bs_bs_banner
Aust. J. Rural Health (2015) 23, 4–18
Special Issue – Rural Cancer Review Article Cancer outcomes for Aboriginal and Torres Strait Islander Australians in rural and remote areas Abbey Diaz, BHlthSc, MAppSc (Research), Lisa J. Whop, BMedSc, MAE, Patricia C. Valery, MD, MPH, PhD, Suzanne P. Moore, BAppSc, MPH, PhD, Joan Cunningham, BA, A.L.M, PhD, Gail Garvey, B.Ed., MEd (Research), and John R. Condon, MBBS, MPH, PhD, FAFPHM, DipRACOG, DTM&H, CertHthEc Epidemiology and Health Services Division, Cancer Epidemiology, Menzies School of Health Research, Casuarina, Northern Territory, Australia
Abstract Objective: To examine the association between residential remoteness and stage of cancer at diagnosis, treatment uptake, and survival within the Australian Indigenous population. Design: Systematic review and matched retrospective cohort study. Setting: Australia. Participants: Systematic review: published papers that included a comparison of cancer stage at diagnosis, treatment uptake, mortality and/or survival for Indigenous people across remoteness categories were identified (n = 181). Fifteen papers (13 studies) were included in the review. Original analyses: new analyses were conducted using data from the Queensland Indigenous Cancer Study (QICS) comparing cancer stage at diagnosis, treatment uptake, and survival for Indigenous cancer patients living in rural/remote areas (n = 627, 66%) and urban areas (n = 329, 34%). Main Outcome Measures: Systematic review: Papers were included if there were related to stage of disease at diagnosis, treatment, mortality and survival of cancer. Restrictions were not placed on the outcome measures reported (e.g. standardised mortality ratios versus crude mortality rates). Original analyses: Odds ratios (OR, 95%CI) were used to compare stage of disease and treatment uptake between the two remoteness groups. Treatment uptake (treated/not treated) was analysed using logistic regression analysis. Survival was analysed using Cox proportional hazards regression. The final Correspondence: Ms Abbey Diaz, Epidemiology and Health Services Division, Menzies School of Health Research, PO Box 41096, Casuarina, Northern Territory 0811, Australia. Email: [email protected] Accepted for publication 3 December 2014. © 2015 National Rural Health Alliance Inc.
multivariate models included stage of cancer at diagnosis and area-level socioeconomic status (SEIFA). Results: Existing evidence of variation in cancer outcomes for Indigenous people in remote compared with metropolitan areas is limited. While no previous studies have reported on differences in cancer stage and treatment uptake by remoteness within the Indigenous population, the available evidence suggests Indigenous cancer patients are less likely to survive their cancer the further they live from urban centres. New analysis of QICS data indicates that Indigenous cancer patients in rural/remote Queensland were less likely to be diagnosed with localised disease and less likely to receive treatment for their cancer compared to their urban counterparts. Conclusion: More research is needed to fully understand geographic differentials in cancer outcomes within the Indigenous population. Knowing how geographical location interacts with Indigenous status can help to identify ways of improving cancer outcomes for Indigenous Australians. KEY WORDS: aboriginal, cancer, indigenous, rural, survival, urban.
Introduction Australians living in rural and remote areas have worse cancer outcomes than those living in major cities. Cancer mortality rates are higher in rural and remote populations,1 access to cancer screening and treatment services is poorer,2–4 and those with cancer have more advanced disease at diagnosis5 and lower survival.5–7 These disparities increase as the level of remoteness from metropolitan centres increases.8 This general pattern is observed when pooling all cancers together, but the pattern appears to differ between cancer types. In New South Wales, cancer patients diagnosed from 1992 to 1996 residing in remote doi: 10.1111/ajr.12169
5
URBAN-RURAL DISPARITIES IN INDIGENOUS CANCER OUTCOMES
What is already known on this subject: • Indigenous Australians compared to other Australians receive less cancer treatment for their cancer and have lower cancer survival rates. • People living in rural and remote areas compared to those living in urban areas of Australia have less access to cancer treatment services and have lower cancer survival rates.
areas had a 25% excess risk of death from cancer compared to their urban counter parts; yet the difference was only significant for cervical, prostate, multiple myeloma and rectum cancers, specifically. The differential was most prominent for cervical and prostate cancers; for both cancer types, patients in remote areas were over three times less likely to survive their cancer than their urban counterparts.5 Compared with other Australians, Aboriginal and Torres Strait Islander (hereafter respectfully referred to as ‘Indigenous’) Australians have higher cancer mortality,9 more advanced cancer at diagnosis10,11 and lower cancer survival.7,9,11–15 Indigenous Australians comprise approximately 3% of the total Australian population,16 but they make up a much higher proportion of the population in remote and very remote areas, comprising almost half of the very remote population (Table 1).17 To date, it is not clear how cancer outcomes vary across remoteness categories within the Indigenous population. Previous work has indicated some variations in cancer incidence by level of remoteness for Indigenous people. Incidence is higher in regional and remote areas than in metropolitan areas for all cancers combined and for some types of cancer (lip, mouth and pharynx, oesophagus, liver, lung, thyroid and unknown primary), but lower for others (melanoma, prostate, testis and kidney).9,18 The higher incidence of some cancers, especially those more likely to be fatal (e.g. oesophagus, liver and lung cancers), suggests that Indigenous cancer mortality may be higher in more remote areas. In contrast, TABLE 1:
What this study adds: • A review of available evidence highlighted the lack of existing evidence regarding urban–rural differentials in cancer outcomes for Indigenous Australians. • There is limited evidence to suggest that, at least for some cancers, Indigenous patients in rural and remote areas have lower survival than those in urban areas. • The QICS study found that Queensland Indigenous cancer patients living in rural and remote areas are diagnosed less with localised disease, have less cancer treatment and are less likely to survive their cancer.
