Patients admitted to Australian intensive care units: impact of remoteness and distance travelled on patient outcome

Transcription

1 Patients admitted to Australian intensive care units: impact of remoteness and distance travelled on patient outcome Arthas Flabouris, Graeme K Hart and Angela Nicholls Understanding spatial patterns of disease and the relationships between disease occurrence and health outcomes is importanber for 2012 health planning. Early examples of Crit Care Resusc ISSN: Decem- the Crit use of geographic Care methods Resusc to explore 2012 patterns in health include Hippocrates About Original air, articles water, and geographic regions (in which he advocates close observation of local conditions to better understand diseases), Finke s map of human diseases, 1 and Snow s depiction of the 1854 cholera outbreak in London. 2 Timely access to specialised services is an important determinant of outcome across a range of acute illnesses. 3-6 Computerised geographical information systems (GISs) enable us to explore health service accessibility 7,8 and geopolitical factors 9-11 as determinants of health outcomes. A number of studies have linked remote residence of patients to poorer outcomes However, remoteness alone is not an accurate surrogate for distance travelled to access health facilities, and evidence for the direct impact of distance (or travel times) on health outcomes has been contradictory. Greater travel distance can be associated with poorer outcomes for acute asthma, 16 ischaemic heart disease 17 and cancer, 18 but may not influence outcomes for ambulance response to trauma 19 and for certain cancers. 20 Moreover, distance may be less important than patient illness and treatment factors for interhospital transfers. 21 In Australia, over the period , access to intensive care unit beds may have fallen, as available ICU beds per head of population has remained relatively constant while ICU occupancy has increased. 22 In addition, the necessity to travel to obtain acute hospital medical care is greater among remotely located patients, and the number of these patients being transferred over some distance to less remotely located hospitals is increasing. 23 These circumstances may ABSTRACT Objective: To use a geographical information system (GIS) to explore the impact of (i) patient remoteness and (ii) distance travelled to an Australian public-hospital intensive care unit on patient outcomes. Design, setting and subjects: We conducted a retrospective study over the period linking intensive care unit resource and clinical datasets with Australian population postcode data and using a GIS for analysis. Data from the Australian and New Zealand Intensive Care Society Centre for Outcome and Resource Evaluation (ANZICS CORE) critical care resources survey (2007), the ANZICS CORE adult patient database ( ) and the Australian Bureau of Statistics were used. Only public-hospital ICUs were included in the study. Classification of remoteness was based on the extended version of the Accessibility/Remoteness Index of Australia (ARIA+). Distance was the distance between centroids of the patient s residential postcode and the postcode of the area in which the admitting ICU was located. ICU admissions were divided into three categories: direct other-hospital ICU admission (patient transferred directly from another hospital), indirect other-hospital ICU admission (patient admitted from a ward, emergency department or operating room after being transferred from another hospital) or home ICU admission (patient not transferred from another hospital). Main outcome measure: Hospital mortality. Results: There were ICU admissions to 76 Australian publichospital ICUs. Of these admissions, (22.7%) were in the indirect group and (8.9%) in the direct group. Over the period of the study, for the indirect and direct groups, remoteness (measured by median ARIA+ rating) increased (from 0.25 to 0.55 [P < 0.01] and from 0.12 to 0.25 [P < 0.01], respectively) as did median distance travelled to the admitting ICU (from 36.4 to 42.5 km [P < 0.01] and from 27.1 to 36.7 km [P < 0.01], respectively), while mortality decreased (from 18.2% to 15.8% [P = 0.01] and from 22.7% to 18.7% [P = 0.01], respectively). ICU length of stay (LOS) and hospital LOS correlated with ARIA+ rating for both the indirect group (R = 0.018, P < 0.01; and R = 0.013, P < 0.01, respectively) and the direct group (R = 0.038, P < 0.01; and R = 0.036, P < 0.01, respectively). The median distance travelled by survivors compared with non-survivors was similar in the direct group (30.8 v 32.1 km [P = 0.66]) but longer in the indirect group (42.8 v 33.8 km [P < 0.01]) and the home admission group (11.5 v 7.6 km [P < 0.01]). Conclusion: For patients who were admitted to the ICU after being transferred from another hospital, greater remoteness and greater distance travelled were not associated with increased mortality, but LOS in the hospital and the ICU was longer. Crit Care Resusc 2012; 14: Critical Care and Resuscitation Volume 14 Number 4 December 2012

2 account for the relatively high level of adverse patient outcomes among critically ill patients who undergo interhospital transfer for ICU admission. 24 The purpose of our study was to use a GIS to explore the impact of (i) patient remoteness and (ii) distance travelled to the admitting ICU on patient outcomes. We also wished to examine changes over time. Methods Data sources We conducted a retrospective study over the period The location of public-hospital ICUs was sourced from the 2007 Australian and New Zealand Intensive Care Society Centre for Outcome and Resource Evaluation (ANZICS CORE) critical care resources survey. 22 Patient demographics (including residential postcode) and illness characteristics were sourced from the ANZICS CORE adult patient database ( ) (readmissions were excluded). Australian postcode population data were based on the Australian Bureau of Statistics (ABS) 2006 census data. Private-hospital ICUs and patients admitted to private ICUs were excluded, as transfers to private ICUs are infrequent. 