% received a PC consultation. PC consultation was associated with decreased use of chemotherapy near death (OR=0.34); and lower risk of ICU admission (OR=0.12), multiple ED visits (OR=0.19), and

ii

multiple hospitalizations near death (OR=0.24). A per unit increase in the monthly rate of PC visits was associated with lower odds of aggressive care.

iii

Acknowledgments Many thanks go to my supervisor Shabbir Alibhai. He has always been extremely supportive (going back to my medical school days!), and it seems like it was only yesterday that we hashed some ideas at ASCO that would eventually become this thesis. I’m very grateful to my fellowship supervisor Monika Krzyzanowska. Besides her great insight on this project, I owe much of my foundation as a medical oncologist to her mentorship. Years from now, when I am confronted with a difficult clinical dilemma, I will still think: “What would Monika do in this situation?” I would like to thank Camilla Zimmermann for bringing her palliative care expertise and for her continued mentorship and Nathan Taback for his statistical and programming expertise. Their inputs have greatly improved this work. Thank you to Craig Earle, Katrina Chan, and Refik Saskin (who patiently answered my endless questions) from ICES for their help through the cd-link program. Although not directly involved with the thesis work, I would like to thank some of the many physicians who have supported me during my training: Gary Naglie, Natasha Leighl, Vincent Chien, Robert Sargeant, John Bryson, Wayne Gold, David Warr, Scott Berry, and Malcolm Moore. Thanks to Alan Fung, Luke Tanur, and Jon Pipitone for the years of friendship. Thank you to my family for their support. Lastly, a heartfelt gratitude goes to my wife, Hyejin, for all her support and love. I couldn’t have done it without you.

iv

This thesis was supported in part by a Young Investigator Award from the Conquer Cancer Foundation of the American Society of Clinical Oncology (ASCO), an award from the University of Toronto Medical Oncology Postgraduate Ontario Student Opportunity Trust Funds (OSOTF), and the Division of Medical Oncology & Hematology, Princess Margaret Cancer Centre, University Health Network.

v

Table of Contents Acknowledgments  ...............................................................................................................................  iv   Table  of  Contents  .................................................................................................................................  vi   List  of  Tables  .........................................................................................................................................  ix   List  of  Figures  ........................................................................................................................................  xi   List  of  Appendices  ..............................................................................................................................  xii   Chapter  1  Introduction  .......................................................................................................................  1   1   Introduction  ....................................................................................................................................  1   1.1   Burden  of  pancreatic  cancer  .............................................................................................................  1   1.2   Quality  in  healthcare  ...........................................................................................................................  2   1.2.1   Oncology  quality  indicators  .........................................................................................................................  3   1.3   Aggressive  care  near  death  as  an  oncology  quality  indicator  ...............................................  4   1.4   Content  validity  of  indicators  of  aggressive  care  near  death  ................................................  5   1.4.1   Chemotherapy  near  death  ............................................................................................................................  5   1.4.2   ICU  admission  near  death  .............................................................................................................................  6   1.4.3   ED  visits  near  death  and  hospitalizations  near  death  ......................................................................  6   1.4.4   Other  indicators:  Hospice  use  near  death,  death  in  a  hospital  .....................................................  7   1.5   Construct  validity  of  indicators  using  administrative  databases  ........................................  8   1.6   Retrospective  studies  of  aggressive  care  near  death  ...............................................................  9   1.6.1   Comparison  between  jurisdictions  ........................................................................................................  11   1.6.2   Trends  over  time  in  aggressiveness  of  care.......................................................................................  11   1.6.3   Factors  that  impact  on  aggressive  care  near  death  ........................................................................  12   1.7   Impact  of  palliative  care  on  aggressive  care  near  death  ......................................................  13   1.7.1   Definition  of  palliative  care  .......................................................................................................................  13   1.7.2   Retrospective  studies  with  palliative  care  and  aggressive  care  ...............................................  13   1.7.3   Trials  with  palliative  care  and  aggressive  care  ................................................................................  14   1.7.4   Early  palliative  care  vs.  the  intensity  of  palliative  care  ................................................................  16   1.8   Prospective  cohort  studies  of  aggressive  care  near  death  ..................................................  16   1.9   Study  objectives  .................................................................................................................................  18  

Chapter  2  Methods  ............................................................................................................................  26   vi