non-Indigenous people in regional and remote compared to metropolitan areas have a lower incidence for all cancers combined and some individual cancers, including liver cancer which is a high-fatality cancer.18 This raises the possibility that the patterns of cancer outcomes by remoteness may differ for the Indigenous and non-Indigenous populations. In this paper, we review the evidence available about differentials in cancer stage at diagnosis, treatment and survival for Indigenous residents of urban compared with rural and remote areas, and report new analyses of data from our previous study of cancer outcomes for Indigenous Australians (The Queensland Indigenous Cancer Study (QICS))15 focusing on variation by remoteness of residence.
Methods Systematic review We conducted a systematic review to identify what is known about differences in Indigenous cancer-related outcomes by level of remoteness. English-language, full-text, peer-reviewed journal articles indexed in PubMed, MEDLINE and CINAHL by 5 December, 2013 were identified. The search terms included
Population distribution by level of remoteness
Factor
Major city
Inner regional
Outer regional
Remote
Very remote
% of total population in each area % of Indigenous population in each area % of each area who are Indigenous Australian
66 30 1
21 20 2
10 23 5
2 9 12
1 18 45
Table adapted from Australian Institute of Health and Welfare, 2007 Rural, Regional and Remote Health: a study on mortality (2nd edition).17 © 2015 National Rural Health Alliance Inc.
6
location (Australia*, Queensland, New South Wales, Victoria*, Tasmania*, Territory), disease (cancer*, Neoplasm*), cohort (Indigenous, Aborigin*, Torres*), outcome (Outcome*, Survival, Mortality, Treatment, Stage, Spread, Advanced Disease, Local*, Regional, TNM, AJCC, Dukes, Metasta*) and risk factor of interest (ARIA, Resident*, Remote*, Rural*, Region*, Metro*, Urban*, Geograph*, Distance, Location and Area). Papers were selected for inclusion if: (i) direct comparisons of Indigenous cancer stage at diagnosis, treatment, mortality and/or survival outcomes were made according to level of remoteness; or (ii) if indirect comparisons were made by comparing Indigenous and nonIndigenous differentials in the outcomes of interest across remoteness categories. Restrictions were not placed on the study methodology (e.g. qualitative versus quantitative), outcome measures reported (e.g. standardised mortality ratios versus crude mortality rates) or the definition of remoteness employed. Papers were excluded if: (i) they were not related to cancer; (ii) were related to a cancer outcome other than stage of disease, treatment, mortality and survival (e.g. screening, incidence and utilisation of support services); (iii) did not include an Indigenous Australian cohort; or (iv) did not indirectly or directly compare Indigenous outcomes by level of remoteness. Ineligible papers were excluded following screening of the title and abstracts (AD and LW) and full text (AD). Uncertainties and conflicting decisions were rare and were resolved through discussion with other authors. To ensure relevant papers were not missed, reference lists and government reports were perused for additional studies, resulting in the inclusion of two additional reports. Two studies required additional information/ clarification. Corresponding authors were contacted and both provided sufficient information for the papers to be included.
The Queensland Indigenous Cancer Study The study methods have previously been described in detail.15 Briefly, all Queensland Indigenous residents diagnosed with cancer during 1998–2004 were identified via the Queensland Cancer Registry. Those who had at least one admission to a state public hospital related to their diagnosis or treatment were included. Diagnostic details, cancer stage, treatment and presence of comorbidities were abstracted from medical records on standard forms. A modified Charlson Comorbidity Index Score19 was calculated and grouped as: ‘0’ score (no comorbidity recorded), 1, 2 + . The Socio-Economic Index for Areas (SEIFA) Index of Relative Socioeconomic Disadvantage20 was determined using patients’ residential postcodes. This index ranks geographical areas into
