Effect of integrated palliative care on the quality of end-of-life care: retrospective analysis of 521 cancer patients

Transcription

1 Effect of integrated palliative care on the quality of end-of-life care: retrospective analysis of 521 cancer patients Isabelle Colombet, 1,2,3 Vincent Montheil, 3 Jean-Philippe Durand, 4 Florence Gillaizeau, 1,2 Ralph Niarra, 2 Cécile Jaeger, 3 Jérôme Alexandre, 1,4 François Goldwasser, 1,4 Pascale Vinant 3 An additional appendix is published online only. To view this fi le please visit the journal online ( content/early/recent). 1 Université Paris Descartes, Sorbonne Paris Cité, Public Health Paris, France 2 INSERM, Centre d Investigation Épidémiologique 4, Paris, France 3 Palliative Medicine, Cochin Teaching Hospital, AP-HP, Paris, France 4 Medical Oncology, Cochin Teaching Hospital, AP-HP, Paris, France Correspondence to Isabelle Colombet, Unité Fonctionnelle de Médecine Palliative, Hôpital Cochin, 27, Rue du Faubourg Saint- Jacques, PARIS Cedex 14, Paris, France; isabelle. colombet@parisdescartes.fr Received 17 October 2011 Accepted 24 April 2012 ABSTRACT Objective To examine the impact of oncologist awareness of palliative care (PC), the intervention of the PC team (PCT) and multidisciplinary decision-making on three quality indicators of end-of-life (EOL) care. Setting Cochin Academic Hospital, Paris, Design and participants A 521 decedent case series study nested in a cohort of 735 metastatic cancer patients previously treated with chemotherapy. Indicators were location of death, number of emergency room (ER) visits in last month of life and chemotherapy administration in last 14 days of life. Multivariable logistic regression models were used to estimate associations between indicators and oncologist s awareness of PC, PCT intervention and case discussions at weekly onco-palliative meetings (OPMs). Results 58 (11%) patients died at home, 45 (9%) in an intensive care unit or ER, and 253 (49%) in an acute care hospital; 185 (36%) patients visited the ER in last month of life and 75 (14%) received chemotherapy in last 14 days of life. Only the OPM (n=179, 34%) independently decreases the odds of receiving chemotherapy in last 14 days of life (OR 0.5, 95% CI 0.2 to 0.9) and of dying in an acute care setting (0.3, 0.1 to 0.5). PCT intervention (n=300, 58%) did not independently improve any indicators. Among patients seen by the PCT, early PCT intervention had no impact on indicators, whereas the OPM reduced the odds of persistent chemotherapy in the last 14 days of life. Conclusion Multidisciplinary decision-making with oncologists and the PCT is the most critical parameter for improving EOL care. INTRODUCTION Considerable attention is currently being focused on the impact of the integration of palliative care (PC) in oncology. 1 In cancer centres, the late referral of patients to specialist PC reflects the persistent barriers to integrated PC in oncology. 2 Temel et al showed that early PC consultation, systematically offered to patients diagnosed with advanced non-small cell lung cancer (NSCLC), increased both their survival and quality of life. 3 This study marks a turning point for the integration of PC into oncology care and highlights the value of the early introduction of the PC team (PCT). However, the factors accounting for these results are unclear. Various models covering the organisation, intervention and interface of specialist PC with standard care have been developed across different countries. Efforts to systematically review the evidence for the effectiveness of PC services led to ambivalent conclusions, mainly due to the inadequate internal and external validity of the original studies. 4 6 In their study focusing on the effectiveness of specialist palliative care teams for people with advanced cancer, Higginson et al list the wide range of services tested and conclude that future studies need to evaluate more specific components of PCT activity. 4 In France, most developed PC services remain hospital based, with inpatient PC centres or hospital PCTs. These teams act as staff counsellors at the patient s bedside in end-of-life (EOL) situations. They concentrate their efforts on guiding the medical and nursing staff directly in charge of patients. In oncology, this mode of intervention respects the oncologist s role, but its impact on the patient is indirect and highly dependant on the degree of organisation and collaboration between both specialties. In our institution, the PCT has developed a collaborative relationship with the oncology ward in order to reduce persistent barriers and improve the quality of EOL care. This collaboration is based on a weekly onco-palliative meeting (OPM), attended by both the PCT and oncology staff, who discuss the cases of any advanced cancer patients who could benefit from joint follow-up care. We hypothesised that this OPM, as a key component of the activity of the PCT, may be as important as on-demand intervention by the PCT to improve indicators of the quality of EOL care. Many indicators have been proposed to measure the quality of the outcomes and processes of EOL care (ie, the use of health services resources). 7 In order to estimate the impact of the different methods of introducing PC in the trajectory of cancer patients, we chose process of care indicators. Those developed and evaluated by Earle et al address three aspects of care 8 : overuse of aggressive treatment, unplanned medical encounters and poor access to PC. In the present study, we evaluated the impact of the oncologist s awareness of PC, the clinical intervention of the PCT and its timing, and multidisciplinary decision-making on the EOL care indicators described by Earle et al. 8 Colombet I, Montheil V, Durand J-P, et al. BMJ Supportive & Palliative Care (2012). doi: /bmjspcare of 9