24 Distance The geospatial reference used was the Australian postcode. Distance was the distance between centroids of each postcode (being, for patients, their residential postcode area and, for ICUs, the postcode area in which the unit was located). For analysis purposes, the distance between a patient s residence and an ICU within the same postcode area was considered to be 0 km. Distances were grouped into categories based on feasible choice of patient transport vehicle, being a road vehicle (0 50 km), helicopter ( km), fixed-wing turbine propeller aircraft ( km) or jet aircraft (> 1500 km). Remoteness classification Patient remoteness categories were based on the extended version of the Accessibility/Remoteness Index of Australia (ARIA+) and Remoteness Area codes, both sourced from the ABS. The ARIA+ classification is an index value (continuous variable) between 0 and 15 that represents the remoteness of a place based on the physical road distance to the nearest town or service centre in each of five population size classes (the higher the ARIA+ value, the more remote the area). The ARIA+ classification groups areas into five categories of remoteness: major city, inner regional, outer regional, remote or very remote. Under this classification, no area in Victoria is very remote, no area in Tasmania is a major city (Hobart is classified as inner regional ), no area in the Northern Territory is a major city or inner regional (Darwin is classified as outer regional ), and the entire Australian Capital Territory is classified as a major city. 25,26 Postcodes Postcode populations were based on 2006 ABS census data. Only postcodes from mainland Australia, Tasmania, the Cocos Islands, Christmas Island and Lord Howe Island were included. After excluding postcodes not listed with Australia Post and those that related to a post office box address, 2478 postcodes remained. Intensive care unit classification ICUs were classified according to College of Intensive Care Medicine of Australia and New Zealand categories: 27 Level 1 (capable of immediate resuscitation and short-term cardiorespiratory support); Level 2 (providing a high standard of general intensive care, including complex multisystem support, in keeping with the hospital s delineated responsibilities); or Level 3 (a tertiary referral ICU, providing multisystem, comprehensive care). Diagnostic categories and severity of illness measures For each ICU admission recorded in the ANZICS CORE adult patient database, an APACHE II (Acute Physiology and Chronic Health Evaluation II) diagnostic category is assigned, and severity-of-illness measures (APACHE II score and predicted risk of death) are calculated. Based on prior evaluation, the following APACHE II diagnostic categories were selected to assess their influence on hospital mortality in our study: 24 surgical condition, trauma, cardiovascular condition, respiratory condition, sepsis, gastrointestinal condition, neurological condition and overdose. Source of hospital and ICU admission In the ANZICS database, a coded schema to describe the source of both hospital and ICU admission is used. Based on this schema, the following descriptors were used in our study: admission : ICU and hospital source of admission is another hospital. admission : ICU source of admission is a ward, emergency department or operating room, and hospital source of admission is another hospital. : ICU source of admission is a ward, emergency department or operating room, and hospital source of admission is not another hospital. Critical Care and Resuscitation Volume 14 Number 4 December

3 Figure 1. Median distance to admitting ICU, and patient s postcode ARIA+ rating, by ICU admission category* Distance (km) Median distance to ICU (direct other-hospital ICU admission) Median distance to ICU (indirect other-hospital ICU admission) Median distance to ICU (home ICU admission) ARIA+ rating (direct other-hospital ICU admission) ARIA+ rating (indirect other-hospital ICU admission) ARIA+ rating (home ICU admission) Year ARIA+ = Accessibility/Remoteness Index of Australia (extended version). ICU = intensive care unit. R = Pearson s correlation coefficient. * Correlation coefficient for slope of line of best fit between ICU admission category and median distance to ICU: direct other-hospital ICU admission: R 2 = 0.726, P = 0.02; indirect other-hospital ICU admission: R 2 = 0.810, P = 0.06; home ICU admission: R 2 = 0.838, P <0.01. Correlation coefficient for slope of line of best fit between ICU admission category and median ARIA+ rating: direct other-hospital ICU admission: R 2 = 0.870, P < 0.01; indirect other-hospital ICU admission: R 2 = 0.962, P < 0.01; home ICU admission: R 2 = 0.750, P =0.01. R 2 is the square of the correlation coefficient for the slope of the line of best fit; P is significance value for R 2. ARIA+ rating that could not be matched to an Australia Post-listed postcode or was related to a post office box address; and 372 patients (0.1%) whose ICU source of admission was missing. This left patients in the final dataset. Patients were admitted to 76 ICUs: seven Level 1 (9% of ICUs; 1.6% of patients), 31 Level 2 (41% of ICUs; 18.9% of patients) and 38 Level 3 (50% of ICUs; 79.5% of patients). The distribution of these ICUs among states and territories was: two in the ACT (3%), 26 in New South Wales (34%), two in the NT (3%), 17 in Queensland (22%), five in South Australia (7%), three in Tasmania (4%), 19 in Victoria (25%) and two in Western Australia (3%). Patients transferred from another hospital For patients (22.7%), the ICU source of admission was indirect transfer from another hospital via the ward, emergency department or operating room; for patients (8.9%), the ICU source of admission was direct transfer from another hospital. The patients who were directly transferred constituted 39.2% of patients whose hospital source of admission was another hospital. Over the period , the median patient ARIA+ rating and median distance to the admitting Patient outcomes and statistical analysis Outcomes were hospital discharge status (alive or dead) and length of stay (LOS). Changes over time were also sought. Statistical analysis was conducted using SPSS 18.0 software (SPSS Inc) and GIS mapping software (MacroHealth Solutions). Categorical data were analysed using the 2 test for comparisons. Continuous data were reported as mean (95% CI) or median (interquartile range [IQR]). Comparative analysis