Introduction
The global population is aging. This trend results from a process referred to as demographic transition, characterized by declines in both fertility and mortality rates [
1]. The probability of survival to older age has improved and the absolute number and proportion of older persons is projected to increase in the next few decades [
1]. The fastest growing age cohort is made up of those aged ≥ 80 years, increasing at an estimated 3.8% per year and projected to represent one-fifth of all older persons by 2050 [
1].
Older age is associated with an increased prevalence of chronic illness and functional impairment [
2,
3]. As a result, the rate of hospitalizations for acute illness among older persons is certain to increase [
4]. Similarly, the demand for critical care services and admissions to intensive care units (ICUs) is also projected to dramatically rise in the next decade [
5]. Data from the United States estimates approximately 55% of all ICU bed-days are incurred by patients aged ≥ 65 years and an estimated 14% of those patients aged ≥ 85 years die in the ICU [
5]. There are conflicting data, however, on the short-term and long-term survival for older patients admitted to the ICU [
6‐
15]. These disparities may reflect differences in the severity and type of illness, length of follow-up, definitions for old age, and treatment intensity for older patients [
12,
16,
17].
Owing to the aging population, an evaluation of how best to provide care for acutely ill older patients and to optimize recovery has become an important issue that may have implications on health resources in terms of triage, decision-making, expansion of ICU capacity, and advanced care planning. Moreover, there is an urgent need to understand the implications on outcomes for older patients after ICU admission, including not only survival but also cognitive impairment, quality-of-life, and functional autonomy [
18‐
23].
Accordingly, we interrogated the Australian and New Zealand Intensive Care Society Clinical Outcomes and Resource Evaluation (ANZICS CORE) Adult Patient Database (APD) to obtain information on very old patients (age ≥ 80 years) from 57 Australian hospitals over a 6-year period. Our primary objectives were to evaluate the cumulative (and annual) change in the proportion of very old patients admitted to the ICU, to evaluate the clinical characteristics and the cumulative (and 6-year trends) outcomes of very old patients compared with those aged < 80 years, to evaluate factors associated with survival for very old patients admitted to the ICU, and to project estimates of ICU admission rates and of ICU and hospital bed-days for this cohort.
Discussion
We performed a 6-year retrospective analysis of over 120,000 ICU admissions to 57 ICUs across ANZ, using a large validated clinical database, to evaluate the rate, clinical characteristics, outcomes and projected resource demand of very old patients (aged ≥ 80 years) admitted to the ICU.
Our study found that very old patients represented 13.0% of all patients admitted to the ICU and this rate increased by an estimated 5.6% annually during the study period. We found similar increases in the annual admission rates for patients aged ≥ 85 and ≥ 90 years. Interestingly, we showed evidence of sex-specific differences in ICU admission rates, with males higher than females, and this was modified by age, with by greater differences in older age strata. We also found that very old patients were more likely to be admitted from chronic care facilities and to have a higher burden of co-morbid illnesses. Similarly, very old patients presented with greater acuity of illness (after accounting for the age points in APACHE II score) yet were less likely to receive mechanical ventilation while in the ICU. This cohort also showed interesting differences in acute physiology and laboratory parameters, including higher serum creatinine, lower urine output, and greater occurrence of early acute kidney injury. Very old patients, when compared with younger age strata, showed consistently lower crude and adjusted ICU and hospital survival. Moreover, these patients were more likely to be admitted from chronic care facilities, have longer durations of stay in the ICU and in the hospital, and were significantly more likely to be discharged to a rehabilitation/long-term care facility.
Importantly, our findings have potential implications for future health resource demand, utilization and planning. These data imply the potential for a 72.4% projected increase in demand for ICU and hospital bed-days for very old patients within a decade.
Survival to older age has improved and contributed to the evolving demographic transition. Current global growth rates of persons aged ≥ 80 years (3.8% per year) are greater than any younger segment of the older population and are twice the growth rate of persons aged ≥ 60 years (1.9%) [
1]. This trend, at present, is largely dominated by growth in more developed countries [
1]. This projected increase will have importance on the demand and delivery of health services; in particular, intensive care.