2 METHODS Design and setting Cochin Hospital is a tertiary care hospital treating 6000 new cancer patients each year, with an oncology department and three other medical specialty wards (gastroenterology, pneumology, dermatology) providing chemotherapy. We conducted a decedent case series study, using a mortality follow-back approach. Decedent cases were identified by following up of all eligible patients recorded in the hospital activity database. Study population All patients diagnosed with a solid tumour and having received their last administration of intravenous chemotherapy at Cochin Hospital between 1 June 2006 and 31 December 2008 were considered eligible for inclusion in our case series. These patients were identified from the electronic record database (CHIMIO) routinely used to manage patients receiving intravenous chemotherapy. Patient vital status was identified in three steps. We first cross-matched the CHIMIO database of eligible patients with the discharge data ( ) of 30 acute care hospitals in the Paris area. Then, for patients not recorded as having died before discharge, we searched these patients medical records for a posteriori notification of death. Finally, we completed the follow-up with requests to civil administrative services and telephone interviews with the patient s referring primary care physician. All patients who died during the study period were included. Processes of care Referent physician Chemotherapy is prescribed by either the five physicians attached to the oncology ward or the 10 specialists from other wards qualified in oncology. None of these physicians are involved in PC practice, but some of them have attended a training program introducing the fundamentals of PC, thereby increasing their awareness of PC services. The patient is then recorded in the CHIMIO database. All 15 prescribers are referred to below as the referent physician. Multidisciplinary decision-making through the OPM In the oncology ward, there is a specific process of care dedicated to patients whose life expectancy is less than 6 months, as estimated by their referent physician, who are still receiving chemotherapy and for whom it is necessary to discuss goals of care. Provision of this care relies on a weekly multidisciplinary OPM. It is attended by both the PCT and the oncology staff, that is physicians, nurses and head nurses, social workers, psychologists and secretaries. The OPM is led by a physician from the PCT, who also records its conclusions in the patient s record. Discussions take into account expected survival benefit of treatment and quality of life, aggressiveness of care and the patient s preferences. Decisions may be to pursue chemotherapy, associated or not with the intervention of the PCT, to propose inclusion in a phase 1 or 2 clinical trial, or to provide PC only. These decisions are then passed to patient for consideration. The patients are followed-up by the referent physician, or jointly by the PCT if decided, through outpatient consultations in a day care setting, or by admission to an acute care setting, as needed. For all patients discussed, the goals and organisation of care are updated at following OPMs, until the patient s death, thus making possible a systematic analysis of practice. Intervention of the PCT The PCT consists of 2.5 full time equivalent physicians, all PC specialists, plus 2.5 full time equivalent nurses, one secretary assistant and one attending psychologist. There are two routes of referral to the PCT. The first is the standard method where the team intervenes following a request by a referent physician from a clinical ward, according to their judgement and without a specific dedicated time for multidisciplinary decision-making. The second method is particular to the oncology ward, where the referent physicians are asked to use the OPM to discuss PCT intervention, except in the case of emergency (ie, acute pain or symptom, or rapid, unanticipated worsening of disease). Whatever the method of referral, the intervening PCT addresses symptom management, EOL care decisions, coping with a life-threatening illness, patient understanding of the illness and its prognosis, organisation of referral to a hospice, coordination with home care services and networks, and outpatient follow-up. Quality indicators We evaluated five indicators based on those elaborated and validated by Earle et al 8 9 : location of death (acute care setting versus home or hospice), number of emergency room (ER) visits in the month preceding death (at least one), the interval between the last administration of chemotherapy and death (less than 14 days), the number of admissions to the intensive care unit (ICU) and the number of acute care admissions, both during the month preceding death. We analysed the first three indicators as outcome dependent variables and the last two as confounding variables. History and process of care characteristics Other data collected included age, gender, cancer primary site and three process of care characteristics: training of the referent physician in PC (reflecting his/her actual degree of awareness of PC), whether or not the patient s case was discussed at the OPM, and whether or not the PCT intervened. For the subgroup of patients who were followed by the PCT, we recorded two other characteristics: performance status (PS) according to the Eastern Cooperative Oncology Group (ECOG) scale 10 at