Evaluating the impact of a home palliative programme for terminally ill non-cancer patients with end-stage organ failure and severe frailty in Singapore

Abstract

Objective

To evaluate Programme IMPACT, a home-based palliative care initiative for terminally ill non-cancer patients with end-stage organ failure and severe frailty at the end-of-life, in achieving palliative outcomes, honouring end-of-life care preferences, and reducing acute healthcare utilisation.

Methods

Patients with non-cancer diagnoses, prognosis of ≤ 6 months, palliative goals of care, and preference for home as a place of care or death were enrolled between October 2017 and December 2021. First, we employed a single-arm prospective cohort design to assess symptom control and concordance between preferences and delivered care. Second, we compared healthcare utilisation, costs and place of death between programme decedents and non-cancer controls without palliative care identified retrospectively. Utilisation was analysed with Poisson regression, costs with generalised linear models (gamma log link), and logistic regression for home deaths, adjusting for baseline confounders.

Results

A total of 387 patients were enrolled in the programme. Patients experienced significant clinical benefit, with mean symptom scores improving by 2.5 points on the Edmonton Symptom Assessment Scale. Concordance between preferences and outcomes exceeded 94% for CPR status and medical treatments and 89% for place of death. Compared with controls, programme decedents incurred significantly lower utilisation and costs: admissions and length of stay (Rate Ratio [RR]: 0.65–0.85, p < 0.005), admission-related and total costs (Mean Ratio [MR]: 0.56–0.75, p < 0.001), and ED visits and costs (RR & MR: 0.69–0.81, p < 0.008). They were also 16.6 times more likely to die at home (95% CI: 10.2–27.0; p < 0.001).

Conclusion

Home-based palliative care improves symptom control, aligns care with preferences, and reduces acute hospital use and costs among this population.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12904-026-02137-6.

Introduction

The global demand for palliative care is increasing substantially, driven by demographic shifts towards an ageing population and the rising prevalence of non-communicable diseases [1–3]. By 2030, one in six people will be aged 60 years or older, with the majority of deaths attributable to conditions such as cancer, cardiovascular diseases, respiratory illnesses, renal diseases and other chronic conditions [4, 5]. Individuals with advanced terminal illnesses could benefit from palliative care, which adopts a holistic, patient-centred approach focused on symptom management, comfort, and quality of life, while addressing physical, psychosocial, and spiritual needs [2, 6–8].

Despite its established benefits, the provision of palliative care has been largely concentrated on patients with advanced cancer. [9–12] This focus is partly attributable to the relatively predictable illness trajectory of many cancers, which facilitates timely referrals. Conversely, patients with end-stage organ failure (ESOF) as well as individuals who are severely frail at end-of-life (FE) typically experience fluctuating “entry-re-entry” trajectories characterised by recurrent acute exacerbations and unpredictable prognosis [2, 13]. These individuals often face high symptom burden, undergo burdensome interventions that may not prolong life, incur greater healthcare expenditures in the last year of life, and are more likely to die in the hospital compared to cancer patients [2, 14–20]. Given their fluctuating illness trajectories and limited benefit from aggressive interventions, medical care in the acute hospital may not always be appropriate for patients at the end-of-life (EoL) [21].

The complex needs of these patients often necessitate an integrated care model that combines hospital-based services with community and home-based hospice support. These models aim to facilitate earlier integration of palliative care, alongside comfort-focused care for terminally-ill patients. However, significant disparity persists, where many patients diagnosed with ESOF and severe FE continue to spend much of their last year of life in hospitals, despite expressing a preference for home care [17]. EoL care remains largely cancer-centric, leaving patients with non-cancer terminal illnesses under-supported in the last months of life. This disparity is reflected in the access to home palliative care services in Singapore, where a retrospective study of decedents (2011—2015) found that only 9.7% of non-cancer decedents (n = 1,049) received such support in the last year of life compared with 40% of cancer decedents (n = 615) [18].