A. DIAZ ET AL.
quintiles of disadvantage, which were then collapsed into three classes for the purposes of this analysis: least (quintiles 4 and 5), moderately (quintile 3) and most (quintile 1 and 2) disadvantaged. Remoteness of residence was determined using the Accessibility/Remoteness Index of Australia (ARIA),21 which assigns localities a score based on their size and road accessibility to service centres, and categorised into urban (ARIA categories ‘major city’ and ‘inner regional’) and rural/remote (categories ‘outer regional’, ‘remote’ and ‘very remote’). Date and cause of death were obtained from the Australian National Death Index. At the time of undertaking this study, National Death Index data were only available until 2007, and we were only able to report on vital status until 31 December, 2006. The study included 956 Queensland Indigenous cancer patients, 329 (34%) living in urban areas and 627 (66%) in rural/remote areas. The cohort was described using counts and percentages for categorical variables and mean and standard deviations (SDs) for continuous variables. Differences in proportions were tested using the Pearson’s Chi-square statistic (nominal data) or the Mantel–Haenszel test for trend (ordinal data). Odds ratios (OR; 95% confidence interval (CI)) were used to compare stage of disease and treatment uptake for the urban and rural/remote groups. Treatment uptake (treated/not treated) was analysed using logistic regression analysis. Survival was analysed using Cox proportional hazards regression. Cox proportional hazards regression was used to calculate crude and adjusted hazard ratios (HR; 95% CI). An HR greater than one indicates a higher risk of death (lower survival) during the study period for those in rural/remote areas compared to urban areas. For both multivariate analyses, the variables of a priori interest were considered, namely cancer stage at diagnosis, arealevel socioeconomic status (SEIFA), cancer type, comorbidity, age group, sex and treatment uptake. Variables were added to the model in a forward step-wise fashion. Likelihood ratio tests were used to select covariates to be included; only stage at diagnosis and SEIFA remained in the final models.
Data considerations The studies identified for inclusion in this review used varying measures to categorise ‘remoteness’. Most studies used the ARIA or ARIA+ and are generally classified into five categories: highly accessible, accessible, moderately accessible, remote and very remote. The new analysis of Queensland Indigenous Cancer Study (QICS) data was conducted using ARIA+. Other studies in the review used the Australian Bureau of Statistics (ABS) Australian Standard Geographical Classification (ASGC) classifications, which use ARIA+ scores to © 2015 National Rural Health Alliance Inc.
7
URBAN-RURAL DISPARITIES IN INDIGENOUS CANCER OUTCOMES
TABLE 2:
State and territory composition of Australian Standard Geographical Classification (ASGC) categories
State/Territory
Major city
Inner regional
Outer regional
Remote
Very remote
New South Wales (NSW) Victoria (Vic) Queensland (Qld) South Australia (SA) Western Australia (WA) Tasmania (Tas) Northern Territory (NT) Australian Capital Territory (ACT)
✓ ✓ ✓ ✓ ✓ x x ✓
✓ ✓ ✓ ✓ ✓ ✓ x ✓
✓ ✓ ✓ ✓ ✓ ✓ ✓ x
✓ ✓ ✓ ✓ ✓ ✓ ✓ x
✓ x ✓ ✓ ✓ ✓ ✓ x
Table adapted from the ABS (2011) Statistical Geographical (Volume 1) ASGC.22
define five similar categories, which are labelled: major city (e.g. Sydney), inner regional (e.g. Toowoomba, Ballarat), outer regional (e.g. Darwin), remote (e.g. Alice Springs, Broome, Cloncurry) and very remote (e.g. Torres Strait, Indulkana). The differences between the two classification systems are slight.21 Other studies used nominal methods of separating the state into categories of remoteness, such as comparing the capital city (urban) against all other areas within a state or territory (regional/remote). The qualitative studies included did not explicitly describe how level of remoteness was determined, but have used ASGC labelling. For consistency and ease of readability, this paper will use the ASGC labels for all studies. Different states are comprised of different levels of remoteness (Table 2).22 For example, the Northern Territory (NT) includes very remote, remote and outer regional areas, and therefore NT studies do not have a ‘major city’ comparison. In these instances, comparisons are made between the levels of remoteness that make up the state/territory. Published data were not of sufficient detail to allow recalculation of results into directly comparable remoteness categories. We have therefore only been able to describe the general pattern of variation between urban and remote areas. The quality of identification of Indigenous people in data sources is also an important consideration, particularly in studies that have used death registrations data, cancer registrations data or clinical and administrative data sources that are not designed for research purposes. The accuracy of Indigenous status data in these data sources varies between states/territories. The Australian Institute of Health and Welfare9 reports that Indigenous status data in cancer registries and the national mortality database are reasonably reliable only for New South Wales (NSW), Queensland (Qld), Western Australia (WA) and the NT, and thus mortality and survival rates derived from these sources can be considered accurate. © 2015 National Rural Health Alliance Inc.
Results The initial literature search yielded 181 reports. After exclusions, 15 eligible papers that reported on 13 individual studies were included for review, relating to cancer treatment (six papers), mortality (3 papers) and survival (6 papers) (Fig. 1); no papers relating to stage at diagnosis were found. Direct evidence of variation in cancer outcomes by remoteness within the Indigenous population is limited; indirect evidence, based on Indigenous : non-Indigenous comparisons within remoteness areas, is also limited. The studies included in the review varied considerably according to: breadth of study (national versus jurisdictional); classification of remoteness; quality of Indigenous identification in data sources; size of the cohort; included cancer types; and the outcome measurements used. As such, the extent to which these studies can be integrated is limited. However, some general patterns are apparent.
