Impact of pneumococcal polysaccharide vaccine in people aged 65 years or older

Invasive pneumococcal disease (IPD) is a major cause of morbidity in very young children and older adults.1 The 23-valent polysaccharide pneumococcal vaccine (23vPPV) has been available in Australia since 1986, and use of it has increased progressively since then. It was recommended and subsidised under the Pharmaceutical Benefits Scheme for Australians aged ≥ 65 years in 1997, provided free of charge for this group in Victoria from 1998, and included in the nationally funded National Immunisation Program from 2005.2 The vaccine’s effectiveness against IPD in immunocompetent older people has been estimated as about 70%,3 but it is generally regarded as not effective in preventing carriage of pneumococcal serotypes against which it is targeted.1,3

Australia is the only country to have introduced nationally funded programs for the 7-valent pneumococcal conjugate vaccine (7vPCV) for infants and the 23vPPV for older people in the same year (2005).4 Unlike the 23vPPV, the 7vPCV has been shown to prevent carriage of vaccine serotypes,1 resulting in herd immunity impacts. Reductions in IPD due to 7vPCV serotypes, in vaccinated and unvaccinated age groups, have been observed in many countries. Increases in non-7vPCV serotypes (referred to as serotype replacement) have also been observed in vaccinated and unvaccinated age groups, with the net impact on total IPD incidence varying from country to country.5

In Victoria, a 36% decrease in IPD incidence occurred in people over 65 years of age following the introduction of funded 23vPPV in that state in 1998;6 this was before widespread use of the 7vPCV in children. 7vPCV coverage among infants rose rapidly to 90% in Australia in the first year of the nationally funded program.7 As all 7vPCV serotypes are also present in 23vPPV, any herd immunity effect from 7vPCV would complicate interpretation of the impact of 23vPPV in older people. Studies from two regions on IPD incidence in Australian adults in the post-7vPCV era have provided conflicting evidence on the changes in total IPD incidence in older people.8,9

In July 2011, the 13-valent pneumococcal conjugate vaccine (13vPCV) was introduced for all Australian infants, replacing the 7vPCV and (in the Northern Territory) the 10-valent pneumococcal conjugate vaccine.10 Unlike the 7vPCV, the 13vPCV has been licensed for use in adults aged ≥ 50 years in the United States and Australia, but has not yet been included in the National Immunisation Program.

To inform policy decisions relating to the use of the 13vPCV and the 23vPPV in Australian adults, we aimed to answer three questions:

Can herd immunity effects in older Australians be observed from 7vPCV use in infants?
Can a direct effect of 23vPPV on IPD incidence in older people be shown in the post-7vPCV era?
Irrespective of ecological trends, has the 23vPPV been effective in preventing IPD in older Australians?

Methods

Notifications of IPD were obtained from the National Notifiable Diseases Surveillance System (NNDSS), which notionally captures all laboratory-confirmed cases of IPD. To distinguish the impact of the 7vPCV and the 23vPPV, notifications were aggregated by serotype category: serotypes contained in both vaccines (7vPCV serotype), those contained in the 23vPPV but not in the 7vPCV (23vPPV–non-7vPCV serotype) and those not contained in either vaccine (non-23vPPV serotype). Notifications for people in Victoria and people recorded as Indigenous were excluded, as funded programs for these populations commenced earlier than the national program (in 1998 and 1999, respectively). The proportion of notifications for which specimens were serotyped increased from 70% in 2002–2003 to 87% in 2006. To ensure that time trends in serotype categories were not distorted by this change, the serotype distribution among untyped cases was inferred from the serotype distribution among typed cases, by jurisdiction and by year.

This study was exempt from the requirement for ethics approval as it was conducted as a quality assurance exercise pertaining to the National Immunisation Program, under the auspices of the Australian Technical Advisory Group on Immunisation. De-identified data were provided for this purpose by the Communicable Diseases Network Australia.

Trends in IPD and vaccination coverage

Population rates of IPD by serotype category in people aged ≥ 65 years by year from 1 January 2002 to 31 December 2011 were plotted with 95% confidence intervals, which were calculated using the Poisson distribution of notification numbers. The estimated residential populations, minus Indigenous population estimates from the 2006 Australian census, were used as denominators for calculations of rates in the non-Indigenous population.11

Vaccination coverage estimates for people aged ≥ 65 years, by jurisdiction, were available from adult vaccination surveys conducted by the Australian Institute of Health and Welfare in 2004, 2006 and 2009.12–14 These were based on respondents’ report in telephone interviews regarding receipt of pneumococcal or pneumonia vaccine within the previous 5 years, without reference to written vaccination records.

IPD rate changes in vaccinated and unvaccinated age groups

Changes in rates from the pre-7vPCV era (2002–2004) to the recent post-7vPCV era (2010–2011) periods by serotype category were measured using incidence rate ratios (IRRs) with 95% confidence intervals. IRRs for the vaccinated age group (≥ 65 years) were compared with those for the unvaccinated age group (50–64 years), in which 23vPPV coverage was low (< 5%).7

Vaccine effectiveness estimates

Estimates of vaccine effectiveness (VE) in the ≥ 65-year age group were calculated using the screening method, a form of case–cohort study. It compares the likelihood of prior vaccination in 23vPPV serotype IPD cases with that for the total population. It uses the formula VE=1−[CV÷(1−CV)]×[(1−PV)÷PV], where CV is the proportion of cases that occurred in vaccinated people, and PV is the proportion of the population that had been vaccinated.15

The proportion of the population vaccinated was obtained from the adult vaccination surveys conducted in 2004, 2006 and 2009.12–14 Separate estimates for the 65–74-year and ≥ 75-year age groups by jurisdiction were available from the 2004 and 2006 surveys.

The proportion of 23vPPV-type IPD cases that were recorded as being in “fully vaccinated” people (ie, those vaccinated within 5 years, according to national recommendations at the time) who were aged ≥ 65 years was derived from the NNDSS for each year that population coverage data were available (2004, 2006 and 2009). For these VE calculations, cases recorded on the NNDSS as having occurred in people in Victoria or Indigenous people were not excluded, as their different dates of 23vPPV funding do not influence VE estimates. In addition, cases that occurred in Indigenous people were included in population coverage data, so the method required their inclusion in the study population.