the time of first intervention, and the time interval between the first intervention of the team and death. History of care data and quality indicators were collected from computerised clinical databases and completed by a search of paper records and an interview with referent physicians. Statistical analysis Location of death (at the hospital (excluding a hospice) versus another location), referral to the emergency services (one or more referrals versus none) and patients receiving chemotherapy in the last 14 days of life were analysed using logistic regression models, both for the whole population (tables 2, 4 and 6) and for the subgroup of patients followed by the PCT (tables 3, 5 and 7). For the whole population, we measured the association of indicators with the process of care characteristics (referent physician trained in PC, patient s case discussed at the OPM and PCT intervention) in both univariable and multivariable analyses using ORs with 95% CIs. In the multivariable analyses, ORs were adjusted for the patient s characteristics (age, sex, primary cancer site). For the analysis of the location of death, ORs were also adjusted for survival after last administration of chemotherapy, ER visit, ICU 2 of 9 Colombet I, Montheil V, Durand J-P, et al. BMJ Supportive & Palliative Care (2012). doi: /bmjspcare

3 admission and acute care admission. For analysis of referral to emergency services, ORs were also adjusted for survival after last administration of chemotherapy, ICU admission and acute care admission. In the subgroup of patients followed up by the PCT, we measured the association of indicators with PS at the first intervention, the time interval between the first intervention and death, the OPM and if the referent physician had previously trained in PC. In the multivariable analyses, ORs were adjusted for the same variables as in the respective models for the whole population. For continuous variables, the linearity of log-odds was checked graphically. The relationships between the quality indicators and the collinearity between categorical variables were checked using Cramer s V coefficient (a measure of association derived from the Pearson χ 2 test). A p value <0.05 was considered significant. The goodnessof-fit of the logistic multivariable regression models was tested using the Hosmer Lemeshow test (small p values indicate a lack of fit of the model). The statistical software SAS release 9.1 was used for all analyses. RESULTS Study population We identified 735 advanced cancer patients who received chemotherapy during the study period (figure 1): 105 (14%) were ineligible because they had not died by the end of the defined study period or because they were followed up and cared for in another hospital shortly after diagnosis (ie, patients choosing to be followed up in another cancer centre). For 109 other patients (14.8%), whose vital status could not be determined from the hospital records, their place of birth was either outside France (n=42) or not included in the hospital admission data (n=67), thus precluding any requests to the civil services concerning their vital status. Consequently, we analysed 521 patients whose characteristics and history of care are described in table 1. A flow chart detailing their trajectory of care is provided in the online supplementary appendix. Description of indicators Location of death was their home for 58 (11%) patients, a hospice for 128 (25%), an ER or ICU for 45 (9%) and acute care wards for 253 (49%). Among the 58 patients who died at home, 17 (34%) died less than 7 days after their last hospital discharge. Of the 128 patients who died in a hospice, 14 (11%) died less than 3 days after admission. Of the 253 patients who died in acute care wards, 126 (42%) had been admitted after visiting the ER (unscheduled hospitalisation). In their last month of life, 185 (36%) patients visited the ER at least once, 49 (10%) were admitted to the ICU, 242 (46%) were hospitalised once in an acute care setting and 97 (19%) were hospitalised twice or more in an acute care setting. Median survival after the last administration of chemotherapy was 55 days (IQR ); 75 (14%) patients received chemotherapy in their last 14 days of life. Description of history of care Patients were followed by 15 referent physicians. Two of these physicians had undertaken PC training and referred 101 and 83 patients, respectively (35% overall), whereas 13 doctors with no PC training referred a median of 16 patients each (min max 1 80). The PCT collaborated with the care of 67/101 (66%) and 51/83 (61%) patients referred by the two PC qualified Table 1 Patient characteristics and history of care All (N=521) Mean age at death, years (SD) 64 (13) Sex, n (%) Male 326 (62.6) Female 193 (37.0) Missing data 2 (0.4) Primary cancer site, n (%) Gastro-intestinal 110 (21) Lung 107 (21) Urological 67 (13) Melanoma 67 (13) Liver and biliary tract 52 (10) Other 118 (23) Location of death, n (%) Acute care hospital 253 (49) Hospice 128 (25) Home 58 (11) Emergency room or ICU 45 (9) Other 34 (7) Missing data 3 (1) Process of care characteristics, n (%) Referent oncologist trained in palliative care Yes 184 (35) No 337 (65) Palliative care team intervention Yes 300 (58) No 221 (42) If yes, patient performance status* at inclusion 1 38 (13) 2 72 (24) 3 78 (26) 4 43 (14) Missing data 69 (23) If yes, median survival from inclusion, 51 (17 114) days (IQR) Onco-palliative meeting Yes 179 (34) No 341 (66) Trajectory of care (indicators) Median survival after last chemotherapy, days (IQR) 55 (25 109) Patients receiving chemotherapy in the last 75 (14) 14 