Differences in survival and healthcare utilization amongst nursing home residents with advanced dementia and frailty compared to other end-of-life conditions: a retrospective cohort study

Abstract

Background

Many nursing home residents suffer from advanced dementia and frailty. Describing their survival patterns and healthcare utilisation while under palliative care enables planning of palliative care services around their needs. We aimed to determine survival differences between nursing home residents with advanced dementia and frailty, incurable cancer and end-stage organ failure, as well as differences in healthcare utilisation in the last 6 months of life.

Methods

A retrospective cohort study of 535 nursing home residents with end-of-life conditions enrolled in a palliative care programme between March 2022 to February 2024 and followed up until February 2025, using Kaplan-Meier survival curve analysis and Cox proportional hazards regression to determine survival and Chi-square test and one way Analysis of Variance (ANOVA) to examine differences in healthcare utilisation.

Results

There were 361 deaths. Median survival was longer for residents with advanced dementia and frailty (16.5 months, 95%CI:13.5–19.5 months, P < 0.001) compared to incurable cancer (8.1 months, 95%CI:5.1–11.1 months, P < 0.001) and end-stage organ failure (7.7 months, 95%CI:3.0-12.4 months, P < 0.001). Residents with incurable cancer (HR 1.62, 95%CI:1.18–2.22, P = 0.003) and end-stage organ failure (HR 1.63, 95%CI:1.15–2.33, P = 0.006) were more likely to die during follow-up. Residents with advanced dementia and frailty had fewer in-person reviews (2.4, P < 0.001) compared to those with incurable cancer (4.5, P < 0.001) and end-stage organ failure (3.2, P < 0.001). Hospitalisations and emergency department visits were similar across groups.

Conclusions

Nursing home residents with advanced dementia and frailty live longer and receive fewer in-person reviews at the last 6 months of life than those with other end-of-life conditions. Providers of palliative care would need to adjust service planning and resource allocation for the longer follow-up period required for this group.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12904-025-01957-2.

Background

With an aging population, more older adults are residing in nursing homes, with a greater proportion of them having severe disability, cognitive impairment [1], and end-of-life conditions [2]. These residents have high mortality rates, with < 50% surviving after 3 years from admission into the nursing home [3] and up to 12.6% die within 6 months [4]. Consequently, nursing homes are becoming de facto hospices, such as in England where 37.6% of those aged 85 and above die in a nursing home, and nursing home deaths contribute to approximately a third of overall mortality in New Zealand and Australia [5].

Providing palliative care in nursing homes is an appropriate strategy to ensure proper end-of-life care [6]. It has been shown to reduce emergency department admission rates decrease healthcare utilisation and improve satisfaction with end-of-life care in the nursing home [7–10]. Many residents also prefer to remain in the nursing home at the end of life [11] and the provision of palliative care respects their choice to do so.

Despite the imperative to provide palliative care in the nursing home, the uptake of palliative care services remains low amongst nursing home residents [12–14]. especially in Singapore. Nursing homes are poorly equipped to meet the palliative care needs of this frail population [15], often have poor nurse to resident ratios, high staff turnover, and routine medical care is provided by general practitioners who may not be able to render palliative care. This leads to high healthcare utilisation and potentially avoidable admissions with up to 25.5% to 69.7% residents being hospitalised in their last month of life [16–18].

Furthermore, the majority of older adults in nursing homes die from chronic and disabling diseases, with the end-of-life becoming a prolonged period marked by poor health and declining function [5]. Residents with dementia and multimorbidity are more strongly associated with dying in the nursing home [19] because it is challenging to prognosticate and identify residents who are at risk of deterioration, resulting in under provision of palliative care services [20]. This impedes timely referral to palliative care especially if access to palliative care is tied to an expected prognosis [21]. As a result, many nursing home residents with advanced dementia and frailty pass away with non-palliative approaches to care [22, 23] and experience burdensome treatment at the end of life such as hospitalisations, parenteral treatments and nasogastric tube placement [24].

The model of palliative care in nursing homes needs to be adjusted to meet the needs of this frail population that is likely to have different care requirements [25, 26] and less predictable dying trajectories [27]. Many studies for persons with advanced dementia and frailty have described survival and derived prognostication tools in very advanced stages of illness or towards the end of life [28–30]. There are fewer studies on earlier palliative care services in the nursing home [31]. While palliative care has been shown to decrease healthcare utilisation, differences in healthcare utilisation amongst nursing home residents with various end-of-life conditions are less frequently analysed. The intensity of palliative care service provision across different disease trajectories towards the end of life is also not well understood.