A logistic regression model was fitted using the GENMOD procedure in SAS 9.1 (SAS Institute) as described previously.16 Data were stratified by year, age group and, for 2004 and 2006, by jurisdiction. A sensitivity analysis of the VE estimates was conducted, recalculating VE using a deviance of ± 10% of the total population in coverage estimates. Statistical analysis was carried out in SAS 9.1.3. Statistical significance was established by non-overlapping 95% confidence intervals, of rates and rate ratios.

Results

From 2002 to 2011, there were 3978 IPD notifications for Australians aged ≥ 65 years who were not in Victoria and were not Indigenous.

Trends in IPD rates and vaccination coverage

Annual IPD notification rates in people aged ≥ 65 years by serotype category are shown in Box 1. There was a substantial and statistically significant decrease in 7vPCV serotype IPD during the post-7vPCV era (2005–2011), as well as significant increases in 23vPPV–non-7vPCV and non-23vPPV serotypes, based on non-overlapping confidence intervals of annual rates.

The serotype most associated with replacement following 7vPCV introduction internationally — 19A — increased from 3% of isolates in 2002–2004 to 22% in 2010–2011.

The range of self-reported vaccination coverage (percentage vaccinated in the previous 5 years) estimates for those aged ≥ 65 years are also shown in Box 1, for all jurisdictions except Victoria. Coverage ranged from 41% to 53% in individual jurisdictions in 2004, increased to 51%–64% in 2006 and decreased to 48%–56% in 2009.

IPD rates in vaccinated and unvaccinated age groups

Pre-7vPCV (2002–2004) to post-7vPCV (2010–2011) changes in IPD rates in the ≥ 65-year age group and the 50–64-year age group are shown in Box 2. In both age groups there were substantial, statistically significant decreases for 7vPCV serotypes and increases for 23vPPV-non-7vPCV and non-23vPPV serotypes, based on IRRs not overlapping 1.0. The magnitude of these changes did not differ significantly between the two age groups. For all serotypes, the IRR point estimate was lower in the ≥ 65-year age group, but confidence intervals for the two age groups overlapped.

Vaccine effectiveness estimates

Numbers of IPD cases and VE estimates for 23vPPV against 23vPPV-type IPD are shown in Box 3. All VE estimates were statistically significantly above zero. The point estimate for 2009 was lower than for 2004, but was compatible with it, as confidence intervals overlapped. A sensitivity analysis to evaluate the impact of varying population coverage estimates for 23vPPV yielded an upper VE estimate of 75.8% (95% CI, 72.1%–79.9%) if true population coverage was 10% higher than estimated in the adult vaccination survey and a lower VE estimate of 40.5% (95% CI, 31.1%–49.9%) if true population coverage was 10% lower than estimated.

Discussion

Changes in IPD rates over time by serotype category presented here provide evidence of a substantial herd immunity impact in older people due to 7vPCV use in infants. However, an impact on IPD rates directly resulting from 23vPPV use in older people, by comparing changes in vaccinated and unvaccinated age groups, was not clearly shown. The VE estimate for 23vPPV against 23vPPV-type IPD was 61.1%. An overall decrease of 35% was observed in total IPD rates in ≥ 65-year-olds 6–7 years after the commencement of the nationally funded programs for 7vPCV and 23vPPV (25.2 notifications/100 000 population/year in 2002–2004 v 16.4 in 2010–2011).

Herd immunity impacts in adults from use of the 7vPCV in children have been shown in many countries. Herd immunity impacts on total IPD rates are heavily dependent on the pre-vaccination serotype distribution in adults and the length of time since 7vPCV introduction, as serotype replacement increases over time.5 Australian non-Indigenous people had one of the highest proportions of 7vPCV-type IPD out of total IPD in the world, similar to that in the US, and these are the only two countries with net decreases in IPD reported for older people following 7vPCV introduction.17

Trends in IPD rates by year in Australia in our study did not show clear evidence of a reduction of disease incidence due to use of 23vPPV in people aged ≥ 65 years. However interpretation of this finding is complicated by two factors: an overall modest level of 23vPPV coverage and a relatively small increase in coverage after national funding began in 2005; and the apparent indirect effects of introducing the 7vPCV for infants at the same time. The comparison of rates in vaccinated and unvaccinated age groups, both subject to herd immunity impacts from infant vaccination, allows the possibility of some impact from the 23vPPV. The absence of impacts on population IPD rates following publicly funded 23vPPV for ≥ 65-year-olds has also been reported in the US18 and United Kingdom.19 Gradual increases in coverage also occurred in those settings, and formal VE assessments in adults have consistently shown significant VE.20–22

All observational methods used to estimate VE are subject to bias. For our application of the screening method, different methods were used to ascertain the vaccination status of the general population (telephone survey) and the vaccination status of 23vPPV-type IPD cases (general practitioner and/or patient interview). However, our sensitivity analysis showed that the VE estimate remained statistically significant even if the true population coverage was 10% lower than the adult vaccination survey estimates. A study of 23vPPV vaccination status in older people in Victoria found that patient recall underestimated vaccination status by 6% compared with medical records.23

During the period of our study, a single revaccination was recommended for people first vaccinated at ≥ 65 years of age. As of December 2011, this is no longer recommended.10 The latest national estimate of the proportion of people aged ≥ 65 years who have ever received 23vPPV is a modest 59%.13 Given the evidence of the vaccine’s effectiveness, higher coverage would be expected to increase the impact of the vaccine in reducing IPD incidence.

Data are yet to emerge on the herd immunity impact from 13vPCV use in Australian children. However, if similar effects are seen from the additional six serotypes as from the 7vPCV, there would be a further reduction in IPD in older people and, therefore, less potential benefit from the 23vPPV.

The appeal of a conjugate vaccine used in older people includes potential, although unproven, benefits such as a superior response to booster doses and impacts on carriage and non-invasive pneumonia. Herd immunity impacts of 7vPCV on non-invasive pneumonia in older people have been reported as being non-existent, very small or extensive.24–26 A randomised controlled trial assessing the impact of 13vPCV use in older people on pneumonia is currently underway.27 However, this trial is not being conducted alongside concurrent use of 13vPCV in infants. The incremental benefits of 13vPCV use in older people in addition to an infant program would be more difficult to evaluate.