days of life, n (%) Emergency room visits, in last 30 days, n (%) (63) (31) 2 25 (5) Missing data 7 (1) Admissions to ICU in last 30 days, n (%) (88) 1 40 (8) 2 9 (2) Missing data 13 (2) If 1, median length of stay, days (IQR) 5 (3 11) Admissions to acute care setting in 30 last days, n (%) (24) (46) 2 97 (19) Missing data 59 (11) If 1, median length of stay, days (IQR) 14 (8 22) *Performance status, as measured by the Eastern Cooperative Oncology Group (ECOG) scale 11 ( 0, fully active, able to carry on all predisease performance without restriction; 1, restricted in physically strenuous activity but ambulatory and able to carry out work of a light or sedentary nature, for example, light house work, offi ce work; 2, ambulatory and capable of all self-care but unable to carry out any work activities. Up and about more than 50% of waking hours; 3, capable of only limited self-care, confi ned to bed or chair for more than 50% of waking hours; 4, completely disabled. Cannot carry on any self-care. Totally confi ned to bed or chair. ICU, intensive care unit. Colombet I, Montheil V, Durand J-P, et al. BMJ Supportive & Palliative Care (2012). doi: /bmjspcare of 9

4 Table 2 Location of death Univariable analysis Hospital (excluding hospice) (n=332) Other (n=186) OR (95% CI) p Value Adjusted OR (95% CI) p Value Hosmer Lemeshow 0.44 goodness-of-fi t test Mean age at death, years (SD) 64 (13) 64 (14) 0.99 (0.87 to 1.14)* (0.83 to 1.21)* 0.98 Sex, n (%) Female 127 (38) 66 (35) Male 204 (61) 119 (64) 0.9 (0.6 to 1.3) (0.4 to 1.1) Primary cancer site, n (%) Global p value 0.08 Global p value 0.90 Melanoma 48 (15) 19 (10) Gastro-intestinal 75 (23) 34 (18) 0.9 (0.4 to 2.0) (0.3 to 2.9) 1.00 Lung 69 (21) 38 (20) 0.7 (0.3 to 1.6) (0.3 to 3.4) 0.99 Urological 33 (10) 33 (18) 0.4 (0.2 to 0.9) (0.4 to 5.5) 0.79 Other 107 (32) 62 (33) 0.7 (0.3 to 1.5) (0.4 to 3.2) 0.99 Process of care characteristics, n (%) Referent oncologist trained in palliative care No 237 (71) 99 (53) Yes 95 (29) 87 (47) 0.5 (0.3 to 0.7) < (0.6 to 2.0) 0.72 Palliative care team intervention No 152 (46) 67 (36) Yes 180 (54) 119 (64) 0.7 (0.5 to 1.0) (0.4 to 1.3) 0.25 Onco-palliative meeting No 259 (78) 81 (44) Yes 73 (22) 105 (57) 0.2 (0.2 to 0.3) < (0.1 to 0.5) <0.001 Trajectory of care (other indicators Global p value <0.001 Global p value 0.18 considered for confounding adjustment), n (%) Survival after last chemotherapy, <14 days 62 (19) 13 (7) days 55 (17) 18 (10) 0.6 (0.3 to 1.6) (0.2 to 1.8) days 96 (29) 35 (19) 0.6 (0.2 to 1.3) (0.3 to 2.5) days 119 (36) 120 (65) 0.2 (0.1 to 0.4) < (0.2 to 1.3) 0.20 Emergency room visit (at least one in last 30 days) No 181 (55) 147 (79) Yes 147 (44) 38 (20) 3.1 (2.1 to 4.8) < (0.8 to 9.8) Intensive care unit admission (at least one in last 30 days) No 277 (83) 181 (97) Yes 46 (14) 3 (2) 10.0 (3.1 to 32.7) (0.1 to 1.3) Acute care admission Global p value <0.001 Global p value <0.001 (in last 30 days) None 27 (8.1) 95 (51) admission 184 (55) 58 (31) 11.2 (6.6 to 18.8) < (4.2 to 16.4) < admissions 86 (26) 11 (6) 27.5 (12.9 to 58.8) < (5.9 to 44.6) <0.001 *For 10 years increase in age. The null hypothesis cannot be rejected, indicating that there is no difference between the observed and predicted values of the response variable (a small p value indicates that the model does not fi t the data). OR for death at hospital (excluding a hospice) versus other. For multiple comparisons testing difference with the reference category (primary cancer site, survival after last chemotherapy, acute care admission), p values were adjusted using Dunnett correction. Reference category for univariable and multivariable logistic regression. For cancer site, we chose the cancer site with the lowest risk of dying at home or in a hospice as the reference category in order to facilitate interpretation (OR >1). In univariable analysis, this category was melanoma and we used it as the reference category for all analyses. physicians versus 182/337 (54%) patients referred by the other 13 doctors. Among the 300 patients followed by the PCT, PS was recorded in 231, with 110 (48%) having a PS of 1 or 2. Median survival after the first referral was 51 days. The first contact with the PCT occurred in the last month of life in 108 (40%) patients. Patients discussed at the OPM (143/300) had a better PS than those not discussed (157/300): PS was 2 in 74/143 (52%) patients discussed at the OPM versus 36/157 (23%) not discussed at the OPM (p<0.001, χ 2 test). Discussion at the OPM resulted in earlier intervention by the PCT: median survival was 75 days (IQR ) in discussed patients versus 29 days (IQR ) in patients not discussed at the OPM (p<0.001, Wilcoxon test). Analysis of location of death In the univariable analyses (table 2), the odds of dying at the hospital (excluding a hospice) was significantly decreased by all three processes of care variables: referent oncologist s PC training, intervention of the PCT and discussion of the patient s case at the OPM. After adjusting for all other variables (age, sex, primary cancer site, ER visits, survival after the last administration of chemotherapy, ICU admission and acute care admission), only 4 of 9 Colombet I, Montheil V, Durand J-P, et al. BMJ Supportive & Palliative Care (2012). doi: /bmjspcare

