Downloaded from http://spcare.bmj.com/ on July 14, 2015 - Published by group.bmj.com
Research
Collaborative palliative care for advanced heart failure: outcomes and costs from the ‘Better Together’ pilot study Jill F Pattenden,1 Anne R Mason,2 R J P Lewin1
1
Health Sciences, University of York, York, UK Centre for Health Economics, University of York, York, UK
2
Correspondence to Jill F Pattenden, Health Sciences, University of York, Seebohm Rowntree Building, Area 4, Alcuin College, Heslington, York YO10 5DD, UK; [email protected] Received 23 March 2012 Accepted 2 July 2012 Published Online First 4 September 2012
To cite: Pattenden JF, Mason AR, Lewin RJP. BMJ Supportive & Palliative Care 2013;3:69–76.
ABSTRACT Background Patients with heart failure often receive little supportive or palliative care. ‘Better Together’ was a 2-year pilot study of a palliative care service for patients with advanced congestive heart failure (CHF). Objective To determine if the intervention made it more likely that patients would be cared for and die in their place of choice, and to investigate its cost-effectiveness. Methods This pragmatic non-randomised pilot evaluation was set in two English primary care trusts (Bradford and Poole). Prospective patientlevel data on outcomes and costs were compared with data from a historical control group of clinically comparable patients. Outcomes included death in preferred place of care (available only for the intervention group) and ‘hospital admissions averted’. Costs included medical procedures, inpatient care and the direct cost of providing the intervention. Results 99 patients were referred. Median survival from referral was 48 days in Bradford and 31 days in Poole. Most patients who died did so in their preferred place of death (Bradford 70%, Poole 77%). An estimated 14 and 18 hospital admissions for heart failure were averted in Bradford and Poole, respectively. The average cost-per-heart failure admission averted was £1529 in Bradford, but the intervention was cost saving in Poole. However, there was considerable uncertainty around these cost-effectiveness estimates. Conclusions This pilot study provides tentative evidence that a collaborative home-based palliative care service for patients with advanced CHF may increase the likelihood of death in place of choice and reduce inpatient admissions. These findings require confirmation using a more robust methodological framework.
Pattenden JF, et al. BMJ Supportive & Palliative Care 2013;3:69–76. doi:10.1136/bmjspcare-2012-000251
INTRODUCTION Heart failure is a progressive and frequently fatal condition with an unpredictable trajectory that affects an increasing number of elderly people with multiple co-morbidities. Despite numerous policy initiatives1–3 that stress the importance of timely access to supportive and palliative care for patients with advanced heart failure, the terminal phase is difficult to determine and patients are significantly more vulnerable to sudden death than cancer patients.4 Clinicians may therefore be reluctant to define palliative status and predict a time to death for patients with congestive heart failure (CHF).5 Uncertainty about prognosis and other features of the trajectory of heart failure make anticipatory management difficult and may lead to ‘prognostic paralysis’.6 7 It may be that these factors and the variation in access to specialist palliative care services have hampered progress in delivering palliative care to people living with heart failure.1 8 Heart failure specialist nurses (HFSNs) have been pivotal to the implementation of disease management programmes for people with heart failure and use evidence-based therapies and selfmanagement to stabilise patients and help them stay out of hospital. Studies have shown that they can reduce readmissions by 30% and be cost-effective.9 However, as the syndrome progresses, although many patients will not require palliative care, others will. A collaborative approach between HFSNs, cardiology, primary and palliative care services may help to identify and manage those who remain symptomatic and unstable despite optimal treatment.
