Population-based palliative care planning in Ireland: how many people will live and die with serious illness to 2046?

Background: All countries face growing demand for palliative care services. Projections of need are essential to plan care in an era of demographic change. We aim to estimate palliative care needs in Ireland from 2016 to 2046. Methods: Static modelling of secondary data. First, we estimate the numbers of people in Ireland who will die from a disease associated with palliative care need. We combine government statistics on cause of death (2007-2015) and projected mortality (2016-2046). Second, we combine these statistics with survey data to estimate numbers of people aged 50+ living and dying with diseases associated with palliative care need. Third, we use these projections and survey data to estimate disability burden, pain prevalence and health care utilisation among people aged 50+ living and dying with serious medical illness. Results: In 2016, the number of people dying annually from a disease indicating palliative care need was estimated as 22,806, and the number of people not in the last year of life aged 50+ with a relevant diagnosis was estimated as 290,185. Equivalent estimates for 2046 are 40,355 and 548,105, increases of 84% and 89% respectively. These groups account disproportionately for disability burden, pain prevalence and health care use among older people, meaning that population health burdens and health care use will increase significantly in the next three decades. Conclusion: The global population is ageing, although significant differences in intensity of ageing can be seen between countries. Prevalence of palliative care need in Ireland will nearly double over 30 years, reflecting Ireland’s relatively young population. People living with a serious disease outnumber those in the last year of life by approximately 12:1, necessitating implementation of integrated palliative care across the disease trajectory. Urgent steps on funding, workforce development and service provision are required to address these challenges.


Introduction
Background Population ageing reflects significant advances in medicine and wider society, but increasing longevity brings new health care challenges in all countries. More people are living longer with serious illness, multimorbidity and functional impairment, creating an urgent need to develop workforce and grow service capacity 1,2 . This is not only a question of upscaling existing services but reconfiguring systems originally designed to provide acute, episodic treatment 3 .
Palliative care is an approach that aims to improve pain and symptom management, communication and planning in care of people with serious medical illness 4 . Palliative care may be provided by specialists, whose core activity is palliative care, and generalists, who are not engaged full time in palliative care but have some relevant training that allows them to practice a palliative care approach as part of usual service provision 5 .
Originating in end-of-life care, palliative care approaches are now widely seen as having potential benefits across the trajectory of serious disease including improved quality of life and, in some circumstances, extended survival 6-8 . The appropriate mix of specialist and generalist involvement in care may be determined by the persistence and complexity of needs, with greater specialist involvement for patients with more complex needs 9 . Some services, notably bereavement support, focus predominantly on the end-of-life and post-death period. Palliative care access is highly variable internationally, and under-supply and unmet need are widely reported even in countries with relatively strong provision 10,11 .
A 2017 study by Etkind et al. used cause-of-death data to project palliative care need in England and Wales to 2040 12 . They estimated that the number of deaths will increase 25% in a 25-year period, and the number of deaths from a disease indicating palliative care need (e.g. cancer, organ failure, Alzheimer's and related dementias) will rise 25-43%, depending on projection method. Expected increases are due to population ageing and growing prevalence of serious chronic disease, with the biggest effects observed in those aged 85+. The United Kingdom has an old population relative to the European Union average 13 . Countries with younger populations, irrespective of high-, middle-and low-income status, can expect faster rates of growth in palliative care need 1 .

Rationale and aim
All countries face inadequate current supply and growing future demand for palliative care services. Accurate projections of future need are essential to inform expansion of current services and development of new models of care in an era of demographic change. Quantifying deaths from a disease indicating palliative care need is a critical first step. Equivalent cause-of-death projections for Ireland to those made in England and Wales do not currently exist, and this study was funded by a Health Service Executive Healthy Ageing Award to address the Irish evidence gap. We detail our methods and data in full to assist researchers in replicating and extending our approach in their own countries.
It is also important to move beyond estimations of need that are based solely on cause of death. This is because cause of death data recorded on death certificates are often subjective where the decedent had one or more condition 14,15 , and risk undercounting palliative care needs if the cause of death does not indicate palliative care need but concurrent unrecorded conditions do. More significantly, policy and clinical guidelines increasingly recommend palliative care across the trajectory of serious disease, meaning that those dying from these diseases are only a subset of the overall population health need 6,16 . To inform decision-making in policy and practice, including the targeting of interventions to those who benefit, groups cannot be defined by their characteristics at death and instead must be identified prospectively on the basis of clinical and other relevant factors 17,18 .