5 Table 3 Location of death in the subgroup of patients followed by the PC team Univariable analysis Hospital (excluding hospice) (n=180) Other (n=119) OR (95% CI) p Value Adjusted OR (95% CI) p Value Performance status at fi rst intervention of PC team, n (%) (31) 53 (45) (40) 49 (41) 1.4 (0.8 to 2.3) (0.3 to 1.6) 0.40 Survival after fi rst intervention 30 (7 to 86) 82 (33 to 179) 1.0 (0.9 to 1.0)* (0.96 to 1.02)* 0.46 of PC team, median (Q1 Q3) OPM, n (%) No 124 (69) 33 (28) Yes 56 (31) 86 (72) 0.2 (0.1 to 0.3) < (0.1 to 0.6) Referent oncologist trained in PC, n (%) No 124 (69) 58 (49) Yes 56 (31) 61 (52) 0.4 (0.3 to 0.7) (0.6 to 3.5) 0.42 *For 14 days increase in survival. Goodness-of-fi t of the multivariable model was tested by the Hosmer Lemeshow test. The null hypothesis cannot be rejected (p=0.84), indicating that there is no difference between the observed and predicted values of the response variable (small p values indicate a lack of fi t of the model). OR for death in hospital (excluding a hospice) versus other. The multivariable logistic regression models included patient characteristics (age, sex, primary cancer site), survival after last chemotherapy, emergency room visit, intensive care unit admission, acute care admission, performance status at fi rst intervention of PC team, survival after fi rst intervention of PC team, OPM and training of referent physician in PC. Reference category for univariable and multivariable logistic regression. For cancer site, we chose the cancer site with the lowest risk of dying at home or in a hospice as the reference category in order to facilitate interpretation (OR >1). In univariable analysis, this category was melanoma and we used it as the reference category for all analyses. OPM, onco-palliative meetings; PC, palliative care. the patient s case discussion at the OPM significantly decreased the odds of patients dying at the hospital (excluding a hospice) by 70% (adjusted OR 0.3, 95% CI 0.1 to 0.5; p<0.001). In the subgroup of 300 patients followed by the PCT (table 3), discussion at the OPM and the time between the first intervention by the PCT and death both significantly decreased the odds of dying at the hospital, whereas PS at the first evaluation by the PCT was not associated with this indicator. In the multivariable analysis, only discussion at the OPM still decreased these odds by 80% (adjusted OR 0.2, 95% CI 0.1 to 0.5; p=0.002). Analysis of the use of emergency resources No process characteristics significantly decreased the proportion of patients visiting the ER at least once in the last 30 days of life, either in univariable or multivariable analyses (table 4). Among the 300 patients followed by the PCT (table 5), survival after the first contact with the PCT was not statistically associated with referral to the emergency services. PS was the only variable associated with the use of emergency services in multivariable analyses: altered PS (3 or 4) doubled the odds of visiting the ER at least once (adjusted OR 2.1, 95% CI 1.0 to 4.4; p=0.038). Analysis of patients receiving chemotherapy in the last 14 days of life Median survival after the last administration of chemotherapy for patients referred by a PC trained or non-trained physician was 76 and 49 days, respectively. In the univariable analyses (table 6), the intervention of the PCT and discussion of the patient s case at the OPM both significantly decreased the odds of receiving chemotherapy in the last 14 days of life. In the multivariable analyses, discussion at the OPM still decreased the odds of receiving chemotherapy in the last 14 days of life by 50% (OR 0.5, 95% CI 0.2 to 0.9; p=0.035). Among the 300 patients followed by the PCT (table 7), discussion at the OPM was the only factor reducing the odds of receiving chemotherapy in the last 14 days of life (by 60%) in the univariable analyses, but this effect did not reach statistical significance in the multivariable analysis (adjusted OR 0.4, 95% CI 0.1 to 1.2; p=0.089). DISCUSSION In this study, we retrospectively analysed a case series of 521 deceased cancer patients to evaluate the impact of the oncologist s awareness of PC, the timing of the PCT intervention and multidisciplinary decision-making on process quality indicators of EOL care. We found that active collaboration between referent physicians and the PCT, through the weekly OPMs, was the only independent factor that decreased the odds of dying at the hospital (excluding a hospice) by 70% and the odds of receiving chemotherapy in last 14 days of life by 60%. These effect sizes are of clinical importance for both PC specialists and oncologists and, to our knowledge, have not previously been published. The clinical intervention of the PCT was not significantly associated with any indicators. However, in the subgroup of patients who were seen by the PCT, patients discussed in the OPM were seen earlier by the PCT, resulting in better PS at first contact and longer follow-up before death. However, in this subgroup early intervention by the PCT had no impact on indicators, whereas discussion of a patient s case at the OPM increased the odds of dying at home or in a hospice, and most significantly, reduced the odds of persistent chemotherapy in the last 14 days of life. Moreover, our study provides a first measure of Earle s indicators in France. Except for the location of death, our results almost meet the published standards. 9 Our choice of a follow-back of decedents does not cover all patients prospectively expected to die, but rather all patients who die during a given period of time Interestingly, retrospective and prospective measures of Earle s indicators Colombet I, Montheil V, Durand J-P, et al. BMJ Supportive & Palliative Care (2012). doi: /bmjspcare of 9