was done using Kruskal Wallis one-way analysis of variance and Mann Whitney non-parametric tests. Associations were explored using Pearson s correlation coefficient (R). Logistical regression analysis (stepwise, conditional backward removal, probability for removal being 0.10, cut-off being 0.5) was undertaken, with hospital mortality as the dependent variable. Statistical significance was set at P < Results There were initially patient records, from which the following exclusions were made: patients (24.4%) from private-hospital ICUs; 109 patients (0.04%) from two ICUs with fewer than 100 patients each; 1518 patients (0.5%) with a postcode Figure 2. Proportion of ICU patients whose source of admission was another hospital and their mortality rate over the study period* % ICU admissions Proportion of ICU admissions (direct other-hospital ICU admission) Proportion of ICU admissions (indirect other-hospital ICU admission) Hospital mortality (direct other-hospital ICU admission) Hospital mortality (indirect other-hospital ICU admission) Year ICU = intensive care unit. R = Pearson s correlation coefficient. * Correlation coefficient for slope of line of best fit between ICU admission category and proportion of ICU admissions that were from another hospital: direct otherhospital ICU admission: R 2 = 0.235, P = 0.27; indirect other-hospital ICU admission: R 2 = 0.390, P =0.13. Correlation coefficient for slope of line of best fit between ICU admission category and hospital mortality: direct other-hospital ICU admission: R 2 = 0.890, P < 0.01; indirect other-hospital ICU admission: R 2 = 0.744, P = R 2 is the square of the correlation coefficient for the slope of the line of best fit; P is significance value for R % Hospital mortality 258 Critical Care and Resuscitation Volume 14 Number 4 December 2012

4 Table 1. Patient remoteness and distance travelled, by state/territory of admitting ICU and ICU admission category* (table continues on following page) State or territory of admitting ICU > 1500 ACT (n = 5863) (n = 5297) admission (n = 204) admission (n = 566) NSW (n = 78486) (n = ) admission (n = 8473) admission (n = ) NT (n = 8863) (n = 7461) admission (n = 462) admission (n = 1402) QLD (n = 46813) (n = ) admission (n = 3707) admission (n = ) SA (n = 15107) (n = ) admission (n = 1620) admission (n = 3512) TAS (n = 4397) (n = 3733) admission (n = 282) admission (n = 664) Distance category (km) Remoteness category Median distance (IQR) (km) 73.7% 24.7% 1.5% 0.1% 12.8 (7.5, 52.7) 40.4% 57.3% 2.2% 0.2% 99.3 (14.7, 133.5) 37.5% 58.1% 4.2% 0.3% (18.0, 147.4) 86.4% 9.4% 4.0% 0.2% 7.7 (3.7, 20.6) 71.7% 21.7% 6.5% 0.1% 23.8 (11.4, 62.7) 69.1% 25.4% 5.4% 0.1% 26.7 (10.9, 69.2) 57.7% 30.3% 7.9% 4.1% 16.9 (0.0, 162.7) 41.3% 13.0% 42.3% 3.4% 252 (7.0, 512.6) 23.7% 22.0% 48.4% 5.9% (162.7, 512.6) 71.1% 19.8% 8.4% 0.6% 18.0 (6.1, 62.3) 48.6% 34.3% 16.4% 0.7% 52.8 (21.4, 185.7) 38.0% 39.4% 19.9% 1.0% 76.8 (25.9, 240.7) 84.5% 8.9% 3.8% 2.8% 7.6 (3.9, 18.5) 47.2% 33.9% 16.7% 2.2% 55.5 (16.2, 263.3) 49.8% 32.4% 14.9% 2.8% 49.4 (15.5, 238.9) 74.5% 23.9% 1.3% 0.2% 27.9 (4.2, 50.1) 30.9% 67.5% 1.2% 0.4% 87.2 (27.9, 156.4) 31.7% 65.6% 1.7% 1.0% 87.2 (38.9, 137.5) Major city Inner regional Outer regional Remote Very remote 66.8% 18.2% 14.2% 0.8% 0.0% 36.7% 27.1% 33.7% 2.4% 0.2% 29.1% 33.1% 35.6% 2.1% 0.1% 76.0% 18.1% 5.4% 0.4% 0.1% 58.5% 30.6% 9.9% 0.9% 0.1% 53.1% 33.7% 11.7% 1.3% 0.2% 2.3% 1.3% 34.6% 25.5% 36.3% 2.0% 0.7% 41.5% 3.7% 52.1% 3.5% 1.2% 23.0% 4.8% 67.4% 55.5% 23.0% 18.1% 2.0% 1.4% 46.7% 27.6% 18.3% 3.1% 4.3% 38.0% 27.1% 23.5% 5.3% 6.0% 78.3% 11.2% 7.0% 2.0% 1.4% 41.7% 19.6% 25.1% 10.2% 3.4% 43.9% 20.0% 23.6% 9.3% 3.2% 4.5% 45.5% 47.2% 2.4% 0.4% 6.0% 26.9% 60.2% 5.6% 1.2% 4.6% 23.6% 64.4% 5.4% 1.9% ACT = Australian Capital Territory. ARIA+ = Accessibility/Remoteness Index of Australia (extended version). ICU = intensive care unit. IQR = interquartile range. NSW = New South Wales. NT = Northern Territory. QLD = Queensland. SA = South Australia. TAS = Tasmania. VIC = Victoria. WA = Western Australia. * The three categories of ICU admission were home ICU admission (hospital source of admission was not another hospital); direct other-hospital ICU admission (ICU source of admission was another hospital, and hospital source of admission was another hospital); and indirect other-hospital ICU admission (ICU source of admission was a ward, emergency department of operating room, and hospital source of admission was another hospital). Median distance travelled to admitting ICU. Based on ARIA+ classification. Critical Care and Resuscitation Volume 14 Number 4 December

5 Table 1. Patient remoteness and distance travelled, by state/territory of admitting ICU and ICU admission category* (continued from previous page) State or territory of admitting ICU > 1500 VIC (n = 54036) (n = ) admission (n = 4203) admission (n = ) WA (n = 5144) (n = 4290) admission (n = 543) admission (n = 854) Australia (n = ) (n = ) admission (n = ) admission (n = ) Distance category (km) Remoteness category Median distance (IQR) (km) 81.4% 16.6% 1.8% 0.2% 10.6 (5.7, 31.3) 59.9% 35.1% 4.9% 0.1% 30.8 (14.0, 116.2) 53.9% 41.5% 4.4% 0.1% 39.6 (14.9, 136.6) 74.2% 16.3% 7.3% 2.2% 20.4 (10.8, 52.8) 49.5% 22.6% 19.0% 8.8% 65.2 (24.6, 386.6) 53.8% 23.7% 14.9% 7.6% 39.6 (24.6, 207.4) 80.3% 14.8% 4.2% 0.7% 10.7 (4.4, 32.1) 59.6% 29.7% 10.0% 0.7% 31.0 (14.3, 122.1) 53.3% 35.1% 10.8% 0.8% 40.3 (15.4, 141.5) Major city Inner regional Outer regional Remote Very remote 67.5% 24.1% 8.1% 0.2% 0.0% 58.0% 30.0% 11.7% 0.3% 0.0% 48.9% 36.1% 14.5% 0.4% 0.1% 68.7% 17.9% 7.6% 2.6% 3.2% 42.8% 27.2% 11.7% 6.3% 12.0% 45.5% 30.7% 8.7% 5.0% 10.2% 64.8% 19.7% 11.4% 1.9% 2.2% 51.6% 27.9% 15.4% 2.4% 2.7% 44.5% 30.5% 18.2% 2.9% 3.8% ACT = Australian Capital Territory. ARIA+ = Accessibility/Remoteness Index of Australia (extended version). ICU = intensive care unit. IQR = interquartile range. NSW = New South Wales. NT = Northern Territory. QLD = Queensland. SA = South Australia. TAS = Tasmania. VIC = Victoria. WA = Western Australia. * The three categories of ICU admission were home ICU admission (hospital source of admission was not another hospital); direct other-hospital ICU admission (ICU source of admission was another hospital, and hospital source of admission was another