Cancer stage at diagnosis No studies were identified via the literature review that compared, directly or indirectly, cancer stage at diagnosis for Indigenous people by remoteness of residence. In new analysis of the QICS data, rural/remote Indigenous residents were less likely to be diagnosed with localised disease (as opposed to regional spread or distant metastasis) than urban residents (36% versus 43%, respectively; OR 0.7 (95% CI 0.5–1.0, P = 0.041) (Table 3).
Cancer treatment No studies were identified that directly compared Indigenous cancer patients’ uptake of treatment across remoteness categories, but there was some indirect evidence of disparities by place of residence. Three
8
A. DIAZ ET AL.
FIGURE 1: Schematic representation of the systematic review process.
Database search
181 papers idenfied
149 Titles and abstracts screened
60 full-paper arcles screened
13 papers idenfied (11 individual studies)
32 Duplicates Removed Excluded n=63 aer tle and n=26 aer abstract screening
47 papers excluded
Government Reports = 2 (2 individual studies)
15 papers (13 individual studies) included in the review
qualitative studies (five papers) suggested that the greater the distance a patient lives from treatment services, the more barriers they face to commence and complete their treatment (Table 4).23–27 Similar themes were reported, regardless of whether data were collected from cancer patients27 and their families24–26 or health care providers.23 Most barriers identified by these studies were related to travelling away from home and community into an unfamiliar city and navigation through intimidating and unfamiliar hospitals and systems. Importantly, these studies identified two facilitators of treatment adherence, namely, Indigenous health workers or Indigenous liaison officers and hospital Aboriginal health units.23–25 In a quantitative study of prevalence of hysterectomy among gynaecological cancer patients, there was little difference among non-Indigenous women in the proportion who underwent hysterectomy by remoteness category.28 Compared with non-Indigenous women from metropolitan areas, hysterectomy was more common for Indigenous women in all regions. While the difference was greater for Indigenous women from regional areas (Relative Risk; RR 1.7, 95% CI 1.3–2.4) than for those from metropolitan (RR 1.3, 95% CI 0.9–1.7) or remote areas (RR 1.2, 95% CI 0.9–1.5), these differentials were not significantly different across remoteness categories.28 In new analysis of QICS data, Indigenous cancer patients were less likely to receive cancer treatment if they were from rural/remote areas (72%) than from urban areas (80%) (OR 0.7, 95% CI 0.5–0.9). The difference remained even after taking into account cancer stage and socioeconomic status (adjusted OR 0.6, 95% CI 0.4–0.8). Among those who received treat-
ment, median time to treatment was similar for rural/ remote and urban patients (15 (95% CI 12–18) versus 13 (95% CI 7–19) days, P = 0.395); there was no significant difference even after adjusting for cancer stage and socioeconomic status (adjusted HR 1.0, 95% CI 0.8–1.1).
Cancer mortality Two studies were identified that directly compared cancer mortality for Indigenous people in remote areas with those in non-remote areas.29,30 One of these,30 along with an additional paper,31 compared the Indigenous : non-Indigenous differential in mortality rates across levels of remoteness (Table 5). The influence of remoteness on mortality may vary by cancer type. For NT Indigenous women in 1996–2006, age-standardised breast cancer mortality rates (per 100 000) in remote/very remote areas (29.9, 95% CI 16.6–43.3) were much lower than those in outer regional areas (60.0, 95% CI 0.0–141.8), although not statistically significant, but for cervical cancer mortality rates were higher in remote/very remote (20.6, 95% CI 8.3–32.9) than outer regional areas (2.5, 95% CI 0.0–7.4).29 An earlier study of cervical cancer mortality in Indigenous women in SA, WA and NT in 1986–1997 found that in comparison to Indigenous people in metropolitan areas, those living in remote areas are significantly more likely to die following a cervical cancer diagnosis (standardised mortality ratio (SMR) 2.1, 95% CI 1.6–2.8), but not those in regional areas (SMR 1.7, 95% CI 0.9–3.3)30 However, Indigenous women were more likely to die than non-Indigenous women in metropolitan areas © 2015 National Rural Health Alliance Inc.