In conclusion, our data show moderate effectiveness of the 23vPPV against IPD in older Australians, consistent with that shown in comparable populations elsewhere. In combination with herd immunity impacts from 7vPCV in children, this resulted in a 35% decrease in IPD in those aged ≥ 65 years. Further benefits could be expected if an increase in 23vPPV coverage in older people could be achieved.

1 IPD notification rates by serotype category for ≥ 65-year-old Australians and pneumococcal vaccination coverage, 2002–2011*

IPD = invasive pneumococcal disease. 7vPCV = serotypes contained in the 7-valent pneumococcal conjugate vaccine and 23-valent polysaccharide pneumococcal vaccine. 23vPPV–non-7vPCV = serotypes contained in the 23-valent polysaccharide pneumococcal vaccine but not the 7-valent pneumococcal conjugate vaccine. Non-23vPPV = serotypes not contained in either vaccine. * IPD notification rates do not include people in Victoria or Indigenous people; serotype categories are adjusted for untyped cases; error bars for IPD notification rates are 95% confidence intervals; and error bars for vaccination coverage are ranges of self-reported vaccination coverage for individual jurisdictions excluding Victoria.

2 Invasive pneumococcal disease notifications and notification rates for unvaccinated (50–64 years) and vaccinated (≥ 65 years) age groups of older Australians by serotype group, 2002–2004 versus 2010–2011*

	Number of notifications		Notifications per 100 000 population per year
Age and serotype group†	2002–2004	2010–2011	2002–2004	2010–2011	Incidence rate ratio (95% CI)‡

50–64-year-olds
7vPCV	490	45	6.59	0.76	0.12 (0.08–0.15)
23vPPV–non-7vPCV	180	311	2.42	5.34	2.21 (1.83–2.67)
Non-23vPPV	53	95	0.71	1.64	2.31 (1.62–3.27)
All	723	451	9.71	7.74	0.80 (0.71–0.90)
≥ 65-year-olds
7vPCV	954	80	17.00	1.84	0.11 (0.09–0.14)
23vPPV–non-7vPCV	317	401	5.65	9.27	1.64 (1.41–1.91)
Non-23vPPV	144	230	2.56	5.30	2.07 (1.67–2.57)
All	1415	711	25.21	16.41	0.65 (0.59–0.71)

7vPCV = serotypes contained in the 7-valent pneumococcal conjugate vaccine and 23-valent polysaccharide pneumococcal vaccine. 23vPPV–non-7vPCV = serotypes contained in the 23-valent polysaccharide pneumococcal vaccine but not the 7-valent pneumococcal conjugate vaccine. Non-23vPPV = serotypes not contained in either vaccine. * Data on people in Victoria and Indigenous people are excluded. † Adjusted for untyped cases. ‡ 2010–2011 : 2002–2004.

3 Numbers of IPD cases and VE estimates for 23vPPV against 23vPPV-type IPD in Australians aged ≥ 65 years*

Year

Cases in people vaccinated with 23vPPV/total cases (%)

Proportion of population vaccinated with 23vPPV†

VE estimate (95% CI)‡

2004

106/339 (31.3%)

51.1%

63.3% (53.1%–71.9%)

2006

132/320 (41.3%)

62.2%

65.6% (56.1%–73.9%)

2009

90/241 (37.3%)

56.0%

50.4% (35.1%–62.9%)

Total

328/900 (36.4%)

61.1% (55.1%–66.9%)

IPD = invasive pneumococcal disease. VE = vaccine effectiveness. 23vPPV = 23-valent pneumococcal polysaccharide vaccine. na = not applicable. * Data include people in Victoria and Indigenous people. † Values are summary proportions of Australians who received the vaccine within the previous 5 years. ‡ VE estimates were calculated using data stratified by jurisdiction, year and age group.

Association between tobacco plain packaging and Quitline calls: a population-based, interrupted time-series analysis

Internationally, Australia is a leader in tobacco control policies that reduce community exposure to tobacco-related harm. In December 2012, Australia became the first country in the world to enact legislation mandating plain packaging for all tobacco products.1 The Tobacco Plain Packaging Act 2011 (Cwlth) is designed to prevent tobacco industry promotion by simultaneously reducing pack attractiveness and increasing the size of graphic health warnings. The legislation required manufacturers to produce plain packs with new warnings from 1 October 2012. From 1 December 2012, plain packaging became compulsory for all tobacco products. The new plain packs are olive green and devoid of brand design. Telephone numbers for the national smoking cessation helpline, Quitline, feature prominently on the packs.

Plain packaging legislation exists to encourage smokers to quit and discourage the uptake of smoking.1 Quitline is a free resource that can be used by smokers who are motivated and seeking support to quit. Therefore, the volume of calls to cessation helplines has frequently been used as one indicator of changes in interest in quitting in response to population-wide cessation policies and programs.2–6

The best level of evidence for evaluating a whole-of-population initiative such as tobacco plain packaging is an interrupted time-series analysis. In exactly the same way in which observational studies have been used to define the association between lung cancer and tobacco (because randomised studies are unethical in such a context), before-and-after evaluation, controlling for secular trends, is the optimal design for assessing the effects of population-wide initiatives.

In this study, we sought to examine behavioural change resulting from the introduction of the Tobacco Plain Packaging Act, complementing a recent report of smokers’ feedback.7 We did this by investigating the impact of the introduction of tobacco plain packaging on Quitline calls. To provide context, we compared the impact on Quitline calls of the introduction of tobacco plain packaging with the nationwide introduction of graphic health warnings on cigarette packaging in 2006.8 The null hypothesis was that there would be no change in call numbers, adjusting for known confounders.

Methods

We used an interrupted time-series analysis to investigate trends in the weekly volume of calls from New South Wales and the Australian Capital Territory to the Quitline.

As plain packs were phased in from 1 October 2012, we considered this date the start of the intervention. Similarly, 1 March 2006 was the date of the introduction of graphic health warnings on cigarette packaging. We looked at Quitline call numbers before and after these dates for the two interventions. Call data from 1 April 2004 to 28 February 2006 were provided by Macquarie Telecom (Sydney, Australia) and from 1 March 2006 to 31 March 2013 by the Telstra Analyser (Telstra, Melbourne, Australia).