hospital); and indirect other-hospital ICU admission (ICU source of admission was a ward, emergency department of operating room, and hospital source of admission was another hospital). Median distance travelled to admitting ICU. Based on ARIA+ classification. ICU increased significantly for both direct and indirect otherhospital ICU admissions (Figure 1). Over the same period, the proportion of patients who were admitted to the ICU, either directly or indirectly, from another hospital did not change significantly. However, hospital mortality fell significantly in both groups (Figure 2). Patients transferred from other hospitals were more often from a remote location and travelled further to their admitting ICU than patients who were not transferred. The proportion and remoteness category of such patients varied significantly among the states and territories (Table 1). Not all patients were admitted to an ICU in their own state or territory: overall, 6.4% were admitted to an ICU in another jurisdiction (proportions in specific states and territories were 1.6% [Tasmania], 1.9% [NSW], 2.1% [WA], 4.2% [Victoria], 8.4% [NT], 8.8% [Queensland], 11.0% [SA] and 64.5% [ACT]). Neither the geographical location nor the population size of a state or territory correlated with the proportion of patients in the direct other-hospital ICU admission category (R = 0.332, P = 0.47; and R = 0.219, P = 0.64, respectively) or the indirect other-hospital ICU admission category (R = 0.258, P = 0.58; and R = 0.333, P = 0.47, respectively). Similarly, neither the geographical location nor the population size of a state or territory correlated with the median distance to the admitting ICU for patients in the direct other-hospital ICU admission category (R = 0.230, P = 0.58; and R = 0.462, P = 0.25, respectively) or the indirect otherhospital ICU admission category (R = 0.265, P = 0.53; and R = 0.412, P = 0.31, respectively). The higher the proportion of patients in the 0 50 km distance to the closest ICU category, the shorter was the distance to the admitting ICU for both direct (R = 0.870, P < 0.01) and indirect (R = 0.864, P < 0.01) other-hospital ICU admissions. The closer any two ICUs were, the shorter was the distance to the admitting ICU for direct (R = 0.974, P < 0.01) and indirect (R = 0.969, P < 0.01) other-hospital ICU admissions. 260 Critical Care and Resuscitation Volume 14 Number 4 December 2012

6 Patient remoteness, distance travelled and associated outcome Mortality was highest in the direct other-hospital ICU admission group (19.5%), followed by the indirect otherhospital ICU admission group (17.8%) and the home ICU admission group (13.1%) (P < 0.01). Hospital mortality was similar across all distance and remoteness categories for direct other-hospital ICU admissions. However, hospital mortality for indirect other-hospital ICU admissions and home ICU admissions was significantly higher among patients in the 0 50 km distance category (compared with other distance categories) and in the major city category (compared with other remoteness categories) (Table 2). Patient ICU admission source, median distance travelled and mortality outcome are shown for each state and territory in Table 3. Patient diagnostic category, ICU admission source, median distance travelled and mortality outcome are listed in Table 4. The median distance travelled differed significantly between survivors and nonsurvivors in some states/territories and some diagnostic categories. However, it was only in the direct transfer groups in Victoria and the NT that the distance travelled was significantly longer for those who died. There were no instances of longer travel distances being linked to mortality among the listed diagnostic categories. Increasing patient remoteness was associated with increasing ICU and hospital LOS for both direct or indirect other-hospital ICU admissions, but with decreasing LOS for home ICU admissions (Table 2). Using regression analysis, the following variables were evaluated as predictors of hospital mortality for patients in the direct other-hospital ICU admission category: age, sex, predicted risk of death, diagnostic category, level of ICU, distance to admitting ICU and ARIA+ rating. In the final model, the only significant predictors of death were age (exp [exponential of the coefficient in the null model], 1.02 [95% CI, 1.01, 1.02]; P < 0.01), predicted risk of death (exp [95% CI, 93.51, ]; P <0.01), and the diagnostic categories of cardiovascular condition (exp 1.56 [95% CI, 1.05, 2.31]; P = 0.03) and neurological condition (exp 2.34 [95% CI, 1.59, 3.46]; P <0.01). Distance to admitting ICU (exp 1.00 [95% CI, 0.87, 1.17]; P = 0.20) and ARIA+ rating (exp 0.97 [95% CI, 0.84, 1.12]; P = 0.64) were not significant predictors of death. The predictors of hospital mortality for patients in the indirect other-hospital ICU admission category were similar, except for the addition of the diagnostic categories of trauma (exp 2.63 [95% CI, 1.84, 3.77]; P <0.01), respiratory condition (exp 1.68 [95% CI, 1.21, 2.31]; P < 0.01) and overdose (exp 2.14 [95% CI, 1.27, 3.6]; P <0.01). Discussion Over the period of our study, patient remoteness (median ARIA+ rating) and median distance travelled to the admitting ICU increased significantly, suggesting that patients were being sent from more remote locations and travelling greater distances to their admitting ICU. The proportion of ICU admissions that were from another hospital also increased over the same period. Distance travelled and remoteness of residence were not significant contributors to mortality, but both these factors were associated with an increase in ICU and hospital LOS for patients admitted from another hospital. These findings were similar across all states and territories and selected diagnostic categories. Factors such as a patient's age, severity of illness and diagnostic category were stronger predictors of outcome than distance and remoteness. Patients from the more remote and distant categories tended to be younger. For patients transferred from another hospital, increasing remoteness and increasing distance to the admitting ICU were associated with increasing severity of illness, a finding consistent with previous studies. 