As palliative care becomes more widely implemented in nursing homes, understanding how to tailor services to resident needs is increasingly important. Characterizing survival trajectories across end-of-life conditions, along with associated patterns of healthcare utilisation, can inform service planning. Therefore, we aimed to analyse the survival differences between residents with advanced dementia and frailty compared to those with other end-of-life conditions. We also aimed to analyse the differences in healthcare utilisation within the last 6 months of life between these groups.

Methods

Study setting and design

We conducted a retrospective cohort study of residents in 9 nursing homes who were enrolled into a palliative care programme (Care@NH) in Singapore between 1 st March 2022 to 29th February 2024 and followed up until 28th February 2025. The study period was selected to coincide with the relaxation of COVID-19 restrictions and minimize confounding effects of pandemic-related shifts toward remote delivery of palliative care. A one-year follow-up was selected because eligibility for palliative services in Singapore requires an estimated prognosis of one year, and most enrolled residents were expected to die within this timeframe.

Residents were included if they were enrolled into the program and had incurable cancer, end-stage organ failure, or advanced dementia and frailty. Individuals were classified into these groups to reflect the three distinct end-of-life illness trajectories [32]. In this study, residents with incurable cancer were defined as residents with locally advanced or metastatic cancer who were not on treatment or receiving treatment with palliative intent. Residents with end-stage organ failure were included based on the Gold Standard Framework [33]. Persons with advanced dementia and frailty were defined as having a Functional Assessment Staging Tool (FAST) score of ≥ 7 A [34] or a Clinical Frailty Scale (CFS) score of ≥ 7 [35] – these residents had total dependency in their activities of daily living. They also had evidence of deterioration such as significant weight loss (≥ 10% weight loss over a 6-month period), rapid functional decline or ≥ 2 admission in the past year. In situations where residents had more than one co-existing end-of-life condition, they were classified according to the dominant disease trajectory. We excluded residents who were enrolled into the programme for complex care coordination but did not have end-of-life conditions. Also excluded were residents enrolled after a compassionate discharge from hospital (terminally ill residents who are transferred back to the nursing home to spend the last days) and those referred by the nursing home for terminal care, this group of terminally unwell residents differed from the rest of the study population and would contravene the proportional hazards assumptions. Residents disenrolled from the program, discharged from the service or discharged from the nursing home were censored in the final data analysis.

The Care@NH programme provides palliative care in partnered nursing homes within the central region of Singapore. Residents could be referred if they had end-of-life conditions or high clinical needs requiring specialist palliative care intervention. Referrals were initiated by the tertiary hospital on discharge, nursing home staff or identified through proactive screening at the nursing home. Care was co-managed by a palliative care team from a tertiary hospital and the nursing home staff which included nurses, allied health therapists and social workers. The palliative care team consisted of a multidisciplinary team of nurses and physicians who delivered care through in-person or telephonic reviews with additional after-hours support for acute deterioration and end-of-life symptom control. Review frequency was guided by the Australia Subacute Non-Acute Patient Classification phases. Nursing home staff were trained to recognise end-of-life symptoms and medication administration, provided nursing care, psychosocial care and physical therapy. All residents completed an Advanced Care Plan (ACP) prior to enrolment to discuss their goals of care and preferences related to medical interventions such as cardiopulmonary resuscitation, hospitalisation, nasogastric tube placement, and their preferred place of death. For residents lacking decision-making capacity, these discussions were conducted with an appointed surrogate decision-maker. At the end of life, residents could receive specialist palliative care from the Care@NH team, including subcutaneous infusions of medication for symptom relief in the nursing home.

Baseline characteristics

We collected baseline demographic data including age, gender, ethnicity, and the nursing home of residence. Functional status, Palliative Performance Scale (PPS) [36], Edmonton Symptom Assessment Scale (ESAS) [37] and ACP outcomes were captured upon enrolment into the programme and at each subsequent in-person review. Based on retrospective review of the available electronic health records upon enrolment into the programme, we measured comorbidities using the Charlson’s Comorbidity Index (CCI) [38] and derived a 30-item electronic frailty index (eFI) score using a score of ≥ 0.25 to determine frailty [39, 40] (Supplement 1).

