Long‐term trends in critical care admissions in Wales *

Summary As national populations age, demands on critical care services are expected to increase. In many healthcare settings, longitudinal trends indicate rising numbers and proportions of patients admitted to ICU who are older; elsewhere, including some parts of the UK, a decrease has raised concerns with regard to rationing according to age. Our aim was to investigate admission trends in Wales, where critical care capacity has not risen in the last decade. We used the Secure Anonymised Information Linkage Databank to identify and characterise critical care admissions in patients aged ≥ 18 years from 1 January 2008 to 31 December 2017. We categorised 85,629 ICU admissions as youngest (18–64 years), older (65–79 years) and oldest (≥ 80 years). The oldest group accounted for 15% of admissions, the older age group 39% and the youngest group 46%. Relative to the national population, the incidence of admission rates per 10,000 population in the oldest group decreased significantly over the study period from 91.5/10,000 in 2008 to 77.5/10,000 (a relative decrease of 15%), and among the older group from 89.2/10,000 in 2008 to 75.3/10,000 in 2017 (a relative decrease of 16%). We observed significant decreases in admissions with high comorbidity (modified Charlson comorbidity index); increases in the proportion of older patients admitted who were considered ‘fit’ rather than frail (electronic frailty index); and decreases in admissions with a medical diagnosis. In contrast to other healthcare settings, capacity constraints and surgical imperatives appear to have contributed to a relative exclusion of older patients presenting with acute medical illness.

in Scotland has raised concerns over rationing of admission according to chronological age and risks of inequitable access [11].
We investigated the potential effects of resource constraints on admission patterns and processes of care in Wales, a UK nation with critical care capacity much lower than the reported European average (5.7 vs. 11.5 per 100,000 population; for comparison with other developed nations, see online Supporting Information Table S1) [12,13]. The purpose of this study was to investigate trends in patient characteristics for adult critical care admissions between 2008 and 2017. We hypothesised that without an increase in capacity, critical care admission characteristics may not follow national population trends, and that with resource constraints there may be a decreasing tendency to admit those with significant underlying illness. As such, although the project was conceived and conducted before the outbreak of COVID-19, the themes could be even more relevant given the recent acute stress on resources.

Methods
We used the Secure Anonymised Information Linkage Databank (www.saildatabank.com) to carry out all analyses.
The development of this Databank has been described previously [14][15][16]. The project received approval from the independent Information Governance Review Panel, Swansea University. independent open data source [17]. Changes in critical care capacity over the period were identified from contemporary Welsh Critical Care and Trauma Network reports.
We restricted inclusion to episodes with high-quality matching from the identity linkage and anonymisation process for individuals who were aged ≥ 18 years on the day of critical care admission and registered to a residential address in Wales. Patients were followed-up until one year after hospital discharge, death or outward migration, whichever occurred first.
We categorised patients according to age as follows: 18-64 (youngest); 65-79 (older); and ≥ 80 y (oldest). We calculated a modified Charlson comorbidity index on the date of critical care admission using the ICD-10 codes [18] within the Patient Episode Database for Wales and a lookback period of one year [19]. We categorised comorbidity according to modified Charlson comorbidity index as: low (-1-0); medium (1-10); and high (> 10). Frailty was determined using the electronic frailty index (eFI) derived from Welsh Longitudinal General Practice data and recently implemented in Wales in those aged ≥ 65 years [20,21].
We calculated the eFI according to date of critical care admission using 10 years of previous general practitioner data for each individual and used this score to categorise as: fit (eFI value 0-0.12); mild (> 0.12-0.24); moderate (> 0.24-0.36); or severely frail (> 0.36).
We explored annual trends in admissions for each age cohort and tested for significant changes over the study period. Differences in proportions of patients according to comorbidity index and eFI category were compared across years using Chi-squared tests for trends. We analysed counts and crude (unadjusted) incidence rates of admissions per 10,000 population using Poisson regression models, with the variable for year of admission added as the independent variable and national population estimates for each age group added as the offset. We converted model coefficients to rate ratios to compare differences across years compared with the baseline year (2008). We used separate models to analyse rates of admissions requiring the following: advanced respiratory support (typically mechanical ventilation); advanced cardiovascular support (multiple vaso-active/anti-arrhythmic drugs and/or cardiac output monitoring, intra-aortic balloon pump or temporary cardiac pacemaker); and renal support (renal replacement therapy). We categorised rates by admission type as medical, surgical or other, and as planned or unplanned.
Proportions of admissions with a recorded death were explored (critical care; post-critical care in-hospital; and total in-hospital mortality) and were tested for significant changes over time for within each age group. One-year mortality was investigated from point of hospital discharge following index critical care admission and again tested for significant changes over time within each group. We considered values of p < 0.05 to be statistically significant.