To address this service gap, Programme Integrated Management and PAlliative Care for the Terminally ill (Programme IMPACT) in 2017 was initiated in one of Singapore’s publicly funded general hospitals. This programme provides 24/7 home-based palliative care to terminally-ill non-cancer patients using a multidisciplinary approach integrated into acute care. The study aims to evaluate Programme IMPACT in providing palliative care for terminally ill non-cancer patients in terms of achieving good palliative outcomes, honouring patients’ EoL care preferences and reducing unnecessary acute healthcare utilisation.

Methodology

Delivery of home palliative care intervention

Programme IMPACT is a hospital-embedded home-based palliative care service, where the primary site of care delivery is the patient’s home. Program IMPACT serves the residents in the central part of Singapore. Patients are referred from inpatient or outpatient services if they meet the following criteria: non-cancer diagnosis, prognosis of ≤ 6 months, care goals focused on palliation, and preference for home as the place of care or death. Organ-specific referral criteria for patients diagnosed with end-stage heart failure (ESHF), end-stage respiratory failure (ESRespF), including end-stage chronic obstructive pulmonary disease (COPD) and end-stage non-COPD respiratory diseases, end-stage renal failure (ESRF), and those with severe FE are provided in Appendix 1. Non-COPD respiratory diseases refer to end-stage lung diseases other than COPD. The multidisciplinary team, comprising doctors, nurses, and medical social workers, provided round-the-clock support, including home-based medical and nursing care, symptom assessments, and psychosocial support, with seamless transitions to the acute palliative ward when clinically required. Care was delivered primarily through scheduled home visits, with additional visits triggered by symptom exacerbations. Between visits, patients and caregivers had access to telephone support, with a 24/7 on-call system enabling timely clinical advice or urgent home visits. As Programme IMPACT is embedded within a hospital ecosystem, it has several distinguishing features:

1. Patients could be directly admitted back to the acute inpatient ward for treatment of potentially reversible conditions, bypassing the usual route, which is through the Emergency Department. This enabled timely access to care.

2. Care is more closely coordinated with the hospital’s organ specialists, allowing for timely inputs when needed. This includes medication optimisation (e.g., titrating diuretics or deprescribing) and interventions such as deactivating an implantable cardiac device when the patient’s condition deteriorates. This integrated linkage facilitates timely communication and continuity of care with specialists.

3. The Program IMPACT team can administer parenteral medications in the home setting to manage acute exacerbations of the patient’s chronic condition.

This integrated model enables patients to be primarily supported in the home setting while maintaining access to hospital care when clinically necessary, thereby facilitating goal-concordant care and reducing unnecessary acute healthcare utilisation near the end of life. Patients were means-tested, with out-of-pocket fees ranging from SGD 150 to 450 per month. Additional programme features are described in Appendix 1.

Study design and population

We developed a logic model a priori to guide the overall evaluation, articulate hypothesised pathways between programme activities and outcomes, and inform the selection of measures across both study components (Appendix 2).

We employed a hybrid design comprising (1) a prospective single-arm cohort and (2) a retrospective cohort comparison to comprehensively assess Programme IMPACT. This two-part design allowed us to evaluate both patient-centred outcomes and system-level effects (Appendix 3).

First, a single-arm prospective cohort design was used to assess real-time clinical processes, palliative outcomes and the concordance between delivered care and patients’ EoL preferences. This approach was ethically necessary as randomising eligible patients to a non-intervention group was not feasible, given that the programme served individuals with significant EoL needs and limited access to alternative care options.

