Regional and rural Australia — that is, those areas that lie outside the major capital cities — is home to over 8 million people (33% of the total population), and comprises about 98% of the Australian land mass. ^{1, 2} The health of regional Australians is poorer than their city counterparts. They die at a younger age, carry a higher burden of chronic disease, and are more likely to engage in unhealthy behaviours. ³ Such findings are not unique to Australian regional and rural populations. ⁴ They also have less access to and are less likely to seek medical services. ⁵ It has recently been hypothesised that some of the misrepresentations of Aboriginal and Torres Strait Islander (Indigenous) Australians in the intensive care unit (ICU) are also due to access barriers to both primary and subspecialty care. ⁶
 

Specialist medical facilities, including ICUs, are clustered in areas of high population density, mostly in and around capital cities. These tertiary and metropolitan ICUs provide highly efficient critical care services to people living in the major cities of Australia, but this aggregation of ICUs may be problematic for critically ill regional and rural Australians. ⁷
 

Despite population and health service differences, comparative data on the provision of critical care services in Australia by regional and rural ICUs are lacking. A description of the characteristics and outcomes of critically ill patients admitted to regional and rural ICUs may help to identify whether equity exists in the provision of ICU care and highlight the role of regional and rural ICUs in the management of critically ill Australians. It was hypothesised that the outcomes of regional Australians admitted for critical illness would mirror the higher mortality observed in regional areas.
 

The Australian and New Zealand Intensive Care Society (ANZICS) Adult Patient Database (APD), one of four clinical quality registries administered by the ANZICS Centre for Outcome and Resource Evaluation (CORE), was used. The APD was used to identify all patients aged 16 years or older admitted to an Australian public ICU between 1 January 2009 and 30 June 2019. Over 90% of ICU admissions in Australia are reported to the APD. ⁸ ICUs are classified by the type of hospital to which they provide services (tertiary, metropolitan, regional/rural or private).
 

Admissions were excluded if there was no admission diagnosis, no in-hospital mortality outcome listed, or if the admission was to a private ICU. To avoid double-counting patient outcomes, ICU readmissions during the same hospital episode were excluded, as were episodes where the admission source was the ICU of another hospital (ie, an interhospital transfer [IHT] defined by the receiving ICU). Patients readmitted to ICU at the same hospital in a separate hospital episode were included and considered as two distinct episodes. It was not possible to identify subsequent ICU admissions at a different hospital. The study was approved by the Central Australian Human Research Ethics Committee (CA-19-3516) and was conducted following the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) standards (Online Appendix, table 1). ⁹
 

Baseline characteristics (age and sex), demographic features (home postcode, regional status), the presence of comorbidities, the need for mechanical ventilation, and outcomes were extracted. ¹⁰ Illness severity was described using the Acute Physiology and Chronic Health Evaluation (APACHE) II and III scoring systems and the Australian and New Zealand Intensive Care Risk of Death (ANZROD) model. ^{11, 12, 13} ANZROD is derived from locally collected variables and components of the APACHE scoring system. It accounts for individual admission diagnoses and applies separate predictive equations for each major physiological system. It provides highly discriminatory mortality prediction for admissions to Australian and New Zealand ICUs. ¹³ The reason for ICU admission was taken from the ANZICS modification of the APACHE III diagnostic coding system. ¹⁰ Individual admission diagnoses were grouped into eight major system-based categories.
 

Remoteness and socio-economic status were determined respectively by the 2011 iteration of the Accessibility and Remoteness Index of Australia (ARIA+) and the 2011 iteration of the Index of Relative Socio-Economic Advantage and Disadvantage (IRSAD) linked to each patient’s postcode. ^{14, 15} ARIA+ is an objective process for classifying remoteness and values vary between 0 (high accessibility) and 15 (high remoteness) based on road distance measurements, population size, and accessibility. ¹⁵ IRSAD is a general socio-economic index that summarises the economic and social conditions within an area. Lower numbers indicate areas of relative greater socio-economic disadvantage and lack of advantage. ¹⁴
 

The primary outcome was in-hospital mortality. Secondary outcomes included in-ICU mortality, measures of resource use (ICU and hospital length of stay, need for mechanical ventilation, need for IHT), and ICU readmissions rate in subsequent hospital admissions.
 