21,24 Despite these differences, the standardised mortality ratio (ie, casemix-adjusted mortality) was highest for patients in the less distant and remote categories. These findings may reflect differences in the demographics of more remotely located populations and/or a patient selection bias for referral and acceptance to another hospital. For example, factors such as early recognition of patient deterioration, 28 the extent to which the severity or complexity of illness exceed the capacity at lower-level facilities, and the risk benefit relationship of transportation (which itself is not without risk) 29 are likely to influence transport decisions.over the period of our study, patient remoteness (median ARIA+ rating) and median distance travelled to the admitting ICU increased significantly, suggesting that patients were being sent from more remote locations and travelling greater distances to their admitting ICU. The proportion of ICU admissions that were from another hospital also increased over the same period. Distance travelled and remoteness of residence were not significant contributors to mortality, but both these factors were associated with an increase in ICU and hospital LOS for patients admitted from another hospital. These findings were similar across all states and territories and selected diagnostic categories. Factors such as a patient s age, severity of illness and diagnostic category were stronger predictors of outcome than distance and remoteness. The distance to the admitting ICU differed among diagnostic categories, but was not significantly longer for nonsurvivors, despite studies demonstrating that time to definitive therapy is critical. 5,6,30 Our findings suggest that, even though distance or remoteness may contribute to the time taken to access definitive care, distance cannot be used as a Critical Care and Resuscitation Volume 14 Number 4 December

7 Table 2. Patient demographics, severity of illness* and hospital outcomes, by distance to admitting ICU, remoteness category and ICU admission category (table continues on following page) Distance to admitting ICU (median) Remoteness category 0 50 km (n = ) admission (n = ) admission (n = ) km (n = ) admission (n = 5785) admission (n = ) km (n = 7171) admission (n = 1951) admission (n = 5299) > 1500 km (n = 1107) admission (n = 138) admission (n = 407) P value Male Age (mean [95% CI]) (years) 60.1% 58.8 (58.6, 59.0) 52.7% 56.8 (56.2, 57.4) 56.2% 58.4 (58.0, 58.8) 65.2% 59.5 (59.1, 59.8) 59.5% 56.8 (55.9, 57.6) 63.1% 58.0 (57.4, 58.5) 64.4% 56.0 (55.1, 56.9) 60.7% 51.4 (49.7, 53.0) 64.8% 52.5 (51.5, 53.5) 65.6% 54.5 (52.2, 56.8) 57.9% 59.4 (52.6, 66.2) 69.0% 50.8 (46.7, 54.9) APACHE II score (mean [95% CI]) 15.0 (14.9, 15.0) 17.1 (17.4, 18.0) 17.0 (16.8, 17.1) 13.8 (13.7, 13.9) 18.3 (17.9, 18.8) 17.0 (16.5, 16.9) 13.9 (13.5, 14.2) 19.7 (18.8, 20.5) 17.6 (17.2, 18.0) 15.0 (14.0, 16.0) 20.7 (15.3, 26.1) 16.7 (14.8, 18.6) Risk of death (mean [95% CI]) (0.154, 0.156) (0.291, 0.307) (0.224, 0.231) (0.092, 0.096) (0.299, 0.324) (0.208, 0.216) (0.092, 0.099) (0.330, 0.377) (0.216, 0.231) (0.125, 0.154) (0.200, 0.479) (0.181, 0.235) ICU LOS (median [IQR]) (days) 3.6 (3.6, 3.6) 3.0 (2.0, 7.0) 2.6 (1.4, 5.0) 2.9 (2.8, 3.0) 4.0 (2.0, 8.0) 2.6 (1.4, 5.5) 3.1 (3.0, 3.2) 5.0 (2.0, 10.0) 2.8 (1.5, 6.2) 3.3 (2.8, 3.8) 8.0 (3.0, 10.0) 2.7 (1.1, 6.5) Hospital LOS (median [IQR]) (days) 16.0 (15.9, 16.2) 8.0 (3.0, 17.0) 9.3 (4.0, 18.8) 13.9 (13.6, 14.1) 10.0 (5.0, 20.0) 11.0 (5.8, 19.9) 15.4 (14.9, 15.9) 12.0 (5.0, 24.0) 13.0 (7.0, 23.5) 13.3 (12.2, 14.3) 15.0 (6.0, 30.0) 11.9 (6.1, 21.8) Hospital mortality < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 admission < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < Hospital source other hospital < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 Remoteness category Major city (n = ) admission (n = ) admission (n = ) Inner regional (n = ) admission (n = 5453) admission (n = ) 60.4% 59.7 (59.3, 60.1) 52.6% 56.4 (55.7, 57.0) 56.9% 58.3 (57.8, 58.7) 63.5% 59.7 (59.3, 60.1) 57.3% 57.4 (56.5, 58.3) 61.2% 59.0 (58.5, 60.0) 15.0 (14.9, 15.1) 17.8 (17.5, 18.1) 17.0 (16.8, 17.2) 13.6 (13.5, 13.8) 17.8 (17.4, 18.2) 16.3 (16.0, 16.5) (0.155, 0.158) (0.292, 0.310) (0.206, 0.216) (0.109, 0.113) (0.286, 0.312) (0.207, 0.216) 3.7 (3.6, 3.7) 4.0 (2.0, 7.0) 2.6 (1.4, 5.0) 3.1 (3.1, 3.2) 4.0 (2.0, 7.0) 2.6 (1.4, 5.4) 16.7 (16.6, 16.9) 8.0 (4.0, 18.0) 9.8 (4.6, 19.1) 13.8 (13.6, 14.1) 9.0 (4.0, 18.0) 10.6 (5.1, 19.0) Standardised mortality ratio 11.2% (0.718, 0.727) 19.0% (0.619, 0.653) 17.2% (0.619, 0.652) 4.9% (0.510, 0.533) 19.9% (0.615, 0.666) 14.9% (0.614, 0.666) 5.5% (0.556, 0.598) 20.7% (0.549, 0.627) 14.7% (0.549, 0.627) 7.8% (0.506, 0.624) 18.8% (0.392, 0.944) 12.0% (0.392, 0.940) 11.4% (0.722, 0.735) 19.6% (0.632, 0.671) 17.1% (0.632, 0.671) 7.4% (0.655, 0.679) 19.6% (0.628, 0.685) 15.5% (0.628, 0.685) APACHE = Acute Physiology and Chronic Health Evaluation. ARIA+ = Accessibility/Remoteness Index of Australia (extended version). ICU = intensive care unit. IQR = interquartile range. LOS = length of stay. * Based on APACHE II score. Based on ARIA+ classification. The three categories of ICU admission were home ICU admission (hospital source of admission was not another hospital); direct other-hospital ICU admission (ICU source of admission was another hospital, and hospital source of admission was another hospital); and indirect other-hospital ICU admission (ICU source of admission was a ward, emergency department of operating room, and hospital source of admission was another hospital). Casemix-adjusted mortality. P values refer to comparisons made within each ICU admission category between each of the distance categories (0 50 km, km, km, > 1500 km). ** P values refer to comparisons made within each ICU admission category between each of the remoteness categories (major city, inner regional, outer regional, remote, very remote). 262 Critical Care and Resuscitation Volume 14 Number 4 December 2012