Outcomes

The primary outcome of death was collected from audited data from the programme. We analysed healthcare utilisation in the last 6 months of life as a suitable benchmark to measure resource utilisation for palliative care [41]. To examine the secondary aim, healthcare utilisation data that was collected included the number of telephonic and in-person reviews conducted by the Care@NH team using audited data. We extracted the number of distinct emergency department visits and hospitalizations in the last 6 months of life from a review of electronic health records.

Statistical analysis

Baseline characteristics were analysed using descriptive statistics with continuous variables recorded as mean and standard deviation, while categorical variables were expressed as absolute numbers and percentages.

Time-to-event analysis using death as the outcome was performed using Kaplan-Meier survival curve analysis to identify median survival times for each group. Log-rank test was used to determine if any difference in survival was statistically significant. In addition, multivariate analysis was performed using Cox proportional hazards regression to determine the hazards ratio between each group with log-minus-log plots to test the proportional hazards assumption. Differences in healthcare utilisation between each of the three groups were analysed using Chi-square test and one way Analysis of Variance (ANOVA), with post hoc Bonferroni correction where applicable.

Statistical analyses were performed using IBM SPSS Statistics version 27.0 (IBM Corporation, Armonk, NY, USA). All statistical tests were two-tailed, with P < 0.05 considered statistically significant.

Results

We analysed 604 referrals during the 2-year period, amongst which 580 residents were enrolled into the programme. 45 residents were excluded, of whom 44 did not have life-limiting conditions and 1 died prior to their first review. 535 residents were included in the final analysis – this comprised 25 residents who were discharged to other services, 361 decedents and 149 survivors (Fig. 1).

Fig. 1.

Flow Diagram

Cohort characteristics

The mean age of the residents was 81.2 ± 10.6 years old. The majority were Chinese (N = 473, 88.4%), men (N = 271, 50.7%) and most were bedbound (N = 272, 50.8%). The cohort was frail (eFI = 0.48 ± 0.08) and had an average CCI of 7.1 ± 2.6. Out of 535 residents, 341 (63.7%) were enrolled due to advanced dementia and frailty, 106 (19.8%) due to incurable cancer and 88 (16.4%) for end-stage organ failure. (Supplement 2). Amongst those who died in the nursing home (N = 192), ESAS was captured at the last physical review prior to death. The average ESAS prior to death was 7.4 ± 5.5 with 103 (53.9%) having at least one ESAS domain scored > 3.

Across the groups (Table 1), a significantly larger proportion of residents with advanced dementia and frailty were bedbound (62.5%), had a poorer PPS, as well as a higher eFI, indicative of their poorer functional status and frailty state on enrolment. The CCI was significantly higher for the advanced cancer cohort (9.3 ± 2.6), because of the higher weight given to metastatic cancer within the index. Amongst the ACP discussions, a greater proportion (N = 140, 41.1%) of the residents with advanced dementia and frailty opted for “Do Not Hospitalize” in the event of deterioration compared to residents with incurable cancer and end-stage organ failure. Amongst those who had died in the nursing home, there was no difference in the ESAS score prior to death between the groups (Incurable cancer 6.96 ± 5.8 vs. end-stage organ disease 7.03 ± 5.4 vs. Advanced Dementia and Frailty 7.63 ± 5.5, P = 0.734).

Table 1.

Comparison of baseline Characteristics, mortality and healthcare utilization between residents with advanced Dementia/Frailty, incurable cancer and end stage organ failure