9
URBAN-RURAL DISPARITIES IN INDIGENOUS CANCER OUTCOMES
TABLE 3: Demographic and clinical characteristics of Indigenous people diagnosed with cancer between 1998 and 2004 by place of residence at diagnosis, from the Queensland Indigenous Cancer Study (QICS)
Total Age 18–39 years 40–59 years 60+ years Sex male female Socioeconomic status (SEIFA) Most disadvantaged Disadvantaged Intermediate advantage Advantaged Most advantaged Stage at diagnosis† Localised cancer Regional spread Distant metastasis Regional and distant metastasis Charlson Comorbidity Index 0 1 2+ Diabetes Renal disease Cardiovascular disease Respiratory disease Any treatment given Treatment No treatment Cancer type Head and neck Colorectal Liver and biliary Lung Breast Cervix Uterus All other cancers
Rural/remote areas n (%)
Urban areas n (%)
Total n (%)
627 (66)
329 (34)
956 (100)
73 (12) 282 (45) 272 (43)
35 (11) 139 (42) 155 (47)
108 (11) 421 (44) 427 (45)
295 (47) 332 (53)
141 (43) 188 (57)
436 (46) 520 (54)
P-value
0.299
0.216
Aust. J. Rural Health (2015) 23, 4–18
Special Issue – Rural Cancer Review Article Cancer outcomes for Aboriginal and Torres Strait Islander Australians in rural and remote areas Abbey Diaz, BHlthSc, MAppSc (Research), Lisa J. Whop, BMedSc, MAE, Patricia C. Valery, MD, MPH, PhD, Suzanne P. Moore, BAppSc, MPH, PhD, Joan Cunningham, BA, A.L.M, PhD, Gail Garvey, B.Ed., MEd (Research), and John R. Condon, MBBS, MPH, PhD, FAFPHM, DipRACOG, DTM&H, CertHthEc Epidemiology and Health Services Division, Cancer Epidemiology, Menzies School of Health Research, Casuarina, Northern Territory, Australia
Abstract Objective: To examine the association between residential remoteness and stage of cancer at diagnosis, treatment uptake, and survival within the Australian Indigenous population. Design: Systematic review and matched retrospective cohort study. Setting: Australia. Participants: Systematic review: published papers that included a comparison of cancer stage at diagnosis, treatment uptake, mortality and/or survival for Indigenous people across remoteness categories were identified (n = 181). Fifteen papers (13 studies) were included in the review. Original analyses: new analyses were conducted using data from the Queensland Indigenous Cancer Study (QICS) comparing cancer stage at diagnosis, treatment uptake, and survival for Indigenous cancer patients living in rural/remote areas (n = 627, 66%) and urban areas (n = 329, 34%). Main Outcome Measures: Systematic review: Papers were included if there were related to stage of disease at diagnosis, treatment, mortality and survival of cancer. Restrictions were not placed on the outcome measures reported (e.g. standardised mortality ratios versus crude mortality rates). Original analyses: Odds ratios (OR, 95%CI) were used to compare stage of disease and treatment uptake between the two remoteness groups. Treatment uptake (treated/not treated) was analysed using logistic regression analysis. Survival was analysed using Cox proportional hazards regression. The final Correspondence: Ms Abbey Diaz, Epidemiology and Health Services Division, Menzies School of Health Research, PO Box 41096, Casuarina, Northern Territory 0811, Australia. Email: [email protected] Accepted for publication 3 December 2014. © 2015 National Rural Health Alliance Inc.
multivariate models included stage of cancer at diagnosis and area-level socioeconomic status (SEIFA). Results: Existing evidence of variation in cancer outcomes for Indigenous people in remote compared with metropolitan areas is limited. While no previous studies have reported on differences in cancer stage and treatment uptake by remoteness within the Indigenous population, the available evidence suggests Indigenous cancer patients are less likely to survive their cancer the further they live from urban centres. New analysis of QICS data indicates that Indigenous cancer patients in rural/remote Queensland were less likely to be diagnosed with localised disease and less likely to receive treatment for their cancer compared to their urban counterparts. Conclusion: More research is needed to fully understand geographic differentials in cancer outcomes within the Indigenous population. Knowing how geographical location interacts with Indigenous status can help to identify ways of improving cancer outcomes for Indigenous Australians. KEY WORDS: aboriginal, cancer, indigenous, rural, survival, urban.
Introduction Australians living in rural and remote areas have worse cancer outcomes than those living in major cities. Cancer mortality rates are higher in rural and remote populations,1 access to cancer screening and treatment services is poorer,2–4 and those with cancer have more advanced disease at diagnosis5 and lower survival.5–7 These disparities increase as the level of remoteness from metropolitan centres increases.8 This general pattern is observed when pooling all cancers together, but the pattern appears to differ between cancer types. In New South Wales, cancer patients diagnosed from 1992 to 1996 residing in remote doi: 10.1111/ajr.12169
5
URBAN-RURAL DISPARITIES IN INDIGENOUS CANCER OUTCOMES
What is already known on this subject: • Indigenous Australians compared to other Australians receive less cancer treatment for their cancer and have lower cancer survival rates. • People living in rural and remote areas compared to those living in urban areas of Australia have less access to cancer treatment services and have lower cancer survival rates.
areas had a 25% excess risk of death from cancer compared to their urban counter parts; yet the difference was only significant for cervical, prostate, multiple myeloma and rectum cancers, specifically. The differential was most prominent for cervical and prostate cancers; for both cancer types, patients in remote areas were over three times less likely to survive their cancer than their urban counterparts.5 Compared with other Australians, Aboriginal and Torres Strait Islander (hereafter respectfully referred to as ‘Indigenous’) Australians have higher cancer mortality,9 more advanced cancer at diagnosis10,11 and lower cancer survival.7,9,11–15 Indigenous Australians comprise approximately 3% of the total Australian population,16 but they make up a much higher proportion of the population in remote and very remote areas, comprising almost half of the very remote population (Table 1).17 To date, it is not clear how cancer outcomes vary across remoteness categories within the Indigenous population. Previous work has indicated some variations in cancer incidence by level of remoteness for Indigenous people. Incidence is higher in regional and remote areas than in metropolitan areas for all cancers combined and for some types of cancer (lip, mouth and pharynx, oesophagus, liver, lung, thyroid and unknown primary), but lower for others (melanoma, prostate, testis and kidney).9,18 The higher incidence of some cancers, especially those more likely to be fatal (e.g. oesophagus, liver and lung cancers), suggests that Indigenous cancer mortality may be higher in more remote areas. In contrast, TABLE 1:
What this study adds: • A review of available evidence highlighted the lack of existing evidence regarding urban–rural differentials in cancer outcomes for Indigenous Australians. • There is limited evidence to suggest that, at least for some cancers, Indigenous patients in rural and remote areas have lower survival than those in urban areas. • The QICS study found that Queensland Indigenous cancer patients living in rural and remote areas are diagnosed less with localised disease, have less cancer treatment and are less likely to survive their cancer.