This study did not require institutional ethics approval as it did not involve data about individuals. The study received no external funding. Reporting of the study complies with the STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) consensus guidelines for reporting observational studies.9

Potential confounders

An increase in anti-smoking advertising in mass media such as television is a potential confounder as it increases the number of calls to the Quitline.2,3,6 We ascertained weekly target audience rating points (TARPs) for advertisements broadcast in NSW and the ACT during the periods of interest, using OzTAM (Australian television audience measurements) for adults aged 18 years and older for free-to-air and cable television using established methods (unpublished report prepared by OzTAM for the Cancer Institute NSW). TARPs are a product of the percentage of the target audience exposed to an advertisement (reach) and the average number of times a target audience member would be exposed (frequency), adjusted for the length of the advertisement. For example, 200 TARPs might represent 100% of the target audience receiving the message twice on average over a specified period, or 50% reached four times.

Another potential confounder is cigarette costliness. We followed Wakefield et al10 and calculated cigarette costliness as the ratio of the average quarterly recommended retail price for a pack of the two top-selling Australian cigarette brands (obtained from the retail trade magazine Australian Retail Tobacconist, volumes 65 to 87) to the average weekly earnings in the same quarter obtained from the Australian Bureau of Statistics.11

Finally, the number of smokers in the community can be a potential confounder. We obtained data on smoking prevalence during the study periods from Health Statistics New South Wales,12 and this was applied to quarterly figures for population size from the Australian Bureau of Statistics13 to calculate the number of smokers in the community. Data for the first quarter of 2012 were not available, so we used results carried forward from the final quarter of 2011.

Statistical analyses

As the data for weekly number of calls to the Quitline were autocorrelated (each value was correlated with the previous value) we used autoregressive integrated moving average (ARIMA) analysis in SAS version 9.3 (SAS Institute Inc). ARIMA models enable the investigation of changes over time while accounting for seasonal variation and background trends in such things as the effects of television anti-tobacco advertising, changes in cigarette pricing relative to weekly earnings and number of smokers in the community. In ARIMA modelling, comprising model investigation, estimation and diagnostic checking, we followed the methods of Box et al (Appendix).14

A single model fitted to the entire 7-year period of Quitline call data did not meet technical criteria for model fit. Therefore, separate models that included data for 12 months before and 6 months after each intervention (1 March 2005 to 1 September 2006 and 1 October 2011 to 1 April 2013) were fitted, as this was the longest duration of follow-up for tobacco plain packaging available at the time of the study.

The same modelling approach was used for fitting models to both data subsets. Indicator terms were created to represent the week of the introduction of the intervention (plain packaging or graphic health warnings). A seasonal New Year’s Eve term was created to allow for an increase in calls around the New Year period. The potential confounders of seasonal variation (New Years’ peaks), anti-smoking advertising activity (TARPs) and number of smokers in the population were included in the models and were retained regardless of statistical significance on the basis of face validity. Cigarette price was included in the plain packaging analysis, but was omitted from the graphic health warnings analysis as there was minimal change in price during the observation period.

Results

Overall, there was a general decrease in the number of calls each week to the Quitline over the period of observation. There were 910 calls in the week before the introduction of graphic health warnings in 2005 compared with 363 calls in the week before the introduction of plain tobacco packaging in 2012 (Box 1, Box 2).

Based on the estimated regression parameters from the ARIMA models (Box 3), the plain packaging intervention resulted in an increase of 288 calls (95% CI, 160–517 calls) to the NSW Quitline, representing a 78% increase from the number of calls in the week the plain packaging legislation was introduced (Box 1). This effect was sustained in subsequent weeks, with an estimated 86% of extra calls retained relative to the previous week (Box 1). As a comparison, the introduction of graphic health warnings increased the number of calls by 763 (95% CI, 473–1053), an estimated increase of 84% relative to the number of calls in the week graphic health warnings legislation was implemented (Box 1). The effect of graphic health warnings diminished more quickly, with each post-peak week retaining only 40% of the previous week’s extra calls (Box 1).

Discussion

We found a significant increase in the number of calls to Quitline coinciding with the introduction of mandatory plain packaging of tobacco after other known confounders had been taken into account. Australia has taken a lead on mandating plain packaging, now supported by evidence of an immediate impact of this legislation. This should encourage other countries that are preparing similar legislation.

We also found that, compared with the introduction of an earlier smoking cessation initiative — graphic health warnings — the effect of plain packaging had a more immediate effect, a slightly smaller peak and a much longer duration of impact.

These changes need to be interpreted in the context of the earlier introduction of graphic health warnings and a raft of other legislated tobacco control policies, and secular trends in smoking rates. It is possible that the magnitude and duration of effect could be different in countries that are implementing aspects of the Framework Convention on Tobacco Control in different ways or different orders.15 For example, in NSW, the most recently legislated changes include the widespread introduction of smoke-free public spaces from 2007,16,17 and banning the display of tobacco products at the point of sale in 2010.18,19 Furthermore, smoking prevalence has declined markedly from 17.7% of NSW residents reporting daily or occasional smoking in 2006 to 14.7% in 2011.12

A key strength of this study was the ability to account for other known influences on quitting behaviour and use of the Quitline:

anti-tobacco advertising activity;
the costliness of tobacco;
the number of smokers in the community; and
seasonal peaks that are observed in the New Year period.

Although the volume of calls to the Quitline is an indirect measure of quitting intentions and behaviour, it does provide an objective outcome measure that is not subject to selection or social desirability biases that may occur in community surveys. Quitline data are available almost in real time so that the immediate impact of community-wide programs or policies can be evaluated soon after implementation. Interrupted time-series analysis provides one of the most robust methods of evaluating the impact of programs and policies that affect the whole population.20 We examined legislation enacted nationally simultaneously, precluding the use of a control group.

A weakness of our study is its inability to differentiate the impact of the increased size of graphic health warnings that happened simultaneously with tobacco plain packaging from the impact of plain packaging itself. Quitline calls are one measure of changes in people’s behaviour in response to legislative change. The rate of calls to Quitline may have been confounded by the Quitline telephone number appearing to be more prominent in the absence of proprietary branding. Also, our study has shown an association but cannot prove causation.

Longer-term investigation of the impact of the legislation should include assessment of more direct measures:

smoking uptake in non-smokers;
prevention of relapse among ex-smokers; and
cessation of smoking among current smokers.