69
Downloaded from http://spcare.bmj.com/ on July 14, 2015 - Published by group.bmj.com
Research Patients are typically admitted to hospital three times during their last year of life. Some of these admissions may be avoidable and so reducing them has potential for increased efficiency.10 The Department of Health ‘End Of Life Care Strategy’ reports that while around 50% of people wish to be cared for at home at the end of life and most would prefer to not to die in hospital, the reality is that most deaths (58%) occur in hospital, with about 18% taking place at home, 17% in care homes, 4% in hospices and 3% elsewhere.11 The preferred and actual place of death for people with heart failure is unknown. In the UK, the place of death from heart failure is difficult to establish, as guidance on death certificates encourages recording the cause of death in CHF patients as ‘coronary heart disease’. Three tools have been developed to improve end of life care: the Gold Standards Framework12 to co-ordinate and plan care for patients in the last year of life, the Liverpool Care Pathway13 now used in any setting for dying patients, and the Preferred Priorities for Care14 which is a patient held document where patients can record and review preferences for care at the end of life, including preferred place of death. These tools are less frequently used with CHF patients than with people dying from other conditions, but are increasingly being employed.8 Some studies have reported that HFSNs have forged links with local supportive and palliative care services, thus giving their patients access to advice from consultants in palliative care, Macmillan nurses, Marie Curie nurses and hospice day or inpatient therapy,15–17 with some recording the place of death.18 To our knowledge, no previous study has investigated whether the costs of using supplementary staff to enable home-based end of life care are offset by reductions in hospital admissions. From September 2006 to August 2008, the British Heart Foundation (BHF) and Marie Curie Cancer Care (MCCC) piloted ‘Better Together’ (BT), a 2-year collaboration to investigate the potential of BHF HFSNs, Marie Curie Cancer Care nurses (MCNs) and Marie Curie Cancer Care healthcare assistants (MCHCAs) working together alongside cardiologists, ‘care of the elderly’ consultants, district nurses and GPs to provide a service for patients with heart failure judged to be in their last year of life. OBJECTIVES Our aim was to evaluate the outcomes and costs of the BT pilot, a joint BHF HFSN and MCN collaboration providing care to people with advanced heart failure. STUDY METHODS This was a prospective, pragmatic, non-randomised pilot evaluation that compared patient outcome and
70
cost data from the intervention group with data from a historical control group of patients. All patients had a New York Heart Association (NYHA) severity classification of III or IV.19
BT intervention
The BHF and MCCC established a supportive and palliative care service. Staff from both organisations underwent joint training to learn about each other’s working practices.20 BHF HFSNs provided selfmanagement education and advice to patients and their carers. They managed symptoms through clinical assessment and regular medication monitoring and review. MCNs provided practical palliative physical nursing care, including the administration of prescribed medications for pain relief and agitation, and psychological support from referral until the end of life. They also liaised with district nurses and other support services for the provision of comfort aids. MCHCAs provided respite care, including basic physical care and psychological support, to patients and carers. Day or night shifts could be booked days or weeks in advance and patients could use the service occasionally (to avoid a readmission), regularly (for respite or last weeks of care) or as a one-off during a particular spell of ill health, but were then discharged until the service was needed again.
Study settings
The nurses and MCHCAs were working in two English primary care trusts (PCTs): Poole, and Bradford and Airedale. Six BHF HFSNs were involved in the study in Bradford and and two in Poole. Poole PCT covers both rural and city areas and has high numbers of elderly people. Bradford and Airedale PCT has a mostly urban population with significant areas of deprivation and high numbers of residents from ethnic minority groups. There were significant differences in the HFSN service configuration at the two sites. In Bradford, the heart failure and palliative care services were already working in partnership with palliative care,15 and the HFSNs had organised a weekly Heart Failure Support Group in the MC hospice day unit. In Poole, prior to establishing BT, the HFSNs had primarily received their caseloads from cardiologists and had fewer severely ill and elderly patients than their counterparts in Bradford, and concentrated more on newly diagnosed CHF patients. However, from the start of BT they began to obtain more referrals from ‘care of the elderly’ wards, GPs and district nurses which increased the proportion of patients in their caseload with an NYHA severity classification of III or IV and multiple co-morbidities.
Pattenden JF, et al. BMJ Supportive & Palliative Care 2013;3:69–76. doi:10.1136/bmjspcare-2012-000251
Downloaded from http://spcare.bmj.com/ on July 14, 2015 - Published by group.bmj.com
Research Study participants BT patients
Knowing when to involve palliative care services has proved problematic for some HFSNs, with no objective criterion to help them, and complex social and contextual circumstances as well as clinical features to consider.7 Most HFSNs aim to use a supportive and palliative approach from the time a patient comes into their caseload, but there will come a point when further support and expert care in the dying phase are needed to achieve the best quality of life for patients and their families. Patients could be referred to the new service by HFSNs, district nurses, community matrons and GPs when they met all the following criteria: ▸ NYHA III or IV ▸ Patient thought to be in the last year of life by their referrer ▸ Repeated hospital admissions ▸ Difficult physical/psychological symptoms despite optimal therapy ▸ Needing extra care or support ▸ Willing to have the service.