non-Indigenous people in regional and remote compared to metropolitan areas have a lower incidence for all cancers combined and some individual cancers, including liver cancer which is a high-fatality cancer.18 This raises the possibility that the patterns of cancer outcomes by remoteness may differ for the Indigenous and non-Indigenous populations. In this paper, we review the evidence available about differentials in cancer stage at diagnosis, treatment and survival for Indigenous residents of urban compared with rural and remote areas, and report new analyses of data from our previous study of cancer outcomes for Indigenous Australians (The Queensland Indigenous Cancer Study (QICS))15 focusing on variation by remoteness of residence.
Methods Systematic review We conducted a systematic review to identify what is known about differences in Indigenous cancer-related outcomes by level of remoteness. English-language, full-text, peer-reviewed journal articles indexed in PubMed, MEDLINE and CINAHL by 5 December, 2013 were identified. The search terms included
Population distribution by level of remoteness
Factor
Major city
Inner regional
Outer regional
Remote
Very remote
% of total population in each area % of Indigenous population in each area % of each area who are Indigenous Australian
66 30 1
21 20 2
10 23 5
2 9 12
1 18 45
Table adapted from Australian Institute of Health and Welfare, 2007 Rural, Regional and Remote Health: a study on mortality (2nd edition).17 © 2015 National Rural Health Alliance Inc.
6
location (Australia*, Queensland, New South Wales, Victoria*, Tasmania*, Territory), disease (cancer*, Neoplasm*), cohort (Indigenous, Aborigin*, Torres*), outcome (Outcome*, Survival, Mortality, Treatment, Stage, Spread, Advanced Disease, Local*, Regional, TNM, AJCC, Dukes, Metasta*) and risk factor of interest (ARIA, Resident*, Remote*, Rural*, Region*, Metro*, Urban*, Geograph*, Distance, Location and Area). Papers were selected for inclusion if: (i) direct comparisons of Indigenous cancer stage at diagnosis, treatment, mortality and/or survival outcomes were made according to level of remoteness; or (ii) if indirect comparisons were made by comparing Indigenous and nonIndigenous differentials in the outcomes of interest across remoteness categories. Restrictions were not placed on the study methodology (e.g. qualitative versus quantitative), outcome measures reported (e.g. standardised mortality ratios versus crude mortality rates) or the definition of remoteness employed. Papers were excluded if: (i) they were not related to cancer; (ii) were related to a cancer outcome other than stage of disease, treatment, mortality and survival (e.g. screening, incidence and utilisation of support services); (iii) did not include an Indigenous Australian cohort; or (iv) did not indirectly or directly compare Indigenous outcomes by level of remoteness. Ineligible papers were excluded following screening of the title and abstracts (AD and LW) and full text (AD). Uncertainties and conflicting decisions were rare and were resolved through discussion with other authors. To ensure relevant papers were not missed, reference lists and government reports were perused for additional studies, resulting in the inclusion of two additional reports. Two studies required additional information/ clarification. Corresponding authors were contacted and both provided sufficient information for the papers to be included.
The Queensland Indigenous Cancer Study The study methods have previously been described in detail.15 Briefly, all Queensland Indigenous residents diagnosed with cancer during 1998–2004 were identified via the Queensland Cancer Registry. Those who had at least one admission to a state public hospital related to their diagnosis or treatment were included. Diagnostic details, cancer stage, treatment and presence of comorbidities were abstracted from medical records on standard forms. A modified Charlson Comorbidity Index Score19 was calculated and grouped as: ‘0’ score (no comorbidity recorded), 1, 2 + . The Socio-Economic Index for Areas (SEIFA) Index of Relative Socioeconomic Disadvantage20 was determined using patients’ residential postcodes. This index ranks geographical areas into