Second, to complement the prospective analysis, we conducted a retrospective cohort comparison focusing on decedents enrolled in Programme IMPACT. This analysis compared healthcare utilisation, associated costs, and place of death between Programme IMPACT decedents and non-cancer ESOF decedents who did not receive any form of palliative care. Appendix 4 details how Programme IMPACT decedents and control group decedents were identified. Briefly, we identified decedents who died between 2017 and 2020 and had at least one admission to the acute hospital in the last year of life. Following this, an ESOF-specific criterion was applied for heart, respiratory, and renal failure decedents using a specific admission-related ESOF diagnosis in the last year of life (Appendix 4.1). To ensure comparability between the groups, control decedents were excluded if they had cancer as a primary cause of death or comorbidity, received renal replacement therapy, or utilised inpatient/day hospice or community-based palliative homecare services in their last year of life. Any control decedent who subsequently crossed over into Programme IMPACT was reclassified as a programme decedent.

Primary outcomes

Overall, the primary outcomes were symptom control, care concordance, and healthcare utilisation and costs in the last 6-, 3-, and 1-month of life.

Prospective study outcomes

We defined good palliative care as achieving successful symptom control. This was assessed using the Edmonton Symptom Assessment Scale (ESAS), which measures 13 symptoms such as anxiety, appetite, confusion, and depression, on a scale of 0 to 10, with higher scores indicating greater severity [22]. The minimally clinically important difference (MCID) was determined to be 1 point for improvement or deterioration [22–24]. This threshold reflects meaningful within-patient change. ESAS was administered to all patients experiencing symptom exacerbations at baseline and repeated at each home visit. Consistent with the literature [22], ESAS scores of 0—1, 1—3, 4—6, and 7—10 correspond to none, mild, moderate, and severe symptoms, respectively. We operationalised symptom exacerbation as any ESAS symptom with a score of 4 or more. To assess the programme’s effectiveness in managing these exacerbations, we examined the changes in ESAS scores between an exacerbation and the following ESAS score for all episodes and stratified by each ESAS symptom. A positive change in scores indicates improvement, reflecting a reduction in symptom severity, whereas a negative change in scores indicates worsening. We also looked at the proportion of episodes that improved, remained the same or worsened.

To assess concordance of care, we retrieved care preferences documented in the official ACP forms and informal discussions, documented in clinical notes but not formally captured through standardised forms. The primary goal here was to determine the concordance of care preferences and actualised outcomes.

Retrospective study outcomes

For the retrospective subset analysis, we obtained data on healthcare utilisation and costs for Programme IMPACT and control decedents from an administrative database, containing secondary, de-identified billing information. From this database, we extracted the cumulative number of inpatient admissions (IP), total length of stay (LOS), emergency department (ED) attendances, and specialist outpatient clinic (SOC) visits. Associated costs for each type of utilisation, as well as total healthcare costs, were also obtained. As Programme IMPACT was a chargeable service, programme charges were included in total healthcare costs. These outcomes were calculated for the 6-, 3-, and 1-month periods prior to death using a look-back approach with the date of death as the reference point. The place and cause of death were retrieved from the same administrative database. All non-residential locations were classified as non-home deaths.

Statistical analysis

For the prospective cohort, descriptive analyses were conducted for all Programme IMPACT patients and stratified by ESOF type and FE. Continuous variables were summarised as means with standard deviations (SD), and categorical variables as percentages. Symptom outcomes were reported using changes in ESAS scores, and concordance between documented care preferences and actual outcomes was described.

For the retrospective subset analysis, baseline characteristics were compared between Programme IMPACT and control decedents. Logistic regression was used to analyse the likelihood of home death. Poisson regression models were applied to cumulative healthcare utilisation outcomes, while generalised linear models (Gamma family with log link) were used to assess cumulative healthcare costs. Rate ratios (RRs) estimated from Poisson regressions represent the proportional difference in expected utilisation rates between the two groups. Mean ratios (MRs) represent the proportional difference in expected mean costs between groups, as estimated from gamma log-link models. All multivariate models were adjusted for ESOF type, gender, race, age at death, year of death, and housing type.

Across both study components, statistical significance was defined as p < 0.05. All analyses were performed using Stata version 18.