Design
Secondary research study of already-collected data. In Analysis 1, we used routinely accessible statutory data in Ireland on recorded deaths 2007-2015, and projected mortality 2016-2046. In Analyses 2 and 3, we retained projected mortality data, and combined these with projected population data 2016-2046 and observed individual-level data from a prospective longitudinal study on ageing.

Setting
Ireland has a relatively young population among high-income countries but faces typical trends in global ageing 19 . Palliative care services are well established by international standards 20 and since 2001 are recognised by a national policy recommending universal provision on the basis of need 21 . However, some aspects of policy are yet to be implemented resulting in gaps in access to aspects of generalist and specialist services in all parts of the country 22 . This is in the context of variable access for non-palliative care services: a means-tested medical card grants free primary and hospital care, and subsidized prescription medicines; people without a medical card pay capped co-payments for hospital care and prescriptions, and full primary care costs out of pocket 23 . Ireland has the second highest level of self-reported unmet needs for specific health care-related services due to financial reasons in the European Union 24 .
Studies have shown that decedents and their close family members still experience suboptimal outcomes that include unmanaged pain and depression, and place of death inconsistent with their preferences 25,26 . An update of the policy is expected to start in 2020 27 . This will take place in the wider context of Sláintecare reforms, a wide-ranging set of recommendations to embed universal entitlements into the Irish health and social care system including palliative care access 28 . We attached values to responses so that 0=not regularly troubled by pain; 0.25=mild pain; 0.5=moderate pain and 1=severe pain.

Data sources
• Health and social care use. General practitioner (GP) visits, emergency department (ED) admissions, overnight acute hospital inpatient admissions and hours of home help received were collected as frequency data: how many times was each category used in the last 12 months.
Each of these outcomes was calculated for all TILDA Wave 3 participants, and for all TILDA decedents to Wave 5 with an end-of-life interview. We calculated proportion of 2015 deaths from a relevant disease and combined this proportion with the total number of projected deaths annually in Ireland to 2046.
• Method 2b: Assuming annual change, prior eight years. The proportion identified in 2a is not assumed to be constant but, rather, increases according to a compound interest rate, r, which is calculated as mean rate of annual change over the prior nine years (i.e., 2007 to 2015 inclusive).
• Method 2c: Assuming annual change, prior three years. A replication of 2b, except that r is calculated as mean rate of annual change over the prior four years (i.e., 2012 to 2015 inclusive).
• Method 2d: Assuming annual change, by age and gender. A replication of 2b, except that r is calculated by gender for each five-year age band.
Analysis 2: estimated total number of people aged 50+ in Ireland living and dying with a disease indicating palliative care need to 2046. We estimated the number of people dying in Ireland in a given year, by age and gender, from the CSO mortality projections (2016-2046). We estimated the number of people alive in Ireland throughout a given year, by age and gender, by subtracting mortality projections from CSO population projections (2016-2046). We calculated the proportion of people aged 50+ living with a disease indicating palliative care need, by age and gender, in TILDA Wave 3. We calculated the proportion of people aged 50+ dying with a disease indicating palliative care need, by age, in TILDA end-of-life interviews.
For each year, we multiplied the number of people projected by the CSO to live through the year by the proportion in TILDA living with a disease indicating palliative care need, and we multiplied the number of people projected by the CSO to die by the proportion in TILDA dying with a disease indicating palliative care need.
Step-by-step calculations can be seen in Underlying/extended data 34 .

