Discussion

This study provides novel analysis and projections of community-based service use for older people living with frailty, addressing an acknowledged evidence gap. Adding to evidence that frailty is associated with higher use of primary, secondary and urgent care services [7, 8, 43], this analysis provides new evidence about population demand for community-based health and social care services in older people living with frailty. [3, 7, 8, 44–48]. Service use increased overall with age and frailty, other than in mental health services, where use was highest in the 50–64 age group, suggesting need being met by other services in older age. Social care use was considerably lower than health service use, but data were only available for publicly funded care under current eligibility criteria; the demand for social care from all sources will be higher.

Model projections suggest that, over the next 16 years, the demand for community-based services will increase by up to 30% as frailty prevalence increases, consistent with previous simulation modelling for primary and secondary care [3]. The scenario experiments explored the impact on future demand if frailty incidence could be reduced or progression of frailty could be slowed, approaches in line with policy recommendations [49, 50] and NHS 10-year plan goals [21]. Scenario projections are illustrative for the purposes of comparison of different approaches, and absolute numbers should be considered in the context of underlying model assumptions, but the overall trends provide important messages for service development. Projected growth in community-based service use could be reduced, compared with baseline, by reducing frailty incidence or progression. However, in both scenarios, demand for care will continue to rise substantially even if these modest reductions are realised, highlighting the importance of service and workforce planning to meet the needs of the ageing population.

A recent audit of frailty identification and support provided in primary and community healthcare indicates there are still significant gaps in diagnosis and support, and therefore unmet needs among the population are likely to either persist or will lead to a steeper increase in caseloads without changes to service provision [51]. These model estimates may provide commissioners and planners with an insight into future demand for health and social care services to 2040. In addition, these analyses, coupled with caseload information, are being used to develop future workforce estimates for frailty [52].

Limitations

These projections are in line with other analyses using eFI, but they should be considered in the context of other studies which include all frailty levels and use eFI [30] for frailty identification. As the eFI is a cumulative deficit index and the simulation model population includes only those aged 50 and over, the progression of frailty and increased prevalence reflect the ageing of this cohort and their accumulation of deficits. In addition, the eFI is generated from routine electronic health records, and so will necessarily reflect local differences in data entry and coding procedures, although validation studies suggest that variability in coding accuracy or completeness does not have a significant impact on the ability of the eFI to identify changes in frailty status in ageing populations [53–56]. These projections should therefore be considered with some caution, with a focus on patterns and trends rather than specific numbers; further research using eFI2 frailty data would be likely to generate somewhat lower peaks in prevalence. Internal and external validation, however, supported confidence in the simulation projections for health and social care services.

Service use data were not derived from national data. We applied average transition rates by age group derived from longitudinal analysis (MSM) of routine data [6], adjusted for sex, ethnicity, deprivation, urban/rural location and applied average yearly rates of service use by age and frailty group across all simulation years; this approach was supported by internal and external validation. This approach also avoided over-estimating frailty prevalence through use of rates from a population that is more deprived and diverse than average. Application of service use rates by frailty group should also have mitigated this possibility for service use data. Service use was compared with previous analyses where possible, suggesting that the differences in service use noted in more deprived populations are due to higher incidence, in line with other studies [57].

The assumption of average service use rates across all years of the simulation could be erroneous, especially in the context of likely changes to service delivery, funding and access over the period of the simulation, including the pandemic. We have attempted to mitigate this by applying average service use rates per age and frailty group derived from the 8-year data extraction period, use of broad service categories rather than specific services and illustrative scenarios that reflect current policy.

These projections use contacts which do not reflect time, intensity or clinical purpose or meaning. However, these administrative categories map against unit cost categories, which do reflect the different intensity of activity for each event type. In addition, estimates of social care demand will underestimate true demand, because data are only available for publicly funded social care. It is assumed that this will continue to be provided in the same way as it is currently, and that individuals will have to satisfy the same eligibility criteria. It is possible that financial eligibility thresholds will change under social care charging reform and as demand increases [58]. In this case, model projections could over-estimate publicly funded care use. The baseline model also assumes that the same health and social care services will continue to be provided in the same way in the future. Whilst this might not be the case, the scenario experiments reflect likely changes in relation to moving more care into the community, based on current health and social care policy. It should be noted that this analysis used only publicly funded social care contacts. The relatively low number of these contacts should not be interpreted as indicating low demand, which is likely to be substantially higher when privately funded and unpaid care and unmet need, are considered. Evidence suggests that ~37% of care home residents are self-funded [59], as are 23% of domiciliary care users [60], giving some indication of the likely scale of demand.