Ethics

Ethical approval was obtained from the National Healthcare Group Domain Specific Review Board (NHG DSRB 2019/00583), and the study adhered to the Declaration of Helsinki. All data were de-identified and anonymised by an independent third party. Informed consent was obtained from all patients and/or their next of kin prior to participation in the study.

Results

Baseline characteristics of programme IMPACT patients enrolled

Between 2017 and 2021, a total of 387 patients (394 admissions) were enrolled in Programme IMPACT (Appendix 5). The mean age at enrolment was 82.3 (SD = 9.8) years, 56.1% were male, and 88.1% were of Chinese ethnicity (Table 1). The majority of patients were diagnosed with ESRF (39.3%), followed by ESHF (24.8%), end-stage COPD (16.3%), severe FE (11.1%), and end-stage non-COPD respiratory diseases (8.5%). Patients were enrolled on average 5.1 (SD 6.1) months with a median of 2.8 months (25th to 75th percentile: 0.70 to 7.4).

Table 1.

Baseline characteristics of all programme IMPACT patients

	All (n = 387)		ESHF (n = 97)		ESRF (n = 153)		FE (n = 43)		COPD (n = 63)		Non-COPD (n = 33)		P value*
Demographics
	mean	SD	mean	SD	mean	SD	mean	SD	mean	SD	mean	SD
Age (years)	82.3	9.8	83.9	9.4	82.6	10.2	84.4	10.3	79.2	9.0	79.4	8.9	< 0.01
Enrolment period (months)	5.1	6.1	5.6	6.1	3.3	3.9	3.0	3.2	8.1	8.4	7.8	7.6	< 0.001
	n	col %	n	col %	n	col %	n	col %	n	col %	n	col %
Male	217	56.1	34	35.4	60	39.5	18	41.9	45	71.4	13	39.4	< 0.001
Ethnicity
Chinese	341	88.1	87	90.6	133	87.5	37	86.0	59	93.6	25	75.8	0.14
Malay	19	4.9	4	4.2	10	6.6	2	4.7	2	3.2	1	3.0
Indian	20	5.2	3	3.1	6	3.9	4	9.3	2	3.2	5	15.1
Others	7	1.8	2	2.1	3	2.0	-	-	-	-	2	6.1
	n	col %	n	col %	n	col %	n	col %	n	col %	n	col %
Discharged from the programme	373	94.7	92	94.8	152	99.3	43	97.7	55	83.3	31	91.2	< 0.001
Mortality data
	n	col %	n	col %	n	col %	n	col %	n	col %	n	col %
No. of decedents who died during the programme	281	71.3	58	59.8	127	83	33	75	38	57.6	25	73.5	< 0.01
	mean	SD	mean	SD	mean	SD	mean	SD	mean	SD	mean	SD
Time to death (months)	3.5	4.8	3.4	5.8	2.9	3.4	1.8	2.2	5.9	5.6	5.7	7.1	< 0.001
	n	col %	n	col %	n	col %	n	col %	n	col %	n	col %
No. of decedents with a post-bereavement survey (PBS) done by caregivers	221	75.4	41	70.7	103	81.1	26	78.8	23	60.5	19	76	0.14
	mean	SD	mean	SD	mean	SD	mean	SD	mean	SD	mean	SD
PBS score	9.2	0.8	9.1	0.9	9.2	0.8	9.5	0.6	9.3	0.6	8.9	0.8	0.15
	n	col %	n	col %	n	col %	n	col %	n	col %	n	col %
No. of decedents with documented EOL care preferences	267	95.0	53	91.4	124	96.9	32	97.0	35	92.1	23	92.0	0.28
Formal ACP	165	58.7	32	55.2	85	66.9	18	54.5	20	52.6	10	40.0	0.16
Informal ACP	102	36.3	21	36.2	39	30.7	14	42.4	15	39.6	13	52.0
No ACP done	14	5.0	5	8.6	3	2.4	1	3.0	3	7.9	2	8.0

Abbreviations: ESHF End-stage heart failure, ESRF End-stage renal failure, FE Frail end-of- life, COPD Chronic obstructive pulmonary disorder, Non-COPD Non-chronic obstructive pulmonary disorder, Eol End-of-Life, ACP Advance care plan

^*P value compared across the individual diseases

Achieving good palliative outcomes

On average, the overall mean improvement after an exacerbation was 2.5 points, and 58.8% of episodes with exacerbations improved by a subsequent visit (Table 2). When stratified by disease groups, similar observations were seen (Appendix 6).