Analysis 3: estimated outcomes among people aged 50+ in Ireland living and dying with a disease indicating palliative care need to 2046.
For each outcome of interest (disability burden, pain prevalence, utilisation categories), we calculated the mean, adjusted for age, gender and palliative care disease, among people aged 50+ in TILDA Wave 3 and end-of-life interviews. We combined these calculations with the population projections from Analysis 2.
To quantify growth in disability and pain burden, and use of different health care services, in a single comprehensible index we set values in each outcome in 2016 to 100. We then created indices to 2046 with this 2016 value as a base (so a 50% increase in any outcome to 2046 gives a 2046 score of 150, etc.). Health service utilisation is therefore based on current patterns, implicitly assuming no changes in policy or access during the projection period.
Step-by-step calculations can be seen in Underlying/extended data 34 .

Results
Analysis 1: estimated total number of people in Ireland dying from a disease indicating palliative care need to 2046 Total number of projected deaths, and estimated proportions of these deaths from a disease associated with palliative care need, are presented in Table 2 and Figure 1.  Using any method that incorporated changing needs (2b, 2c, 2d), total number of projected deaths from a disease associated with palliative care need in 2046 ranged from 39,081 (=80% of all deaths in 2046; method 2d) to 40,355 (=83% of all deaths of all deaths in 2046, method 2b). These represent a 2016-2046 increase in absolute numbers of deaths from a disease associated with palliative care need between 78% and 84%.
Projected numbers of deaths from a disease associated with palliative care need by age are presented in Figure 2. Both absolute numbers of these deaths at each time point, and the projected growth in numbers over time, are heavily driven by the oldest age groups (85 years and over). Analysis 3: estimated outcomes among people aged 50+ in Ireland living and dying with a disease indicating palliative care need to 2046 Estimated increases in outcomes for older people dying with a disease indicating palliative care need are presented in Figure 4. Disability burden was projected to increase 96% between 2016 and 2046, and pain burden is projected to increase 68%. Large increases in health care use were also predicted for all categories: GP visits (74%), emergency department admissions (65%), inpatient hospital admissions (62%) and home health hours (107%).
Estimated increases in outcomes for older people living through the year with a disease indicating palliative care need is presented in Figure 5. Disability burden is projected to increase 173% between 2016 and 2046, and pain burden is projected to increase 83%. Large increases in health care use are also predicted for all categories: GP visits (101%), emergency  department admissions (90%), inpatient hospital admissions (100%) and home health hours (106%).
The proportion of estimated outcomes accounted for by those living and dying with diseases indicating palliative care are presented in Table 4.
People in the last year of life with a palliative care need account heavily for observed outcomes. This group was 2% of those aged 50+ in 2016, yet they are estimated to account for 25% of the disability burden and 3% of the pain burden. Under current patterns of service use, they also account for 5% of GP visits, 15% of ED admissions, 10% of inpatient admissions and 11%   People living with a palliative care need also account disproportionately for outcomes, although this disproportionality is not so pronounced. This group was 20% of those aged 50+ in 2016, yet they are estimated to account for 39% of the disability burden, 25% of the pain burden, 29% of GP visits, 37% of ED admissions, 34% of inpatient admissions and 46% of home help hours. All proportions increase moderately to 2046, as this group increases to 24% of all people aged 50+.

Key results
The number of people in Ireland dying from a disease associated with palliative care need is estimated to increase 68-84% between 2016 and 2046, depending on projection method employed (    For full details of each outcome, see Methods>Outcome variables. 'Living with' = live the entirety of a given year; 'Dying with' = die within a given year; all data points represent the outcome total in a given year for a given population as a proportion of total outcome in all people aged 50+ in Ireland. (from 76% in 2016 to 80-83% in 2046) ( Table 2). Increases are most observable in the 85+ age group (Figure 2).
The number of people living with a disease associated with palliative care need outnumber those in the last year of life with a relevant diagnosis by 12:1, and this ratio is projected to remain fairly constant as both groups increase in absolute size over the next 30 years (Table 3, Figure 3). People living and dying with a disease associated with palliative care need account disproportionately for disability burden, pain prevalence and health care utilisation (Table 4). Per person burden is highest among those who are in the last year of life; total burden is larger among those with life expectancy of more than a year because they are a much larger group. Both groups, and so their associated health burdens and care needs, will increase significantly in size to 2046 (Figure 4, Figure 5).