	Advanced Dementia/Frailty N = 341	Incurable Cancer, N = 106	End Stage Organ Failure N = 88	P
Demographics
Age, years	81.7 ± 11.0	80.0 ± 10.2	80.6 ± 9.6	0.339
Gender, Male	169 (49.6)	58 (54.7)	44 (50.0)	0.645
Ethnicity, Chinese	309 (90.6)	92 (86.8)	72 (81.8)	0.067
Functional Status and Comorbidities
Function, Bedbound	213 (62.5)	36 (34.0)	23 (26.1)	< 0.001
Palliative Performance Scale	36.2 ± 8.4	42.0 ± 9.7	42.3 ± 9.3	< 0.001
Charlson Comorbidity Index	6.1 ± 2.0	9.3 ± 2.6	8.1 ± 2.4	< 0.001
Electronic Frailty Index	0.49 ± 0.07	0.45 ± 0.08	0.50 ± 0.09	< 0.001
Advanced Care Plan Outcomes
Do Not Hospitalise*	140 (41.1)	35 (33.0)	23 (26.1)	0.023
Change in Preferred Place of Care†	43 (12.6)	13 (12.3)	15 (17.0)	0.519
Mortality
Death	217 (63.6)	78 (73.6)	66 (75.0)	0.029
Healthcare Utilization at Last 6 Months of Life
Number of Hospitalisation	1.1 ± 1.4	1.3 ± 1.3	1.2 ± 1.3	0.480
Number of ED Visits	0.1 ± 0.6	0.1 ± 0.3	0.0 ± 0.2	0.815
Number of Telephonic Reviews	8.3 ± 9.0	8.1 ± 7.2	6.9 ± 6.3	0.440
Number of In-person Reviews	2.4 ± 2.8	4.5 ± 3.1	3.2 ± 3.0	< 0.001

All values expressed as Mean ± Standard Deviation or N (%)

*Decision for preferred place of care following advance care planning (ACP)

^†Change in ACP decision during course of enrolment

Mortality

There were 361 decedents during the follow-up period – 78 (73.6%) deaths occurred amongst residents with incurable cancer, 66 (75.0%) occurred amongst residents with end-stage organ failure and 217 (63.6%) occurred amongst residents with advanced dementia and frailty. The overall median survival of all residents was 12.8 ± 1.1 months.

On univariate time-to-event analysis using Kaplan-Meier survival curve analysis, the median survival of residents with advanced dementia and frailty (16.5 ± 1.5 months) was significantly longer (P < 0.001) than incurable cancer (8.1 ± 1.5 months) and end-stage organ failure (7.7 ± 2.4 months) based on the Log-rank test (Fig. 2).

Fig. 2.

Kaplan-Meier Curve Comparing Survival Between Residents with Advanced Dementia/Frailty, Incurable Cancer and End Stage Organ Disease

In the multivariate cox regression analysis (Table 2), Having advanced dementia and frailty was independently associated with a reduced risk of mortality compared to other disease profiles, even after adjusting for age, gender, CCI, eFI, PPS and functional status. A decision for “Do Not Hospitalize” following ACP was significantly associated with an increased risk of mortality (HR 1.72, 95% CI 1.35–2.21, P < 0.001). Higher CCI contributed to a greater risk for mortality (HR 1.10, 95% CI 1.04–1.17, P < 0.001) compared to eFI (HR 2.28, 95% CI 0.42–12.46, P = 0.34) and functional status (HR 1.94, 95% CI 0.94–4.00, P = 0.72). The log-minus-log plots show parallel lines for the different covariate groups, indicating that the proportional hazards assumption is met.

Table 2.

Cox proportional hazards analysis

Outcome Variables	Adjusted HR (95% CI)*	P
Age	1.02 (1.01–1.03)	< 0.001
Sex, Male	1.41 (1.12–1.76)	0.003
Ethnicity, Chinese	1
Malay	1.20 (0.71–2.01)	0.500
Indian	1.32 (0.78–2.24)	0.300
Eurasian	0.71 (0.31–1.65)	0.427
Do Not Hospitalise Decision	1.72 (1.34–2.21)	< 0.001
Charlson Comorbidity Index (CCI)	1.10 (1.04–1.17)	< 0.001
Electronic Frailty Index (eFI) †	2.28 (0.42–12.46)	0.343
Ambulatory status, Bedbound†	1.94 (0.94–4.00)	0.072
Palliative Performance Scale (PPS) †	0.98 (0.97–1.00)	0.044
Advanced dementia and frailty	1
End Stage Organ Failure	1.63 (1.15–2.33)	0.006
Advanced Cancer	1.62 (1.18–2.22)	0.003

*Adjusted for age, ambulatory status, CCI, eFI, ethnicity, gender, nursing home, PPS

^† Variance Inflation Factor (VIF) analysis was performed for Ambulatory status, eFI and PPS to check for multicollinearity. VIF < 5 for these domains

Healthcare utilization at the last 6 months of life

The number of hospitalisations and visits to the emergency department were similar across groups. In the last 6 months of life, residents with advanced dementia and frailty had significantly fewer physical reviews (2.4 ± 2.8, P < 0.001) compared to residents with incurable cancer (4.5 ± 3.1, P < 0.001) and end-stage organ failure (3.2 ± 3.0, P < 0.001). 192 residents passed away in the nursing home. In this subgroup, encounters with the Care@NH team increased sharply in the last month of life (Fig. 3). Additionally, residents with incurable cancer had more in-person reviews at 4 months (0.8 ± 0.0 vs. 0.2 ± 0.1, P < 0.001) and 3 months (1.1 ± 0.4 vs. 0.4 ± 0.2, P = 0.001) before death compared to those with advanced dementia and frailty.