Table 2.

Change in ESAS score (overall)

Symptoms	No. of exacerbations	Moderate to high ESAS score		Next ESAS score		Δ Scores (95% CI)+	% Change#
		Mean	SD	Mean	SD		Decrease	No change	Increase
Constipation	390	6.7	1.3	4.1	3.1	2.5 (95% CI: 2.2 to 2.8)	6.2	37.7	56.2
Appetite	280	6.7	1.4	4.6	3.2	2.1 (95% CI: 1.8 to 2.5)	7.9	42.9	49.3
Dyspnea	229	6.1	0.9	3.0	2.4	3.1 (95% CI: 2.8 to 3.4)	2.2	21.4	76.4
Tiredness	227	6.3	1.0	4.0	2.9	2.3 (95% CI: 1.9 to 2.7)	6.2	37.0	56.8
Lower limb oedema	204	6.2	1.1	4.4	2.8	1.8 (95% CI:1.5 to 2.2)	6.9	43.6	49.5
Drowsiness	189	6.8	1.4	5.0	3.3	1.8 (95% CI:1.4 to 2.2)	7.9	47.1	45.0
Well-being	141	6.3	1.1	3.0	2.7	3.3 (95% CI:2.8 to 3.8)	3.5	25.5	70.9
Confusion	94	6.5	1.1	4.7	2.6	1.8 (95% CI: 1.3 to 2.3)	3.2	53.2	43.6
Pain	76	6.2	1.0	2.6	2.9	3.6 (95% CI: 2.9 to 4.2)	5.3	11.8	82.9
Itch	64	6.4	1.3	3.3	2.6	3.0 (95% CI:2.3 to 3.7)	6.3	21.9	71.9
Depression	38	6.2	1.0	3.8	2.6	2.4 (95% CI:1.6 to 3.3)	5.3	31.6	63.2
Anxiety	35	6.3	1.5	3.3	2.6	3.0 (95% CI: 2.1 to 3.8)	0.0	25.7	74.3
Nausea	20	6.2	0.7	1.1	1.8	5.1 (95% CI: 4.7 to 6.1)	0.0	5.0	95.0

Abbreviations: ESAS Edmonton symptom assessment scale

 + Δ Score is defined as the difference between the baseline moderate-to-high ESAS score and the subsequent ESAS score. Mean Δ Scores were calculated across all symptom exacerbation episodes. A positive Δ Score reflects improvement (reduced symptom severity), whereas a negative Δ Score indicates worsening

^# For the percentage categories, “decrease” refers to episodes where symptoms worsened, “no change” indicates stability, and “increase” denotes episodes with improved symptom control

Honouring patients’ preferences

During this period, 281 patients died, with a mean time to death from enrolment of 3.5 months (Table 1). Among the decedents, 95% had either a formal (n = 165) or informal (n = 102) ACP. Among the 165 patients with a formally documented ACP, 64% preferred to die at home, 21.8% had no specific preference, and 14.6% indicated multiple acceptable locations. For resuscitation preferences, 98.2% opted for do-not-resuscitate (DNR). Regarding the level of medical intervention, 83% preferred limited intervention, 15.8% comfort-focused care, and 1.2% full treatment.