The NYHA classification system is a ‘disease severity’ classification system based on self-reported symptom load, disability and quality of life. Class III criteria include “a marked limitation of physical activity. Comfortable at rest, but less than ordinary activity causes fatigue, palpitation, or dyspnoea”. Class IV is defined as “Unable to carry out any physical activity without discomfort. If any physical activity is undertaken, discomfort is increased”.19 Control patients
Control groups were a convenience sample identified retrospectively by the HFSNs from their service caseloads in Poole and Bradford. The nurses reviewed their caseloads from September 2004 to August 2006 and selected all NYHA level III and IV patients who would have been considered eligible for a palliative care service such as BT. Data on patient outcome (date and place of death) were sourced separately and matched to resource use estimates using the patient’s NHS number.20 Data
This was a pragmatic evaluation using a convenience sample of all eligible patients who agreed to take part in the BT study. Patient data were obtained from the HFSNs through an MCCC ‘care manager’ referral form. Patient clinical and demographic data, hospital readmissions and length of stay, and type and place of death (sudden cardiac, heart failure deterioration, non-cardiac) were entered into the Statistical Package for Social Sciences (SPSS) version XV by JFP. The patient-level data obtained for the intervention and control groups are described below.
Demographic data: age, gender, ethnicity and date of death. Data on NYHA scores were not available for the intervention groups, and so clinical comparability could not be demonstrated. However, all patients met the inclusion criteria for the study, and so had a baseline NYHA score of III or IV. Resource use data: date of initiation into BT or control start date (ie, date of referral to the specialist nurse service), admissions between initiation and death, type of admission (elective vs emergency), date of admission, date of discharge, length of stay, principal diagnosis for admission, and principal intervention following admission. Cost data: cost data on outpatient, primary and community care use were not available for either group, so the analysis focused on secondary care costs. Inpatient costs and the cost of procedures undergone while in hospital were estimated using the appropriate healthcare resource group (HRG), identified on the basis of diagnosis, age and intervention data. Reference cost data for 2006/7 (the latest available at the time of analysis) were used to cost these HRGs, and the costs of any additional procedures were added to give a total inpatient cost for each patient.21 Admissions from September 2006 to August 2008 (intervention groups) and from September 2004 to August 2006 (control groups) were costed. For Poole PCT, these data were sourced from patients’ electronic records. In Bradford, the PCT supplied the data from their administrative databases. Costs were estimated for all patients included in the study, that is, regardless of whether the intervention was received. Taking a conservative approach, the analysis of benefits included only admissions with a principal diagnosis of heart failure. BT salary costs: the full cost of employing the specialist nurses and the costs of sessions per month provided by MCNs and MCHCAs. Patient centred outcome: preferred place of care (PPC). This was available only for the intervention group, and was recorded in the patients’ notes by HFSNs based on discussions with patients and their carers about their care preferences as their condition deteriorated. There were no data on the PPC for the control groups. The HFSNs did not solicit this information because there was no home-based or hospice supportive and palliative care they could offer. Instead, the number of ‘hospital admissions averted’ was used as a proxy outcome of value to patients. The cost-effectiveness of the intervention could therefore be assessed using an ‘incremental cost-per-admission averted’. ‘Averted admissions’ are the difference between the observed and expected admissions for heart failure in the intervention group. The expected admissions are those we would observe if the admission rate in the intervention group were the same as that observed in the control group (table 2).
Pattenden JF, et al. BMJ Supportive & Palliative Care 2013;3:69–76. doi:10.1136/bmjspcare-2012-000251
71
Downloaded from http://spcare.bmj.com/ on July 14, 2015 - Published by group.bmj.com
Research Ethical considerations
Comparability of control and intervention patients
The study received ethics approval from the Multi Centre Research Ethics Committee based in Leeds and the research governance departments of the participating hospital trusts.
In Bradford, patients in the intervention group were significantly older than their control group, with a mean difference of 3.8 years (95% CI 0.24 to 7.56). There were no other statistically significant differences in the demographic characteristics of the control and intervention groups in either site (Table 1).