Prior data have estimated ICU admission rates in the range 3.0% to 16.5% for patients aged ≥ 80 years [
9,
13,
15,
20,
29]. These apparent differences are probably attributed to variations in study design, cultural/geographic variations and differences in the study populations (that is, severity of illness, definitions for old age, treatment intensity/triage). In comparison with these studies, the proportion of ICU admissions for very old patients in ANZ is relatively high. In addition, our large observational data have clearly shown an increasing trend in admission rates for this cohort. Moreover, this trend would be anticipated to continue and/or accelerate, and would certainly be expected to impact on health resources. For example, assuming a stable estimated average cost per patient ICU bed-day for nonchargeable patients in New South Wales (2004/2005) of AUS$3,999 [
30] and a stable median ICU length of stay (that is, 2.76 days), the estimated increment in resource expenditures for ANZ (population~25 million) due to ICU bed-days for very old patients alone would be AUS$134 million per year by 2015 (AUS$681 million cumulative). This estimation, however, does not address the projected need for increased capacity.
Several investigations have shown that very old patients admitted to the ICU receive less aggressive treatment, including mechanical ventilation, when compared with younger cohorts, despite comparable illness severity scores [
13,
15,
16]. Boumendil and colleagues conducted a matched-cohort study of over 6,000 patients comparing those aged 65 to 79 years and those aged ≥ 80 years [
16]. Patients were matched by sex, surgical status, co-morbid disease, and illness severity. They found that fewer very old patients received mechanical ventilation or renal replacement therapy [
16]. Hamel and colleagues found reduced resource intensity for older patients and a greater likelihood of having renal replacement therapy, vasopressors, tube feeding, and major surgical interventions withheld compared with younger patients [
31]. Likewise, there were similar findings in our data – with very old patients less likely to receive mechanical ventilation.
Whether these differences represent active or passive therapeutic limits recommended by clinicians, the preferences by families or patients, or the result of unaccounted confounding factors remains uncertain. This may, however, also indicate that there is a sizeable selection bias for which very old patients are triaged access to ICU support [
12,
16]. During a 20-month prospective, single-centre study of 180 very old patients triaged for the ICU, Garrouste-Orgeas and colleagues found that 73% were refused admission [
12]. The explanations cited included being too well (28%) or too sick (44%) to benefit from ICU support. Further, Boumendil and colleagues found that those very old patients admitted were in reasonably good health, had fewer fatal underlying diseases, and had fewer complicated surgical interventions. This was also supported by the lower total costs of ICU admission for very old patients, largely attributable to a shorter ICU stay and less invasive therapies (that is, mechanical ventilation, tracheostomy, renal replacement therapy) [
16,
32]. These data potentially imply that a comprehensive assessment and careful selection of very old patients for ICU support may deliver positive short-term clinical outcomes.
Similar to prior observational studies, we found evidence of sex-specific differences in admission rates to the ICU [
33‐
35]. Moreover, this difference was notably modified by increasing age, with males consistently more likely than females to be admitted to the ICU. In a retrospective Canadian analysis of 24,778 consecutive adult ICU admissions over a 2-year period, Fowler and colleagues found significant age-specific and sex-specific differences in ICU admission rates [
33]. While a greater number of older females (aged ≥ 50 years) were admitted to hospital, older males were more likely – after covariate adjustment for the diagnostic category and illness severity – to be admitted to the ICU and to receive other ICU-related interventions. Older females also had shorter ICU durations of stay and higher adjusted ICU and hospital mortality. Likewise, in a prospective European of 25,998 consecutive ICU admissions to 31 centres over 3 years, Valentin and colleagues showed similar disparities in sex-specific admission rates to the ICU and in ICU-related interventions; however, there were no differences in survival [
35].
Whether these findings represent genuine age-related and/or sex-related differences in access to the ICU and related interventions, represent variation in patient preferences for provision of ICU support, represent other health policy or decision-making processes, or were unaccounted for confounding remains uncertain [
33]. This was a secondary finding of our study, and as such has not been specifically examined in detail. Owing to the consistency of these data across diverse health jurisdictions, however, further prospective evaluation is warranted.