In terms of care concordance, among those with formal or informal ACP (n = 267), 70.8% died at home, followed by acute hospital (28.5%), and community hospital/hospice (< 1%). High concordance was achieved for preferences regarding cardiopulmonary resuscitation (94.8%), medical interventions (96.3%) and place of death (89.1%) (Fig. 1). Discordance for CPR and medical intervention stemmed from the reversal of decisions by caregiver or next-of-kin during crisis events or the healthcare staff's unawareness of care preferences. For the place of death, the primary reason for discordance was dying in the hospital.

Fig. 1.

Overall ACP concordance. Abbreviations: Advance care planning (ACP); End-of-life (EoL)

Reducing healthcare utilisation and associated costs outcomes

In the sub-set retrospective analysis, 148 Programme IMPACT and 853 control decedents were analysed, and their mean age-at-death was 82.3 (SD = 9.4) and 77.2 (SD = 11.5) years, respectively (Appendix 7). Unadjusted healthcare utilisation and costs over study look-back periods of Programme IMPACT and control decedents in the sub-set analysis are depicted in Fig. 2.

Fig. 2.

Unadjusted mean cumulative healthcare utilisation and associated cost outcomes over study look-back periods, by groups in sub-set analysis. Abbreviations: Programme IMPACT (IP); Length of stay (LOS); Emergency department (ED); Specialist outpatient clinic (SOC)

In our multivariate analysis examining healthcare utilisation and associated costs, we found sizable differences between Programme IMPACT decedents and control decedents across all time-points. After adjusting for ESOF type, gender, race, age at death, year of death, and housing type, Programme IMPACT decedents demonstrated consistently lower healthcare utilisation patterns. They had fewer hospital admissions and shorter length of stay (RR = 0.65–0.85; all p < 0.005), reduced admission-related and total associated costs (MR = 0.56–0.75; all p < 0.001), and fewer emergency department visits with lower associated costs (RR = 0.73–0.81 and MR = 0.69–0.76; all p < 0.008). Programme IMPACT patients had fewer SOC visits 3- & 6-months prior to death (RR = 0.80—0.86; all p:0.022—0.027), compared to controls. No between-group differences in costs associated with SOC visits at all look-back periods were found. Results of RRs, MRs and 95% CIs from these analyses are depicted in Fig. 3 and Table 3.

Fig. 3.

Plots of incident risk ratios and 95% confidence intervals for Programme Impact descendants from healthcare utilisation and associated cost models across study look-back periods (Reference = Control patients: Incident risk ratio = 1). Abbreviations: Programme IMPACT (IP); Length of stay (LOS); Emergency department (ED); Specialist outpatient clinic (SOC)

Table 3.

Summary of healthcare utilisation and cost outcomes: programme IMPACT vs historical controls (Reference)

Programme IMPACT Patients
Healthcare Utilisation Outcome	1-month RR (95% CI)	3-month RR (95% CI)	6-month RR (95% CI)
Inpatient Admissions	0.71 (0.56, 0.90)	0.78 (0.66, 0.91)	0.81 (0.71, 0.91)
Inpatient Length of Stay	0.65 (0.60, 0.71)	0.78 (0.74, 0.81)	0.85 (0.82, 0.89)
Emergency Department Visits	0.73 (0.58, 0.93)	0.76 (0.65, 0.88)	0.81 (0.71, 0.91)
Specialist Outpatient Clinic Visits	0.70 (0.46, 1.05)	0.80 (0.66, 0.97)	0.86 (0.76, 0.98)
Healthcare Utilisation Cost Outcome	1-month MR (95% CI)	3-month MR (95% CI)	6-month MR (95% CI)
Inpatient Costs	0.56 (0.43, 0.73)	0.65 (0.54, 0.79)	0.75 (0.64, 0.88)
Emergency Department Costs	0.69 (0.55, 0.85)	0.71 (0.61, 0.82)	0.76 (0.67, 0.87)
Specialist Outpatient Clinic Costs	0.97 (0.44, 2.13)	0.92 (0.66, 1.30)	0.88 (0.69, 1.12)
Total Healthcare Costs	0.57 (0.44, 0.73)	0.66 (0.55, 0.79)	0.75 (0.64, 0.88)

Abbreviations: CI Confidence Interval, RR Rate Ratio, MR Mean Ratio

In the ESOF-specific analyses, we found that ESRF decedents benefited most from the programme, followed by ESHF decedents to some extent. However, Programme IMPACT did not benefit ESRespF decedents for all healthcare utilisation & cost outcomes examined. These results are included in Appendix 8A-C.