Analysis
There were significant between-site differences in the configuration of care, so we analysed results separately for the two study sites. For each site, between-group differences in patient costs and outcomes were estimated using standard boot strapping techniques. Sample sizes matched the number of patients in each group, and 10 000 replications were performed, generating non-parametric estimates of the mean and 95% CI. Statistical tests were undertaken using StatsDirect Statistical Software version 2.7.2.22 Differences in continuous data were tested using unpaired t tests; differences in proportions were tested using the χ2 or Fisher’s exact test as appropriate. Full details of the analytical methods are reported elsewhere.20 RESULTS Study patients
Ninety-nine patients agreed to take part, 62 in Bradford and 37 in Poole. Across the two sites the average referral rate per nurse per year was eight patients. Six Bradford patients subsequently declined the service, four were admitted to hospital and died, two died suddenly at home and one was admitted to a nursing home before care could be received. Thus 49 patients in Bradford received care from MCCC staff. In Poole four patients died suddenly at home, one patient died in hospital, and one patient cancelled as he was admitted to hospital and discharged to a nursing home. Thus 31 patients received home care. No patient was lost to follow-up (figure 1). The median survival time from referral to the palliative service was 31 days in Poole and 48 days in Bradford.
Figure 1
72
Pathway of the study patients
Episodes of inpatient care
Table 2 shows the hospital admissions for each group. At both sites, a lower proportion of intervention patients were admitted to hospital compared with the control groups, but only in Bradford was the difference statistically significant. In Poole, the mean number of admissions per patient was significantly lower in the intervention group (mean difference −1.02, 95% CI −1.72 to −0.28). When considering only patients who were admitted with a principal diagnosis of heart failure, the proportion of admissions was significantly lower in the intervention groups at both sites. Costs of care
Figure 2 shows how the total cost of patient care varied within and between each of the four study groups. The total per-patient costs are presented in table 3. These costs include inpatient costs and the cost of the BT intervention. For both study sites, the mean costs per patient were higher in the control group than in the intervention group. For the Bradford site, the mean difference in patient cost was significantly lower in the intervention group: −£1190 (95% CI −£2367 to −£74). In Poole, the difference was smaller than that of Bradford and not statistically significant, but there was a trend towards lower per-patient costs in the intervention group (mean difference −£850; 95% CI −£2381 to £665). Benefits of care
Data on death in PPC were available only for the intervention group: in Bradford, 31 of 44 deaths were in PPC (70%); in Poole, the corresponding figure was 77% (24/31). A summary of place of death is given in figure 3. In Bradford, the distribution of place of death in the intervention group was significantly different from that of the respective control group (Fisher-Freeman-Halton exact test: p
Research
Collaborative palliative care for advanced heart failure: outcomes and costs from the ‘Better Together’ pilot study Jill F Pattenden,1 Anne R Mason,2 R J P Lewin1
1
Health Sciences, University of York, York, UK Centre for Health Economics, University of York, York, UK
2
Correspondence to Jill F Pattenden, Health Sciences, University of York, Seebohm Rowntree Building, Area 4, Alcuin College, Heslington, York YO10 5DD, UK; [email protected] Received 23 March 2012 Accepted 2 July 2012 Published Online First 4 September 2012
To cite: Pattenden JF, Mason AR, Lewin RJP. BMJ Supportive & Palliative Care 2013;3:69–76.
ABSTRACT Background Patients with heart failure often receive little supportive or palliative care. ‘Better Together’ was a 2-year pilot study of a palliative care service for patients with advanced congestive heart failure (CHF). Objective To determine if the intervention made it more likely that patients would be cared for and die in their place of choice, and to investigate its cost-effectiveness. Methods This pragmatic non-randomised pilot evaluation was set in two English primary care trusts (Bradford and Poole). Prospective patientlevel data on outcomes and costs were compared with data from a historical control group of clinically comparable patients. Outcomes included death in preferred place of care (available only for the intervention group) and ‘hospital admissions averted’. Costs included medical procedures, inpatient care and the direct cost of providing the intervention. Results 99 patients were referred. Median survival from referral was 48 days in Bradford and 31 days in Poole. Most patients who died did so in their preferred place of death (Bradford 70%, Poole 77%). An estimated 14 and 18 hospital admissions for heart failure were averted in Bradford and Poole, respectively. The average cost-per-heart failure admission averted was £1529 in Bradford, but the intervention was cost saving in Poole. However, there was considerable uncertainty around these cost-effectiveness estimates. Conclusions This pilot study provides tentative evidence that a collaborative home-based palliative care service for patients with advanced CHF may increase the likelihood of death in place of choice and reduce inpatient admissions. These findings require confirmation using a more robust methodological framework.