Limitations
All projections use static modelling to extrapolate future trends.
In prevalence projections we implicitly assume no change in disease profile by age and gender, but long-term changes are expected; e.g. smoking prevalence is on the decline so in the future older people should have fewer smoking-related problems than the current cohort, obesity prevalence is on the rise so future cohorts will have more related problems, future disability burden may grow more slowly than disease prevalence due to assistive technologies 37 . Palliative care need is defined by prevalence of specific conditions but broader definitions of eligibility, e.g. to include multimorbidity 38 , frailty 39 and pain prevalence 40 , could be defended and would deliver larger projected increases 12 . Future research may wish to revisit the definition of palliative care need in routine data 41 and longitudinal ageing surveys 42 , particularly among the oldest old for whom a single diagnosis is not typically the best indicator of need.
In outcome projections we implicitly assume no exogenous changes in these; e.g. reduced pain through improved prescribing policies, increased service use due to ageing among unpaid family carers, reduced hospital use through health service reconfiguration 28 . Health care use is estimated based on past use not need, so unmet need may be uncounted and unnecessary use over-counted. Dynamic modelling that takes account of the interaction of different variables over time, changing health profiles across generations and an evolving policy landscape, as well as cross-validating predictive accuracy and quantifying the uncertainty associated with all stages of projections, would offer a more sophisticated picture of future population numbers and their associated outcomes. Such analyses are planned.
All CSO population projections are based on assumptions of future fertility (F) and net migration (M) 43 ; our estimates rely on the CSO's mid-range assumptions 'F2' (total fertility rate to decrease to 1.8 by 2026 and to remain constant thereafter) and 'M2' (net migration returning to positive by 2018 and rising slowly thereafter to plus 10,000 by 2021). Since deaths occur overwhelmingly among older age groups, and fertility and immigration impact demographics mainly among younger people, our headline conclusions are robust to alternative assumptions.
In Analysis 1, all data are reliant on death certificates whose reliability as a source of death is sub-optimal and variable 14,15 . In Analyses 2 and 3, disease prevalence and outcome data were available only for people aged 50+. While this group accounts for the large majority of people living and dying with diseases relevant to palliative care, there are also needs in younger groups that future analyses must address.
TILDA uses self-report data in regular waves, and proxy endof-life interviews, both of which are subject to biases 44,45 . TILDA does not ask specifically about all causes of death listed in Table 1, risking under-estimate of relevant deaths. We reviewed CSO cause-of-death data, counting only diseases recorded by TILDA, and found a 1% discrepancy (Underlying/extended data: 20191101 Appendix TILDA prevalence 34 ), so we believe this risk to be low. However, neurological diseases are usually associated with high burden of disability so future work should investigate this potential undercounting further. TILDA does not ask about pain specifically associated with terminal illness, so reported pain burden is as a proportion of all chronic pain self-reported by older people.
Analysis 1 quantifies people dying from a given set of diseases, Analyses 2 and 3 people dying with those diseases. The former are a subset of the latter. Our estimated 74% increase in those dying with a relevant disease ( While a small proportion of those living with relevant diseases require specialist care throughout their disease trajectory, many will benefit from some specialist palliative care input. Although occasional or episodic involvement of specialist palliative care services is proposed as the best practice model of care for Ireland 47 , significant barriers to providing integrated palliative care exist 48 . Focused efforts will be required to overcome these obstacles.
Notably the disability burden -as defined by the number of everyday tasks people require help with -will nearly treble among older people in the next 30 years unless novel treatment modalities are associated with improved disability-free survival rates. Lack of appropriate supports in everyday tasks is an established risk for health deterioration and avoidable service use 39,49,50 . Reconfiguring care provision and medical training for an age of multimorbidity and complexity is critical to the needs of this group being met. System-wide projections suggest large increases in demand for services in all sectors even under optimistic assumptions around healthy ageing and service reconfiguration 51 .
These trends are imposing, particularly given the challenges already facing the Irish health service, but those responsible for planning health care services are forewarned. Previous studies have found that both specialist and generalist palliative care capacity and funding in Ireland are well short of levels required to meet current need 22,52 . Universal access to generalist and specialist services will require multiplicative increases in available budgets 28,52 . Resource allocation is necessary but not the sole requirement to meet this growing demand. One foremost supply-side challenge is the identification, training and retention of staff to provide services on a scale outlined in our analyses, although this challenge is faced by the whole health care system. While workforce planning and development is always a complex task, there are particular issues related to palliative care services. For example, healthcare professionals have previously reported lacking the skills or confidence required to provide palliative care 53,54 . These deficits have been linked primarily to historical gaps in education and ongoing training in palliative care 47 . However, variation in knowledge or experience in Ireland may also be attributable to regional differences in service provision 55 .
Sláintecare project milestones for 2019 include the development of a revised national palliative care policy 56 . Ireland is recognised to have a high standard of palliative care provision 10,11,20 and previous policy has done much to enable this. However, it is timely to reflect on the observation that historically key policy goals in palliative care were not realised in Ireland because of "large resource commitments required; the competition for resources from other, better-established healthcare sectors; and challenges in expanding workforce and capacity" 52 . There is a growing literature focused on policy implementation that recognises that policies do not succeed or fail on their own merits, and that endeavours to close the policy-implementation gap. Chief among strategies recommended is improving policy design 57 . The process of policy design should include the use of a good quality evidence base and the detailed population-focused analysis of need presented in this paper aims to serve this purpose in order to advance palliative care service development in Ireland.