Results
We identified 85,629 admissions aged ≥ 18 years admitted   A high degree of comorbidity was present in 63.8% of all admissions and was most prevalent in the oldest group (Table 1). During the study period, we observed a significant decreasing trend in the proportion of admissions with high comorbidity across all groups. There was a relative decrease       (Fig. 5), in post-critical care hospital mortality in all three age groups, and in post-discharge one-year mortality in older age groups, but not in the youngest or oldest groups, though numbers were relatively low (online Supporting Information Table S5).

Discussion
Critical care capacity decreased slightly over the study period, while the national population increased, particularly among those aged > 65 years. There was a significant decrease in overall admissions per 10,000 population, with a 15% relative decrease among the  which made by far the biggest contribution to these data, there was an increase in critical care bed capacity of 35% between 1999 and 2006 [24], and a further increase of 15% between 2008 and 2016 according to other data sources [25,26].
Our findings more closely resemble trends observed by investigators from Canada [10], the Netherlands [8] and Scotland [11].  College of Surgeons Report on the Peri-operative Care of the Higher Risk General Surgical Patient [27]. Targetfocused administrative concerns over the progress of elective surgery (vs. harder to measure non-elective medical demands) and critical care benchmarking processes that at present do not directly account for frailty may have added impetus. Notably, the recent national initiative to improve care of the critically ill in Wales has primarily focused on enhanced care following surgery rather than core critical care capacity [30].
Importantly, our study also adds to the limited literature describing longitudinal trends for comorbidity. We previously reported the predictive value of the modified Charlson comorbidity index [19] in determining long-term survival following discharge from Welsh critical care units [31]. Applying this same method to a larger, less selected cohort, we observed a prevalence of high comorbidity (modified Charlson comorbidity index > 10) among Welsh patients (63.8%), greater than for Danish (modified Charlson comorbidity index 3 or more, 16.8%) [32] and Scottish patients (three or more comorbidities, 2.8%) [11] to a degree that warrants further investigation. However, examining trends, the significant decrease in proportions of patients with high comorbidity in the study period has not previously been reported, contrasts with data from other parts of the UK [9], and must be considered in the context of capacity constraint. We are unaware of other reports of longitudinal trends in critical care admission according to frailty; we applied the eFI, which was developed and validated in a UK population [20] and implemented in Wales [21]. Using this methodology, the proportion of patients aged ≥ 65 years identified as 'non-fit' (60.6%) was similar to the proportions identified using frailty indices among those aged ≥ 65 years in a Chinese geriatric ICU (60%) [33], and those aged ≥ 16 years in Brazilian ICUs (68.6%) [34]. Among those aged 65-79 years, we found an increased proportion of 'fit' patients and decreases in those with mild frailty over time. Further work is required to explore the potential effect of 'look-back' on eFI trends, restricted to those with a complete primary care record for the period under review, but our initial data do not currently support expectations voiced in the literature of "increased numbers of frail patients being admitted to intensive care units" [35].
Although not our primary aim, significant improvements in critical care and post-critical care hospital unadjusted mortality were seen in all age groups with time, and in one-year mortality among those aged 65-79 years.
This is consistent with other large-scale studies reporting improvements in short-term [6] and long-term mortality [36] particularly among older patients, but requires further exploration of the contributions of changing case-mix and illness severity.

Supporting Information
Additional supporting information may be found online via the journal website.