A. DIAZ ET AL.
quintiles of disadvantage, which were then collapsed into three classes for the purposes of this analysis: least (quintiles 4 and 5), moderately (quintile 3) and most (quintile 1 and 2) disadvantaged. Remoteness of residence was determined using the Accessibility/Remoteness Index of Australia (ARIA),21 which assigns localities a score based on their size and road accessibility to service centres, and categorised into urban (ARIA categories ‘major city’ and ‘inner regional’) and rural/remote (categories ‘outer regional’, ‘remote’ and ‘very remote’). Date and cause of death were obtained from the Australian National Death Index. At the time of undertaking this study, National Death Index data were only available until 2007, and we were only able to report on vital status until 31 December, 2006. The study included 956 Queensland Indigenous cancer patients, 329 (34%) living in urban areas and 627 (66%) in rural/remote areas. The cohort was described using counts and percentages for categorical variables and mean and standard deviations (SDs) for continuous variables. Differences in proportions were tested using the Pearson’s Chi-square statistic (nominal data) or the Mantel–Haenszel test for trend (ordinal data). Odds ratios (OR; 95% confidence interval (CI)) were used to compare stage of disease and treatment uptake for the urban and rural/remote groups. Treatment uptake (treated/not treated) was analysed using logistic regression analysis. Survival was analysed using Cox proportional hazards regression. Cox proportional hazards regression was used to calculate crude and adjusted hazard ratios (HR; 95% CI). An HR greater than one indicates a higher risk of death (lower survival) during the study period for those in rural/remote areas compared to urban areas. For both multivariate analyses, the variables of a priori interest were considered, namely cancer stage at diagnosis, arealevel socioeconomic status (SEIFA), cancer type, comorbidity, age group, sex and treatment uptake. Variables were added to the model in a forward step-wise fashion. Likelihood ratio tests were used to select covariates to be included; only stage at diagnosis and SEIFA remained in the final models.
Data considerations The studies identified for inclusion in this review used varying measures to categorise ‘remoteness’. Most studies used the ARIA or ARIA+ and are generally classified into five categories: highly accessible, accessible, moderately accessible, remote and very remote. The new analysis of Queensland Indigenous Cancer Study (QICS) data was conducted using ARIA+. Other studies in the review used the Australian Bureau of Statistics (ABS) Australian Standard Geographical Classification (ASGC) classifications, which use ARIA+ scores to © 2015 National Rural Health Alliance Inc.
7
URBAN-RURAL DISPARITIES IN INDIGENOUS CANCER OUTCOMES
TABLE 2:
State and territory composition of Australian Standard Geographical Classification (ASGC) categories
State/Territory
Major city
Inner regional
Outer regional
Remote
Very remote
New South Wales (NSW) Victoria (Vic) Queensland (Qld) South Australia (SA) Western Australia (WA) Tasmania (Tas) Northern Territory (NT) Australian Capital Territory (ACT)
✓ ✓ ✓ ✓ ✓ x x ✓
✓ ✓ ✓ ✓ ✓ ✓ x ✓
✓ ✓ ✓ ✓ ✓ ✓ ✓ x
✓ ✓ ✓ ✓ ✓ ✓ ✓ x
✓ x ✓ ✓ ✓ ✓ ✓ x
Table adapted from the ABS (2011) Statistical Geographical (Volume 1) ASGC.22
define five similar categories, which are labelled: major city (e.g. Sydney), inner regional (e.g. Toowoomba, Ballarat), outer regional (e.g. Darwin), remote (e.g. Alice Springs, Broome, Cloncurry) and very remote (e.g. Torres Strait, Indulkana). The differences between the two classification systems are slight.21 Other studies used nominal methods of separating the state into categories of remoteness, such as comparing the capital city (urban) against all other areas within a state or territory (regional/remote). The qualitative studies included did not explicitly describe how level of remoteness was determined, but have used ASGC labelling. For consistency and ease of readability, this paper will use the ASGC labels for all studies. Different states are comprised of different levels of remoteness (Table 2).22 For example, the Northern Territory (NT) includes very remote, remote and outer regional areas, and therefore NT studies do not have a ‘major city’ comparison. In these instances, comparisons are made between the levels of remoteness that make up the state/territory. Published data were not of sufficient detail to allow recalculation of results into directly comparable remoteness categories. We have therefore only been able to describe the general pattern of variation between urban and remote areas. The quality of identification of Indigenous people in data sources is also an important consideration, particularly in studies that have used death registrations data, cancer registrations data or clinical and administrative data sources that are not designed for research purposes. The accuracy of Indigenous status data in these data sources varies between states/territories. The Australian Institute of Health and Welfare9 reports that Indigenous status data in cancer registries and the national mortality database are reasonably reliable only for New South Wales (NSW), Queensland (Qld), Western Australia (WA) and the NT, and thus mortality and survival rates derived from these sources can be considered accurate. © 2015 National Rural Health Alliance Inc.
Results The initial literature search yielded 181 reports. After exclusions, 15 eligible papers that reported on 13 individual studies were included for review, relating to cancer treatment (six papers), mortality (3 papers) and survival (6 papers) (Fig. 1); no papers relating to stage at diagnosis were found. Direct evidence of variation in cancer outcomes by remoteness within the Indigenous population is limited; indirect evidence, based on Indigenous : non-Indigenous comparisons within remoteness areas, is also limited. The studies included in the review varied considerably according to: breadth of study (national versus jurisdictional); classification of remoteness; quality of Indigenous identification in data sources; size of the cohort; included cancer types; and the outcome measurements used. As such, the extent to which these studies can be integrated is limited. However, some general patterns are apparent.