6 Table 4 Use of emergency services in last 30 days Univariable analysis At least one visit to the ER (n=185) None (n=329) OR (95% CI) p Value Adjusted OR (95% CI) p Value All patients Hosmer Lemeshow goodness-of-fi t test 0.26 Mean age at death (years) (SD) 64 (13) 64 (13) 0.99 (0.86 to 1.13)* (0.79 to 1.10)* 0.40 Sex, n (%) Female 59 (32) 132 (40) Male 125 (68) 196 (60) 1.4 (0.98 to 1) (0.7 to 9) 0.48 Primary cancer site, n (%) Global p value 0.04 Global p value Melanoma 14 (8) 53 (16) Gastro-intestinal 35 (19) 73 (22) 1.8 (0.8 to 4.3) (0.8 to 5.9) Lung 45 (24) 61 (19) 2.8 (1.2 to 6.6) (1.2 to 8.8) Urological 26 (14) 39 (12) 2.5 (1.0 to 6.5) (1.7 to 17.3) Other 65 (35) 103 (31) 2.4 (1.1 to 5.4) (1.3 to 8.8) Process of care characteristics, n (%) Referent oncologist trained in PC No 116 (63) 217 (66) Yes 69 (37) 112 (34) 1.2 (0.8 to 1.7) (1.1 to 3.2) PC team intervention No 79 (43) 136 (41) Yes 106 (57) 193 (59) 1.0 (0.7 to 1.4) (0.5 to 1.3) 0.37 Onco-palliative meeting No 125 (68) 212 (64) Yes 60 (32) 117 (36) 0.9 (0.6 to 1.3) (0.5 to 1.4) 0.45 Trajectory of care (other indicators Global p value 0.11 Global p value considered for confounding adjustment), n (%) Survival after last administration of chemotherapy <14 days 24 (13) 49 (15) days 30 (16) 43 (13) 1.4 (0.6 to 3.2) (0.7 to 4.4) days 56 (30) 74 (22) 1.5 (0.8 to 3.1) (1.1 to 6.0) days 75 (41) 163 (50) 0.9 (0.5 to 1.8) (0.8 to 3.7) 0.26 Intensive care unit admission (at least one in last 30 days) No 154 (83) 305 (93) Yes 25 (14) 24 (7) 2.1 (1.1 to 3.7) (0.5 to 1.9) 0.86 Acute care admission (in last Global p value <0.001 Global p value < days) None 18 (10) 105 (32) visit 101 (55) 141 (43) 4.2 (2.2 to 7.8) < (2.6 to 10.3) < visits 58 (31) 39 (12) 8.7 (4.2 to 17.8) < (6.0 to 32.5) <0.001 *For 10 years increase in age. The null hypothesis cannot be rejected, indicating that there is no difference between the observed and predicted values of the response variable (a small p value indicates that the model does not fi t the data). OR for at least one visit to ER in last 30 days versus none. For multiple comparisons testing difference with the reference category (primary cancer site, survival after last chemotherapy, acute care admission), p values were adjusted using Dunnett correction. Reference category for univariable and multivariable logistic regression. For cancer site, we chose the cancer site with the lowest risk of dying at home or in a hospice as the reference category in order to facilitate interpretation (OR >1). In univariable analysis, this category was melanoma and we used it as the reference category for all analyses. ER, emergency room; PC, palliative care. identified similar patterns of EOL care. 14 In our study, for the 14.8% of patients whose vital status could not be determined from hospital records, incomplete administrative data made it impossible to request the civil services for vital status. Some of these patients could have died during the study period and it is difficult to make any assumptions concerning their location of death. However, home PC services are not highly developed in France and have to be solicited from hospital-based PCTs at patient discharge. Patients who have not been referred to and followed up by a PCT during their hospital stay are therefore unlikely to be referred to home care services. Conversely, when patients are lost to follow-up by the hospital teams, they can die at home or in another acute care hospital. Underestimation of the proportion of patients who died at home, in the remaining data, should therefore be slight if at all. Improving the quality of the outcome in EOL patients requires clinical research in two different directions. The first task is to improve the ability of physicians to evaluate the prognosis and life expectancy of advanced cancer patients. Several clinical and biological parameters have been shown to correlate with very poor survival and several simple scores have been described and are under prospective evaluation. The second task is to define and evaluate the organisation and procedures most likely to impact on the quality of care and patient outcomes. Rigorous evaluation requires both actual implementation of these procedures and the use of validated 6 of 9 Colombet I, Montheil V, Durand J-P, et al. BMJ Supportive & Palliative Care (2012). doi: /bmjspcare

7 Table 5 Use of emergency services in last 30 days in the subgroup of patients followed by the PC team At least one visit to ER (n=106) None (n=193) Univariable analysis OR (95% CI) p Value Adjusted OR (95% CI) p Value Performance status at fi rst intervention of PC team, n (%) (34) 73 (38) (48) 70 (36) 1.5 (0.9 to 2.5) (1.0 to 4.4) Survival after fi rst intervention of 36 (13 to 88) 56 (19 to 129) 0.99 (0.97 to 1.02)* (0.98 to 1.04)* 0.59 PC team, median (Q1 Q3) OPM, n (%) No 56 (53) 101 (52) Yes 50 (47) 92 (48) 1.0 (0.6 to 1.6) (0.5 to 2.6) 0.71 Referent oncologist trained in PC, n (%) No 63 (59) 119 (62) Yes 43 (41) 74 (38) 1.1 (0.7 to 1.8) (0.7 to 3.2) 0.30 *For 14 days increase in survival. Goodness-of-fi t of the multivariable model was tested by the Hosmer Lemeshow test. The null hypothesis cannot be rejected (p=0.94), indicating that there is no difference between the observed and predicted values of the response variable (small p values indicate a lack of fi t of the model). OR for at least one visit to the emergency room in last 30 days versus none. The multivariable logistic regression models included patient characteristics (age, sex, primary cancer site), survival after last chemotherapy, intensive care unit admission, acute care admission, performance status at fi rst intervention of PC team, survival after fi rst intervention of PC team, OPM and training of referent physician in PC. Reference category for univariable and multivariable logistic