8 Table 2. Patient demographics, severity of illness* and hospital outcomes, by distance to admitting ICU, remoteness category and ICU admission category (continued from previous page) Remoteness category Outer regional (n = ) Direct other-hospital ICU admission (n = 2998) Indirect other-hospital ICU admission (n = 8971) Remote (n = 3278) Direct other-hospital ICU admission (n = 458) Indirect other-hospital ICU admission (n = 1466) Very remote (n = 3642) Direct other-hospital ICU admission (n = 138) Indirect other-hospital ICU admission (n = 1918) P value** Male Age (mean [95% CI]) (years) 63.2% 56.7 (56.1, 57.2) 59.5% 57.0 (55.9, 58.2) 62.9% 57.3 (56.5, 58.1) 59.7% 51.0 (50.0, 51.9) 66.9% 52.9 (49.4, 56.3) 70.7% 55.5 (53.3,57.8) 55.9% 44.3 (43.4, 45.1) 55.6% 44.7 (42.0, 47.4) 57.5% 43.8 (42.4, 45.2) APACHE II score (mean [95% CI]) 14.7 (14.5, 15.0) 18.7 (18.1, 19.3) 17.3 (17.0, 17.6) 14.3 (13.8, 14.7) 19.3 (17.6, 20.9) 18.5 (17.5, 19.6) 15.8 (15.4, 16.3) 22.1 (20.6, 23.5) 19.0 (18.3, 19.8) Risk of death (mean [95% CI]) (0.114, 0.121) (0.308, 0.344) (0.212, 0.224) (0.139, 0.155) (0.286, 0.384) (0.219, 0.249) (0.169, 0.185) (0.351, 0.437) (0.247, 0.274) ICU LOS (median [IQR]) (days) 3.0 (2.9, 3.0) 4.0 (2.0, 8.0) 2.6 (1.3, 5.7) 2.9 (2.8, 3.1) 5.0 (3.0, 10.0) 2.5 (1.2, 5.9) 3.3 (3.2, 3.5) 5.0 (3.0, 10.0) 3.0 (1.6, 6.0) Hospital LOS (median [IQR]) (days) 13.6 (13.3, 14.0) 10.0 (4.0, 21.0) 10.9 (5.3, 20.4) 12.6 (11.7, 13.5) 13.5 (7.8, 23.3) 11.0 (5.7, 20.0) 14.6 (13.7, 15.4) 14.0 (6.0, 25.0) 11.6 (6.0, 22.0) Hospital mortality < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 Direct other-hospital < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < ICU admission Indirect other-hospital ICU admission <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 Pearson correlation with patient postcode ARIA+ rating Direct other-hospital ICU admission Indirect other-hospital ICU admission Pearson correlation with distance to admitting ICU Direct other-hospital ICU admission Indirect other-hospital ICU admission (P =0.34) (P =0.25) (P =0.01) Standardised mortality ratio 7.5% (0.620, 0.658) 18.6% (0.541, 0.604) 14.7% (0.541, 0.604) 7.2% (0.465, 0.518) 20.5% (0.534, 0.717) 16.2% (0.534, 0.717) 8.3% (0.449, 0.491) 20.5% (0.469, 0.584) 15.4% (0.469, 0.584) APACHE = Acute Physiology and Chronic Health Evaluation. ARIA+ = Accessibility/Remoteness Index of Australia (extended version). ICU = intensive care unit. IQR = interquartile range. LOS = length of stay. * Based on APACHE II score. Based on ARIA+ classification. The three categories of ICU admission were home ICU admission (hospital source of admission was not another hospital); direct other-hospital ICU admission (ICU source of admission was another hospital, and hospital source of admission was another hospital); and indirect other-hospital ICU admission (ICU source of admission was a ward, emergency department of operating room, and hospital source of admission was another hospital). Casemix-adjusted mortality. P values refer to comparisons made within each ICU admission category between each of the distance categories (0 50 km, km, km, > 1500 km). ** P values refer to comparisons made within each ICU admission category between each of the remoteness categories (major city, inner regional, outer regional, remote, very remote). Critical Care and Resuscitation Volume 14 Number 4 December

9 Table 3. Distance to admitting ICU, by state/territory of admitting ICU, hospital outcome and ICU admission category* State or territory of admitting ICU ACT (n = 5863) NSW (n =78486) Hospital outcome Home ICU admission Alive 7131 (90.6%) Dead 744 (9.4%) Alive (85.1%) Dead 8167 (14.9%) Median distance (IQR) (km) P 14.0 (7.8, 64.2) 9.8 (5.2, 18.0) 8.1 (3.7, 23.5) 6.4 (3.6, 12.8) Direct otherhospital ICU admission (78.0%) 112 (22.0%) < (78.8%) 1657 (21.3%) Median distance (IQR) (km) P 98.9 (14.7, 133.5) (14.7, 115.4) 23.8 (11.5, 64.4) 23.7 (11.1, 55.3) Indirect otherhospital ICU admission (81.4%) 392 (18.6%) (79.3%) 3509 (20.6%) Median distance (IQR) (km) P (18.0, 147.4) 98.9 (17.7, 135.2) 28.7 (11.3, 70.7) 23.7 (9.9, 57.6) 0.05 NT (n = 8863) QLD (n =46813) SA (n =15107) Alive 6418 (88.2%) Dead 861 (11.8%) Alive (90.5%) Dead 3433 (9.6%) Alive 9580 (80.8%) Dead 2282 (19.2%) 16.9 (0.0, 162.7) 16.9 (0.0, 162.7) 19.2 (6.7, 65.9) 11.2 (3.3, 22.9) 7.9 (4.2, 21.6) 5.9 (3.5, 11.9) (85.0%) 61 (15.0%) < (83.9%) 641 (16.1%) < (78.0%) 395 (22.0%) (4.2, 509.8) (252.0, 607.2) 50.5 (21.0, 174.4) 65.1 (24.6, 248.9) 55.5 (16.2, 263.3) 54.5 (16.0, 263.3) < (86.5%) 177 (13.4%) (86.9%) 1588 (13.1%) (76.3%) 917 (23.7%) (162.7, 512.6) (162.7, 512.6) 77.3 (25.9, 240.7) 71.1 (26.9, 248.9) 51.2 (16.2, 245.4) 30.0 (13.2, 218.1) <0.01 TAS (n = 4397) VIC (n =54036) WA (n = 5144) Alive 3092 (86.2%) Dead 496 (13.8%) Alive (86.8%) Dead 5505 (13.1%) Alive 3620 (87.8%) Dead 502 (12.2%) 27.9 (4.2, 55.7) 10.8 (0.0, 43.0) 11.1 (5.7, 34.0) 8.4 (4.4, 16.2) 20.4 (11.1, 57.0) 15.5 (9.7, 36.9) (84.3%) 39 (15.7%) < (80.9%) 809 (19.1%) (84.7%) 80 (15.3%) 87.3 (27.9, 153.4) 87.2 (38.9, 168.2) 29.8 (14.0, 110.8) 40.8 (14.5, 147.5) 48.5 (26.1, 386.6) 73.8 (22.0, 386.6) (82.5%) 103 (17.5%) (82.8%) 2032 (17.2%) (83.5%) 128 (16.5%) 87.3 (38.9, 130.4) 87.2 (42.6,137.5) 39.7 (15.2, 135.6) 40.1 (15.8, 135.9) 39.6 (15.2, 135.6) 39.9 (15.1, 136.4) All patients (n = ) Alive (86.9%) Dead (13.1%) 11.5 (4.7, 35.5) 7.6 (3.7, 17.3) < (80.5%) 3794 (19.5%) 30.8 (14.4, 121.5) 32.1 (14.0, 131.4) (82.2%) 8846 (17.8%) 42.8 (16.2, 145.8) 33.8 (13.4, 124.4) <0.01 ACT = Australian Capital Territory. ICU = intensive care unit. IQR = interquartile range. NSW = New South Wales. NT = Northern Territory. QLD = Queensland. SA = South Australia. TAS = Tasmania. VIC = Victoria. WA = Western Australia. * The three categories of ICU admission were home ICU admission (hospital source of admission was not another hospital); direct other-hospital ICU admission (ICU source of admission was another hospital, and hospital source of admission was another hospital); and indirect other-hospital ICU admission (ICU source of admission was a ward, emergency department of operating room, and hospital source of admission was another hospital). Figures are number of patients (%). P values relate to comparisons of distance for hospital outcome being alive or dead. surrogate for time to definitive care. For example, the time to notification for transport of patients in need of time-critical therapy can vary according to the diagnosis. 31 We suggest further investigation to better clarify differences in distance travelled based on diagnosis and overall timing. In keeping with previous studies, 24,32 our results showed that hospital mortality was higher in patients admitted from another hospital with trauma, overdose, or neurological, cardiovascular or respiratory conditions than in patients not admitted from another hospital. There were significant variations between states and territories in patient distance to admitting ICU and remoteness categories. Overall, 80% of patients postcode areas were less than 50 km from the admitting ICU, and 264 Critical Care and Resuscitation Volume 14 Number 4 December 2012