Data were analysed with STATA version 15.1 (Statacorp, Texas, USA). All data were initially assessed for normality. Group comparisons were performed using χ² tests, Student t test for normally distributed data, and Wilcoxon rank sum tests for non-normally distributed data. Results are reported as n (%), mean (standard deviation [SD]) or median (interquartile range [IQR]) respectively. Hierarchical logistic regression was used to account for baseline severity of illness using ANZROD methodology, adjusting for hospital type, region, year of admission (treated as a continuous variable) and socio-economic status, with patients clustered by site, and site treated as a random effect. ¹³ To reduce the risk of confounding due to an IHT (since attribution of mortality may be transferred from the regional or rural ICU to a tertiary ICU), further sensitivity analyses were undertaken excluding patients who had an ICU discharge destination of another ICU and also without excluding any IHT. For comparative analyses, P values are provided for perspective rather than for statistical significance; and for regression, results were considered significant at P < 0.05.
 

Between January 2009 and July 2019, there were nearly 1.4 million admissions to adult ICUs within Australia reported to the ANZICS APD. After exclusions (Figure 1), the study dataset comprised 876 522 episodes of care in 103 distinct ICUs, of which 172 444 (19.7%) occurred in one of 36 regional/rural units (Table 1). The proportion of admissions to regional/rural units over the sampling period was relatively stable (Figure 2). Patients admitted to regional/rural units were less likely to originate from a major city, had a higher ARIA+ score, indicating a residential postcode from more remote centres, and were more likely to come from areas that were relatively disadvantaged and had less advantage.
 

Patients admitted to a regional/rural ICU had a similar age (62.5 ν 63.9 ν 64.6 years for tertiary, metropolitan and regional/rural ICUs respectively), and were more frequently admitted emergently (64.5% ν 81.7% ν 81.1% respectively). They more frequently had chronic cardiovascular or chronic respiratory disease and less frequently had metastatic cancer, leukaemia, or were immunosuppressed (Table 1).

Outcomes

Nearly 20% of ICU admissions in Australia occur to regional/rural ICUs. Admissions are more frequently emergencies than to tertiary ICUs and illness severity is slightly less than that of metropolitan and tertiary ICUs, while ICU readmissions in subsequent hospital episodes is nearly twice that of tertiary ICUs. Unadjusted mortality following an ICU admission in a regional/rural ICU is lower than both metropolitan and tertiary ICUs, and this finding persists after adjusting for illness severity and other potential confounders. These findings are topical, given media reporting about the quality of health care delivered at regional hospitals. ¹⁶
 

The lower mortality rate seen in regional/rural units may have several explanations. First, it may be as a result of residual confounding. Second, there could be a real relationship between remoteness and mortality such that there is a protective effect, although the evidence is contradictory and the association between remoteness and any mortality benefit are likely complex and multifactorial. ^{3, 4, 17, 18, 19, 20, 21} This signal, at least within the critical care literature, appears to be robust and warrants further research. ²¹ Third, it is possible that the sickest patients in remote regions die before admission to an ICU because of longer transfer times or through a lack of timely access to appropriate critical care services. ^{22, 23, 24} Fourth, there may be less pressure to admit patients who are unlikely to survive to regional/rural ICUs. Fifth, the higher mortality reported in regional/rural areas appears to be mostly attributable to chronic disease progression. ³ These data support the hypothesis that critical illness does not contribute markedly to the observed mortality difference between regional/rural populations and their major city counterparts. ^{3, 25} There is some emerging evidence of this, albeit mostly in small single centre observational studies, seemingly suggesting that once admitted to an ICU mortality, equivalence exists. ^{21, 23, 26, 27} Sixth, it is possible that patients admitted to a regional/rural ICU receive high quality largely consultant-delivered care, which results in improved outcomes. Finally, it is possible that patients with a higher risk of mortality are being transported to a tertiary centre after initial stabilisation at a regional/rural ICU. However, the latter explanation is unlikely to explain the lower mortality, given the robustness of the adjusted mortality outcome on sensitivity analyses.
 

The higher frequency of emergency admissions to both regional/rural and metropolitan ICUs are thought to reflect the rates of high risk elective surgery undertaken in tertiary centres (predominantly elective cardiac surgery), and through the relatively lower rate of elective post-operative admission diagnosis, probably also contributed to the lower rates of mechanical ventilation in both regional/rural and metropolitan ICUs. The high proportion of patients admitted emergently, however, highlights the lack of surge capacity that could be recruited through cancellation of elective surgery in the setting of pandemic illness. The surprisingly high number of patients admitted to regional/rural ICUs with a capital city residential postcode may reflect tourists. Nevertheless, this emphasises the difference in caseload that regional/rural and metropolitan ICUs encounter compared with tertiary ICUs.
 