regression. For cancer site, we chose the cancer site with the lowest risk of dying at home or in a hospice as the reference category in order to facilitate interpretation (OR >1). In univariable analysis, this category was melanoma and we used it as the reference category for all analyses. OPM, onco-palliative meetings; PC, palliative care. Table 6 Patients receiving chemotherapy in the last 14 days of life Univariable analysis Chemotherapy in the No chemotherapy in last 14 days of life (n=75) the last 14 days of life (n=446) OR (95% CI) p Value Adjusted OR (95% CI) All patients Hosmer Lemeshow goodnessof-fi t test 0.93 Mean age at death (years) (SD) 66 (11) 64 (13) 1.0 (0.9 to 1.0)* (0.8 to 1.3)* 0.79 Sex, n (%) Female 30 (40) 163 (37) Male 44 (59) 282 (63) 1.2 (0.7 to 2.0) (0.7 to 1.9) 0.64 Primary cancer site, n (%) Global p value 0.35 Global p value 0.34 Melanoma 5 (7) 62 (14) Gastro-intestinal 18 (24) 92 (21) 2.4 (0.9 to 6.9) (0.6 to 8.0) 0.33 Lung 17 (23) 90 (20) 2.3 (0.8 to 6.7) (0.7 to 10.2) Urological 7 (9) 60 (13) 1.5 (0.4 to 4.8) (0.4 to 8.3) 0.70 Other 28 (37) 142 (32) 2.4 (0.9 to 6.6) (0.8 to 9.5) Process of care characteristics, n (%) Referent oncologist trained in palliative care No 52 (69) 285 (64) Yes 23 (31) 161 (36) 0.8 (0.5 to 1.3) (0.5 to 1.7) 0.85 Palliative care team intervention No 43 (57) 178 (40) Yes 32 (43) 268 (60) 0.5 (0.3 to 0.8) (0.4 to 1.1) Onco- palliative meeting No 60 (80) 282 (63) Yes 15 (20) 164 (37) 0.4 (0.2 to 0.8) (0.2 to 0.9) *For 10 years increase in age. The null hypothesis cannot be rejected, indicating that there is no difference between the observed and predicted values of the response variable (small p values indicate a lack of fi t of the model). OR for receiving chemotherapy versus no chemotherapy in the last 14 days of life. For multiple comparisons testing difference with the reference category (primary cancer site, survival after last chemotherapy, acute care admission), p values were adjusted using Dunnett correction. Reference category for univariable and multivariable logistic regression. For cancer site, we chose the cancer site with the lowest risk of dying at home or in a hospice as the reference category in order to facilitate interpretation (OR >1). In univariable analysis, this category was melanoma and we used it as the reference category for all analyses. p Value Colombet I, Montheil V, Durand J-P, et al. BMJ Supportive & Palliative Care (2012). doi: /bmjspcare of 9

8 Table 7 Patients receiving chemotherapy in the last 14 days of life in the subgroup of patients followed by the PC team Chemotherapy in the last 14 days of life No chemotherapy in the last 14 days of life Univariable analysis, (n=32) (n=268) OR (95% CI) p Value Adjusted OR (95% CI) p Value Performance status at fi rst intervention of PC team, n (%) (31) 100 (37) (47) 106 (40) 1.4 (0.6 to 3.3) (0.3 to 2.3) 0.76 Survival after fi rst 27 (6 to 52) 53 (19 to 126) 0.96 (0.90 to 1.01)* (0.9 to 1.0)* 0.50 intervention of PC team, median (Q1 Q3) OPM No 23 (72) 134 (50) Yes 9 (28) 134 (50) 0.4 (0.2 to 0.9) (0.1 to 1.2) Referent oncologist trained in PC, n (%) No 21 (66) 161 (60) Yes 11 (34) 107 (40) 0.8 (0.4 to 1.7) (0.5 to 3.8) 0.58 *For 14 days increase in survival. Goodness-of-fi t of the multivariable model was tested by the Hosmer Lemeshow test. The null hypothesis cannot be rejected (p=0.61), indicating there is no difference between the observed and predicted values of the response variable (small p values indicate a lack of fi t of the model). OR for receiving chemotherapy versus no chemotherapy in the last 14 days of life. The multivariable logistic regression models included patient characteristics (age, sex, primary cancer site), performance status at fi rst intervention of PC team, survival after fi rst intervention of PC team, OPM and training of referent physician in PC. Reference category for univariable and multivariable logistic regression. For cancer site, we chose the cancer site with the lowest risk of dying at home or in a hospice as the reference category in order to facilitate interpretation (OR >1). In univariable analysis, this category was melanoma and we used it as the reference category for all analyses. OPM, onco-palliative meetings; PC, palliative care. Figure 1 Flow diagram for selection of decedents. indicators to ensure the quality of care. Temel et al demonstrated that early intervention by the PCT in newly diagnosed metastatic NSCLC increased the quality of life and survival However, the prediction of life expectancy is not always as accurate as in metastatic NSCLC. Additionally, the effect of early consultation and actual sharing of decision-making between PC specialists and oncologists remained unexplored. The role of the oncologist is known to be crucial for integrating PC into oncology practice. 20 Kao et al 21 found that the oncologist was the only predictor for continuing chemotherapy in the last 4 weeks of life. Despite recent progress, 22 oncologists still refer their patients to the PCT late in the trajectory of illness. 2 This paper continues this research by providing the first measure of such indicators in a French centre and describing the critical role of structured collaboration including a shared decision-making process. Our results indicate that a weekly meeting between the PCT and the oncology staff offering multidisciplinary decision-making for non-curable patients is a critical factor in improving quality indicators. This suggests that in addition to early clinical intervention by the PCT, the quality of collaboration and the structuring of discussion may be necessary for integrating PC into oncology. Moreover, such structured collaboration can easily be organised in the oncology setting and adjusted for the trajectory of illness and cancer therapy innovation, which is known to be a barrier in early PC intervention. 