Fig. 3.

Number of Encounters by Care@NH Team in the Last 6 Months of Life

Discussion

We evaluated survival differences between nursing home residents with varying end-of-life conditions who received end-of-life care from the Care@NH team. Specifically, we observed that residents with advanced dementia and frailty survived longer despite being frailer and more functionally impaired compared to other residents with other end-of-life conditions. In addition, they received fewer in-person reviews at the end of life by the Care@NH team, indicating differences in palliative care delivery in the last 6 months of life.

Our findings reinforce the difficulty of prognostication for individuals with advanced dementia and frailty [42]. Whilst functional ability indicates the severity of dementia, it is not a reliable predictor for mortality; the FAST score of 7C was widely used as a criteria for hospice eligibility in the United States but was found to be a poor predictor of 6-month mortality [43]. It also assumes a specific sequence of functional deterioration, which may not be the case in the face of multimorbidity. Our study demonstrates similar findings, as functional status was not associated with mortality in residents with advanced dementia and frailty. Despite common understanding that frailty predicts mortality [44, 45], worsening frailty as determined by the eFI was not associated with mortality in our study. This is consistent with a previous study that showed that a single time point eFI is a poor predictor of mortality [46] due to its inability to account for individual variations in frailty progression and the equal weighting of all deficits discounts the proportional effect of comorbidities on mortality. The eFI may also have poorer discriminative power amongst nursing home residents given that they were frail upon admission to the nursing home irrespective of their diagnosis.

The findings suggest that disease-specific prognostic scores, such as ADEPT [28], PalS-DEM [29] or PRO-MADE [30] may need to be considered for accurate prognosis and to guide eligibility for palliative care programmes in nursing home residents with advanced dementia and frailty. However, these tools have only moderate predictive ability and may require blood tests, making them less applicable in the nursing home setting. Furthermore, many of these tools are highly specific and perform better at predicting high risk individuals with shorter prognosis; they may be inappropriate for an early palliative care approach and may underrepresent individuals with high care needs who need palliative care. An alternative is to use nursing home-specific predictive tools such as the MDS-CHESS [47–49] or the Flacker Mortality Index [50]. These tools are more feasible as they are derived using routinely collected clinical information from nursing home care. However, these tools were developed using the United States Minimum Data Set and their predictive validity in nursing homes across various geographical settings is yet to be ascertained because of variations in the makeup of nursing home residents and differing standards of care. In summary, real-world challenges exist when relying on prognostic estimates to determine eligibility for palliative care services for nursing home residents with advanced dementia and frailty.

Our study also showed that there were fewer in-person reviews to deliver end-of-life care for residents with advanced dementia and frailty compared to other end-of-life conditions, suggesting that residents were stable or experienced fewer clinically complex palliative care needs. In usual community-based palliative care services, in-person reviews by a multidisciplinary team are usually arranged for patients who are actively deteriorating, requiring closer follow-up, or needing acute symptom control. Given that advanced dementia and frailty is inherently difficult to prognosticate, it is challenging to predict future deterioration and assess the need for in-person follow up. Furthermore, symptoms for residents with advanced dementia and frailty may not be obvious until death is imminent because of a terminal event. Palliative care needs in this group may focus more on good nursing care and effective communication with residents and their next-of-kin to prepare them for potential deterioration, which may not require frequent in-person visits by the specialist palliative care team. However, we also acknowledge that cognitive impairment may also lead to underreporting of symptoms [51], as in-person reviews are often triggered by symptoms reported by nursing home staff. Our findings suggest that a model of care with less emphasis on in-person review or a telehealth approach can be appropriate to deliver palliative care for these residents [52].