Cancer stage at diagnosis No studies were identified via the literature review that compared, directly or indirectly, cancer stage at diagnosis for Indigenous people by remoteness of residence. In new analysis of the QICS data, rural/remote Indigenous residents were less likely to be diagnosed with localised disease (as opposed to regional spread or distant metastasis) than urban residents (36% versus 43%, respectively; OR 0.7 (95% CI 0.5–1.0, P = 0.041) (Table 3).
Cancer treatment No studies were identified that directly compared Indigenous cancer patients’ uptake of treatment across remoteness categories, but there was some indirect evidence of disparities by place of residence. Three
8
A. DIAZ ET AL.
FIGURE 1: Schematic representation of the systematic review process.
Database search
181 papers idenfied
149 Titles and abstracts screened
60 full-paper arcles screened
13 papers idenfied (11 individual studies)
32 Duplicates Removed Excluded n=63 aer tle and n=26 aer abstract screening
47 papers excluded
Government Reports = 2 (2 individual studies)
15 papers (13 individual studies) included in the review
qualitative studies (five papers) suggested that the greater the distance a patient lives from treatment services, the more barriers they face to commence and complete their treatment (Table 4).23–27 Similar themes were reported, regardless of whether data were collected from cancer patients27 and their families24–26 or health care providers.23 Most barriers identified by these studies were related to travelling away from home and community into an unfamiliar city and navigation through intimidating and unfamiliar hospitals and systems. Importantly, these studies identified two facilitators of treatment adherence, namely, Indigenous health workers or Indigenous liaison officers and hospital Aboriginal health units.23–25 In a quantitative study of prevalence of hysterectomy among gynaecological cancer patients, there was little difference among non-Indigenous women in the proportion who underwent hysterectomy by remoteness category.28 Compared with non-Indigenous women from metropolitan areas, hysterectomy was more common for Indigenous women in all regions. While the difference was greater for Indigenous women from regional areas (Relative Risk; RR 1.7, 95% CI 1.3–2.4) than for those from metropolitan (RR 1.3, 95% CI 0.9–1.7) or remote areas (RR 1.2, 95% CI 0.9–1.5), these differentials were not significantly different across remoteness categories.28 In new analysis of QICS data, Indigenous cancer patients were less likely to receive cancer treatment if they were from rural/remote areas (72%) than from urban areas (80%) (OR 0.7, 95% CI 0.5–0.9). The difference remained even after taking into account cancer stage and socioeconomic status (adjusted OR 0.6, 95% CI 0.4–0.8). Among those who received treat-
ment, median time to treatment was similar for rural/ remote and urban patients (15 (95% CI 12–18) versus 13 (95% CI 7–19) days, P = 0.395); there was no significant difference even after adjusting for cancer stage and socioeconomic status (adjusted HR 1.0, 95% CI 0.8–1.1).
Cancer mortality Two studies were identified that directly compared cancer mortality for Indigenous people in remote areas with those in non-remote areas.29,30 One of these,30 along with an additional paper,31 compared the Indigenous : non-Indigenous differential in mortality rates across levels of remoteness (Table 5). The influence of remoteness on mortality may vary by cancer type. For NT Indigenous women in 1996–2006, age-standardised breast cancer mortality rates (per 100 000) in remote/very remote areas (29.9, 95% CI 16.6–43.3) were much lower than those in outer regional areas (60.0, 95% CI 0.0–141.8), although not statistically significant, but for cervical cancer mortality rates were higher in remote/very remote (20.6, 95% CI 8.3–32.9) than outer regional areas (2.5, 95% CI 0.0–7.4).29 An earlier study of cervical cancer mortality in Indigenous women in SA, WA and NT in 1986–1997 found that in comparison to Indigenous people in metropolitan areas, those living in remote areas are significantly more likely to die following a cervical cancer diagnosis (standardised mortality ratio (SMR) 2.1, 95% CI 1.6–2.8), but not those in regional areas (SMR 1.7, 95% CI 0.9–3.3)30 However, Indigenous women were more likely to die than non-Indigenous women in metropolitan areas © 2015 National Rural Health Alliance Inc.
9
URBAN-RURAL DISPARITIES IN INDIGENOUS CANCER OUTCOMES
TABLE 3: Demographic and clinical characteristics of Indigenous people diagnosed with cancer between 1998 and 2004 by place of residence at diagnosis, from the Queensland Indigenous Cancer Study (QICS)
Total Age 18–39 years 40–59 years 60+ years Sex male female Socioeconomic status (SEIFA) Most disadvantaged Disadvantaged Intermediate advantage Advantaged Most advantaged Stage at diagnosis† Localised cancer Regional spread Distant metastasis Regional and distant metastasis Charlson Comorbidity Index 0 1 2+ Diabetes Renal disease Cardiovascular disease Respiratory disease Any treatment given Treatment No treatment Cancer type Head and neck Colorectal Liver and biliary Lung Breast Cervix Uterus All other cancers
Rural/remote areas n (%)
Urban areas n (%)
Total n (%)
627 (66)
329 (34)
956 (100)
73 (12) 282 (45) 272 (43)
35 (11) 139 (42) 155 (47)
108 (11) 421 (44) 427 (45)
295 (47) 332 (53)
141 (43) 188 (57)
436 (46) 520 (54)
P-value
0.299
0.216