This would include the need to investigate variations in impact on subgroups of the population. A comprehensive evaluation of the policy will require ongoing monitoring to assess factors such as changes in brand recognition, awareness of the health risks of smoking, the social acceptability of smoking as well as smoking prevalence.21

In conclusion, we showed a significant increase in the number of calls to the Quitline that coincided with the introduction of plain packaging nationally in Australia. Future evaluations of other smoking-related outcomes need to include the impact on smoking prevalence.

1 Autoregressive integrated moving average (ARIMA) modelling results for the plain packaging and graphic health warning interventions

Intervention

Number of calls in
week intervention
was introduced

Peak call volume attributable to
intervention (calls/week)

Time
to
peak

Duration
of effect

Exponential decay
rate of call volume following peak

Estimate (95% CI)

Relative increase

Plain packaging

363

651 (523–780)

78%

4 weeks

43 weeks

14% per week

Graphic health warnings

910

1673 (1383–1963)

84%

12 weeks

20 weeks

60% per week

2 Weekly calls to Quitline, target audience rating points (TARPs) and cigarette price relative to income, before and after the introduction of plain tobacco packaging and graphic health warnings

3 Change in numbers of call to Quitline after the introduction of plain packaging and graphic health warnings, adjusted for
anti-tobacco advertising, seasonal trend, smoking prevalence and tobacco costliness*

	Number of Quitline calls
	Graphic health warnings		Plain packaging
Parameter	Estimate (SE)	P	Estimate (SE)	P

Intervention	12-week time to peak		4-week time to peak
Peak number of additional calls	762.88 (147.88)	< 0.001	288.18 (65.48)	< 0.001
Weekly decay in number of additional calls	0.60 (0.24)	0.095	0.14 (0.10)	< 0.001
Number of calls to Quitline attributable to confounders
Target audience rating points (TARPs)	1.47 (0.20)	< 0.001	0.08 (0.11)	0.47
New Year period	199.38 (169.49)	0.24	127.18 (43.78)	0.004
Smoking prevalence (per 100 000 smokers)	− 275.26 (230.15)	0.23	90.54 (595.37)	0.88
Cigarette costliness (per 1% increase relative to average earnings)			428.98 (546.25)	0.43

* Full model details can be supplied by the authors on request.

Medical graduates becoming rural doctors: rural background versus extended rural placement

To encourage the development of a sustainable rural medical workforce, the Australian Government provides incentives for Australian medical schools to recruit students with a rural background and funds rural clinical schools to allow rural exposure via extended placements to students of all backgrounds.

Both recruitment of rural students and uptake of extended rural placements (ERPs) have been shown to be associated with choosing a rural career.1–4 We aimed to compare the apparent association of these two factors on students’ expressed intentions to undertake rural internships and their acceptance of rural internships after finishing medical school.

The Medical Schools Outcomes Database (MSOD) and Longitudinal Tracking Project was established by the Deans of Australian and New Zealand medical schools in 2005.5–6 For the MSOD, all Australian and New Zealand medical students are asked to complete a commencing medical students questionnaire on entry and an exit questionnaire on leaving their university courses. Among other things, the questionnaires cover career aspirations, including the types of setting in which students aspire to practise. The MSOD also follows medical students’ postgraduate trajectories. Our comparison used longitudinal data from three cohorts of students in the Sydney Medical Program (SMP), the 4-year graduate-entry program of the University of Sydney.

Methods

We analysed MSOD data5 for the student cohorts that commenced the SMP in 2005, 2006 and 2007 and completed it in 2008, 2009 and 2010, respectively. The variables studied were rural versus urban background, ERP (yes/no), preference for a rural or an urban internship, acceptance of a rural or an urban intern position, and the preferred type of location of future practice.

We used χ2 analysis to assess associations between categorical variables and the McNemar test for repeated data.7 Where cell sizes were too small for valid analysis, the Fisher exact test (FET) P is reported. P < 0.05 was considered significant. Data were analysed using SPSS version 20.0 (SPSS Inc).

Definitions and data sources

An ERP was defined as a 32-week placement in the School of Rural Health (with clinical schools at Dubbo and Orange, in central-west New South Wales) or one of the university’s two Departments of Rural Health (in Lismore, on the NSW North Coast, or Broken Hill, in far-west NSW) over the course of a single 37-week academic year. Students voluntarily applied to undertake an ERP and all students who applied were able to be accommodated. Data on whether individual students had undertaken an ERP were obtained from rolls kept by Sydney Medical School.

Rural or urban background was determined by the item in the commencing questionnaire asking students to nominate the type of location in which they had lived the longest within Australia. Locations were categorised according to the Rural, Remote and Metropolitan Areas (RRMA) classification8 and further grouped as urban or rural (Box 1). We used an answer of yes to “self-perception of rural background” as another measure of rural background.

Future practice location preferences were determined from responses in both the commencing and the exit questionnaires. Locations were mapped to RRMA categories and grouped as urban or rural (Box 1).

Internship first preferences were determined from the exit questionnaire. Students could indicate up to four hospital preferences for their intern year. We used the first preference. Hospital postcodes were categorised using the Australian Standard Geographical Classification — Remoteness Area9 and grouped as urban or rural (Box 1).

Intern positions were determined from MSOD follow-up data and routine enquiries made by clinical schools regarding the locations of the internships taken up by their alumni.

Ethics approval was obtained from the University of Sydney Human Research Ethics Committee (Ref. No. 11436).

Results

Our analysis was confined to students who remained in their original entry cohort throughout the program and completed both the commencing and the exit questionnaires (“respondents”). We excluded students who left their original entry cohort and those who did not complete both questionnaires (Box 2). We followed a total of 448 students from the three cohorts (54.6% of the total number of students enrolling in the 2005, 2006 and 2007 entry cohorts; N = 821). While we did not have access to demographic variables for the cohorts overall, we note that the proportion of respondents who undertook an ERP (98/448; 21.9%) was similar to the proportion overall (156/786; 19.8%).

Of the 448 students followed, 426 (95.1%) responded to the item on self-perception of rural background and 76 of these (17.8%) considered themselves to be from a rural background. The item on the “location of longest residency within Australia” was answered by 415/448 respondents (92.6%), with 73 (17.6%) having lived in a rural location for the longest part of their lives. The difference between classifying rural or urban background based on these two items was not significant (paired data; n = 397; McNemar test χ12 = 0.03; P = 0.86). We based subsequent analyses on location of longest residency within Australia.