10 Table 4. Distance to admitting ICU, by APACHE II diagnostic category, hospital outcome and ICU admission category* Diagnostic category Surgical condition (n =73671) Trauma (n =11061) Hospital outcome Home ICU admission Alive (93.6%) Dead 4762 (6.5%) Alive (91.1%) Median distance (IQR) (km) P 15.7 (6.4, 59.6) 9.1 (4.1, 21.9) 15.3 (6.4, 37.7) Direct otherhospital ICU admission < (84.7%) 243 (15.3%) (88.8%) Median distance (IQR) (km) P 24.8 (10.5, 95.7) 23.7 (10.9, 94.2) 90.5 (27.0, 263.1) Indirect otherhospital ICU admission (88.7%) 1517 (11.3%) (89.6%) Median distance (IQR) (km) P 54.6 <0.01 (17.2, 161.7) 32.3 (12.8, 117.8) (28.7, 261.6) 0.20 Cardiovascular condition (n =43217) Respiratory condition (n =26551) Dead 990 (9.0%) Alive (86.2%) Dead 5987 (13.9%) Alive (83.3%) 12.0 (5.5, 31.7) 18.0 (6.8, 77.5) 7.4 (3.7, 16.2) 8.9 (3.8, 23.8) 159 (11.2%) < (68.9%) 978 (31.1%) < (83.0%) 74.9 (23.5, 234.0) 32.7 (15.1, 119.7) 29.9 (14.4, 104.2) 29.3 (14.2, 117.7) 470 (10.4%) (79.9%) 2204 (20.1%) (81.0%) 90.6 (24.6, 257.0) 54.9 (17.9, 161.5) 31.4 (13.5, 104.2) 33.2 (14.3, 118.0) <0.01 <0.01 Sepsis (n = 8900) Gastrointestinal condition (n =28854) Dead 4443 (16.7%) Alive 6343 (71.2%) Dead 2557 (28.7%) Alive (87.2%) 6.7 (3.4, 15.4) 7.8 (3.7, 19.7) 8.0 (3.7, 18.9) 9.5 (3.8, 25.9) 692 (17.0%) (72.1%) 563 (27.9%) < (81.8%) 27.8 (12.8, 102.8) 32.3 (14.7, 151.3) 30.8 (13.7, 150.0) 21.8 (9.7, 68.4) 1552 (19.0%) (71.0%) 1147 (29.0%) (81.8%) 29.4 (11.9, 97.3) 37.4 (15.1, 142.5) 33.2 (12.9, 150.0) 30.3 (12.8, 102.1) Neurological condition (n =14943) Overdose (n = 6834) Dead 3691 (12.8%) Alive (82.8%) Dead 2565 (17.2%) Alive 6694 (97.7%) 7.6 (3.7, 16.8) 11.7 (5.2, 31.9) 8.5 (4.1, 18.7) 6.6 (2.4, 14.6) 389 (18.2%) < (73.5%) 796 (26.6%) (97.7%) 29.3 (12.2, 117.4) 43.9 (16.4, 147.4) 49.1 (16.2, 175.6) 29.5 (15.0, 73.5) 1196 (18.2%) (75.0%) 1677 (25.0%) (97.5%) 31.8 (12.6, 112.0) 47.8 (16.7, 162.7) 40.7 (14.4, 151.8) 30.6 (14.8, 76.5) Dead 140 (2.0%) 5.5 (2.1, 10.4) 39 (2.4%) 26.7 (10.9, 106.5) 59 (2.6%) 29.2 (13.5, 114.9) APACHE = Acute Physiology and Chronic Health Evaluation. ICU = intensive care unit. IQR = interquartile range. * The three categories of ICU admission were home ICU admission (hospital source of admission was not another hospital); direct other-hospital ICU admission (ICU source of admission was another hospital, and hospital source of admission was another hospital); and indirect other-hospital ICU admission (ICU source of admission was a ward, emergency department of operating room, and hospital source of admission was another hospital). Figures are number of patients (%). P values relate to comparisons of distance for hospital outcome being alive or dead. thus within road vehicle capability. This ranged from 58% in the NT to 86% in NSW. However, the need for alternative vehicle transport was significant, as 35% of patients travelled between 51 and 300 km (appropriate for helicopter transport) and 12% travelled more than 300 km (appropriate for fixed-wing aircraft transport). It was a state or territory s population proximity to an ICU, and the proximity of one ICU to another, rather than the jurisdiction s geographical size, that correlated with the occurrence of interhospital transfers and distance to the admitting ICU. These findings have implications for planning of ICU resources and the selection of and demand for patient transport services. Our study highlights the utility of a GIS to integrate and illustrate (qualitatively and quantitatively) geocoded critical-care patient and illness-specific information with that Critical Care and Resuscitation Volume 14 Number 4 December

11 of geocoded ICU resource information. The accuracy of such analyses is influenced by the accuracy, precision and timeliness of the geocoded information. Our ICU dataset was large and reliable. 33 The distance measurements were based on centroids of postcodes, which are less accurate than other more precise measures of physical location (such as postcode of actual patient location at the time of needing hospital admission), but this is an accepted method, particularly for whole-of-population measures. 34 Most illnesses and injuries are likely to occur close to a patient s residential location. 35,36 Another potential confounder was the assumption that the referring hospital was close to the patient s residential postcode area. This may have biased our distance calculations, but less so our overall conclusions, as the measured effect of distance on outcomes was consistent across all jurisdictions. Finally, the findings of studies like ours may alter with time, in line with changes in population distribution, changes in a postcode s ARIA+ rating and remoteness category, 25 and changes in ICU Levels. Conclusion In summary, our findings suggest that severity of illness, diagnostic category and the need for transfer to another hospital are significant predictors of mortality, but patient remoteness and distance travelled to the admitting ICU are not. Referral patterns and mode of transport vary by state and territory, and it is the relationship of the population to ICU resources that correlates with distance to the admitting ICU. These findings have implications for planning of ICU and patient transport services. Future evaluation of factors other than distance is required to more clearly identify preventable factors that may potentially contribute to the increased mortality observed among patients admitted to an ICU from another hospital. Author details Arthas Flabouris, Staff Specialist in Intensive Care, 1 and Associate Professor 2 Graeme K Hart, Director, Intensive Care Unit 3 Angela Nicholls, Geographical Information Systems Specialist 4 1 Royal Adelaide Hospital, Adelaide, SA, Australia. 