Pattenden JF, et al. BMJ Supportive & Palliative Care 2013;3:69–76. doi:10.1136/bmjspcare-2012-000251
INTRODUCTION Heart failure is a progressive and frequently fatal condition with an unpredictable trajectory that affects an increasing number of elderly people with multiple co-morbidities. Despite numerous policy initiatives1–3 that stress the importance of timely access to supportive and palliative care for patients with advanced heart failure, the terminal phase is difficult to determine and patients are significantly more vulnerable to sudden death than cancer patients.4 Clinicians may therefore be reluctant to define palliative status and predict a time to death for patients with congestive heart failure (CHF).5 Uncertainty about prognosis and other features of the trajectory of heart failure make anticipatory management difficult and may lead to ‘prognostic paralysis’.6 7 It may be that these factors and the variation in access to specialist palliative care services have hampered progress in delivering palliative care to people living with heart failure.1 8 Heart failure specialist nurses (HFSNs) have been pivotal to the implementation of disease management programmes for people with heart failure and use evidence-based therapies and selfmanagement to stabilise patients and help them stay out of hospital. Studies have shown that they can reduce readmissions by 30% and be cost-effective.9 However, as the syndrome progresses, although many patients will not require palliative care, others will. A collaborative approach between HFSNs, cardiology, primary and palliative care services may help to identify and manage those who remain symptomatic and unstable despite optimal treatment.
69
Downloaded from http://spcare.bmj.com/ on July 14, 2015 - Published by group.bmj.com
Research Patients are typically admitted to hospital three times during their last year of life. Some of these admissions may be avoidable and so reducing them has potential for increased efficiency.10 The Department of Health ‘End Of Life Care Strategy’ reports that while around 50% of people wish to be cared for at home at the end of life and most would prefer to not to die in hospital, the reality is that most deaths (58%) occur in hospital, with about 18% taking place at home, 17% in care homes, 4% in hospices and 3% elsewhere.11 The preferred and actual place of death for people with heart failure is unknown. In the UK, the place of death from heart failure is difficult to establish, as guidance on death certificates encourages recording the cause of death in CHF patients as ‘coronary heart disease’. Three tools have been developed to improve end of life care: the Gold Standards Framework12 to co-ordinate and plan care for patients in the last year of life, the Liverpool Care Pathway13 now used in any setting for dying patients, and the Preferred Priorities for Care14 which is a patient held document where patients can record and review preferences for care at the end of life, including preferred place of death. These tools are less frequently used with CHF patients than with people dying from other conditions, but are increasingly being employed.8 Some studies have reported that HFSNs have forged links with local supportive and palliative care services, thus giving their patients access to advice from consultants in palliative care, Macmillan nurses, Marie Curie nurses and hospice day or inpatient therapy,15–17 with some recording the place of death.18 To our knowledge, no previous study has investigated whether the costs of using supplementary staff to enable home-based end of life care are offset by reductions in hospital admissions. From September 2006 to August 2008, the British Heart Foundation (BHF) and Marie Curie Cancer Care (MCCC) piloted ‘Better Together’ (BT), a 2-year collaboration to investigate the potential of BHF HFSNs, Marie Curie Cancer Care nurses (MCNs) and Marie Curie Cancer Care healthcare assistants (MCHCAs) working together alongside cardiologists, ‘care of the elderly’ consultants, district nurses and GPs to provide a service for patients with heart failure judged to be in their last year of life. OBJECTIVES Our aim was to evaluate the outcomes and costs of the BT pilot, a joint BHF HFSN and MCN collaboration providing care to people with advanced heart failure. STUDY METHODS This was a prospective, pragmatic, non-randomised pilot evaluation that compared patient outcome and
70
cost data from the intervention group with data from a historical control group of patients. All patients had a New York Heart Association (NYHA) severity classification of III or IV.19
BT intervention
The BHF and MCCC established a supportive and palliative care service. Staff from both organisations underwent joint training to learn about each other’s working practices.20 BHF HFSNs provided selfmanagement education and advice to patients and their carers. They managed symptoms through clinical assessment and regular medication monitoring and review. MCNs provided practical palliative physical nursing care, including the administration of prescribed medications for pain relief and agitation, and psychological support from referral until the end of life. They also liaised with district nurses and other support services for the provision of comfort aids. MCHCAs provided respite care, including basic physical care and psychological support, to patients and carers. Day or night shifts could be booked days or weeks in advance and patients could use the service occasionally (to avoid a readmission), regularly (for respite or last weeks of care) or as a one-off during a particular spell of ill health, but were then discharged until the service was needed again.