The characteristics, experience, preferences and internship locations of the 448 students in the three cohorts are summarised in Box 3. The proportion preferring a rural career dropped by more than one-third between entry into and exit from the program (from 20.7% to 12.5%). ERPs were undertaken by 21.9% (98/448). Ultimately, 8.1% (35/434) of the students accepted a rural internship, although 14.5% (60/415) had indicated in the exit questionnaire that they had a first preference for a rural post.

Students who had a rural background responded differently from those who undertook an ERP (Box 4 and Box 5). Compared with students from an urban background, students from a rural background were significantly more likely to prefer a rural future practice at the beginning of the program (χ2 = 64.11; P < 0.001) and at the end (χ2 = 25.28; P < 0.001) (Box 4). Significant differences were also found when students who undertook an ERP were compared with those who did not (Box 5) with regard to preference for rural future practice at the beginning of the program (χ2 = 10.50; P = 0.001) and at the end (χ2 = 26.02; P < 0.001). Respondents who undertook an ERP were more than three times as likely as those from rural backgrounds to express a first preference for a rural internship (23.9% v 7.7%; χ2 = 7.04; P = 0.008), and more than twice as likely to accept a rural internship (21.3% v 9.9%; χ2 = 3.85; P = 0.05).

Among respondents with a rural background (Box 4), 58.7% (37/63) began the SMP with rural career intentions, but this proportion decreased to 30.6% (22/72) at the end of the program, an absolute reduction of 28%. Among the group who undertook an ERP (Box 5), 33.0% (29/88) expressed rural career intentions upon entry to the program, but this decreased to 28.0% (26/93) by the end. While the final proportions of students with rural career intentions were similar for both groups, the ERP group showed a smaller absolute decrease (5%) between the start and the end of the program.

Rural backgrounds were over-represented among the 98 students who undertook an ERP: the group comprised 22 of the 73 students with a rural background (30.1%) and 72 of the 342 students who did not have a rural background (21.1%) (Box 4). Among students with a rural background, there was a significant association between undertaking an ERP and rural internship preference and acceptance (P = 0.034 and P = 0.021, respectively [FET]; Box 6), however numbers were small. There were also highly significant differences between students from an urban background who undertook an ERP and those from an urban background who did not, in rural internship preference and acceptance (χ2 = 5.43, P = 0.020 and χ2 = 22.04, P < 0.001, respectively; Box 6).

Of students who had a rural background and did an ERP, 4/21 (19.0%) had a first preference for a rural internship, and 5/21 (23.8%) accepted a rural internship. These proportions did not differ significantly for those students without a rural background who undertook an ERP, of whom 17/68 (25.0%) had a first preference for a rural internship, and 14/69 (20.3%) accepted a rural internship (P > 0.8 for both comparisons), although numbers are too small to exclude a small difference.

Discussion

Our results clearly point to an association between undertaking an ERP and early postgraduate adoption of rural medical practice. Extended rural clinical placements appear to have a stronger association than that of rural background with students’ preference for a rural internship and their acceptance of rural intern posts upon completion of their medical course. We observed that students with rural backgrounds were overrepresented among the students who undertook an ERP, so rural background may have contributed to rural internship preference and uptake in this group. The small numbers of students with a rural background who undertook an extended rural clinical placement preclude a definitive conclusion on whether the two attributes would be additive above the influence of undertaking an extended rural clinical placement. These findings echo the results of cross-sectional studies of student cohorts11,12 and retrospective studies of factors that influenced doctors to enter rural practice in Australia13 and Canada.14

The strength of our conclusions derives from the fact that we used longitudinal follow-up data for three successive cohorts of Australian medical students and obtained information on the end point of taking a rural intern post. By following these cohorts, we were able to compare the apparent influence of an extended rural clinical training experience on students who had a rural background with its influence on those who did not. Our study thus demonstrated the value of the MSOD, which was set up at the time when the first cohort began their medical course. A weakness of our results is that they rely on the experience of the students of only one medical school. Also, the end point of rural internship is an important short-term outcome, but it will be interesting to observe the longer-term career choices of these students through longitudinal MSOD follow-up, or possibly through future linkage with Medical Board of Australia data.

Many possible reasons can be advanced to explain the association between extended rural clinical placements and both the preference for, and the acceptance of, rural internships. These include prior interest in rural medicine among students who choose to take an ERP; appreciation of the educational opportunities that are more likely to be available in rural settings than metropolitan settings, such as interprofessional learning; the smaller numbers of students in rural clinical schools, enabling more direct interaction between teaching staff and students; students’ experience of a closer involvement with the health of rural communities than is possible with metropolitan communities; the broad appeal of the rural environment; and the quality of educational infrastructure that has resulted from consistent funding of rural clinical teaching facilities. The MSOD provides a vehicle for analysis of these possible reasons for the stimulating effect of extended rural placements.

In the meantime, our findings support the Australian Government’s commitment to the funding of rural clinical schools and the provision of financial incentives for medical schools to ensure that at least 25% of each graduating cohort spends one academic year in a rural setting.

1 Classifications used to categorise students’ background locations and practice preferences as urban or rural

Rural, Remote and Metropolitan Areas (RRMA) Classification8*

Australian Standard Geographic
Classification — Remoteness Area9†

Urban‡

Capital city

RA1

Major cities

Major urban centre

Rural‡

Regional city or large town

RA2

Inner regional

RA3

Outer regional

Smaller town

RA4

Remote

Small community

RA5

Very remote

* Used to categorise students’ backgrounds and future practice preferences. † Used to categorise internship first preferences and accepted placements. ‡ Urban/rural distinction for RRMA categories is consistent with Gerber and Landau.10

2 Distribution of student responses

3 Characteristics, experience, preferences and internships of respondents from the 2005, 2006 and 2007 Sydney Medical School entry cohorts (N = 448)

	n*	Yes	No

Rural background	415	73 (17.6%)	342 (82.4%)
Rural future practice preference (CMSQ)	382	79 (20.7%)	303 (79.3%)
Extended rural placement	448†	98 (21.9%)	350 (78.1%)
Rural future practice preference (EQ)	433	54 (12.5%)	379 (87.5%)
Rural internship first preference	415	60 (14.5%)	355 (85.5%)
Accepted a rural internship	434	35 (8.1%)	399 (91.9%)
CMSQ = commencing medical students questionnaire. EQ = exit questionnaire. * Number of respondents (students who remained in their original entry cohort throughout the Sydney Medical Program and completed both the commencing and the exit questionnaires) who answered the question. † Data obtained for all respondents from Sydney Medical School rolls.