We observed that a higher proportion of residents with advanced dementia and frailty chose not to be hospitalised in the event of deterioration. This may be due to a preference to age in place, focus on comfort care, and perceived distress or lack of benefit from recurrent hospitalisations [53, 54]. With the overall longer survival for this group of residents, providers of palliative care need to adjust service planning to account for the longer follow-up period required, yet al.so be ready to allocate resources if the need arises to ensure that symptoms are adequately treated to support aging in place, especially in the last month of life when there is demand for greater intensity of care.

Another notable finding is that residents with end-stage organ failure had a similar survival pattern to that of residents with cancer, contrary to concurrent evidence which shows otherwise. We postulate this is because approximately half in the end-stage organ group comprises residents with stage V chronic kidney disease who declined renal replacement therapy (Supplement 2), which has a trajectory of illness similar to terminal cancer [55].

In summary, our study indicates that the design of palliative care services for advanced dementia and frailty needs to be re-examined. These is evidence that an early palliative care approach is beneficial for patients living with dementia [56], and we need to re-evaluate how we can deliver this service in a sustainable manner. A narrow, prognosis-centred criterion for access to palliative care services may limit the number of patients who can benefit from the service, while a broader criterion could result in a longer follow-up period and strain resources amongst hospice providers. This supports development of other paradigms, such as a needs-based approach, to determine eligibility for palliative care for residents in nursing homes with advanced dementia and frailty [42, 57]. The ability to deliver palliative care with fewer in-person visits also suggest that a less resource-intensive model of delivering palliative care (such as through telehealth) is feasible.

Strengths and limitations

Our study had a large cohort with minimal loss to follow-up. The overall duration of follow-up was longer, describing survival patterns for a timely palliative care approach rather than just provision of terminal care.

Our study had the following limitations. As a retrospective cohort study, we could not identify all possible confounders affecting mortality at the end-of-life. However, we adjusted for frailty status using the eFI and incorporated factors known to affect prognosis such as comorbidity and overall performance. We also lacked access to national healthcare records, limiting our ability to capture all hospitalisations and emergency visits from other healthcare clusters. Nonetheless, given that all healthcare attendances were documented in the Care@NH medical records, the numbers of missed healthcare encounters are likely to be minimal. For residents with multiple terminal illnesses, clinical judgement was required to determine the primary life-limiting condition and the effect of having multiple end-of-life conditions on survival could not be examined. Finally, ESAS data was not available for residents who passed away in the hospital, limiting symptom burden comparison for all enrolled residents.

Recommendation

We propose that future studies could compare disease-specific and nursing home-specific prognostic tools to determine which is the best for prognostication in residents with advanced dementia and frailty. Additionally, an assessment of needs and complexity should be used alongside prognostication tools to guide the design and allocation of specific palliative care services for this population to maximize service uptake and sustainability. Further studies can also investigate the perspectives of nursing home staff on their ability to provide palliative care and the quality of end-of-life care rendered.

Conclusion

In conclusion, nursing home residents with advanced dementia and frailty live longer under palliative care and receive fewer in-person reviews at the end of life compared to other life-limiting conditions. Providers of palliative care need to adjust service planning and allocate resources adequately for the longer follow-up period required for this group of residents. This suggest that novel care models of palliative care are needed to meet the unique end-of-life needs for these residents.

Abbreviations

FAST

Functional Assessment Staging Tool

CFS

Clinical Frailty Scale

ACP

Advanced Care Plan

PPS

Palliative Performance Scale

ESAS

Edmonton Symptom Assessment Scale

CCI

Charlson’s Comorbidity Index

eFI

electronic frailty index

ANOVA

Analysis of Variance

Authors’ contributions

Y.C and J.S.L were involved in the conceptualization, design of the study, data collection, data analysis, data interpretation and writing of the manuscript. B.Y.P, G.O.S and L.L.S were involved in the data collection. W.S.L was involved in data analysis and interpretation. All authors contributed to the writing and revision of the manuscript.

Funding

No funding was received for the conduct of this study.

Data availability

All data generated or analysed during this study are included in this published article [and its supplementary information files].

Declarations

Ethics approval and consent to participate

Routinely audited and anonymised information was used in this retrospective study. In accordance with the Declaration of Helsinki and the Belmont Report, the National Healthcare Group Institutional Review Board did not require ethics approval for the study based on institutional policy. Informed consent was also not required.

Competing interests

The authors declare no competing interests.