Study settings
The nurses and MCHCAs were working in two English primary care trusts (PCTs): Poole, and Bradford and Airedale. Six BHF HFSNs were involved in the study in Bradford and and two in Poole. Poole PCT covers both rural and city areas and has high numbers of elderly people. Bradford and Airedale PCT has a mostly urban population with significant areas of deprivation and high numbers of residents from ethnic minority groups. There were significant differences in the HFSN service configuration at the two sites. In Bradford, the heart failure and palliative care services were already working in partnership with palliative care,15 and the HFSNs had organised a weekly Heart Failure Support Group in the MC hospice day unit. In Poole, prior to establishing BT, the HFSNs had primarily received their caseloads from cardiologists and had fewer severely ill and elderly patients than their counterparts in Bradford, and concentrated more on newly diagnosed CHF patients. However, from the start of BT they began to obtain more referrals from ‘care of the elderly’ wards, GPs and district nurses which increased the proportion of patients in their caseload with an NYHA severity classification of III or IV and multiple co-morbidities.
Pattenden JF, et al. BMJ Supportive & Palliative Care 2013;3:69–76. doi:10.1136/bmjspcare-2012-000251
Downloaded from http://spcare.bmj.com/ on July 14, 2015 - Published by group.bmj.com
Research Study participants BT patients
Knowing when to involve palliative care services has proved problematic for some HFSNs, with no objective criterion to help them, and complex social and contextual circumstances as well as clinical features to consider.7 Most HFSNs aim to use a supportive and palliative approach from the time a patient comes into their caseload, but there will come a point when further support and expert care in the dying phase are needed to achieve the best quality of life for patients and their families. Patients could be referred to the new service by HFSNs, district nurses, community matrons and GPs when they met all the following criteria: ▸ NYHA III or IV ▸ Patient thought to be in the last year of life by their referrer ▸ Repeated hospital admissions ▸ Difficult physical/psychological symptoms despite optimal therapy ▸ Needing extra care or support ▸ Willing to have the service.
The NYHA classification system is a ‘disease severity’ classification system based on self-reported symptom load, disability and quality of life. Class III criteria include “a marked limitation of physical activity. Comfortable at rest, but less than ordinary activity causes fatigue, palpitation, or dyspnoea”. Class IV is defined as “Unable to carry out any physical activity without discomfort. If any physical activity is undertaken, discomfort is increased”.19 Control patients
Control groups were a convenience sample identified retrospectively by the HFSNs from their service caseloads in Poole and Bradford. The nurses reviewed their caseloads from September 2004 to August 2006 and selected all NYHA level III and IV patients who would have been considered eligible for a palliative care service such as BT. Data on patient outcome (date and place of death) were sourced separately and matched to resource use estimates using the patient’s NHS number.20 Data
This was a pragmatic evaluation using a convenience sample of all eligible patients who agreed to take part in the BT study. Patient data were obtained from the HFSNs through an MCCC ‘care manager’ referral form. Patient clinical and demographic data, hospital readmissions and length of stay, and type and place of death (sudden cardiac, heart failure deterioration, non-cardiac) were entered into the Statistical Package for Social Sciences (SPSS) version XV by JFP. The patient-level data obtained for the intervention and control groups are described below.
Demographic data: age, gender, ethnicity and date of death. Data on NYHA scores were not available for the intervention groups, and so clinical comparability could not be demonstrated. However, all patients met the inclusion criteria for the study, and so had a baseline NYHA score of III or IV. Resource use data: date of initiation into BT or control start date (ie, date of referral to the specialist nurse service), admissions between initiation and death, type of admission (elective vs emergency), date of admission, date of discharge, length of stay, principal diagnosis for admission, and principal intervention following admission. Cost data: cost data on outpatient, primary and community care use were not available for either group, so the analysis focused on secondary care costs. Inpatient costs and the cost of procedures undergone while in hospital were estimated using the appropriate healthcare resource group (HRG), identified on the basis of diagnosis, age and intervention data. Reference cost data for 2006/7 (the latest available at the time of analysis) were used to cost these HRGs, and the costs of any additional procedures were added to give a total inpatient cost for each patient.21 Admissions from September 2006 to August 2008 (intervention groups) and from September 2004 to August 2006 (control groups) were costed. For Poole PCT, these data were sourced from patients’ electronic records. In Bradford, the PCT supplied the data from their administrative databases. Costs were estimated for all patients included in the study, that is, regardless of whether the intervention was received. Taking a conservative approach, the analysis of benefits included only admissions with a principal diagnosis of heart failure. BT salary costs: the full cost of employing the specialist nurses and the costs of sessions per month provided by MCNs and MCHCAs. Patient centred outcome: preferred place of care (PPC). This was available only for the intervention group, and was recorded in the patients’ notes by HFSNs based on discussions with patients and their carers about their care preferences as their condition deteriorated. There were no data on the PPC for the control groups. The HFSNs did not solicit this information because there was no home-based or hospice supportive and palliative care they could offer. Instead, the number of ‘hospital admissions averted’ was used as a proxy outcome of value to patients. The cost-effectiveness of the intervention could therefore be assessed using an ‘incremental cost-per-admission averted’. ‘Averted admissions’ are the difference between the observed and expected admissions for heart failure in the intervention group. The expected admissions are those we would observe if the admission rate in the intervention group were the same as that observed in the control group (table 2).