4 Respondent characteristics, experience, preferences and internships (N = 448), by urban or rural background (n = 415)

					Statistical comparison§
	n*	Rural background, no. (%)† (n = 73)	Urban background no. (%)† (n = 342)	Total,‡ no. (%)	χ2	P

Rural future practice preference (CMSQ)	368	37/63 (58.7%)	41/305 (13.4%)	78 (21.2%)	64.11	< 0.001
Extended rural placement (ERP)	415	22/73 (30.1%)	72/342 (21.1%)	94 (22.7%)	2.83	0.092
Rural future practice preference (EQ)	402	22/72 (30.6%)	29/330 (8.8%)	51 (12.7%)	25.28	< 0.001
Rural internship first preference	387	5/65 (7.7%)	51/322 (15.8%)	56 (14.5%)	2.90	0.089
Accepted a rural internship	402	7/71 (9.9%)	24/331 (7.3%)	31 (7.7%)	0.56	0.455

CMSQ = commencing medical students questionnaire. ERP = extended rural placement. EQ = exit questionnaire. * Total number of respondents who answered the question. † Denominators vary according to number of students who answered each question. ‡ Positive response to item for the subsample of 448 who provided information on urban or rural background. § df = 1.

5 Respondent characteristics, experience, preferences and internships (N = 448), by extended rural placement (ERP) (n = 448)

					Statistical comparison§
	n*	Undertook ERP, no. (%)† (n = 98)	Did not undertake ERP, no. (%)† (n = 350)	Total,‡ no. (%)	χ2	P

Rural background	415	22/94 (23.4%)	51/321 (15.9%)	73 (17.6%)	2.83	0.092
Rural future practice preference (CMSQ)	382	29/88 (33.0%)	50/294 (17.0%)	79 (20.7%)	10.50	0.001
Rural future practice preference (EQ)	433	26/93 (28.0%)	28/340 (8.2%)	54 (12.5%)	26.02	< 0.001
Rural internship first preference	415	22/92 (23.9%)	38/323 (11.8%)	60 (14.5%)	8.55	0.003
Accepted a rural internship	434	20/94 (21.3%)	15/342 (4.4%)	35 (8.1%)	28.25	< 0.001

CMSQ = commencing medical students questionnaire. EQ = exit questionnaire. * Total number of respondents who answered the question. † Denominators vary according to number of students who answered each question. ‡ Overall positive response to item. § df = 1.

6 Student characteristics, experience, preferences and internships (N = 448), by extended rural placement (ERP) split by rural background

Statistical comparison§

Undertook ERP, no. (%)†

Did not undertake ERP, no. (%)†

Total,‡ no. (%)

χ2

Rural background students (n = 73)

(n = 22)

(n = 51)

Rural future practice preference (CMSQ)

15/21 (71.4%)

22/42 (52.4%)

37 (58.7%)

2.10

0.148

Rural future practice preference (EQ)

10/21 (47.6%)

12/51 (23.5%)

22 (30.6%)

4.07

0.044

Rural internship first preference

4/21 (19.0%)

1/44 (2.3%)

5 (7.7%)

–

0.034¶

Accepted a rural internship

5/21 (23.8%)

2/50 (4.0%)

7 (9.9%)

–

0.021¶

Urban background students (n = 342)

(n = 72)

(n = 270)

Rural future practice preference (CMSQ)

305

14/63 (22.2%)

27/242 (11.2%)

41 (13.4%)

5.26

0.022

Rural future practice preference (EQ)

330

15/69 (21.7%)

14/261 (5.4%)

29 (8.8%)

18.26

< 0.001

Rural internship first preference

322

17/68 (25.0%)

34/254 (13.4%)

51 (15.8%)

5.43

0.020

Accepted a rural internship

331

14/69 (20.3%)

10/262 (3.8%)

24 (7.3%)

22.04

< 0.001

CMSQ = commencing medical students questionnaire. EQ = exit questionnaire. * Total number of respondents who answered the question. † Denominators vary according to number of students who answered each question. ‡ Overall positive response to item. § df = 1. ¶ Fisher exact test.

Pointers for pandemic planning

As the World Health Organization regularly reminds us, neither the timing nor the severity of the next influenza pandemic can be predicted. So, when avian influenza A(H5N1) virus emerged and fatal human cases were detected in a number of countries from 2003 onwards, pandemic planning took centrestage. At that time, those tasked with writing and implementing pandemic plans had no easy reference or experience to draw upon as the most recent pandemic had occurred 35 years earlier, in 1968, when technologies such as antiviral drugs, split and subunit vaccine preparations and computer-aided disease surveillance systems for outbreak detection did not exist. This made the first edition of Van-Tam and Sellwood’s book Pandemic influenza, published in April 2009, a useful resource. The book brought together leading experts to summarise the epidemiology, virology and clinical aspects of influenza as well as public health surveillance, emergency response and risk aspects of pandemic preparedness.

This second edition of the book was published in February 2013 and has built in the experience and lessons learnt from the 2009 influenza A(H1N1pdm09) pandemic. It consists of 20 chapters and nine country case studies. The editors are eminently qualified to oversee such a project: Van-Tam is Professor of Health Protection at the University of Nottingham in the United Kingdom, and Sellwood is Pandemic and Seasonal Influenza Resilience Manager at the National Health Service in London. Once again, they have selected some of the world’s most respected and influential practitioners in the field to present summaries of current pandemic influenza knowledge, preparedness and response plans. Each chapter presents an authoritative and balanced overview.

Without doubt, this book fills a gap as no other publication addresses the plethora of issues faced by pandemic planners. The book will greatly benefit those advising decisionmakers on the rationale for various recommendations including on topical issues such as the utility of the anti-influenza drug oseltamivir, which has been recently questioned. The book’s layout and limited use of colour does detract from its shelf appeal but the content is not compromised. The book will serve public health practitioners and students in Australia and other developed countries well on both technical and operational issues.