(i) Home deaths

The percentages of home deaths for Programme IMPACT and control decedents are 63.5% and 10.3%, respectively. Percentages of home deaths with ESOF type breakdown, by groups, can be found in Appendix 9. In the multivariate logistic regression for home deaths with adjustments for ESOF type, gender, race, death age, death year, and housing type, we found that Programme IMPACT decedents had higher odds (16.6 times) of passing away at home [95% confidence interval (CI) of odds ratio: 10.2–27.0; p < 0.001].

Discussion

Key findings

This study demonstrates that Programme IMPACT delivers measurable benefits across clinical, patient, and system outcomes, underscoring the need to expand home-based palliative care to meet the demands of an ageing population and rising prevalence of chronic diseases. There are three key findings. First, symptom control was effective, with an average decrease of 2.5 ESAS points, and more than half of symptom exacerbations showed clinically significant improvement. Second, the programme demonstrated high concordance between patients’ expressed EoL preferences and actualised outcomes. Third, when compared with control decedents who did not receive palliative care, programme decedents showed substantially lower healthcare utilisation and costs. Notably, 70.8% of patients enrolled in Programme IMPACT died at home and were 16.6 times more likely to do so compared to controls. Although the patient population differs, similar trends were observed in other local home palliative programmes. For example, a study involving advanced dementia patients reported 78.9% home deaths. This is in contrast with the 2019 national statistics where home death rates are substantially lower at 22.9% [25].

While these findings are largely positive, it is important to acknowledge variation across patient subgroups. Respiratory failure patients demonstrated significantly longer inpatient LOS (IRR: 1.19, p < 0.001) compared to controls, likely reflecting the inherent complexity of managing acute respiratory crises and the time required for adequate symptom stabilisation and caregiver preparation. This finding underscores that certain patient populations may require more intensive resource allocation despite the programme’s overall effectiveness. Collectively, these findings highlight the potential value of structured home palliative care programmes in facilitating improved EoL care, enabling home death as per the patient’s preference, and reducing healthcare cost and utilisation, though the intensity of resource requirements may vary across different patient subgroups.

Comparison with global and local evidence

Our findings are aligned with both international and local evidence on the benefits of home-based palliative care. A systematic review of nine studies, including patients with severe, progressive, and terminal illnesses, demonstrated significant reductions in healthcare utilisation and shorter length of stays. However, no significant differences in symptom control were observed [26]. Our results are further reinforced by local evidence. A Singapore study on advanced dementia patients receiving home-based palliative care similarly demonstrated improved symptom control, better quality of life, and reduced acute care use [27]. Despite differences in underlying conditions, the alignment of outcomes across disease groups underscores the broader effectiveness of home-based palliative care, including for non-cancer populations [27]. Taken together, these converging lines of evidence suggest that community-anchored palliative care models not only relieve symptom burden and reduce unnecessary hospital use but also ensure that the care delivered is aligned with patient preferences.