Conclusion
Annual deaths in Ireland from a disease associated with palliative care need are estimated to increase up to 84% between 2016 and 2046. The number of people in a given year living through the year with a disease indicating palliative care need is estimated to increase 89% in the same period. These increases are large compared to many other high-income countries, and reflect Ireland's relatively young population that is in the early stages of demographic ageing. People living with a disease indicating palliative care need outnumber those in the last year of life with such a diagnosis by about 12:1, and this ratio is steady across the period of analyses. To meet population health needs requires urgent strategic steps on funding, workforce development and service provision in specialist and generalist palliative care. It is hoped that these data will be of value in informing the upcoming review of Irish national palliative care policy. Please add content to reference 34 link -further details could not be accessed.

Data availability
Please check page 4, last paragraph -last sentence -should this say ICD-10 codes instead of CSO codes?
In analysis 2 and 3, please justify why those with cancer (as opposed to those who died with living cancer) can be classified as having a disease associated with palliative care need, given increased survivorship in those with a cancer diagnosis.
Please add detail on step-by-step calculations for Analysis 2 & 3 -reference 34 contains no data.

Results
Please recheck percentages in Results, paragraph 2. A 30-year increase of 68% is reported. Is this correct? Should this be 61% over 31 years? Please clarify.
Please provide detail on how the increase in absolute numbers of deaths from a disease associated with palliative care need was calculated (78% to 84%) as the percentages are not presented in the table. Please recheck all percentages reported. Is 84% correct?
Very interesting data in Table 3 and related figures, with important implications for policy and service commissioning.

Discussion
Please check first line which states that the number of people in Ireland dying from a disease associated with palliative care need is estimated to increase between 68% to 84%. However in the results section, end of paragraph 3, the percentages mentioned are 78% to 84%. Please clarify.
Please Limitations -last paragraph -should Table 1 read Table 2?   " ". Is there something missing from this and that endeavours to close the policy-implementation gap sentence?
Important points for future approaches to estimating future need are described and will be helpful to other researchers working in this area.
Overall the conclusion summarises the findings well and highlights important implications for Overall the conclusion summarises the findings well and highlights important implications for palliative care policy and service delivery in Ireland.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound? Yes

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Yes
No competing interests were disclosed.