2   Methods  ..........................................................................................................................................  26   2.1   Overview  of  study  design  ................................................................................................................  26   2.2   Data  sources  ........................................................................................................................................  26   2.3   Defining  a  cohort  of  patients  with  advanced  pancreatic  cancer  .......................................  28   2.4   Defining  palliative  care  ...................................................................................................................  29   2.5   Outcome  measures  ............................................................................................................................  31   2.6   Statistical  analysis  .............................................................................................................................  33   2.6.1   Model  diagnostics  .........................................................................................................................................  36   2.7   Ethics  and  confidentiality  ...............................................................................................................  37  

Chapter  3  Results  ...............................................................................................................................  43   3   Results  ............................................................................................................................................  43   3.1   Cohort  description  and  patient  demographics  .......................................................................  43   3.2   Survival  times  .....................................................................................................................................  44   3.3   Palliative  care  characteristics  .......................................................................................................  44   3.4   Frequency  of  aggressive  care  near  death  ..................................................................................  46   3.4.1   Frequency  of  aggressive  care  near  death  and  palliative  care  ....................................................  46   3.4.2   Frequency  of  aggressive  care  near  death  and  other  covariates  ...............................................  48   3.5   Chemotherapy  near  death  ..............................................................................................................  49   3.5.1   Chemotherapy  near  death  -­‐  Univariable  analysis  (Table  3.13  &  Table  3.14)  .....................  49   3.5.2   Chemotherapy  near  death  –  Multivariable  analysis  (Table  3.13  &  Table  3.14)  ................  50   3.5.3   Chemotherapy  near  death  (Chemotherapy  subcohort)  (Table  3.15)  ....................................  51   3.6   ICU  Admission  Near  Death  ..............................................................................................................  51   3.6.1   ICU  admission  near  death  -­‐  Univariable  analysis  (Table  3.16  &  Table  3.17)  ......................  51   3.6.2   ICU  admission  near  death  -­‐  Multivariable  analysis  (Table  3.16  &  Table  3.17)  ..................  52   3.6.3   ICU  admission  near  death  (Chemotherapy  subcohort)  (Table  3.18)  .....................................  53   3.7   Multiple  ED  Visits  Near  Death  .......................................................................................................  53   3.7.1   Multiple  ED  visits  near  death  -­‐  Univariable  analysis  (Table  3.19  &  Table  3.20)  ...............  53   3.7.2   Multiple  ED  visits  near  death  -­‐  Multivariable  analysis  (Table  3.19  &  Table  3.20)  ...........  54   3.7.3   Multiple  ED  visits  near  death  (Chemotherapy  subcohort)  (Table  3.21)  ...............................  54   3.8   Multiple  hospitalizations  near  death  ..........................................................................................  55   3.8.1   Multiple  hospitalizations  near  death  –  Univariable  analysis  (Table  3.22  &  Table  3.23)  55   3.8.2   Multiple  hospitalizations  near  death  –  Multivariable  analysis  (Table  3.22  &  Table  3.23)  

56   vii

3.8.3   Multiple  hospitalizations  near  death  (Chemotherapy  subcohort)  (Table  3.24)  ................  56   3.9   Model  Diagnostics  .............................................................................................................................  56   3.9.1   Model  Diagnostics  –  Model  predictive  ability  and  goodness  of  fit  ...........................................  56   3.9.2   Model  Diagnostics  –  Overfitting  ..............................................................................................................  57   3.9.3   Model  Diagnostics  –  Collinearity  ............................................................................................................  57   3.10   Interaction  between  palliative  care  and  age  .........................................................................  57   3.11   Adjustment  for  survival  duration  ..............................................................................................  58  

Chapter  4  Discussion  ........................................................................................................................  91   4   Discussion  .....................................................................................................................................  91   4.1   Major  findings  from  the  analysis  ..................................................................................................  91   4.1.1   The  elements  of  palliative  care  ...............................................................................................................  92   4.1.2   Early  palliative  care  ......................................................................................................................................  94   4.1.3   Intensity  of  palliative  care  .........................................................................................................................  96   4.1.4   The  use  of  composite  end  points  /  comparison  of  different  end  points  ...............................  97   4.2   Secondary  findings  ..........................................................................................................................  100   4.2.1   Age  and  aggressiveness  of  care  ............................................................................................................  100   4.2.2   Other  factors  that  impact  on  aggressive  care  ................................................................................  100   4.3   Survival  differences  with  palliative  care  .................................................................................  101   4.4   Major  limitations  .............................................................................................................................  101   4.5   Strengths  of  the  study  ....................................................................................................................  107   4.6   Future  research  and  questions  ...................................................................................................  108   4.7   Clinical  implications  .......................................................................................................................  110  