Strengths and limitations

Programme IMPACT was one of the first programmes piloted to provide home-based palliative care to non-cancer terminal patients. In addition to the clinical and cost benefits of the programme, pertinent information, such as the composition, skills, and operational structure of a multidisciplinary team, and the collaborative framework established with community providers, was vital to inform policymakers so that this care model can be adopted in other settings. The limitations of the study should be noted, particularly regarding study design. While the retrospective component included a control group, it was not feasible to establish a control group for between-group comparisons in the prospective cohort component. Given that the primary goal of the programme was to provide essential home care services to non-cancer patients with limited alternative care options, withholding this service to eligible patients assigned to a control group would have been ethically inappropriate. Furthermore, comparisons with other palliative care models were limited, as no alternative services for non-cancer patients were available in the local context at the time of Program IMPACT’s pilot study. To address this, we conducted within-group comparisons, such as pre-post evaluations of symptom control, assessed for clinically meaningful improvements, and established benchmarks in collaboration with policymakers. Nevertheless, future randomised trials are necessary to validate these findings, establish definitive causal relationships, and evaluate the effectiveness of structured home palliative care programmes, building on the preliminary evidence provided by this study. In the retrospective study, ACP information was not available in the administrative databases. Therefore, we were unable to compare the rates of ACP completion and care concordance between Program IMPACT and decedents who did not receive palliative care. Despite statistical adjustments, residual unmeasured confounding and bias may persist in the retrospective analysis and influence results to an extent. Lastly, cost savings may have been underestimated because the control group's costs for community-based healthcare services were not captured.

Policy implications

With Singapore’s ageing population, a holistic approach is critical to building a sustainable and patient-centred EoL care system. This requires strong hospital and community partnerships, timely assessment of patient needs and prognosis, and appropriate placement in affordable care settings. The overarching aim is to provide high-quality EoL care in the patient’s preferred location, minimising avoidable hospitalisation and upholding their wishes for a dignified death. Achieving this will depend on dedicated financing and resources to expand home and community-based services, complemented by workforce training in prognostication and communication, and interoperable IT systems to support continuity of care. Embedding non-cancer palliative care indicators into national performance frameworks would strengthen accountability, while early advance care planning and public education can empower patients and families to make informed decisions, [19] reduce reliance on acute hospitals, and ensure care aligns with individual values at the EoL.

Conclusions

Findings from this study add weight to existing evidence that home-based palliative care may be beneficial for patients with end-stage organ failure and severe frailty, including possible improvements in care alignment with patient preferences, and reduction in acute healthcare utilisation and costs. Programme IMPACT offers a scalable model that bridges hospital and community care, showing how palliative services can be extended effectively to non-cancer populations. As Singapore faces rapid population ageing and rising non-cancer disease burden, embedding such approaches into the health system will be critical.

Abbreviations

WHO

World Health Organisation

ESAS

Edmonton symptom assessment scale

MCID

Minimally clinical important difference

ACP

Advanced care planning

PI

Programme IMPACT

IP

Inpatient

LOS

Length of stay

ED

Emergency department

SOC

Specialist outpatient clinic

SD

Standard deviation

CI

Confidence interval

ESOF

End-stage-Organ Failure

FE

Frail end-of- life

EoL

End of Life

ESHF

End stage heart failure

ESRF

End-Stage Renal Failure

COPD

Chronic obstructive pulmonary disorder

Non-COPD

Non-chronic obstructive pulmonary disorder

HF

Heart failure

RespF

Respiratory failure

RF

Renal failure

Authors’ contributions

WHY secured the funding for the study, conceived the study, organised the project, and provided guidance throughout execution of the project. JM, PK and MJP designed the evaluation, extracted and analysed the data. LKS, PK, MJP and AM wrote the initial draft of the manuscript. All authors give final approval of the revision to be published.

Funding

This was funded by Singapore’s Ministry of Health (MOH Reference 24:42/5—16). The funders had no role in the collection, analysis, or interpretation of the data, in the writing of the report, or in the decision to submit the manuscript for publication. The researchers had full access to the data in the study and took responsibility for the integrity of the data and the data analysis.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval was obtained from the National Healthcare Group Domain Specific Review Board (NHG DSRB 2019/00583), and the study adhered to the Declaration of Helsinki. All data were de-identified and anonymised by an independent third party. Informed consent was obtained from all patients and/or their next of kin prior to participation in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.