Competing Interests:
Reviewer Expertise: palliative care, complex interventions, mixed-methods I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Author Response 21 Jan 2020 , Trinity College Dublin, Dublin, Ireland

Peter May
We are grateful to the reviewer for their detailed and thoughtful report. We have made a number of changes to the manuscript following these comments. The most substantive change is the addition of a new Figure comparing Ireland's population age distribution to the rest of the European Union ( Figure 1, version 2). We also add clarifying text on Ireland's population ageing and the interpretation of our main results tables.
For the most part changes were minor. Each specific change is documented in responses to the relevant reviewer report. For each reviewer comment requiring action, we use the notation [R1/C1, R1/C2, etc], where R indicates which reviewer and C numbers each comment. We mark our responses to each comment >>>.
For those outside of Ireland, perhaps clarify if data refers to the island of Ireland or Republic R2/C1 of Ireland.
>>>Abstract changed to Republic of Ireland to avoid confusion.
>>>In the main manuscript we continue to use 'Ireland' and provide the following explanation under Methods>Setting: >>>In the main manuscript we continue to use 'Ireland' and provide the following explanation under Methods>Setting: [Ireland] is sometimes called the Republic of Ireland to distinguish the sovereign state from the island of Ireland (incorporating Northern Ireland), but the country's official name is Ireland and we use this form throughout.
Please clarify whether the study is specific to the Republic of Ireland or the whole island of R2/C2 Ireland? >>>See R2/C1.
What is the population and number of deaths in recent year(s)?

R2/C3
>>>Added to Methods>Setting In 2015, the baseline year for our data, the estimated population was 4.7million people with 30,127 (0.006%) recorded deaths.
Is it possible to add a comparison graph comparing age distribution of the Irish population R2/C4 (current and projected if available) with that of the EU or other relevant comparison countries (e.g. England/Wales; Scotland). This might provide useful detail on the extent to which the age profile of Ireland's population differs from other high-income countries. This could be presented in the Introduction, or in the Methods (setting) section. Perhaps clarify what are 'exposure' variables where they are first mentioned.

R2/C6
>>>We have reworded the end of paragraph 2 in Methods>Data sources, so 'exposure variables' are now mentioned for the first time in their own sub-section, where they are defined.
Please add content to reference 34 link -further details could not be accessed.

R2/C7
>>>This reference mistakenly redirects the reader to the OSF registration, not the OSF content. We have asked editorial to correct. We have also revised how appendices are described in the text so that readers know which specific file to open.
Please check page 4, last paragraph -last sentence -should this say ICD-10 codes R2/C8 instead of CSO codes? >>>Revised to: For further details on concordance between the CSO's use of ICD-10 codes… In analysis 2 and 3, please justify why those with cancer (as opposed to those who R2/C9 living died with cancer) can be classified as having a disease associated with palliative care need, given died with cancer) can be classified as having a disease associated with palliative care need, given increased survivorship in those with a cancer diagnosis.
>>>Explanation provided in Introduction>Rationale and aim: policy and clinical guidelines increasingly recommend palliative care across the trajectory of serious disease, meaning that those dying from these diseases are only a subset of the overall population health need (refs). To inform decision-making in policy and practice, including the targeting of interventions to those who benefit, groups cannot be defined by their characteristics at death and instead must be identified prospectively on the basis of clinical and other relevant factors (refs).
Please add detail on step-by-step calculations for Analysis 2 & 3 -reference 34 contains R2/C10 no data. Please provide detail on how the increase in absolute numbers of deaths from a disease R2/C12 associated with palliative care need was calculated (78% to 84%) as the percentages are not presented in the table. Please recheck all percentages reported. Is 84% correct? >>>These numbers are correct and correspond to the relevant cells. We have added more explanatory text prior to Table 2 to assist interpretation: We quantify changing need over time in three ways in Table 2: absolute number of projected deaths with palliative care need in a given year (denoted n ), this absolute number as a proportion of all projected deaths in a given year (denoted p(N )), and this absolute number relative to absolute number of projected deaths with palliative care need at baseline in 2016 (denoted 'Rate of change index').
Please check first line which states that the number of people in Ireland dying from a R2/C13 disease associated with palliative care need is estimated to increase between 68% to 84%. However in the results section, end of paragraph 3, the percentages mentioned are 78% to 84%. Please clarify. >>>As well as changes under R2/C12, we revise the opening section of the discussion as follows: >>>Our 43% comes from the abstract (an increase of 161842 to 537240) but in their Discussion they do report 47% which can be sourced to a different projection method (which gives an increase of 175,336 to 550,734). Since we also used this latter method, you are correct that this is a valid y y