23 Early seed of the PC team is necessary but not sufficient. Support by the oncologic soil is needed and leads to integrative medicine. Acknowledgements The authors thank Marie-Yvonne Guillard, Nathalie Moreau and Françoise Travers (PC team nurses), Hélène de la Ménardière (psychologist), all referent physicians and the health professionals who participated to the oncopalliative meetings. Funding This research was supported by the Foundation Martine Midy through a grant to VM. Competing interests None. Provenance and peer review Not commissioned; externally peer reviewed. REFERENCES 1. Bruera E, Hui D. Integrating supportive and palliative care in the trajectory of cancer: establishing goals and models of care. J Clin Oncol 2010;28: Hui D, Elsayem A, De la Cruz M, et al. Availability and integration of palliative care at US cancer centers. JAMA 2010;303: Temel JS, Greer JA, Muzikansky A, et al. Early palliative care for patients with metastatic non-small-cell lung cancer. N Engl J Med 2010;363: Higginson IJ, Evans CJ. What is the evidence that palliative care teams improve outcomes for cancer patients and their families? Cancer J 2010;16: Lorenz KA, Lynn J, Dy SM, et al. Evidence for improving palliative care at the end of life: a systematic review. Ann Intern Med 2008;148: of 9 Colombet I, Montheil V, Durand J-P, et al. BMJ Supportive & Palliative Care (2012). doi: /bmjspcare

9 6. Zimmermann C, Riechelmann R, Krzyzanowska M, et al. Effectiveness of specialized palliative care: a systematic review. JAMA 2008;299: Pasman HR, Brandt HE, Deliens L, et al. Quality indicators for palliative care: a systematic review. J Pain Symptom Manage 2009;38: Earle CC, Park ER, Lai B, et al. Identifying potential indicators of the quality of end-of-life cancer care from administrative data. J Clin Oncol 2003;21: Earle CC, Neville BA, Landrum MB, et al. Evaluating claims-based indicators of the intensity of end-of-life cancer care. Int J Qual Health Care 2005;17: Oken MM, Creech RH, Tormey DC, et al. Toxicity and response criteria of the Eastern Cooperative Oncology Group. Am J Clin Oncol 1982;5: Bach PB, Schrag D, Begg CB. Resurrecting treatment histories of dead patients: a study design that should be laid to rest. JAMA 2004;292: Barnato AE, Farrell MH, Chang CC, et al. Development and validation of hospital end-of-life treatment intensity measures. Med Care 2009;47: Earle CC, Ayanian JZ. Looking back from death: the value of retrospective studies of end-of-life care. J Clin Oncol 2006;24: Setoguchi S, Earle CC, Glynn R, et al. Comparison of prospective and retrospective indicators of the quality of end-of-life cancer care. J Clin Oncol 2008;26: Durand JP, Mir O, Coriat R, et al. Validation of the Cochin Risk Index Score (CRIS) for life expectancy prediction in terminally ill cancer patients. Support Care Cancer 2012;20: Maltoni M, Nanni O, Pirovano M, et al. Successful validation of the palliative prognostic score in terminally ill cancer patients. Italian Multicenter Study Group on Palliative Care. J Pain Symptom Manage 1999;17: Morita T, Tsunoda J, Inoue S, et al. The Palliative Prognostic Index: a scoring system for survival prediction of terminally ill cancer patients. Support Care Cancer 1999;7: Temel JS, Jackson VA, Billings JA, et al. Phase II study: integrated palliative care in newly diagnosed advanced non-small-cell lung cancer patients. J Clin Oncol 2007;25: Jacobsen J, Jackson V, Dahlin C, et al. Components of early outpatient palliative care consultation in patients with metastatic nonsmall cell lung cancer. J Palliat Med 2011;14: Maltoni M, Amadori D. Palliative medicine and medical oncology. Ann Oncol 2001;12: Kao S, Shafi q J, Vardy J, et al. Use of chemotherapy at end of life in oncology patients. Ann Oncol 2009;20: Osta BE, Palmer JL, Paraskevopoulos T, et al. Interval between fi rst palliative care consult and death in patients diagnosed with advanced cancer at a comprehensive cancer center. J Palliat Med 2008;11: Mintzer DM, Zagrabbe K. On how increasing numbers of newer cancer therapies further delay referral to hospice: the increasing palliative care imperative. Am J Hosp Palliat Care 2007;24: Colombet I, Montheil V, Durand J-P, et al. BMJ Supportive & Palliative Care (2012). doi: /bmjspcare of 9

10 Effect of integrated palliative care on the quality of end-of-life care: retrospective analysis of 521 cancer patients Isabelle Colombet, Vincent Montheil, Jean-Philippe Durand, et al. BMJ Support Palliat Care published online June 29, 2012 doi: /bmjspcare Updated information and services can be found at: Data Supplement References P<P alerting service These include: "Web Only Data" This article cites 23 articles, 10 of which can be accessed free at: Published online June 29, 2012 in advance of the print journal. Receive free alerts when new articles cite this article. Sign up in the box at the top right corner of the online article. Notes Advance online articles have been peer reviewed, accepted for publication, edited and typeset, but have not not yet appeared in the paper journal. Advance online articles are citable and establish publication priority; they are indexed by PubMed from initial publication. Citations to Advance online articles must include the digital object identifier (DOIs) and date of initial publication. To request permissions go to: To order reprints go to: To subscribe to BMJ go to: