Arthroscopy to treat osteoarthritis of the knee?

To the Editor: In their editorial, Buchbinder and Harris conclude that “The use of arthroscopy for knee osteoarthritis has been allowed to continue, exposing patients to an intervention that is at best ineffective, and at worst, harmful”.1

Each year, HCF funds more than 5000 knee arthroscopies in private hospitals alone, and the primary diagnosis is osteoarthritis (coded
as gonarthrosis [arthrosis of the knee]) in more than 1000 of these procedures.

As such, HCF will endeavour to contribute to the debate by surveying members who have a primary diagnosis of gonarthrosis to collect data on self-assessed benefits of knee arthroscopies. It would be fair to say that the patient’s view of the benefits of the procedure is a leading indicator and should form an integral part of assessing the success of knee arthroscopies for osteoarthritis.

Arthroscopy to treat osteoarthritis of the knee?

In reply: We thank Adams for providing private health insurance data that confirm the continued use
of arthroscopic surgery for patients with osteoarthritis.

We agree that the patient’s view of the benefits of the procedure is fundamental to assessing treatment success. We do not doubt that many patients are happy with the results of arthroscopic knee surgery, but this does not necessarily imply that the surgery has had any specific effect, as satisfaction rates are high after many ineffective placebo treatments. Indeed, high-quality randomised controlled trials have consistently failed to demonstrate clinically relevant self-assessed benefits of arthroscopy compared with sham surgery1 or non-surgical comparators.2–4 Potential risks of arthroscopy are also an important consideration. These include thrombosis, infection, complications of anaesthesia, and increased progression of osteoarthritis and likelihood of joint replacement.5

Satisfaction surveys do not justify the ongoing use of ineffective interventions. While some may find it acceptable to fund care based on perceived effectiveness, in this instance to do so might be doing more harm than good.

Can sleep contribute to “closing the gap” for Indigenous children?

Relatively simple interventions could make a significant difference

The wellbeing of Australian Indigenous children has long been an issue of concern and the subject of numerous national partnerships, action plans and government policies. This is primarily because of the high incidence of health problems and academic deficits among Indigenous children in comparison with non-Indigenous children.1 The aim of these government policies is to bring about a general increase in Indigenous children’s health and academic outcomes. We propose that poor sleep health may be a significant and, to date, poorly addressed factor that should be considered within the discourse around closing the gap in the health and wellbeing of Indigenous children and young people.

The body of literature on this issue provides very clear evidence that sleep problems in children (whether they have a physiological or non-physiological cause) have strong and causal associations with secondary deficits in academic performance, attention and learning, emotional regulation, behaviour and mood regulation, with increased likelihood of obesity, diabetes, high blood pressure, somatic health and psychological health.2 While there is a paucity of comparable data for Indigenous children, some studies are beginning to report similar findings. Recent findings on the sleep of Indigenous children suggest that this group may also be encumbered with a higher prevalence of sleep problems.3–7

Among physiological sleep disturbances, secondary sleep disturbance due to asthma has been reported in non-Indigenous children, but has yet to be fully explored in Indigenous children. This is despite the greater incidence of asthma among Indigenous children compared with non-Indigenous children.4 Sleep disordered breathing (ranging from primary snoring to obstructive sleep apnoea accompanied by nocturnal hypoxaemia) has known associations with daytime deficits in neuropsychological and psychosocial domains, and has also been found in one study to have a prevalence of 14.2% in Indigenous children.3 This study, one of the first to investigate sleep-disordered breathing in Indigenous children, found high prevalences of snoring, wheezing and restless sleep. Despite this, no further studies have been undertaken since 2004.4 Associations between all these conditions therefore remain to be explored in Indigenous children.

Not only must we consider the physiological aetiology of poor sleep, but also the impact it has on both the physiological and psychosocial development of Indigenous children. Recent findings suggest links between obesity and reduced sleep duration,2 and with the increasing and worrying prevalence of obesity among Indigenous children in Australia, their sleep profiles should be considered. In addition, there is a growing body of research showing associations between diabetes and sleep quality that have not been sufficiently explored in Indigenous children and young people.8

Some efforts to understand sleep in Indigenous children have been undertaken. In summary, data from various studies show that, compared with non-Indigenous children, Indigenous children report poorer sleep quality (eg, sleep scheduling, sleep fragmentation),5,7 decreased sleep duration,7 worse sleep hygiene,5 increased sleepiness,6 and more instability and irregularity in their sleep–wake patterns,5 particularly in “get up” times. Furthermore, these sleep problems were related to aggression,6 withdrawn behaviours,6 thought problems and internalised behaviours,6 reduced reading ability and numerical skills.7

What now?

Poor sleep, whether inferior in quality or quantity, is essentially modifiable. There are currently few data on which to base any assessment of how much poor sleep might contribute to poor health, wellbeing and academic performance in Indigenous children, but evidence in non-Indigenous children and young people suggests that not only is it significant, but also that it is amenable to treatment regardless of whether the sleep problem has a physiological cause.2 Treatment can have significant and positive outcomes. Considering that sleep is one of the key requirements of good health, it is only logical that it should be explored, investigated and improved, and that doing this might have positive impacts on these children’s lives. This may seem simplistic, but health-related lifestyle interventions have been shown to be successful in the past.9 Such interventions can be targeted at an individual or community level, and if they have a positive impact on even a single child, this would be an improvement on what is happening at present.

Clearly, there are considerable challenges to intervening to try to close the gap between Indigenous and non-Indigenous health, including socioeconomic and demographic factors, cultural differences, preferences about sleep and sleep hygiene and parenting, and Indigenous scepticism about “white fella” interventions. However, exploring whether sleep interventions would be an acceptable method to bridge our divides might be worth the effort. Certainly careful and sensitive negotiations have previously allowed researchers to engage and work with community elders to facilitate the first objective investigation of children’s sleep in a remote Indigenous community.7

Poor sleep is inherently modifiable. Therefore, any contribution sleep has to downstream factors (eg, health, wellbeing, academic performance, behaviour) is also potentially modifiable. For this reason, research funding and cross-institutional and multidisciplinary research efforts into understanding Indigenous sleep are necessary if we are serious about investigating not if but how much sleep is a contributor to Indigenous wellbeing so we can attempt, through sleep, to close the gap.

The quality of health research for young Indigenous Australians: systematic review

There are major incentives to invest in the health of Aboriginals and Torres Strait Islanders who are adolescent or young (used interchangeably throughout this article to refer to 10–24-year-olds).1 An estimated 31.7% of Indigenous Australians were aged 10–24 years in 2006, compared with 20.4% of the non-Indigenous population,2 and they experience an excess burden of preventable and treatable disease at a life stage when opportunities for education, employment, reproduction and independent living are at their peak.3 Births to Indigenous teenagers represent about one-fifth of all births to Australian Indigenous women;4 Indigenous young people have high rates of risk factors for the development of non-communicable diseases in adulthood (eg, obesity, tobacco-related disease);3 and the mortality gap between Indigenous and non-Indigenous Australians widens during adolescence and persists into adulthood.5–7 Indeed, improving young people’s health would appear to be critical to the success of the government’s National Indigenous Reform Agreement,8 particularly with regard to life expectancy, reading, writing and numeracy, secondary school completion and employment attainment.9

Indigenous young people’s health is a critical target for health system reform; however, the evidence base to inform health policy and the provision of programs to respond to specific needs remains poorly described.3,10 Measuring young people’s health is complicated by a lack of agreed indicators, and the problem is compounded by the publication of adolescent health data throughout the diverse paediatric and adult literature.11 Significant deficits of Indigenous health research have also been described.12–14 However, specific research policy reform by the National Health and Medical Research Council (NHMRC) has provided guidance for improved quality of Indigenous health research,15,16 and there was an increase in the quantity of Australian Indigenous health research during 1987–2003.17

The overarching aim of our project was to establish the health status of young Indigenous Australians and identify opportunities to improve their health outcomes. The starting point was to systematically document the existing good-quality literature and the limitations of this evidence base, which we describe here. A synthesis of health outcomes and effective interventions for young Indigenous Australians is forthcoming.

Methods

We systematically searched the literature from 1 Jan 1994 – 1 Jan 2011 for peer-reviewed studies reporting health data for Indigenous Australians aged 10–24 years, identifying both descriptive data and evaluations of health interventions.17 We included studies that exclusively reported data for Indigenous young people, as well as those reporting disaggregated data (ie, studies sampling Indigenous Australians and reporting data for young people, or studies sampling young people and reporting data for Indigenous Australians). Our review process is outlined in Box 1. Appendix 1a provides details of our search strategy and inclusion criteria.

Critical appraisal

We critically appraised the included studies, considering studies to be of good quality if they reliably ascertained Indigenous status, included samples that were representative of the target population and employed well defined measures of exposure and outcome (Appendix 1b). We identified whether descriptive studies reported determinants of health outcomes or health-risk exposures.

Demography: age and location

We defined three age subgroups (10–14 years, 15–19 years and 20–24 years) and described the source and location (eg, urban, rural) of the study sample as reported in the publication. Most studies did not report the geographical detail necessary for analysis by remoteness index8 as we had intended. Instead, we mapped the sample locations from each study manually on an Australian Bureau of Statistics map and compared it with the population distribution of Indigenous Australians,2,19 identifying locations using Google Maps Australia (Google Inc). We used the population distribution of all Indigenous Australians as our comparator, as the population distribution of young Indigenous Australians was not available.2,3

Study focus

We categorised the focus of each study as either health outcome or health-risk exposure. We further categorised health outcomes using the burden of disease framework as: Group Ia, Communicable diseases; Group Ib, Maternal, perinatal and nutritional conditions; Group II, Non-communicable diseases; and Group III, Injury6,20 (Appendix 2). Health-risk exposures were further categorised using an index of health risks which included the 11 main risk factors defined in the Australian Indigenous burden of disease report,6 risk factors as identified by the Global Burden of Disease Study expert group21 and risk factors identified for Australian Indigenous children (Appendix 2).22 We included teenage pregnancy as a health-risk exposure as it is a significant social-role transition with implications for individual and population health.23 While health outcomes could be categorised against a single burden of disease category, some studies of health risk measured multiple exposures (eg, alcohol and tobacco). We allowed multiple categorisation for studies measuring the 11 key risk factors. Studies measuring mortality rate (not cause-specific) and population-based studies measuring multiple indicators of health outcome and risk were considered separately.

Data handling and analyses

The review process was managed using an Excel (Microsoft Corporation) datasheet imported into Stata 10.0 (StataCorp) for analysis. Summary statistics were calculated using the χ2 test and linear and logistic regression. Hierarchical burden of disease groupings were reported using tree maps24 generated using the visualisation tool Many Eyes (IBM) and manually traced using Illustrator CS6 (Adobe). The level of statistical significance was set at P < 0.05.

Results

Our search strategy identified 3788 citations; 1509 full texts were reviewed, and we identified 360 peer-reviewed studies reporting data for young Indigenous Australians (Box 1). Ninety studies (25.0%) exclusively sampled Indigenous young people. Two-hundred and two studies (56.1%) were prospective in design. Studies were mostly descriptive (306 [85.0%]) and included 14 case series, 37 qualitative studies, 93 surveillance studies, 149 cross-sectional studies, 4 case–control studies and 9 cohort studies. Ninety-three of the 306 descriptive studies (30.4%) reported a determinant. Of the 54 evaluation studies, 34 (63.0%) were quantitative in design and there were no randomised trials. The number of studies published increased over time. The proportion focusing exclusively on Indigenous young people and the proportion with prospective study design increased significantly (P = 0.04 and P = 0.03, respectively) (Box 2). While the number of studies evaluating interventions also increased during 1994–2010, the proportion of all studies that were evaluative in design did not change (P = 0.56).

Critical appraisal

Two hundred and fifty studies (69.4%) were graded as good quality. Data quality improved significantly over time (P < 0.01) (Box 2). Quality was no different for descriptive compared with evaluative design (P = 0.42), or according to state or territory of the study sample (P = 0.46), urban compared with rural location (P = 0.37), study focus (P = 0.18) or for studies exclusively sampling Indigenous young people compared with those reporting disaggregated data (P = 0.36).

Demography: age and location

Most studies included young people aged 15–19 years (290 [80.6%]); 204 studies (56.7%) included 10–14-year-olds and 141 (39.2%) included 20–24-year-olds; 173 studies sampled two age subgroups and 51 sampled all three.

Overall, 61 studies (16.9%) reported data for young people from urban areas, 116 (32.2%) reported data for young people from rural areas, 139 (38.6%) presented mixed data and 44 (12.2%) studies were unclear. Forty-one of the 61 studies focusing on urban populations were published in 2005–2010 (P = 0.02). The community, town or city from which the young people were recruited was clearly identified in 144 studies (40.0%). Most studies included young people from the Northern Territory, Western Australia and Queensland (Box 3). Compared with studies including disaggregated data, studies that exclusively included Indigenous young people were more likely to have sampled urban populations (41% v 25%; P = 0.03); however, there was no difference by state or territory (P = 0.22). Indigenous young people were recruited or data were obtained from communities (113 [31.4%]), schools (58 [16.1%]), population datasets (50 [13.9%]), hospitals (48, 13.3%), clinics (47, 13.1%) and correctional facilities (27, 7.5%). Seventeen studies (4.7%) included data from mixed sources.

Study focus

Overall, 163 (45.3%) studies focused on health-risk exposures and 197 (54.7%) focused on outcomes: 65 of these (18.1%) on communicable diseases, 3 (0.8%) on maternal and nutritional conditions, 97 (26.9%) on non-communicable diseases, 16 (4.4%) on injury, and 16 (4.4%) on non-cause-specific mortality or population morbidity. The focus of research did not change during 1994–2010 (P = 0.6). Studies that exclusively included Indigenous young people did not focus on injury at all. Study focus did not differ by study quality (P = 0.59), and we restricted further analysis to good-quality studies.

Of the 250 good-quality studies, three reported multiple indicators of health for Indigenous young people25–27 and seven reported mortality (not cause-specific) for the NT, WA, South Australia and Queensland. The remaining 240 good-quality studies included 124 studies of health outcomes (Box 4) and 116 studies of exposure to health risks.

Health outcomes

The 124 good-quality studies of health outcomes comprised 109 descriptive studies and 15 that evaluated an intervention.

Group Ia: Communicable diseases: Thirty-seven descriptive and eight evaluative studies focused on communicable diseases. Descriptive studies mostly provided prevalence estimates from surveillance studies: 13 provided good-quality data for sexually transmitted infections and nine reported tuberculosis notifications. The eight evaluations focused on testing (4) and treatment (4) approaches.

Group Ib: Maternal, perinatal and nutritional conditions: There was a paucity of data on maternal diseases; one retrospective hospital-based study included some measures of obstetric risk among Indigenous teenage mothers and associated perinatal outcomes and was categorised under health-risk exposure.28

Group II: Non-communicable diseases: Fifty-seven descriptive studies and six evaluative studies focused on non-communicable diseases. Eighteen studies focused on mental disorders, nine of these on harmful substance use. Data on depression, anxiety and psychosis were limited. Fifteen descriptive studies focused on oral health; however, we did not identify any studies evaluating interventions. Four studies focused on cardiovascular disease (all on rheumatic heart disease), including a study on mortality.29 One study measured mortality related to cervical cancer.30

Group III: Injury: Thirteen descriptive studies focused on injury, and included three descriptive studies of suicide and one study of fatal unintentional injury.31–34 One study evaluated a community project to address both intentional and unintentional injury.35

Health-risk exposures

The 116 good-quality studies of health-risk exposures comprised 90 descriptive studies and 26 that evaluated an intervention. Eleven studies focused on nutrition and lifestyle, 11 on social–emotional wellbeing, eight on education and nine on justice. Twenty studies focused on sexual and reproductive health, including nine studies of adolescent pregnancy and 11 of related perinatal outcomes. Ten studies focused on health access and included four evaluations. A further eight studies focused on housing, community or parenting. Thirty-nine studies focused on the 11 key risk factors, with eight of these focused on more than one risk factor (Box 5). There were good data for substance use, and while data for unsafe sex were limited, two studies evaluated interventions.36,37

Discussion

Most health-outcome data for Indigenous young people published during 1994–2010 focused on communicable diseases, oral health and substance use. There were also some good-quality data for health-risk exposures related to adult non-communicable diseases (such as substance use, physical activity and diet) and for adolescent pregnancy, including some data on perinatal outcomes. These data illustrate how the health of today’s Indigenous young people will affect the health of Indigenous Australian adults and their children in the years to come. The quality of data improved over time, but there were still some important gaps. Data for urban locations were limited, as were data for mental disorders and injury. Data on the health of adolescent mothers were limited, despite fertility peaking in this age group.4 Mortality data were not available for all jurisdictions of Australia, and cause-specific estimates were limited. Overall, there was also a paucity of evaluation of programs and interventions. Identification of these gaps may assist in considering future research needs.

It is important not to interpret these findings as over-investment of research in some health areas. For example, sexually transmitted infection causes a disproportionate burden of poor health among Indigenous young people,38 and relevant good-quality data are likely to have informed the national strategy prioritising control of sexually transmitted infections among Indigenous Australians, particularly young people.39

We mapped the focus of studies using the burden of disease framework because using a predefined schedule of health outcome and risk allowed for objective assessment of any data gaps. It is not possible, however, to directly compare the scope of these publications to estimates of the burden of disease among Indigenous young people. Age-disaggregated data allowing calculation of disease burden for 10–24-year-olds were not available.6 Furthermore, estimates reported in the Australian Indigenous burden of disease study are dependent on quality input data, largely missing for Indigenous people living in urban settings13 and in areas such as mental health.40 With these limitations in mind, mental disorders and injury together account for 56% of the burden of disease among Indigenous people aged 15–34 years (calculated from annex tables).6 This proportion is similar for mental disorders and injury in young people aged 10–24 years globally.41 Burden of disease, however, is only one marker of research priority; priorities identified by Indigenous young people and their communities are a fundamental consideration. For example, a qualitative study of Indigenous adolescents in a remote area identified substance use, violence, boredom and racism as significant issues.25 Placing the findings of our study in this context, there appears to be a mismatch between estimates of the burden of disease, needs reported by young people, and the focus of research studies in Indigenous young people. The potential for implementing effective (and cost-effective) prevention and intervention is an additional consideration.42

The findings of our study are similar to a review of Canadian Indigenous health research that found research was primarily focused on health conditions amenable to “curative services”.43 One possible explanation is that competitive research funding and policy (such as Closing the Gap) encourages investment in health targets that are easily measured and amenable to simple intervention: this is a well recognised challenge of global health policy where there has been acceptable progress towards goals targeting individuals (ie, child survival), while goals such as gender equality and maternal health (shaped by socioeconomic and cultural determinants) have lagged.44,45 Despite social determinants playing a central role in Indigenous health inequality,8 we identified few studies that focused on social exclusion or socioeconomic inequality (justice, education, employment or housing). Less than a third of descriptive studies measured a determinant.

Three-quarters of all young Indigenous people live in urban and regional areas.3 Despite this, overall investment in research involving this population was limited, although it is improving. Challenges of conducting urban Indigenous health research are distinct from the cost and methodological challenges of rural research.46 There are no geographical boundaries or readily identifiable target populations that allow identification of Indigenous Australians living in urban settings.13 Appropriate consultation and collaboration (essential principles of Indigenous health research) in this context may involve a large number of communities and organisations and requires time, flexibility, and resourcing,47–50 which are increasingly at odds with research grants that focus on timely outputs. Developing a robust evidence base may also require non-standard approaches. While Indigenous health research is often conducted in collaboration with Aboriginal community-controlled health services or Aboriginal medical services,13 the Western Australian Aboriginal Child Health Survey showed that only 12% of Indigenous 12–17-year-olds had contact with an Aboriginal medical service in the preceding 6 months, with contact lowest among those living in the most urbanised settings.51 Data linkage may provide one mechanism to better understand patterns of health and social service access.

Health needs and opportunities for intervention vary significantly across the 10–24-years age band. We did not disaggregate our analysis by age subgroup because almost two-thirds of studies included multiple age subgroups. Additionally, we only included peer-reviewed literature; the grey literature may fill some of the identified gaps. The Western Australian Aboriginal Child Health Survey, for example, is published in four volumes and includes mental health data for 12–17-year-olds.52

The findings of our study serve two main purposes. First, we have identified some good-quality literature that can be used to inform health policy and programs to improve the health of young Indigenous Australians; a synthesis of this literature is forthcoming. Second, the findings provide a framework to allow Indigenous communities, researchers, funding bodies and the NHMRC to consider priorities for future research. Marriage of health research to need can only occur with consultation, engagement and the trust of Indigenous communities. This is essential for priority setting, dealing with sensitive health issues (such as mental health, injury and sexual and reproductive health) and appropriately and sustainably engaging young Indigenous Australians in understanding and addressing their health needs.53

1 Systematic review process

ERIC = Education Resources Information Center. CINAHL = Cumulative Index to Nursing and Allied Health Literature. ATSIhealth = Aboriginal and Torres Strait Islander Health Bibliography. * These were sorted by study design and include publications that did not sample Indigenous young people.

2 Frequency of publication and quality of data for young Indigenous Australians over time, 1994–2010

3 Sampling of young Indigenous Australians compared with the population distribution of all Indigenous Australians*

ACT = Australian Capital Territory. NSW = New South Wales. NT = Northern Territory.
QLD = Queensland. SA = South Australia. TAS = Tasmania. VIC = Victoria. WA = Western Australia.

* The graph shows the population distribution of Indigenous Australians, one dot representing 100 Indigenous Australians.2,19 Each red dot represents a single sampling site for a study (studies reporting sampling from five distinct sites are indicated by five individual dots; studies reporting sample location at a state or nation level are represented separately by a numeral).

4 Tree map plots* of 124 good-quality published studies on health outcomes among Australian Indigenous young people aged 10–24 years, colour-coded by Burden of Disease Framework category

M = musculoskeletal. Ca = malignancy. GIT = gastrointestinal. ME = meningoencephalitis. STI = sexually transmitted infection. TB = tuberculosis.

* Plots are drawn to scale to demonstrate the proportion of descriptive to evaluative studies.

5 Studies of health-risk exposure focusing on 11 key risk factors

Stent insertion for palliation of advanced oesophageal carcinoma symptoms by level of socioeconomic disadvantage in urban New South Wales

To the Editor: For patients with advanced oesophageal carcinoma, palliation of debilitating symptoms such as dysphagia and odynophagia is important for improving quality
of life.1 Owing to the possibility
of complications, it is generally recommended that stents be used as a palliative measure when expected survival is less than 3 months.2 We analysed linked records from the NSW Central Cancer Registry, the NSW Admitted Patient Data Collection, NSW Registry of Births, Deaths and Marriages death registrations data and Australian Bureau of Statistics mortality data to investigate the association between socioeconomic disadvantage and palliation of advanced (stage IV) oesophageal carcinoma symptoms by stent insertion in urban-dwelling patients in New South Wales, from July 2001 to December 2007. The study was approved by the NSW Population and Health Services Research Ethics Committee.

Of the 479 patients who were diagnosed with primary advanced oesophageal carcinoma before death and for whom linked hospital records were available, 28.4%
(136 patients) received stents. The proportion who received stents decreased with increasing disadvantage (P = 0.006 for association; P < 0.001 for trend). After adjustment for patient and tumour characteristics, greater disadvantage remained significantly associated with decreasing odds of stent insertion (Box) (P = 0.01 for association; P = 0.001 for trend). Dysphagia or oesophageal obstruction was reported for some patients and was associated with receiving a stent (χ2 test, P < 0.001) but not associated with level of socioeconomic disadvantage (χ2 test, P = 0.37). As this information is more likely to be reported for those who received a stent, it was not included in the multivariable analysis of association between stenting and socioeconomic disadvantage.

Increased socioeconomic disadvantage has been associated with reduced access to treatment for many cancers.3 Further, research on access to palliative care (generally and for advanced cancer in Australia) has identified barriers such as lack of standardised referral processes and lack of consensus about appropriate timing for access to palliative care.4,5 These factors, plus others that could not be measured reliably or at all from the data we used (such as indication for stenting, patient choice, and use of chemotherapy and radiotherapy), are likely to have contributed to the variation we observed.

For patients with expected survival of less than 3 months, stenting is recommended over alternative treatments such as brachytherapy.2 Because the proportion of patients with less than 3 months between diagnosis and death increased with increasing disadvantage (P = 0.004 for trend), immediate palliation of symptoms by stenting should be a priority for more disadvantaged patients. Later diagnosis for more disadvantaged patients may contribute to the observed variation in stenting, but is at odds with recommended treatment.2 Although patients who die soon after diagnosis may have reduced opportunity to receive a stent, excluding patients who survived
1–2 months after diagnosis made no difference to the findings.

Given the variation in use of stents by level of socioeconomic disadvantage that we observed and the possible role of other factors, further research is required to fully understand patient and health system factors that affect access to palliative care for patients with advanced oesophageal carcinoma. Understanding treatment pathways for more disadvantaged patients should be a priority because stent insertion can provide patients with immediate improvement to quality of life.

Association between stent insertion and socioeconomic disadvantage in 479 urban-dwelling patients with advanced oesophageal carcinoma, New South Wales, 2001–2007

Quintile of socioeconomic disadvantage*	No. (%) of patients	No. (%) of patients who died ≤ 3 months after diagnosis	No. (%) of patients who had stent inserted	Adjusted odds ratio† (95% CI)

Least disadvantaged	87 (18.2%)	26 (29.9%)	37 (42.5%)	2.00 (1.10–3.64)
Second least disadvantaged	102 (21.3%)	38 (37.3%)	32 (31.4%)	1.22 (0.67–2.23)
Middle	122 (25.5%)	56 (45.9%)	32 (26.2%)	1.00‡
Second most disadvantaged	96 (20.0%)	46 (47.9%)	22 (22.9%)	0.85 (0.45–1.60)
Most disadvantaged	72 (15.0%)	32 (44.4%)	13 (18.1%)	0.60 (0.29–1.26)
All	479 (100.0%)	198 (41.3%)	136 (28.4%)	—

* Quintiles of the Index of Relative Socio-Economic Disadvantage, based on the local government area of each patient’s residence at the time of diagnosis. † Model included: age group, sex, country of birth, year of diagnosis, diagnostic group (oesophageal squamous cell carcinoma, oesophageal adenocarcinoma, and other specified or unspecified oesophageal carcinoma), number of comorbidities, and time between diagnosis and death (≤ 3 months or > 3 months). The Hosmer–Lemeshow test indicated that the model was a reasonable fit (χ2 = 7.02, df = 8, P = 0.54). ‡ Reference category.

Human papillomavirus vaccine in boys: background rates of potential adverse events

Cervical cancer is the most common cancer affecting women in developing countries. It is caused by persistent infection with specific types of human papillomavirus (HPV).1 Quadrivalent human papillomavirus (4vHPV) vaccine is a recombinant vaccine administered as a three-dose course to provide protection against four types of HPV (6, 11, 16 and 18).2 The vaccine is highly efficacious for the four included types, of which 16 and 18 are reported to cause 70% of cervical cancers and 6 and 11 cause anogenital warts.1,3 4vHPV vaccination was introduced under the Australian National Immunisation Program (NIP) in April 2007 for adolescent girls, with an initial catch-up program including women up to 26 years of age. The current ongoing funded program is only for girls in the first year of high school (aged 12–13 years). Recent data suggest that the 4vHPV vaccination program has caused a rapid decline in genital wart presentations in females,4,5 and there are early indications of a reduction in high-grade cervical dysplasias.6

Following advice from the Australian Technical Advisory Group on Immunisation, vaccination of males was recommended as a cost-effective intervention by the Pharmaceutical Benefits Advisory Committee in November 2011.7 Accordingly, 4vHPV vaccination for boys has been added to the Australian NIP, commencing in 2013 and targeting boys aged 12–13 years in a school-based program, with a catch-up program over 2 years for boys aged 14–16 years.7,8 The program aims to reduce the incidence of HPV disease in males, such as anogenital warts and anal intraepithelial neoplasia,9 and reduce sexual pathways of virus transmission. Australia will be the first nation to implement HPV vaccination for boys in a national program.

Vaccines, as with any medicine, have potential adverse reactions varying from mild and expected to rare and/or serious events. Vaccination may cause such events — the nature of adverse events following immunisation (AEFI) and the timing of onset after vaccination are important factors when assessing causation. Adverse events may also coincide temporally with vaccine administration by chance. To interpret postlicensure surveillance data, it is useful to know the background rates of common and rare potential adverse events before introduction of the vaccine.10,11 With this understanding, increases above background rates can be rapidly identified, which can assist with the evaluation and reporting of potential vaccine-associated adverse event rates.

The mass school-based introduction of female 4vHPV vaccination raised a number of well publicised initial safety concerns, including “scares” regarding potential episodes of anaphylaxis and multiple sclerosis after vaccination.12–14 In addition, a mass psychogenic reaction was seen in a Melbourne school vaccination environment,15 with syncope and syncopal seizures occurring in response to the vaccination process.16 Such spurious events may arise from the psychological impact of the vaccination process, particularly when using mass vaccination strategies in a school-based teenaged population.

Release of the 4vHPV vaccine to boys has the advantage of adverse event information from prelicensure clinical trials and postlicensure surveillance of adverse events arising from administration to adolescent girls. However, additional information on the background rates of potential adverse events in teenaged boys is critical for assessing the safety of this vaccination program.

Our aim was to explore the use of routinely collected information for estimating potential adverse event rates. We used population-level health outcome administration data to describe the background rates of potential AEFI before the introduction of 4vHPV vaccination for boys into the NIP in Australia, and to estimate numbers of a range of neurological, allergic and other events that can be expected following vaccination, assuming temporal association with administration of vaccine but no other association.

Methods

Two statewide Victorian datasets were accessed — the Victorian Admitted Episodes Dataset (VAED; hospital discharge data) and the Victorian Emergency Minimum Dataset (VEMD; emergency department visit data) — both of which include International Classification of Diseases 10th revision Australian modification (ICD-10-AM) codes. The data included a unique identifier that enabled linking of individuals across the datasets, but were otherwise non-identifying, according to Victorian Department of Health data linkage protocols.17 Ethics approval for the study was provided by the VAED and VEMD data custodians.

Multiple records of the same event within a dataset or across datasets — for example, a person presenting at emergency who is subsequently admitted, or a person admitted to hospital who is then discharged to a different hospital or to home and who later returns with continuation of the same episode (with each presentation recorded as a separate event) — were linked via the unique identifier. All events occurring within 28 days of a previous event were combined into a single episode.

The data that we analysed comprised all episodes that occurred in boys aged 12 to < 16 years and were recorded in the VAED and/or VEMD with one of the ICD-10-AM codes listed in Box 1 and an admission or presentation date from 1 July 2004 to 30 June 2009.18 Conditions selected for inclusion are rare adverse events, conditions that patients are likely to present to hospitals with after vaccination, and conditions that have previously been raised as potential sources of concern in Australia and overseas.10,19

Age was taken to be the youngest age at which an episode occurred, and records were excluded from the analysis if sex was recorded inconsistently among records with the same unique identifier. Some records had more than one ICD-10-AM code, and these were preserved. Events with an interstate or overseas postcode were excluded, but those with “unknown” (8888 and 9988) and “of no fixed abode” (1000) postcodes were preserved under the assumption that these occurred in Victoria. Episodes that were ongoing from the 3 months before the study period, the washout period (31 March to 30 June 2004), were also excluded.

Events were described as the number of episodes and the number of first events. An episode was considered a discrete event if it occurred more than 28 days after a prior event in the same individual, as patients were deemed to still be “at risk” of the same event during their recovery from an acute condition. First events were defined as the first time a condition was diagnosed in each patient during the study period. First events are more relevant for chronic conditions and episodes are more relevant for acute conditions.

We calculated background annual incidence rates as the number of events during the 5-year study period divided by the population at risk during this period, using Australian Bureau of Statistics 2006 mid-year resident population data for males.20

The analysis was restricted to boys aged 12 to < 16 years — the target age range for vaccination. We used these background rates to estimate the number of events expected within 1 day, 1 week and 6 weeks of vaccination per 100 000 vaccinees. We then estimated the expected number of events for each condition 1 day, 1 week and 6 weeks after vaccination across Australia following the introduction of 4vHVP into the NIP, assuming there is no association (other than temporal) with the vaccine.

Seasonal variation was analysed by graphing the number of first events or episodes by month of presentation. As the numbers of chronic neurological presentations in the study group were small, they were combined and compared with numbers of all-age presentations in males for individual neurological conditions. For multiple sclerosis, data were also presented omitting presentations in the first 12 months of the study period to assess the effectiveness of the study’s 3-month washout period.

Results

The numbers of and incidence rates for potential AEFI in boys aged 12 to < 16 years are shown in Box 2, and the estimated numbers of cases of potential AEFI per 100 000 adolescent boys that would occur, even in the absence of vaccine, are shown in Box 3. Assuming an 80% vaccination rate with three doses per person — which equates to about 480 000 boys vaccinated and a total of 1 440 000 doses administered nationally per year in the first 2 years of the program — about 2.4 episodes of Guillain-Barré syndrome would be expected to occur within 6 weeks of vaccination. In addition, about 3.9 seizures and 6.5 acute allergy presentations would be expected to occur within 1 day of vaccination, including 0.3 episodes of anaphylaxis.

There was minimal seasonal variation in the occurrence of potential AEFI (Box 4, Box 5). However, repeating this analysis with a larger number of neurological presentations (using data for all age groups) revealed a notable peak in the number of multiple sclerosis presentations in July. This peak was reduced but not eliminated when the washout period was increased to 15 months (Box 4).

Discussion

Using statewide morbidity data, we estimated background rates of neurological and allergic events in adolescent boys in Victoria to be 252.9 and 175.2 per 100 000 person-years, respectively. Such adverse events may be mistakenly assumed to be caused by vaccination, owing to temporal association, when the 4vHPV vaccination program is expanded to include adolescent boys.10 Postlicensure safety assessments of 4vHPV vaccine programs in adolescent girls have shown little evidence of increased risk of neurological and allergic adverse events after vaccination.3,21,22

Expected rates of potential AEFI in recent studies vary widely, but direct comparisons are restricted because of differences in methods, health care systems and data collection and analyses.10,11,23 In particular, caution is required when using emergency presentation databases as these may record preliminary diagnoses, rather than final diagnoses. Studies limited to analysis of ICD-10 coded data, such as ours, lack the rigour of diagnosis verification and conformity to standardised case definitions, although coding standards are maintained. Our study identified higher reporting rates for anaphylaxis compared with similar studies.10,11 While data aberrations are possible, marked increases in anaphylaxis rates Australia and the United States over the past two decades may play a part.24,25

Background rates of potential AEFI and consequent thresholds for safety flags should not be informed merely using data on adolescent girls because sex-related differences could cause misinterpretation of potential signals.10,11 For example, the rate of adolescent boys presenting with a first multiple sclerosis event in the 6 weeks following vaccination would be expected to be one-third of the rate seen for adolescent girls assuming no relationship with vaccine other than temporal.26

In our study, we used a 3-month washout period to attempt to remove the risk of categorising events as incident cases when they were part of a pre-existing illness than was ongoing from before the study period. However, the 3-month washout period did not remove this issue for multiple sclerosis. While our study showed little seasonal variation in potential AEFI, school-based vaccination programs are conducted in blocks (as convenient to the vaccine schedule and the school year), which may give rise to false signal detection. Specific investigation of appropriate washout periods, as well as seasonal variation in the occurrence potential AEFI and implementation of the vaccine program, must therefore be explored before conducting in-depth analyses for specific conditions or extrapolating data to other jurisdictions.

In Victoria, first-dose 4vHPV vaccine coverage for adolescent girls has reached 80%,27 but challenges of uptake and course completion by males may be anticipated.28 If coverage for boys is less than 80%, the expected rates in our study should be recalculated to avoid erroneous alert thresholds.

The background rates of potential AEFI that we have estimated can be used to inform surveillance systems, health care providers and the community regarding health care events that may be temporally related to vaccination. In mass vaccination programs, where vaccine exposure is a common event in the target group, many incident acute health conditions will occur following vaccination, irrespective of causal association. While current passive surveillance system reporting is likely to underascertain postvaccination events, prior knowledge of expected numbers of events are valuable in helping determine whether reports or clusters of reports represent real safety flags that require urgent investigation.26

Our data highlight the value of statewide and nationwide health datasets in providing information that can improve public safety. In addition to establishing background rates of diseases, international systems such as those in Denmark and the US, have been used to link vaccination databases to health care event databases, enabling direct investigation of potential associations with adverse events.29–31 These methods, conducted in accordance with state and federal privacy protections, offer a promising future for further improving vaccine safety in Australia.32

Routinely collected state health outcome data can enable informed postlicensure safety surveillance of conditions that may be perceived as AEFI. When the 4vHPV vaccine program is expanded to adolescent boys, such data can be used for targeted active surveillance of potential vaccine safety flags.

1 Conditions included in the study

ICD-10-AM codes

Neurological

Guillain-Barré syndrome*

G61.0

Transverse myelitis*

G37.3

Multiple sclerosis*

G35

Optic neuritis*

H46, G36.0

ADEM

G04.0

Bell’s palsy

G51.0

Syncope

R55

Seizures

R56, R56.0, R56.8

Allergic

Anaphylaxis

T78.2, T88.6

Urticaria

L50.0, L50.1, L50.9

Serum sickness

T80.6

Adverse effect of drug or medication†

T88.7

Other

Adverse events

T78.8, T78.9, T88.1, T78.3

ICD-10-AM = International Classification of Diseases 10th revision Australian modification. ADEM = acute disseminated encephalomyelitis. * Conditions considered chronic. † Not otherwise specified.

2 Numbers of and incidence rates for potential AEFI in boys aged 12 to < 16 years (Victoria, July
2004 – June 2009)

First events

Episodes

No. of events

Incidence rate (95% CI) per 100 000 person-years

No. of events

Incidence rate (95% CI) per 100 000 person-years

Neurological

Guillain-Barré syndrome

1.46 (0.56 to 2.37)

1.61 (0.66 to 2.56)

Transverse myelitis

0.29 (− 0.11 to 0.70)

0.44 (− 0.06 to 0.94)

Multiple sclerosis

0.29 (− 0.11 to 0.70)

Optic neuritis

0.59 (0.01 to 1.16)

0.88 (0.18 to 1.58)

ADEM

0.17 (0.45 to 1.90)

1.61 (0.66 to 2.56)

Bell’s palsy

8.78 (6.56 to 11.00)

Syncope

807

118.0 (109.9 to 126.2)

831

121.5 (113.3 to 129.8)

Seizures

666

97.4 (90.0 to 104.8)

830

121.4 (113.1 to 129.7)

Total

1516

221.7 (210.6 to 232.9)

1729

252.9 (241.0 to 264.8)

Allergic

Anaphylaxis

7.17 (5.16 to 9.17)

7.46 (5.41 to 9.51)

Urticaria

620

90.7 (83.6 to 97.8)

647

94.6 (87.3 to 101.9)

Serum sickness

3.4 (2.0 to 4.7)

Allergic reaction

495

72.4 (66.0 to 78.8)

517

75.6 (69.1 to 82.1)

Total

1125

164.6 (154.9 to 174.2)

1198

175.2 (165.3 to 185.1)

Other

Total

1.02 (0.27 to 1.78)

AEFI = adverse events following immunisation. ADEM = acute disseminated encephalomyelitis.

3 Estimated numbers of cases of potential AEFI in vaccinated boys aged 12 to < 16 years, assuming no relationship with vaccine*

No. of first events per 100 000 population

No. of episodes per 100 000 population

1 day

1 week

6 weeks

1 day

1 week

6 weeks

Neurological

Guillain-Barré syndrome

0 (0.00–0.01)

0.03 (0.01–0.05)

0.17 (0.06–0.27)

0 (0.00–0.01)

0.03 (0.01–0.05)

0.19 (0.08–0.29)

Transverse myelitis

0 (0.00–0.00)

0.01 (0.00–0.01)

0.03 (0.00–0.08)

0 (0.00–0.00)

0.01 (0.00–0.02)

0.05 (0.00–0.11)

Multiple sclerosis

0 (0.00–0.00)

0.01 (0.00–0.01)

0.03 (0.00–0.08)

0 (0.00–0.00)

0.01 (0.00–0.01)

0.03 (0.00–0.08)

Optic neuritis

0 (0.00–0.00)

0.01 (0.00–0.02)

0.07 (0.00–0.13)

0 (0.00–0.00)

0.02 (0.00–0.03)

0.10 (0.02–0.18)

ADEM

0 (0.00–0.01)

0.02 (0.01–0.04)

0.15 (0.05–0.25)

0 (0.00–0.01)

0.03 (0.01–0.05)

0.19 (0.08–0.29)

Bell’s palsy

0.02 (0.02–0.03)

0.17 (0.13–0.21)

1.01 (0.75–1.26)

0.02 (0.02–0.03)

0.17 (0.13–0.21)

1.01 (0.75–1.26)

Syncope

0.32 (0.30–0.35)

2.26 (2.11–2.42)

13.57 (12.64–14.51)

0.33 (0.31–0.36)

2.33 (2.17–2.49)

13.98 (13.03–14.93)

Seizures

0.27 (0.25–0.29)

1.87 (1.73–2.01)

11.20 (10.35–12.05)

0.33 (0.31–0.35)

2.33 (2.17–2.48)

13.96 (13.01–14.91)

Total

0.61 (0.58–0.64)

4.25 (4.04–4.46)

25.50 (24.22–26.78)

0.69 (0.66–0.72)

4.85 (4.62–5.07)

29.08 (27.71–30.45)

Allergic

Anaphylaxis

0.02 (0.01–0.03)

0.14 (0.10–0.18)

0.82 (0.59–1.05)

0.02 (0.01–0.03)

0.14 (0.10–0.18)

0.86 (0.62–1.09)

Urticaria

0.25 (0.23–0.27)

1.74 (1.60–1.87)

10.43 (9.61–11.25)

0.26 (0.24–0.28)

1.81 (1.67–1.95)

10.88 (10.04–11.72)

Serum sickness

0.01 (0.01–0.01)

0.06 (0.04–0.09)

0.39 (0.23–0.54)

0.01 (0.01–0.01)

0.06 (0.04–0.09)

0.39 (0.23–0.54)

Allergic reaction

0.20 (0.18–0.22)

1.39 (1.27–1.51)

8.33 (7.59–9.06)

0.21 (0.19–0.22)

1.45 (1.32–1.57)

8.70 (7.95–9.44)

Total

0.45 (0.42–0.48)

3.15 (2.97–3.34)

18.92 (17.82–20.03)

0.48 (0.45–0.51)

3.36 (3.17–3.55)

20.15 (19.01–21.29)

Other

Total

0 (0.00–0.00)

0.02 (0.01–0.03)

0.12 (0.03–0.20)

0 (0.00–0.00)

0.02 (0.01–0.03)

0.12 (0.03–0.20)

AEFI = adverse events following immunisation. ADEM = acute disseminated encephalomyelitis. * Data are based on one dose of vaccine per vaccinee.

4 Numbers of first events of chronic conditions, by month (Victoria, July 2004 – June 2009)

* Data are numbers of first events for chronic neurological conditions analysed in boys aged 12 to
< 16 years and numbers of presentations for individual neurological conditions in males of all ages.
† Conditions included were Guillain-Barré syndrome, transverse myelitis, multiple sclerosis and
optic neuritis.

5 Number of episodes of acute conditions by month in boys aged 12 to < 16 years (Victoria, July 2004 – June 2009)

ADEM = acute disseminated encephalomyelitis.

Where is the next generation of medical educators?

To the Editor: The arguments put forward by Hu and colleagues for recognition of medical education as a specialty are persuasive, especially considering current national requirements for accreditation to ensure delivery of high-quality programs.1

In recent decades, medical educationalists have flooded journals and other publications with articles (Box). Many have resulted from, and led to, innovative educational programs. Unlike other areas of scientific research, however, the true impact of these educational programs may only be appreciated after 10 years.2 Such research needs to be done to demonstrate the effectiveness of educational interventions.

Medical educators need to prove to the community that they have clearly improved the quality of both teaching and learning, and their graduate doctors. Very few education providers assess the quality of the clinical care their graduates provide, instead using evaluation of the course as a surrogate marker, which can be modified by successful statistical manipulation of data.3 This evaluation needs to be done by external organisations such as colleges of postgraduate training, and by clinical supervisors in prevocational and vocational training. It is relevant to note that of the four Australian-based authors of the editorial,1 two come from medical schools whose graduates rate the overall teaching quality as worse than average.4

After 25 years, I know which medical school graduates I will happily employ, because they are clinically competent, reliable, keen to learn and show compassion to patients and colleagues. Focused discussions with these graduates consistently show that it is skilled clinicians and excellent mentors, not specialist medical educators, who have honed their abilities.

Number of publications on medical education over time*

* The figures were derived in February 2013 from a search of OneSearch (http://libguides.library.uwa.edu.au/onesearch), a research repository of journal articles, book chapters, newspaper articles, conference papers and broad cross-disciplinary databases.

Where is the next generation of medical educators?

In reply: We thank Hart and Pearce for supporting the views raised in our editorial, noting the unmet demand for medical education expertise.

We also thank Kandiah for his response, and agree that medical graduates should be “clinically competent, reliable, keen to learn and show compassion to patients and colleagues”. We believe this outcome is best achieved by strong collaborations among “skilled clinicians and excellent mentors” and medical educators, many of whom are also practising clinicians. Clinicians provide critical input to ensure the validity and authenticity of what is taught and assessed, and are an essential element of the “triad” of patient, student and clinician in clinical learning.1 Collaboration between clinicians and medical educators is not difficult because they are often embodied within the same people.

The question of proof in medical education is the subject of much activity and, as Kandiah notes, there is an increasing output of scholarship in medical education. Moreover, the quality and rigour of this output is increasing, with a growing evidence base for medical educational practice.2 Generating new knowledge and applying it to medical student education is a key goal of an increasingly professionalised medical education community.

Medical education research is confounded by multiple factors, not the least being the powerful and uncontrolled effects of the diverse clinical environments in which students learn and practise as graduates.3 These make causal pathways difficult to unpick. While researching the effect of medical educators may be desirable, we believe that researching the effects of medical education interventions is more fruitful. For example, if one were to substitute medical educators with radiologists, how could one “prove” that radiologists have improved the health of the Australian population? Yet we are convinced that radiology does play an important role, based on multiple individual studies showing contributory evidence for this claim.

We welcome opportunities to work with health services and the community to examine the long-term performance of our students and their impact on the health system. Collaboratively defining and answering specific questions is likely to be much more productive than making artificial distinctions between clinicians and educators.

Supply and demand mismatch for flexible (part-time) surgical training in Australasia

Surgical training follows an apprenticeship model, traditionally involving long hours of full-time mentored practice over several years. Recently, this model has been challenged by several trends, including working-hour restrictions, falling case exposure, and a desire for work–life balance.1–3 Another challenge is an increasing demand for flexible (part-time) training.4,5

The Royal Australasian College of Surgeons (RACS) supports flexible training by allowing trainees to accredit part-time work, but it mandates a time commitment of at least 50% during any training year.6 However, the opportunity to train part-time in surgery is also influenced by hospital employers and supervisors, and the supply of part-time surgical training posts is limited.

There are currently no data from our region regarding the number of surgical trainees undertaking flexible training. Our primary aim was therefore to define current flexible surgical training uptake and demand in Australia and New Zealand. A secondary aim was to identify demographic and work-related factors motivating interest in flexible training.

Methods

All 1191 trainees enrolled in an RACS program during 2010 were identified through the College’s database and invited by email to complete an anonymous online questionnaire, with weekly reminders over 4 weeks. Approval was granted by the RACS Ethics Committee.

The survey comprised four sections with option buttons. The first section defined demographic characteristics, including age, sex, specialty and Surgical Education and Training (SET) program year. Demographic data were also analysed to determine whether the responding sample was representative of all RACS trainees. The second section defined current working hours, and the third section rated work-related fatigue. Trainees were asked if they worked part-time, full-time or were not currently in active clinical training (ie, interrupted or deferred, such as for research or parenting). Results concerning trainee working hours and impacts of fatigue have been reported elsewhere.3,7 The fourth section used Likert scales to ascertain respondents’ perceptions of their work–life balance and interest in flexible training. This section included the question, “Are you interested in the option of applying for flexible (less than full time) training during your surgical training?”. Responses were assessed for differences between sexes and specialties. Further analyses were undertaken to identify whether interest in flexible training was correlated with working hours or fatigue.

Analyses were performed using SPSS version 19.0 (IBM), using cross tabulations with the χ2 test (threshold P < 0.05).

Results

Of the 1191 trainees, 659 responded (response rate, 55.3%), and 587 respondents (89.1%) completed all relevant questions. Respondents were similar to all trainees in terms of specialty (P = 0.22) and sex (P = 0.09). Of the 659 respondents, 187 (28.4%) were female, with the proportion of women differing between specialties (P = 0.02), ranging from 2/19 in cardiothoracic surgery to 9/16 in paediatric surgery (Box 1). The median age of respondents was 32 years (range, 24–50 years).

Most of the 659 respondents (627, 95.1%) were engaged in full-time clinical training, with 30 (4.6%) not in active clinical training, and only two (0.3%) in a part-time clinical training position. Both respondents who were working part-time reported working 40 hours per week, compared with a median of 60 hours per week for those in full-time clinical work. The small number of part-time trainees precluded further comparisons with full-time trainees.

An interest in flexible training was reported by 208 respondents (31.6%), being more common among women than men (54.3% [94/173] v 25.9% [114/441]; P < 0.001) (Box 2, A). There was no statistically significant difference in interest in flexible training by state or country (P = 0.82) or by hospital setting (rural, regional, tertiary) (P = 0.07) (data not shown). There was also no difference in interest in flexible training by age (P = 0.21), but junior trainees were more likely to express an interest than senior trainees (45.8% [60/131] and 38.4% [56/146] for SET years 1 and 2, respectively, v 24.8% [26/105] and 26.4% (23/87) for SET years 4 and 5–6, respectively; P = 0.002). Trainees interested in part-time training were more likely to express concerns regarding fatigue impairing their work performance and limiting their social or family life, inadequate work–life balance, and insufficient time for things outside surgical training, including study or research (Box 3 and Appendix).

General and orthopaedic surgery trainees were most likely to report an interest in flexible training (41.6% [116/279] and 32.5% [37/114], respectively), while cardiothoracic and vascular surgery trainees were least likely (6.7% [1/15] and 8.0% [2/25], respectively). Female general surgery trainees were more likely to be interested (65.2% [58/89]) than female trainees in other specialties (P = 0.04).

There was no significant difference in work–life balance satisfaction between male and female surgical trainees with respect to working ≥ 60 or < 60 hours per week (P = 0.48). About two-thirds of trainees reported currently working “too many” or “far too many” hours in terms of their preferred work–life balance (men, 62.0% [259/418] and women, 65.1% [110/169]). Trainees’ opinions on whether they had satisfactory time in their lives for things outside of surgical training are shown in Box 2, B.

Discussion

This study demonstrates a striking mismatch between interest in flexible training among Australasian surgical trainees and the number of trainees currently in a part-time post. Although 32% of trainees were interested in flexible training, less than 1% were engaged in part-time clinical training.

These results show a previously undocumented high level of interest in flexible training among both male and female trainees. The rate of interest among men was higher than the 13% rate reported among male general surgical residents and medical students in the United States.4 The leading factor known to motivate interest in flexible training is time for parenting,4,8 and the mean age of trainees in our study (32 years) coincides with the age when Australians would typically choose to become parents.9 Impact on family life has been found to be among the biggest regrets of US surgeons regarding their time in residency,4 and the American Surgical Association has previously called for increased flexibility in training to facilitate parenting.5 Our results show that similar efforts are required in Australasia. Limited opportunities for flexible training may discourage graduates from considering a surgical career.10–12

Our study also found that work-related factors are associated with interest in flexible training. On average, Australasian surgical trainees work more than 60 hours per week, and around 75% also perform on-call duties for a further 28 hours per week.7 A previous study of Younger Fellows of the RACS showed that those working more than 60 hours per week were at higher risk of “burnout”,13 which could perhaps be mitigated by increased work flexibility. Factors other than family and fatigue, such as health and academic interests, are also likely to contribute to interest in flexible training.4,8,14

Barriers to flexible surgical training that may explain the low uptake in our study include clinical, supervisory, trainee and employment factors.12,14,15 From a clinical and supervisory perspective, the potential impact of part-time training on continuity of care, and the associated need for additional handovers, is a concern.16 As part-time training occurs at a lower intensity and over a longer period, it may also be an impediment to gaining technical skills.15 Few educational data are currently available to assess this, but limited Australasian experience suggests quality outcomes can be achieved within the right model.14 Factors deterring trainees from flexible training include prolongation of training, a trade-off in salary and benefits, complexities in negotiating a part-time hospital contract, a perception of receiving “second-class training”, and discouragement from supervisors and other trainees.4,17 In addition, the limited availability of part-time hospital appointments is a key barrier.14 It may be difficult in some rotations to provide trainees with the necessary range of clinical experience (spanning acute, elective, operative and non-operative experience) in a flexible capacity.

Despite these challenges, opportunities exist to enhance the supply of flexible surgical training posts. Two possible models are job-sharing and stand-alone posts, both of which have successful precedents in Australia.12,14 Job-sharing can be facilitated by allowing trainees to “match” into suitable posts, but this typically must be planned months in advance and may be difficult in smaller specialties or regions. A possible path to advancing flexible training through a stand-alone model is through private hospital rotations, which are a focus of increasing interest in Australasia.18 Private sector training imposes opportunity costs on surgeon and hospital income,19 which could be partly offset by these positions being part-time, if quality operating exposure could be assured. Job-sharing in acute surgical units offers an opportunity to mitigate impacts on continuity of care, as scheduled handovers occur continuously during the 24-hour acute care model.20,21

Female students now outnumber male students in Australasian medical schools,22 and our study suggests the proportion of female trainees in surgery is also growing. Although only around 8% of qualified Australasian surgeons are women,23 we found that 28% of RACS trainees are women, indicating that a workforce transition is occurring. As we also found that more women than men are interested in flexible training, these demographic trends are likely to increase pressure for part-time training opportunities.

A limitation of our study is the sourcing of data from self-survey responses; however, the response rate was satisfactory.24 Further, a reported interest in flexible training may not translate into uptake of flexible training, even if the opportunities are available.

In conclusion, we believe efforts should be made to facilitate part-time surgical training in our region.

1 Sex of respondents by surgical specialty

2 Comparisons of male and female trainee responses to questions about interest in flexible training (A) and time outside training for other things (B)

* P < 0.001 for male v female. † P = 0.06 for male v female.

3 Factors associated with an increased interest in flexible training

Factor

Fatigue is impairing concentration or performance at work

0.009

Fatigue is limiting participation in social or family life

< 0.001

Current working hours are in excess of preferred work–life balance

< 0.001

Perceived insufficient time in life for things outside of surgical training

< 0.001

Perceived insufficient time for surgical study and research needs

0.003

* Each question was assessed on a five-point Likert scale. P values reflect χ2 analyses of Likert response and interest in flexible training. The full datasets for these analyses are provided as an Appendix (online at mja.com.au).

Socioeconomic area disparities in tobacco retail outlet density: a Western Australian analysis

While Australia has been applauded internationally for its lead on plain packaging for cigarettes1 and, previously, for being at the forefront of tough tobacco advertising restrictions, tobacco remains as readily available as dietary staples like bread and milk. There are an estimated 35 000 tobacco retail outlets in Australia,2 and two states (Queensland and Victoria) do not even require a licence to sell tobacco. The pervasive availability of tobacco products is at stark odds with the harm caused by tobacco2 and with the progress that has been made in most other areas of tobacco control.

In public health more broadly, there is growing research and policy interest in the relative availability of unhealthy products (ie, tobacco, alcohol, fast food) in more socio-economically disadvantaged areas. A number of United States studies have reported higher densities of tobacco outlets in neighbourhoods with lower socioeconomic status (SES),3–5 or in areas with lower household incomes and a greater proportion of residents from minority groups.6–8 By contrast, the only published Australian study on this subject to date found no relationship between SES and tobacco outlet density in the Hunter region of New South Wales (comprising two regional cities and rural towns).9 However, this study also suggested a relationship between perceived tobacco availability and consumption, with 85.7% of smokers reporting that they were within walking distance of a tobacco outlet during the course of day-to-day activities. Its authors noted that reducing the availability of tobacco stands benefited smokers who wished to quit,9 a point that has been made in other articles calling for regulation of the retail environment.10

Methods

We used an ecological cross-sectional design to investigate the relationship between local area SES and the density of retail outlets selling tobacco in Western Australia in 2011. Local areas were defined by Australian Bureau of Statistics (ABS) suburb boundaries,11 with the “suburb” equating spatially to a single town outside of metropolitan Perth and larger regional centres. Using ABS classifications, suburbs were classified as metropolitan if they fell within the Perth Statistical District, and non-metropolitan if they fell outside this boundary.11 Analysis was undertaken at several levels, including whole of state, metropolitan Perth (where 71.6% of the state’s population resides), regional WA and five larger regional centres.11

Measures

Area-level socioeconomic status: this was determined using the 2006 ABS Index of Relative Socioeconomic Advantage and Disadvantage (IRSAD), where lower values indicate more disadvantage.12 The SES of suburbs and towns were classified into quartiles, using the IRSAD percentiles.

Suburb or town population: this was sourced from the 2006 census and was used as a denominator in the computation of a per capita tobacco outlet density rate, to account for the likelihood of suburbs or towns with larger populations having more tobacco outlets.

Tobacco outlet data: these were sourced from the WA Department of Health (DoH) by Cancer Council WA in May 2011. The tobacco retailer data were geocoded by the DoH, which allowed the number of outlets per suburb or town to be identified using a geographic information system (ArcGIS 10.0, Esri). These counts were then joined back to the IRSAD and population data for analysis.

Statistical analysis

The “tobacco outlet rate” was calculated as the number of tobacco outlets per 10 000 residents. We investigated the association between the per capita tobacco outlet rate and suburb SES (IRSAD) using a negative-binomial model with offset to account for the usual residential population. The negative-binomial model was used to calculate rate ratios, representing the ratio of the number of tobacco outlets per capita in the comparison group to the number per capita in the reference category. Rate ratios and their 95% confidence intervals were calculated for each IRSAD quartile. Suburbs in the highest quartile of IRSAD were used as the reference category.

Ethics approval

Approval for this study was obtained from the Human Research Ethics Committees of the WA DoH and the University of Western Australia.

Results

There were 911 suburbs and towns in WA overall. The number of tobacco outlets per capita for suburbs and towns with a low IRSAD was more than twice as high as that for suburbs or towns with a very high IRSAD (Box). Suburbs and towns with a very low IRSAD had more than four times the number of tobacco outlets per capita than those with a very high IRSAD (P < 0.001).

For the 296 metropolitan suburbs, the number of tobacco outlets per capita for suburbs with a very low IRSAD was almost 50% higher than the number per capita for those with a very high IRSAD (P < 0.001).

The strongest associations between SES and tobacco outlet density were observed for the 608 suburbs and towns outside Perth (ie, regional WA). The number of tobacco outlets per capita for suburbs and towns with a very low IRSAD was more than five times higher than for those with a very high IRSAD (P < 0.001). The effect size decreased markedly with increasing IRSAD quartiles.

The five major regional centres in which further analysis was undertaken were Albany (19 suburbs or towns), Bunbury (16), Busselton (6), Geraldton (17), and Kalgoorlie–Boulder (12). In all of these regional centres except Bunbury, tobacco outlet density was inversely associated with SES.

Discussion

This is the first Australian study to confirm an inverse relationship between tobacco outlet density and area SES. The excess of tobacco outlets in lower SES areas is of public health concern for a number of reasons.

First, there are already marked socioeconomic disparities in Australia in smoking prevalence rates, barriers to smoking cessation and cessation success,13 and tobacco use can contribute to the financial hardship experienced by smokers in disadvantaged circumstances.13 It has been argued that environments that support easy access to tobacco products can undermine people’s intentions to quit or cut down tobacco consumption.14

Second, the concentration of tobacco outlets in more disadvantaged neighbourhoods may accentuate the scope for addiction, making it harder for people to quit or not relapse.

Third, overall vulnerability to poor health is exacerbated if tobacco outlets are more concentrated in areas where people at higher risk of negative health outcomes live.15 This is the case in Australia, where behavioural risk factors such as alcohol consumption, smoking and poor nutrition are more likely to cluster among populations with lower SES, and where lower SES groups are overrepresented in preventable mortality and morbidity.16

Fourth, from an economic perspective, a higher density of tobacco retailers creates a competitive market that may stimulate price discounting, in turn influencing consumption levels as smokers with lower incomes may be particularly sensitive to the price of tobacco products. This is similar to the economic rationale for restricting alcohol outlet density.17 The price elasticity of demand for tobacco products is more pronounced in lower SES groups.18

Finally, despite the significant progress made in Australia by banning visible point-of-sale display of tobacco products and advertising, cigarette dispensing cabinets are typically located at the front of stores, where payments are made, so tobacco availability is still in effect “on display”. The normalisation of tobacco as perpetuated by its widespread availability is of concern across the socioeconomic spectrum, but this is exacerbated when outlets are overly represented in lower SES neighbourhoods, where smoking prevalence and acceptability are already higher.

Our findings underscore the merits of considering whether the time has come to regulate the number of outlets able to sell a product that is known to kill around 15 500 Australians each year.19 In Australia, tobacco retail licensing is the remit of state and territory governments, and none have implemented any restrictions on the number of licences they grant. This is in stark contrast to the processes applied to alcohol, for which decisions to grant new liquor licences in a given area require the public interest to be considered.20 Internationally, California is the only notable example where cities and towns have the power to enact ordinances on land use through licensing and zoning regulations and where the location of tobacco retailers is regulated.21

While any move to restrict the number of tobacco outlets in Australia will no doubt evoke some opposition, as noted previously,10 various precedents exist for governments restricting either the number or location of other types of commercial activity. Sometimes this is on the grounds of public health or the public good, as in the case of alcohol outlets or restrictions on nightclubs or adult stores.10 Given our observation of the far higher number of outlets per capita in areas of lower SES, advocacy for restricting the number of tobacco outlet licences granted has evidence-based merit.

One limitation of our study is the lack of data on smoking behaviour. However, our primary aim was to examine the relationship between area SES and tobacco retail outlet density. Further, 2006 census data were linked with tobacco outlet information from 2011; while it would have been preferable to use 2011 census data, most of our analyses were complete before these data were released.

Further research linking geospatial retail outlet data to smoking prevalence data is needed to investigate the relationship between tobacco outlet density and smoking prevalence in Australia, as has been done in some US studies. Added insights would be gained if this could be linked to data on quit attempts, smoking cessation success and triggers for relapse, given the hypothesised association between ease of access and opportunities for unplanned purchases and relapse. More broadly, there is merit in investigating the spatial clustering of tobacco, alcohol and unhealthy food outlets in lower SES areas, as the dense collocation of such outlets can compound disparities in health behaviours and health outcomes in more disadvantaged areas.

Rate ratio of number of tobacco outlets per capita by Index of Relative Socioeconomic Advantage and Disadvantage (IRSAD) category for suburbs and towns in Western Australia, 2011

All WA suburbs and towns*(n = 911)

Perth metropolitan area*†(n = 296)

Regional WA*(n = 608)

IRSAD (percentile)

Rate ratio (95% CI)

Very low (< 25th)

4.14 (3.00–5.71)

< 0.001

1.48 (1.17–1.87)

< 0.001

5.51 (4.12–7.36)

< 0.001

Low (25th to < 50th)

2.12 (1.55–2.92)

< 0.001

1.19 (0.94–1.52)

0.157

2.60 (2.00–3.38)

< 0.001

High (50th to < 75th)

1.43 (1.15–1.79)

0.002

1.10 (0.87–1.40)

0.430

1.87 (1.34–2.60)

< 0.001

Very high (≥ 75th)

1.00

* For all WA suburbs and towns, Perth metropolitan area and regional WA, the P value for linear trend was < 0.001. † Perth metropolitan area analysis (296 suburbs) excluded seven suburbs with very high tobacco outlet counts or that were not generally representative of residential suburbs (eg, Rottnest Island, Naval Base).