2 Discipline of Acute Care, School of Medicine, University of Adelaide, Adelaide, SA, Australia. 3 Department of Intensive Care, Austin Hospital, Melbourne, VIC, Australia. 4 MacroHealth Solutions, Woody Point, QLD, Australia. Correspondence: Arthas.Flabouris@health.sa.gov.au References 1 Barrett FA. Finke s 1792 map of human diseases: the first world disease map? Soc Sci Med 2000; 50: Brody H, Rip MR, Vinten-Johansen P, et al. Map-making and mythmaking in Broad Street: the London cholera epidemic, Lancet 2000; 356: Batmanian JJ, Lam M, Matthews C, et al. A protocol-driven model for the rapid initiation of stroke thrombolysis in the emergency department. Med J Aust 2007; 187: Kumar A, Roberts D, Wood KE, et al. Duration of hypotension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock. Crit Care Med 2006; 34: Nallamothu B, Fox KA, Kennelly BM, et al. Relationship of treatment delays and mortality in patients undergoing fibrinolysis and primary percutaneous coronary intervention. The Global Registry of Acute Coronary Events. Heart 2007; 93: Marler JR, Tilley BC, Lu M, et al. Early stroke treatment associated with better outcome: the NINDS rt-pa stroke study. Neurology 2000; 55: Fortney J, Rost K, Warren J. Comparing alternative methods of measuring geographic access to health services. Health Services and Outcomes Research Methodology 2000; 1: Higgs G. A literature review of the use of GIS-based measures of access to health care services. Health Serv Outcomes Res Methodol 2004; 5: Hall S, Holman CD, Sheiner H, Hendrie D. The influence of socioeconomic and locational disadvantage on survival after a diagnosis of lung or breast cancer in Western Australia. J Health Serv Res Policy 2004; 9 Suppl 2: Hall S, Holman CD, Platell C, et al. Colorectal cancer surgical care and survival: do private health insurance, socioeconomic and locational status make a difference? ANZ J Surg 2005; 75: Glover J, Tennant S, Page A. The impact of socioeconomic status and geographic location on Indigenous mortality in Australia, Adelaide: Public Health Information Development Unit, (Occasional Paper Series No. 1.) 12 Baker SP, Whitfield RA, O Neil B. Geographic variations in mortality from motor vehicle crashes. N Engl J Med 1987; 316: Jong KE, Smith DP, Yu XQ, et al. Remoteness of residence and survival from cancer in New South Wales. Med J Aust 2004; 180: Public Health Division, New South Wales Department of Health. The health of the people of New South Wales report of the Chief Health Officer. Sydney: NSW Department of Health, (accessed Aug 2012). 15 Australian Institute of Health and Welfare. Rural, regional and remote health: a study on mortality. 2nd ed. Canberra: AIHW, (AIHW Cat. No. PHE 95; Rural Health Series No. 8.) 16 Jones AP, Bentham G. Health service accessibility and deaths from asthma in 401 local authority districts in England and Wales, Thorax 1997; 52: Sexton P, Sexton TL. Excess coronary mortality among Australian men and women living outside the capital city statistical divisions. Med J Aust 2000; 172: Baade PD, Dasgupta P. Distance to the closest radiotherapy facility and survival after a diagnosis of rectal cancer in Queensland. Med J Aust 2011; 195: Jones AP, Bentham G. Emergency medical service accessibility and outcome from road traffic accidents. Public Health 1995; 109: Celaya MO, Berke EM, Onega TL, et al. Breast cancer stage at diagnosis and geographic access to mammography screening. Rural Remote Health 2010; 10: Durairaj L, Will JG, Torner JC, Doebbeling BN. Prognostic factors for mortality following interhospital transfers to the medical intensive care unit of a tertiary referral center. Crit Care Med 2003; 31: Drennan K, Hicks P, Hart GK. Intensive care resources and activity: Australia and New Zealand 2007/2008. Melbourne: Australian and New Zealand Intensive Care Society, Australian Department of Health and Ageing. The state of our public hospitals, June 2006 report. Canberra: Commonwealth of Australia, 266 Critical Care and Resuscitation Volume 14 Number 4 December 2012

12 health-ahca-sooph-index06.htm (accessed Aug 2012). 24 Flabouris A, Hart GK, George C. Observational study of patients admitted to intensive care units in Australia and New Zealand after interhospital transfer. Crit Care Resusc 2008; 10: Australian Bureau of Statistics. ASGC remoteness classification: purpose and use (Census Paper No. 03/01). Canberra: Commonwealth of Australia, Glover J, Tennant S. Remote areas statistical geography in Australia: notes on the Accessibility/Remoteness Index for Australia (ARIA+ version). Adelaide: Public Health Information Development Unit, (Working Paper Series No. 9.) 27 College of Intensive Care Medicine of Australia and New Zealand. Minimum standards for intensive care units. Policy document IC-1. Melbourne: CICM, Hillman KM, Bristow PJ, Chey T, et al. Duration of life-threatening antecedents prior to intensive care admission. Intensive Care Med 2002; 28: Flabouris A, Runciman WB, Levings B. Incidents during out of hospital patient transportation. Anaesth Intensive Care 2006; 34: Lerner EB, Moscati RM. The golden hour: scientific fact or medical urban legend? Acad Emerg Med 2001; 8: Leira EC, Lamb DL, Nugent AS, et al. Feasibility of acute clinical trials during aerial interhospital transfer. Stroke 2006; 37: Flabouris A. Patient referral and transportation to a regional tertiary ICU: patient demographics, severity of illness and outcome comparison with non-transported patients. Anaesth Intensive Care 1999; 27: Stow PJ, Hart GK, Higlett T, et al. Development and implementation of a high quality clinical database: the Australian and New Zealand Intensive Care Society adult patient database. J Crit Care 2006; 21: Jones SG, Ashby AJ, Momin SR, Naidoo A. Spatial implications associated with using Euclidean distance measurements and geographic centroid imputation in health care research. Health Serv Res 2010; 45: Wang HE, Weaver MD, Shapiro NI, Yealy DM. Opportunities for emergency medical services care of sepsis. Resuscitation 2010; 81: Margey R, Browne L, Murphy E, et al. The Dublin cardiac arrest registry: temporal improvement in survival from out-of-hospital cardiac arrest reflects improved pre-hospital emergency care. Europace 2011; 13: Critical Care and Resuscitation Volume 14 Number 4 December