Pattenden JF, et al. BMJ Supportive & Palliative Care 2013;3:69–76. doi:10.1136/bmjspcare-2012-000251
71
Downloaded from http://spcare.bmj.com/ on July 14, 2015 - Published by group.bmj.com
Research Ethical considerations
Comparability of control and intervention patients
The study received ethics approval from the Multi Centre Research Ethics Committee based in Leeds and the research governance departments of the participating hospital trusts.
In Bradford, patients in the intervention group were significantly older than their control group, with a mean difference of 3.8 years (95% CI 0.24 to 7.56). There were no other statistically significant differences in the demographic characteristics of the control and intervention groups in either site (Table 1).
Analysis
There were significant between-site differences in the configuration of care, so we analysed results separately for the two study sites. For each site, between-group differences in patient costs and outcomes were estimated using standard boot strapping techniques. Sample sizes matched the number of patients in each group, and 10 000 replications were performed, generating non-parametric estimates of the mean and 95% CI. Statistical tests were undertaken using StatsDirect Statistical Software version 2.7.2.22 Differences in continuous data were tested using unpaired t tests; differences in proportions were tested using the χ2 or Fisher’s exact test as appropriate. Full details of the analytical methods are reported elsewhere.20 RESULTS Study patients
Ninety-nine patients agreed to take part, 62 in Bradford and 37 in Poole. Across the two sites the average referral rate per nurse per year was eight patients. Six Bradford patients subsequently declined the service, four were admitted to hospital and died, two died suddenly at home and one was admitted to a nursing home before care could be received. Thus 49 patients in Bradford received care from MCCC staff. In Poole four patients died suddenly at home, one patient died in hospital, and one patient cancelled as he was admitted to hospital and discharged to a nursing home. Thus 31 patients received home care. No patient was lost to follow-up (figure 1). The median survival time from referral to the palliative service was 31 days in Poole and 48 days in Bradford.
Figure 1
72
Pathway of the study patients
Episodes of inpatient care
Table 2 shows the hospital admissions for each group. At both sites, a lower proportion of intervention patients were admitted to hospital compared with the control groups, but only in Bradford was the difference statistically significant. In Poole, the mean number of admissions per patient was significantly lower in the intervention group (mean difference −1.02, 95% CI −1.72 to −0.28). When considering only patients who were admitted with a principal diagnosis of heart failure, the proportion of admissions was significantly lower in the intervention groups at both sites. Costs of care
Figure 2 shows how the total cost of patient care varied within and between each of the four study groups. The total per-patient costs are presented in table 3. These costs include inpatient costs and the cost of the BT intervention. For both study sites, the mean costs per patient were higher in the control group than in the intervention group. For the Bradford site, the mean difference in patient cost was significantly lower in the intervention group: −£1190 (95% CI −£2367 to −£74). In Poole, the difference was smaller than that of Bradford and not statistically significant, but there was a trend towards lower per-patient costs in the intervention group (mean difference −£850; 95% CI −£2381 to £665). Benefits of care
Data on death in PPC were available only for the intervention group: in Bradford, 31 of 44 deaths were in PPC (70%); in Poole, the corresponding figure was 77% (24/31). A summary of place of death is given in figure 3. In Bradford, the distribution of place of death in the intervention group was significantly different from that of the respective control group (Fisher-Freeman-Halton exact test: p