References  ..........................................................................................................................................  113   Appendices  ........................................................................................................................................  124  

viii

List of Tables Table 1.1: Summary of retrospective studies examining indicators of aggressive care near death ....................................................................................................................................................... 19   Table 1.2: Predictors of aggressive care in retrospective studies ................................................. 23   Table 2.1: Definitions of important variables ............................................................................... 38   Table 2.2: OHIP codes not used for the definition of palliative care ........................................... 41   Table 3.1: Most frequent secondary cancers in patients with pancreatic cancer .......................... 60   Table 3.2: Pancreatic cohort demographics .................................................................................. 61   Table 3.3: Survival duration of different subgroups ..................................................................... 63   Table 3.4: Palliative Care (PC) characteristics ............................................................................. 64   Table 3.5: Patient characteristics according to palliative care status ............................................ 65   Table 3.6: Unadjusted frequency of each outcome for all patients and by receipt of PC ............. 67   Table 3.7: Adjusted frequency of each outcome by receipt of PC ............................................... 67   Table 3.8: Unadjusted frequency of each outcome for all patients and by receipt of PC (Chemotherapy subcohort)............................................................................................................ 68   Table 3.9: Overlap of different measures of aggressive care near death ..................................... 69   Table 3.10: Frequency of aggressive care by intensity of palliative care ..................................... 70   Table 3.11 Frequency of aggressive care by different clinical characteristics for all patients ..... 71   Table 3.12: Summary of the effects of different predictors on aggressive care near death .......... 73   Table 3.13: Logistic Regression Analysis – Chemotherapy near death (Advanced cohort) ........ 74   Table 3.14: Logistic Regression Analysis – Chemotherapy near death (Intensity of PC) ........... 75   Table 3.15: Logistic Regression Analysis – Chemotherapy near death (Chemotherapy subcohort) ....................................................................................................................................................... 76   Table 3.16: Logistic Regression Analysis – ICU near death (Advanced cohort) ......................... 77   Table 3.17: Logistic Regression Analysis – ICU near death (Intensity of PC) ............................ 78   Table 3.18: Logistic Regression Analysis – ICU near death (Chemotherapy subcohort) ............ 79   Table 3.19: Logistic Regression Analysis – Multiple ED visits near death (Advanced cohort) .. 80   ix

Table 3.20: Logistic Regression Analysis – Multiple ED visits near death (Intensity of PC) ..... 81   Table 3.21: Logistic Regression Analysis – Multiple ED visits near death (Chemotherapy subcohort) ..................................................................................................................................... 82   Table 3.22: Logistic Regression Analysis – Multiple hospitalizations visits near death (Advanced cohort) ........................................................................................................................................... 83   Table 3.23: Logistic Regression Analysis – Multiple hospitalizations visits near death (Intensity of PC) ............................................................................................................................................ 84   Table 3.24: Logistic Regression Analysis – Multiple Hospitalizations near death (Chemotherapy subcohort) ..................................................................................................................................... 85   Table 3.25: C-statistics, heuristic shrinkage estimates, and Hosmer-Lemeshow statistics for the full models .................................................................................................................................... 86   Table 3.26: Impact of PC on the likelihood of ICU near death for different age groups ............. 86  

x

List of Figures Figure 1.1: Conceptual model of the trajectory of patients from diagnosis to death .................... 25   Figure 2.1: Time Frame Definitions ............................................................................................. 42   Figure 3.1: Flow diagram of pancreatic cancer patients ............................................................... 59   Figure 3.2: ROC curve for chemotherapy near death ................................................................... 87   Figure 3.3: ROC curve for ICU near death ................................................................................... 88   Figure 3.4: ROC curve for multiple Emergency Department visits near death ............................ 89   Figure 3.5: ROC curve for multiple hospitalizations near death .................................................. 90  

xi

List of Appendices Appendix 1: Data de-identification through cd-link ................................................................... 124   Appendix 2: Ethics approvals ..................................................................................................... 125   Appendix 3: Survival histograms (in days) ................................................................................ 127   Appendix 4: Frequency of aggressive care by different clinical characteristics for patients who received any PC consultation before the outcome of interest ..................................................... 128   Appendix 5: Multivariable logistic regression analyses with and without survival duration as a covariate ...................................................................................................................................... 130   Appendix 6: Variance Inflation Factors for the full models ....................................................... 131  

xii

1

Chapter 1 Introduction

1

Introduction

1.1 Burden of pancreatic cancer Pancreatic cancer is the 5th leading cause of cancer death in Canada, accounting for nearly 4,000 deaths each year (1). Surgery is the only curative treatment, but only around 20% patients are candidates for surgery because of the frequently advanced presentation at diagnosis. Even among those who have potentially curative surgical resection, the 5-year survival is only 20% (2). Treatment for advanced incurable pancreatic cancer consists largely of palliative chemotherapy, with radiation therapy being reserved for selected cases. The goals of palliative chemotherapy are to alleviate symptoms such as pain, increase quality of life, and if possible, increase survival. Median survival is 8 to 12 months for patients with locally advanced disease and only three to six months for those who present with metastases (3). As recently as 2011, single agent gemcitabine represented the standard of care and the only proven line of chemotherapy. Gemcitabine became the standard based on a modest survival benefit seen in a randomized controlled trial (RCT) over fluorouracil (5-FU) chemotherapy and because it conferred a clinical benefit, defined as a composite endpoint of measurements of pain (analgesic consumption and pain intensity), Karnofsky performance status, and weight (4). More recently, FOLFIRINOX (5-FU, leucovorin, irinotecan, and oxaliplatin) was approved for pancreatic cancer because of a substantial survival benefit compared to gemcitabine in a different

2

RCT (5). However, even in this trial, which generally included patients with the best performance status (and hence the best prognosis), the median survival for the FOLFIRINOX group was 11.1 months vs. 6.8 months for the gemcitabine group. Furthermore, the survival benefit of FOLFIRINOX was counterbalanced by the increased toxicity of this regimen (e.g. 12.7% vs. 1.8% of grade 3 or 4 diarrhea). Regardless of the type of treatment, pancreatic cancer patients also have a high burden of symptoms, including pain from the primary tumour or metastasis, gastric outlet obstruction, biliary tract obstruction, ascites, venothromboembolism, and malnutrition. Among patients with cancer, patients with pancreatic cancer are high users of the Emergency Department (ED), being second only to lung cancer in the frequency of ED visits in the last six months and last two weeks of life (6). Therefore, given that the majority of patients die in a short amount of time and with a high symptom burden, how these patients die, their quality of life near death, and the quality of care that they receive are all of utmost importance (1). For these reasons, pancreatic cancer is also a good prototype cancer to study the quality of care that patients receive near death.

1.2 Quality in healthcare Quality in health care has a specific meaning. The Institute of Medicine has defined quality as “the degree to which health services for individuals and populations increase the likelihood of desired health outcomes and are consistent with current professional knowledge” (7). In practical terms, poor quality has been defined to occur when “practices of known effectiveness are being underutilized, practices of known ineffectiveness are being overutilized,

3

and when services of equivocal effectiveness are being utilized in accordance with provider rather than patient preferences” (8).

1.2.1

Oncology quality indicators Quality indicators have been developed to measure the quality of care that patients

receive with the goal that they will lead to improvements in quality. Within oncology, quality indicators have been created in multiple areas, including core domains (e.g. a pathology report confirming malignancy on the chart), tumour-type specific domains (e.g. recommendation of trastuzumab for Her2 positive breast cancer), as well as end-of-life and symptom management domains (e.g. assessment of pain on either of the last two visits before death) (9–11). These quality indicators have been used for research purposes, but they have also been deployed in other ways. One example is the American Society of Clinical Oncology Quality Oncology Practice Initiative (ASCO QOPI), which is a practice-based quality assessment tool in the United States (12). Quality indicators have also been used from a health policy and planning perspective. For example, the province of Ontario has developed an evidence-based Cancer System Quality Index that tracks the province’s progress against cancer and points out where cancer service providers can make quality and performance improvements (13). These indicators are tracked at both the institution level and health region level.

4

1.3 Aggressive care near death as an oncology quality indicator More specifically, quality indicators near the end-of-life have been developed and studied. Of these, overly aggressive care for patients dying of cancer has been found to be an important area, starting from the late 1990s and early 2000s (10). Different concepts of aggressive care have been identified and studied, based on literature review, and focus groups with oncology health professionals, patients, and families (14,15). The indicators’ measurability within different jurisdictions has been evaluated (16,17). However, it was not until later that much of the evidence showing that aggressive care leads to worse quality of life for the patient and worse bereavement adjustment for their caregivers became available (18,19). Aggressive care near death has been measured and tracked in both the ASCO QOPI program (12) and Ontario’s Cancer System Quality Index (13). The most common indicators studied include the administration of chemotherapy near death, frequent ED visits near death, hospitalizations near death (including death in a hospital), and intensive care unit admissions near death. These indicators have been most often studied retrospectively using health administrative databases, and, to a lesser extent, prospectively using clinical trials or cohort studies. Table 1.1 lists all the papers that studied these outcomes retrospectively, including the slight differences in definition, and the prevalence of aggressive care. Other end-of-life quality indicators include the assessment of pain and symptoms (9,11,20), but will not be discussed further here.

5

1.4 Content validity of indicators of aggressive care near death The content validity of these indicators was initially based on a couple of studies using a modified Delphi approach among health professionals (N=8 and N=21) as well as focus groups with patients with advanced cancer and their families (N=16 and N=24) (14,15). Most of the health professionals were oncologists, palliative care physicians, and family physicians, but also social workers, nurses, and an infectious disease specialist.

1.4.1

Chemotherapy near death Chemotherapy near death was initially proposed as an indicator of poor quality care as

this may be a result of oncologists delaying discussion of prognosis and advance care directives, and choosing the easier route of ordering more chemotherapy (14). This indicator had some disagreement about both its meaningfulness and importance among both health professionals and patients. In one study, this indicator was ranked as the most important indicator by experts (14), while in a second study more disagreement was found among experts (15). Similarly, some patients and families thought that patients should have a right to choose more chemotherapy as long as they are informed about the usefulness (or lack thereof), while others thought that the continuation of chemotherapy until very late masked how advanced the disease was, leaving them unprepared for the eventual death (14). Continuation of chemotherapy near death has also been found to be inversely associated with home care, palliative care assessment, and physician house calls (21).

6

Chemotherapy near death has been usually defined as any chemotherapy within 14 days of death (see Table 1.1), but other definitions that have been used include chemotherapy within 30 days of death (22), or a new chemotherapy regimen started within 30 days of death (11,23).

1.4.2

ICU admission near death Broad agreement was found among the expert panels that ICU admission near death was

a meaningful indicator of poor quality care (14,15). Patients with advanced cancer have very poor survival following an ICU admission (24), have high symptom burden during the admission (25), have more emotional distress and worse quality of life as a result of the admission (19), and their caregivers have worse bereavement adjustment after an ICU admission (19). ICU admission near death has usually been defined as occurring within the last month of life (23,26), however a few studies have defined it as occurring within the last 2 weeks of life (21,27)

1.4.3

ED visits near death and hospitalizations near death High agreement was found among experts that frequent ED visits near death was an

important and meaningful indicator (14,15). Frequent ED visits near death are thought to represent poor quality care because they are thought to be a result of lack of planning, lack of symptom control, and/or lack of resources (14). A previous study in Ontario showed that many ED visits near death may have been avoidable in patients who died of cancer (6).

7

The two studies that measured the importance of end-of-life quality indicators among different stakeholders did not comment specifically on the importance of frequent hospitalizations near death (14). One study grouped frequent hospitalizations near death with frequent ED visits near death (14). Given that most patients with advanced cancer who visit ED near death are admitted (6), the assumption that the root causes of frequent hospitalizations near death are likely to be the same causes as frequent ED visits is likely a sound one. Frequent ED visits near death are usually defined as more than one ED visit in the last month of life, although some studies have defined it as at least one ED visit in the last two weeks of life (21,27,28) or at least one ED visit in the last month of life (29).

1.4.4

Other indicators: Hospice use near death, death in a hospital The other indicator that has been frequently studied is that of death in a hospital. This is

predicated on the fact that the majority of patients want to die at home (30). High rates of death in hospital was deemed an important indicator by one set of experts in the United States (14), while there was more disagreement among another group of experts in Canada (15). The latter opinion was a reflection of the belief that patient preferences should inform end-of-life decisions such as place of death (15). However, more recent literature suggests that a hospital death leads to more caregiver psychiatric illness (19). In operationalizing the definition, some studies have used death as an inpatient in a hospital (22,29) whereas other studies have used death in an acute care bed (27,28,31).

8

Other indicators that have been used less frequently include home care use (28), physician house calls (28), and hospice use near death (14,23). As Canada does not have a comprehensive hospice system compared to the United States, this last indicator cannot be used in Canada.

1.5 Construct validity of indicators using administrative databases Validation studies to examine the accuracy of administrative databases have been performed previously, both with American and Canadian (Ontario and Nova Scotia) databases. In both studies encompassing the three jurisdictions, a high agreement rate was found between the administrative database source (e.g. chemotherapy billings) and the medical record (considered to be the gold standard) (16,17). Specifically, in Ontario, for three of the indicators used in our study, the percentage agreement rates for chemotherapy within 14 days of death, multiple ED visits within 30 days, and multiple hospitalizations within 30 days were: 0.95, 0.87, 0.77 (17). The databases used in Ontario were the same as the ones used in our study. Agreement rates for ICU admission near death were not available in this study. The American study found that administrative databases were generally accurate when compared to the medical charts at one teaching hospital, although administrative databases tended to underestimate aggressive care (16).

9

1.6 Retrospective studies of aggressive care near death The various retrospective studies of aggressive care near death differed with respect to the number of patients, tumour sites included, location of study, goals of study, and operationalization of the definitions of aggressive care (see Table 1.1). As all but two of these studies were based on administrative data, the number of patients included was large. The smallest population-based study included more than 2,000 patients (28), and the largest study had over 30,000 patients (11). The papers that relied on chart abstraction had 298 patients (32) and 319 patients (33), respectively. Two studies included all cancer types (21,26), while the rest included either only one tumour site or a group of sites. Only one paper focused on pancreatic cancer exclusively (22). None of the papers that included pancreatic cancer patients reported on the outcomes of pancreatic cancer separately. No paper focused exclusively on those with hematologic cancers. All the studies were either from the United States or Canada, with the exception of one Korean (32) and one Portuguese chart review papers (33). All American studies, except for one, included patients over the age of 65, as they relied on Medicare data. The sole exception was an American study that relied on the participation and data abstraction from medical oncology practices enrolled in a continuous quality improvement program (9). The majority of studies did not use a cohort of patients with advanced cancer who had died, but rather a decedent cohort of patients with cancer. This study design has been criticized, as some of the patients who died may not have been those dying of advanced cancer, but rather died because of treatment complications during adjuvant treatment or due to a completely unrelated reason (34). For this reason, one of the more recent studies used a cohort of patients

10

with advanced cancer (35), while a different study did separate analyses for patients with metastatic pancreatic cancer (22). In calculating the frequencies of the indicators of aggressive care, different studies used different denominators. Some studies included in the denominator all patients who died after being diagnosed with the cancer of interest (9,11,22,35), while others required that the cause of death be cancer (21,26), or even more strictly that the cause of death was the cancer of interest (17,23,27). For the outcome of ED visits near death, some studies required the subtraction of patients who had been hospitalized for the entire month before death, as these patients would not have had the chance to have an ED visit (21,28,29). The goal of many of the retrospective studies was to describe the frequencies of each indicator, thereby providing a “snapshot” of the aggressiveness of care within a jurisdiction. The frequency of chemotherapy within 14 days of death ranged from 2% to 8% for most studies, although some American studies had higher rates, especially if the denominator included only patients who had received some chemotherapy (15.7%). The frequency of ICU admission within 1 month of death ranged from 2% to 8% in Canada while it was 9 to 20% in the United States. The frequency of multiple ED visits within 1 month of death ranged from 5% to 15% in North America, although it was 33.6% in a Korean study. Lastly, the frequency of multiple hospitalizations within 1 month of death ranged from 7% to 16%. However, other studies have gone beyond the reporting of frequencies to examine differences between jurisdictions, trends over time, and multivariable analysis to determine factors that impact on the aggressiveness of care. No study has previously looked at whether a single predictor is associated with a higher frequency of aggressive care using retrospective data.

11

1.6.1

Comparison between jurisdictions Two studies compared the care in different jurisdictions (17,29), although for different

purposes. The purpose of one study was to simply demonstrate the feasibility of measuring different indicators using the provinces’ databases (17), while the purpose of the other study was to see if differences in the structure of health care between the United States and Ontario impact on the aggressiveness of care provided (29). Chemotherapy near death was more frequent, but rates of hospitalization were lower in the United States than Ontario (29). The authors cited differences in the perceived benefit of chemotherapy, the financial incentives for administering chemotherapy, and the availability of hospice care as possible reasons for the discrepant rates. Within the United States, a different study specifically examined the issue of payment incentives by studying a cohort of older patients with metastatic colorectal and lung cancer treated in the Veterans Health Administration (VHA) vs. the private sector, and found that patients treated in the VHA received less aggressive care near death (35).

1.6.2

Trends over time in aggressiveness of care Two studies have examined how end-of-life cancer care has changed over time (23,26).

The first study examined the Medicare claims for end-of-life care for older patients who died between the years of 1993 to 1996. It found that the frequency of aggressive care near death increased during this time period: chemotherapy 13.8% to 18.5%; ICU admissions 7.1% to 9.3%; multiple ED visits 7.2% to 9.2%; multiple hospitalizations 7.8% to 9.1%; all p-values < 0.05) (23). The second study reached similar conclusions in Ontario, Canada between the years of

12

1993 and 2004, although the magnitude of increasing aggressiveness of care with each successive year was quite small (OR 1.01; 95% CI 1.01-1.02) (26).

1.6.3

Factors that impact on aggressive care near death Three retrospective studies have examined the different factors that predict for aggressive

care (21,23,26). Of these, only one study examined the impact of palliative care, and will be discussed later in this chapter. Table 1.2 lists the predictors and their effects on aggressive care. Two of the studies used a composite outcome for aggressive care (23,26). However, a composite outcome may not be appropriate for certain predictors. For example, rural patients were found to have more frequent ED visits near death but fewer ICU admissions near death in one study (21), and this may be due to the availability of fewer resources to prevent ED visits and also fewer ICU beds in rural areas. Older age (OR=0.98, 95% CI 0.97-0.99 and OR=0.97, 95% CI 0.97-0.99 for each subsequent year) (23,26) and female sex (OR=0.80, 95% CI 0.73-0.87) (23) were both associated with less aggressive care. Higher comorbidity (OR=1.14; 95% CI 1.06-1.23 and OR=1.25, 95% CI 1.24-1.26 for each point in the Charlson score) (23,26) was associated with more aggressive care. Many predictors were only available in one study, and therefore it was difficult to generalize across different studies (Table 1.2).

13

1.7 Impact of palliative care on aggressive care near death 1.7.1

Definition of palliative care Palliative care has been defined by the World Health Organization (WHO) as “an

approach that improves the quality of life of patients and their families facing the problem associated with life-threatening illness, through the prevention and relief of suffering by means of early identification and impeccable assessment and treatment of pain and other problems, physical, psychosocial and spiritual” (36). More specifically, specialized palliative care service has been defined as a service of professionals that provides or coordinates comprehensive care for patients with a terminal illness (37). Because of its patient-centred approach, palliative care may be an ideal intervention to reduce aggressiveness of care near the end-of-life. A conceptual model of the trajectory of patients with advanced pancreatic cancer from diagnosis to death, and the role of pallitiave care in decreasing aggressive care near death is shown in Figure 1.1. For the purposes of this thesis, a palliative care consultation and/or visit requires the assessment of a physician for the purposes of providing palliative care as defined by the WHO.

1.7.2

Retrospective studies with palliative care and aggressive care Only one previous study has studied the possible impact of palliative care on

aggressiveness of care using administrative databases. In that Ontario study, all patients who died of a cancer diagnosis in 2001 were included (21). Three outcomes of aggressive care (chemotherapy, ICU admission, and at least one ED visit, all within 14 days) were studied. Palliative care within 14 days of death was associated with less aggressive care (OR=0.562, 95% CI 0.476-0.662, for chemotherapy; OR=0.603, 95% CI 0.548-0.662, for ED visits; and

14

OR=0.774, 95% CI 0.674-0.889, for ICU admissions). No specifications were made as to whether the palliative care was before or after the outcome. Palliative care was defined using Ontario Health Insurance Program (OHIP) physician billing information. Palliative care consultation has been defined in only one other study in Ontario using administrative databases (29), largely because in the United States there is no Medicare billing code for palliative care services (29). No validation studies were found on the use of palliative care using administrative databases.