The higher ICU readmission rate seen in regional/rural ICUs may reflect the geographical isolation of the units and the lack of other health care facilities to which a patient can present. An alternative explanation, however, is that the high readmission rate points to a lack of primary health access and/or engagement and subsequent poorly managed chronic disease results in a frequent need for ICU readmission. ^{5, 21, 28}
 

The high rates of IHTs, while not unexpected, underline the costs (both personal, in terms of displacement from home and family, and economic) that regional Australians bear when faced with a health crisis. While this has been previously described in surgical subspecialty fields and for patients with cancer, this is the first study to highlight this issue in the Australian critical care literature. ^{29, 30, 31, 32, 33, 34, 35}

There is little Australian literature describing regional/rural critical care services, and the support they provide to regional hospitals. Two exceptions are a before and after observational study examining the introduction of an ICU telemedicine intervention in regional Victoria, which demonstrated a reduced need for IHTs, and a prospective observational study of emergency department (ED) presentations, also in rural Victoria, which demonstrated the breadth of presentation and interventions that were required in regional EDs was not dissimilar to that in tertiary hospitals. ^{36, 37} Their conclusion — that regional/rural hospitals provide a surprising amount of critical care support — is not dissimilar to the findings in our data. ³⁶
 

There are implications in these data for the training of intensive care specialists. Despite the differences between regional/rural ICUs and tertiary ICUs that these data highlight, current trainees are required to spend only 3 months in a regional/rural hospital. This time is not necessarily within an ICU. Many trainees are seeking, and receiving, retrospective recognition for time spent as a junior doctor in a regional hospital, yet it seems unlikely that this is adequate preparation. ³⁸ The need for intensive care medicine to expand into regional and regional/rural Australia to support both acute hospital servicing and to accommodate an increasing number of trainees has been recognised and is strengthened by these data. ^{31, 39} The issues around ensuring adequate training time in regional/rural centres are not unique to intensive care medicine and there are calls from other specialties, both within Australia and internationally, to ensure that specialist clinicians receive an appropriately broad training experience. ^{40, 41, 42}
 

The major strength of this study is the large, multicentre database from which the data are drawn. The APD data covered a sampling period exceeding 10 years, and incorporated admission data from 103 ICUs representing every state and territory, providing strong external validity. It is the first study to describe the characteristics and outcomes of regional/rural ICUs and provides important benchmarking.
 

The study has limitations. First, the population described is unique to Australia and may not represent the clinical practices, hospital resourcing, or geographical dispersion of population in other countries. Second, in a dataset of this size, it is inevitable that differences between groups may be statistically significant without necessarily being clinically relevant nor translate into an important patient-centred outcome. Third, the APD does not include cardiac care beds that may be co-located with ICUs and which regional/rural intensivists may oversee, which is thought to be more common in regional areas where fewer resources exist. As a result the contribution of a regional/rural ICU may be underestimated in this dataset. Finally, it is not possible with the current dataset to match those patients transferred from a regional/rural unit to a higher level of care, which may affect mortality outcomes, since the sickest patients who have a higher mortality risk are probably more likely to be transferred. We have attempted to mitigate this bias through multiple sensitivity analyses variably excluding IHTs and the result appears robust.
 

Future studies should focus on further elucidation of the apparent survival benefit of living in a remote area which necessarily includes describing workforce characteristics, resourcing and processes of care in regional/rural ICUs, to better quantify the impediments to accessing ICU admission, teasing out the reasons for the high readmission rate to regional/rural ICUs, the role that regional/rural ICUs play in resuscitating the critically ill patient before IHT, exploring further the impact of tourists on resource demands in regional/rural ICUs and, finally, identifying the barriers to training future intensivists and in recruiting intensivists to fill positions in regional/rural ICUs.
 

These data highlight the importance of regional/rural ICUs in supporting regional Australia, and they are topical, given media reporting around the quality of health care delivered at regional hospitals. ¹⁶ It is the first study to describe the role of regional/rural ICUs in Australia in supporting critically ill patients and highlights the important role that regional/rural ICUs perform in supporting the regional hospital where they are sited. Mortality outcomes in regional/rural ICUs are at least as good and possibly better than those in tertiary centres.