Background

The UK population is ageing, with projected increases in those aged 50+ from 22 million in 2022 to 26.3 million in 2044 [1–3]. Ageing is associated with frailty, a clinical state of compromised ability to cope with stressors, resulting from age-associated declines in physiological reserve and function [4]. It is estimated that 1.8 million people in the UK aged ≥60 were living with frailty in 2016 [5] and prevalence increased from 27% (2006) to 39% (2017) in England, with 11% of those aged 50–64 already living with frailty [6]. Prevalence of mild to severe frailty (measured via the eFI) in those aged 50 and over is projected to rise further, to 76% in 2040 [3]. Primary and secondary care service use and associated costs are higher in adults living with frailty and increase with the severity of frailty [7, 8]. Projected growth in frailty will be associated with increased service use costs in these services of £10 billion by 2040 [2, 3] and numbers of people with complex care needs are predicted to increase substantially by 2035 [9]. There is, however, evidence of unmet need for health and social care, with significant health inequalities related to deprivation [10, 11] and this gap between need and capacity will continue to expand. In this context, robust service and workforce planning to meet future health and social care needs of those living with frailty is essential to better match service demand and workforce capacity to meet the needs of the ageing population [12–15].

Providing care for older people with frailty is complex, requiring an integrated approach from primary, secondary, community and social services [16, 17]. Although there is a growing body of evidence on primary and secondary health care utilisation in older people with frailty, less is known about their community-based health, mental health and publicly funded social care use. There is some evidence that increased frailty in ageing populations will result in increased demand for social care in those aged 85 and over [18, 19]. Although it is recognised that new ways of working will be needed to address demand for community-based services, particularly in deprived areas [20, 21], the shift to home-based care will require better understanding of trends in demand for community-based care and prevention and early intervention services [22]. This study aimed to address the evidence gap around the impact of frailty on community-based health and social care for people aged 50 and over, to provide projections of demand as the population ages and to inform service and workforce planning.

Aim

To fill the knowledge gap around community-based services in older people with frailty, by estimating future demand for community health, mental health and publicly funded social care for those aged ≥50 with frailty, in England, through the use of a simulation model informed by linked routine data analysis.

Methods

Study design

Retrospective analysis of electronic health records, combined with System Dynamics (SD) simulation modelling. SD modelling has often been used to model healthcare systems to provide a strategic view of a system [23–26]. It is ideally suited to model large populations where individual variability is not a major consideration.

Data source

The Discover-NOW Research Environment hosts a de-identified dataset linking depersonalised, contemporary, primary, acute, community and mental healthcare and social care electronic patient records from over 2.8 million patients in North-West London (NWL) registered at 344 GP practices [27, 28]. The Discover-NOW dataset is one of Europe’s largest linked longitudinal datasets capturing a population of around a third of London, an ethnically highly diverse population. The number of participating GP practices remained stable during the data collection period, varying between 351 and 357.

Participants

Patients aged 50 and over registered at NWL General Practitioner (GP) practices contributing to the Discover-NOW databank between 2015 and 2022 were eligible. This age range allowed consistency with previous simulation modelling [3], and service use to be tracked in frailty in the ageing population; previous analyses demonstrated that frailty is already present over 10% of those aged 50–64. An open cohort design enabled the addition of patients who turned 50 or moved to a participating practice and were present on 1 January of a calendar year in the study period. Patients left the cohort by leaving the participating practice or dying.

Age was categorised into four groups, reflecting groupings reported in literature relating to older adults’ healthcare, and cut-offs for services recommended by the Stakeholder Engagement Group (SEG) in a linked study: 50–64, 65–74, 75–84 and ≥85 [2, 29]. Frailty was categorised using electronic Frailty Index (eFI) score, a ‘cumulative deficit’ index, measuring frailty through accumulation of a range of deficits [30]. The eFI score indicates the number of deficits present out of a possible total of 36, with higher scores indicating more severe frailty [30]. An eFI score was calculated on 1 January for each patient and the relevant frailty category assigned (fit, mild, moderate, severe).