1.7.3

Trials with palliative care and aggressive care Previous trials have reported on the positive impact of palliative care on symptom control

(38,39), quality of life (39,40), patient satisfaction (38,39,41,42), and mood in cancer patients (38,40). However, some of the trials have been hampered by methodological problems as outlined in a systematic review of the effect of palliative care (37). Studies done on the impact of palliative care on health resource utilization have been mostly retrospective single institution reviews, showing a decrease in acute care hospitalizations (43,44), including ICU admissions (45,46). One prospective trial that studied health resource utilization showed no difference in the rate of re-hospitalizations (after the index hospitalization) but showed decreased subsequent ICU admissions (42). Unlike aggressiveness of care studies, the studies that studied health resource utilization did not specify the time frame of ICU admission (e.g. within 1 month of death). Only one prospective randomized controlled trial has examined the impact of palliative care on aggressiveness of care near the end of life (40). The trial used early palliative care as an intervention for a cohort of patients with metastatic lung cancer. The cohort of 151 patients from a single American academic centre was randomized to early palliative care (defined as within 2

15

months of diagnosis of metastatic lung cancer and with an ECOG performance status of 2 or better) or to standard oncologic care. The end-points were quality of life, mood, and health care use including aggressiveness of care. Patients were deemed to have received aggressive care if they met any of the following three criteria: chemotherapy within 14 days before death, no hospice care, or admission to hospice 3 days or less before death. Early palliative care increased quality of life, improved mood, and was associated with less aggressive care (33% vs. 54%, P = 0.05), but increased survival (11.6 months vs. 8.9 months, P = 0.02). Although not used to define aggressive care, indicators such as any ED visit within the last month of life and any hospitalization within the last month of life were reported in the appendix. The authors reported that the numbers were too small to have adequate power, but the reported differences in frequencies were strongly suggestive. Patients in the early palliative care group received less chemotherapy within 14 days of death [17.5% (7/40) vs. 24% (12/50)], had fewer ED visits within 30 days of death [22.4% (11/49) vs. 30.4% (17/56)], and had fewer hospitalizations within 30 days of death [36.7% (18/49) vs. 53.6% (30/56)]. Other trials with palliative care have been recently summarized in an ASCO provisional guideline of palliative care (47). Most of these trials had a mixed population of cancer and noncancer patients. Of the studies that used a mixed population, one of them showed that an interdisciplinary palliative care service (IPCS) comprising of a palliative care physician, nurse, hospital social worker, and chaplain, led to more advance care directives at hospital discharge (91.1% v 77.8%, P= 180 days

131

Appendix 6: Variance Inflation Factors for the full models     (Advanced cohort; N=5,831)

Chemotherapy ICU near near death death PC Age 50-59 60-69 70-79 80-high Sex Male Rural Primary care physician Med onc consult Charlson index 0 1 2+ Chemotherapy Income Quintile 1 2 3 4 5 Year of Diagnosis 2005 2006 2007 2008 2009 2010 LHIN 1 2 3 4 5 6 7

1.060935

1.040347

Multiple ED visits near death 1.070272

Multiple hospitalizations near death 1.070750

2.717493 3.030575 3.094649 1.520561

2.776048 3.727642 4.177383 3.377713

3.583648 4.669290 5.119883 5.016516

3.047960 3.947955 3.964093 3.792707

1.027090 1.175227 1.028356

1.051465 1.186973 1.028271

1.055073 1.248581 1.019335

1.052808 1.231012 1.022567

---

1.428539

1.393013

1.408992

1.044161 1.047864 ---

1.088440 1.116007 1.387853

1.059298 1.082016 1.377980

1.055986 1.065929 1.390590

1.740956 1.774838 1.864543 1.848892

1.636648 1.615005 1.612853 1.603152

1.569148 1.549180 1.547958 1.521195

1.627202 1.557393 1.590402 1.475003

1.811410 1.757633 1.911529 1.864856 1.655270

1.749238 1.727388 1.851993 1.717023 1.575211

1.630689 1.776237 1.738927 1.685362 1.577981

1.558565 1.693713 1.652915 1.567240 1.515595

4.299925 4.979032 3.386583 6.227408 2.641817 3.420828

1.756046 2.161840 1.378349 2.966967 2.053098 1.913991

1.831800 2.579923 2.523655 2.495585 1.739044 1.917434

1.711736 2.471542 2.422859 2.432018 1.715885 1.860033

132

8 9 70 74 76 77 79

 

4.052763 5.282083 1.612640 3.406229 3.358936 3.017268 1.963277

2.077770 2.883217 1.783311 1.603665 1.452014 2.511978 1.465964

2.063695 2.832669 2.126371 1.629459 2.185628 2.223963 1.477575

1.887515 2.419702 1.867272 1.654276 2.122680 1.838259 1.609078