Substantial effort has been dedicated to determining whether age is a primary determinant of ICU outcome for very old patients [
6‐
9,
11,
13‐
15,
20,
29,
36‐
39]. In our cohort, when compared with younger subgroups, older patients had greater odds of death in the ICU and the hospital after covariate adjustment that included co-morbid disease and severity of illness. Hamel and colleagues found that age was independently associated with lower short-term survival in older patients admitted to the ICU that was not attributable to older patients receiving less intensive therapy [
13]. Moreover, patients aged ≥ 80 years had the highest 6-month mortality rates when compared with other age strata. Boumendil and colleagues found very old patients had comparably greater ICU and hospital-adjusted odds of death [
16]. This has been similarly shown with the majority of deaths in older patients occurring either in the ICU or early following ICU discharge [
9,
15,
20]. Whether these observed differences in short-term survival for very old patients are, in part, attributed to reduced treatment intensity, to the number of received interventions or to other factors remains uncertain. In our study, several nonmodifiable factors were predictive of hospital mortality including medical admission status, emergency surgery, primary neurologic, cardiac and gastrointestinal admission diagnoses, and admission from a chronic care facility.
Our study and prior available data, however, suggest that chronological age alone is probably insufficient to discriminate triage decisions on ICU admission. Rather, age probably represents an additive factor when coupled with frailty, physiologic reserve, burden of co-morbid illness, primary diagnosis, and illness severity. Prehospital disposition and/or functional status have been shown in numerous investigations to predict a worse clinical outcome [
15,
23,
40]. This constellation of clinical factors probably has important bearing not only on short-term survival but also on long-term survival, neurocognitive performance, functional autonomy, and quality of life [
6,
12,
15,
20,
29,
41‐
43]. Accordingly, very old patients developing critical illness – who are characterized by a low burden of co-morbid disease, good function status, and no measurable frailty – are likely to benefit from ICU support. These relatively fit older patients may be characterized by a unique phenotype that portends greater physiologic reserve and resilience during episodes of critical illness [
44,
45]. We need more accurate and robust methods, however, for predicting quality-adjusted long-term survival and for optimizing therapeutic recovery in all patients ultimately admitted to the ICU – particularly very old patients, given their higher observed short-term mortality.
The present study has several limitations that should be considered. First, our study is prone to selection bias due to not capturing data on rates of ICU admission refusal and/or patient/surrogate preferences for not foregoing ICU support. Accordingly, we were not able to estimate the incremental gain of ICU support compared with ICU refusal in very old patients. Likewise, our study was not able to capture data on the process of triage, or the occurrence of bed shortages that may have prompted rationing of ICU beds.
Second, the APD does not collect data on additional variables that we would consider valuable, including cause of death, long-term (that is, 1-year) survival, health-related quality of life, neurocognitive outcomes, functional status, and measures of frailty. Our analysis is therefore largely descriptive and used the hospital discharge location (that is, rehabilitation/long-term care facility) as a surrogate. Third, we retrieved data only from centres that had consistently contributed to the APD during the study period; however, small variations in consistency and/or efficiency of data contribution may bias estimates and may impact generalizability across ANZ or other jurisdictions.
Fourth, we are unable to comment on additional clinical outcomes that are clearly important in this cohort, such as long-term functional status and cognitive decline. Finally, our estimated projections for resource demand are based on a number of assumptions that are prone to change (that is, annual changes in admission rate, median ICU/hospital lengths of stay, average cost per ICU patient-day, linear growth rate).
While crude, however, we contend that the data from this large cohort clearly identify the need for additional formal evaluations of ICU health resources to match the epidemiologic trends.
Authors' contributions
SMB, SARW, AD, and RB participated in the conception and design of the study. CG, GKH, and DP contributed to data acquisition. SMB, SARW, AD, and RB performed and interpreted the data analysis. SMB, SARW, and RB drafted the manuscript. SMB, SARW, AD, CG, GKH, DP, and RB participated in critical revision of the manuscript. All authors read and approved the final manuscript.