Service use data

All mental health, community health trusts and local authority social care services in NWL contributed data. Mental health contacts included psychiatry, older persons mental health services, crisis resolution and liaison teams, with a contact representing a face-to-face or telephone appointment, day case or inpatient admission. Mental health service use, although not directly attributable to frailty, was included following consultation with the SEG due to prevalence of anxiety and depression in older people and associated service use [31] [32, 33]. Community health contacts covered physiotherapy, telehealth, rehabilitation and community nursing. For social care, a contact was defined as either an assessment, new care package or a change in care package, and included services such as personal home care, domestic support, home adaptations and respite care. Service use contacts for each service type reflect heterogenous administrative records, collated to capture broad activity within different services; they do not reflect time or intensity of input. Social care assessments are important to service provision in frailty, and have therefore been included, though they may not always lead to provision of a care package. Service use data were used to generate average yearly service use rates, which were applied to all simulation years.

Analysis

For each of the services (mental health, community health, publicly funded social care), the proportion of follow-up years with a contact was calculated for each of the 16 age and frailty combination categories. For each category, the average (standard deviation) number of contacts per year was calculated for patients with at least one contact in a year.

Simulation model development

The Frailty Dynamics Model provided population-level estimates of frailty incidence, prevalence and rates of frailty transitions (adjusted for deprivation, sex, ethnicity and urban location) for England [29, 34]. Estimated transitions from fit to any level of frailty were 48/1000 person-years aged 50–64, 130/1000 person-years aged 65–74, 214/1000 person-years aged 75–84 and 380/1000 person-years aged ≥85 [2]. This model was developed to provide projections of frailty prevalence in the English population [3]. The population structure within the model aged ≥50 is considered as 16 separate subgroups, categorised according to age group (50–64, 65–74, 75–84 and 85+) and frailty category (fit, mild, moderate and severe) [29].

Discover-NOW data analyses provided average community health and social care service use rates for each of the 16 age/frailty groups over the study period (2015–22). Average rates were applied to the national-level estimates of people within each frailty category to provide projections of the number of contacts in each of the three services in England during 2025–40. The service use data were derived from an area with higher deprivation and diversity than the average for England; applying national frailty transition rates and service use rates per age and frailty sub-group reduced the risk of over-estimating service use for the whole population. The simulation assumes that service use rates are constant for each of the age/frailty subgroups throughout the simulation.

Validation of the simulation model

The simulation model projections were internally validated against service use data from Discover-NOW and externally validated against available NHS Digital data for the development period to the present. Comparison between the mental health projections and national dashboard figures was close (See Appendix 1 in the Supplementary Data Section for details). The comparison for community health was less close, but national data were more complete in later years for which the model provides much closer estimates (See Appendix 2 in the Supplementary Data Section for details). Confidence in the future projections is therefore high for community and mental health care, and moderate for social care, for which the simulation model was closer to national dashboard data for all adults (See Appendix 3 in the Supplementary Data Section for details).

Model estimation and scenario testing

Projected service use contacts for the simulation period (16 years inclusive) for each frailty category are presented in Table 3. Two illustrative ‘what-if’ scenarios are considered alongside the baseline (no change to services or frailty trends) experiment to estimate impact of broad policy and practice shifts. The scenarios were identified through consultation with the SEG [2] for their potential to be useful to commissioners and service planners. Following prioritisation of the scenario options, the parameters for the scenarios were informed by review of literature evidence as per a previous study [7]. The scenario experiments were as follows:

• Reducing Incidence: Fit-to-Mild transition rates for all age groups were reduced by 5%, thus reducing incidence of frailty in the population, representing the potential impact of general public health measures or the introduction of services targeting those that are pre-frail or at risk of developing frailty. Whilst studies have suggested potential reductions of up to 20%, this was considered a more feasible reduction [2, 35–40].
• Reducing progression: frailty progression slowed by reducing Mild-to-Moderate transition rates by 10% and Moderate-to-Severe by 5% in each age group [2, 35, 41, 42]. The aim was to consider the impact of clinical or public health intervention.