Carrying weapons and intent to harm among Victorian secondary school students in 1999 and 2009

Youth violence (for example, carrying a weapon or attacking someone with intent to harm) is one of the most important social and public health problems worldwide.1–3 The costs of youth violence stem from harm caused to victims, as well as policing and criminal justice responses and community perceptions of reduced safety.3,4 Violence was listed in the top three issues concerning young Australians in 2010.5 Eight per cent of Victorian students in Years 7 and 9 engaged in violent behaviour in 2002, with higher rates among boys than girls.6 In addition, rates of youth violence are higher in disadvantaged communities7 and in regional communities with unstable populations and economic structures as well as high unemployment,8 and lower among youths from immigrant families.9

There have been reports of increases in violent offences perpetrated by youth in Australia10 and media reports suggest more youth are violent.11,12 National population-based surveys could measure whether rates of self-reported violence among youth have changed over time, but Australia does not have such surveys. In this article, we analyse data from Victorian surveys of secondary school students in 1999 and 2009 to examine whether rates of carrying a weapon and attacking someone with intent to harm have increased, after adjusting for sample demographic characteristics.

Methods

Data for this study were drawn from two Victorian cross-sectional surveys; the 1999 Adolescent Health and Wellbeing Survey13 and the 2009 Victorian Adolescent Health and Wellbeing Survey (HOWRU).14 Both surveys used a modified version of the Communities That Care Youth Survey15,16 to measure behavioural and mental health outcomes, as well as risk and protective factors.

Both surveys used a two-stage cluster sampling approach. The first stage consisted of a stratified random sample of government (public) and non-government (Catholic and independent) schools from each of 36 areas across Victoria — the 31 metropolitan government areas and the five state government regions outside the metropolitan area. Schools were selected randomly with a probability proportional to the number of Year 7, 9 and 11 students in the school. In the second stage, a random sample was taken of one class at each year level.

Data collection covered the first three school terms (February to September) in 1999 and the second and third terms (April to November) in 2009. Responses to questionnaires were anonymous. Students completed the questionnaire in class, taking about 40 minutes in 1999 for a pen-and-paper survey and about 60 minutes in 2009 to complete a longer online survey (although computers were not available for 19% of surveys, so paper versions were used). Students were supervised by trained research staff while they completed the survey. In 1999, copies of the questionnaire with a prepaid return envelope were left at the school to be returned to researchers by students who were absent on the day of the survey. For 2009, there was no follow-up to collect surveys from absent students.

Measures

Carrying a weapon was measured by asking students whether in the past 12 months they had carried a weapon. Attacking someone with intent to harm was measured by asking students whether in the past 12 months they had attacked someone with the idea of seriously hurting them. In the 1999 survey, participants responded to both questions on an 8-point scale ranging from “No” to “Yes, more than 40 times”. In the 2009 survey, items were rated on a 5-point scale ranging from “Never” to “10 or more times”. For both surveys, responses to these items were recoded and dichotomised to 0 (“No/Never”) or 1 (“At least once”).

Demographic measures controlled in the analyses included age, sex, language spoken at home, and geographic location of school. Student residential socioeconomic disadvantage was measured using four quartiles: (0–25th quartile [most disadvantaged], 26th–50th quartile [disadvantaged], 51st–75th quartile [advantaged] and 76th-100th quartile [most disadvantaged]), on the index of relative socioeconomic disadvantage from the Australian Bureau of Statistics’ Socio-Economic Indices for Areas (SEIFA) for 1996 and 2006.17,18

Statistical analysis

Prevalence rates with 95% CIs were estimated for the outcomes of interest in each survey. In unadjusted analyses, the relationship between year of survey and the outcomes was analysed by logistic generalised estimating equations (GEE) to take account of the clustering of students in schools. These analyses were then repeated, controlling for all demographic measures. A complete case analysis was performed with Stata, version 11 (StataCorp), that included only students who responded to all the variables analysed here.

Ethics approval

Ethics approval for both surveys was obtained from the Royal Children’s Hospital Human Research Ethics Committee and relevant education authorities. For the 1999 survey, active informed consent was required from parents for students to participate. In the 2009 survey, passive informed consent from parents was required in most schools, but active informed consent from parents was required at some Catholic schools. For both surveys, informed assent was sought from students on the day of the survey.

Results

The student participation rate was 70% (8984 students) in the 1999 survey and 76% (10 273 students) in the 2009 survey. There were complete case data for 92% of students in the 1999 survey (7998) and 91% of students in the 2009 survey (9364). Box 1 describes the school and student samples for each survey.

The prevalence rate for carrying a weapon was about 15.0% in both the 1999 and 2009 surveys and the prevalence rate for attacking someone with intent to harm was about 7.0% in both surveys. Rates for both outcomes did not differ over time. More boys than girls engaged in these behaviours (Box 2).

Discussion

We found no change over time in the self-reported rates of student violent behaviour based on two indicators, carrying weapons or attacking someone with intent to harm, even after controlling for demographic changes. Our findings suggest that changes in policing and court policies rather than in youth behaviour may explain increases in violence offences.19 Consistent with previous research,7 more boys than girls carried weapons or attacked another with intent to harm.

A strength of our study is that the methods used in the 1999 and 2009 surveys were virtually identical, enabling comparison of rates of behaviour across a 10-year period. However, response rates in the two surveys differed, with a lower response rate in 1999, most likely most likely because that survey required active parental consent. The survey measures, which originated in the United States, have been used in other high-income countries, and had been extensively tested before their use with Victorian students. Although the survey relies on youth self-report, this is considered a reliable data source for behaviours not always visible to adults, such as violence, and the reliability of reporting is unlikely to have changed over the decade.20,21

Violence in adolescence remains an important social and health issue. Yet our study challenges trends in offence data and recent media reports of increasing youth violence by finding no shifts in two self-report indices between 1999 and 2009. Our findings illustrate the need for sound self-report data for planning balanced policy responses and to challenge some negative media portrayals that can create erroneous and damaging stereotypes of young people.22

1 Description of the Victorian secondary school samples in 1999 and 2009

Descriptor	1999 Survey	2009 Survey

School sample
Number of government areas
Metropolitan	31	31
Regional	5	5
Number of schools
In sampling frame*	535	740
Approached	309	357
Recruited	194	189
Type of school
Government	126 (65.0%)	112 (59.3%)
Catholic	41 (21.1%)	40 (21.2%)
Independent	27 (13.9%)	37 (19.6%)
Student sample
Total number in analysed sample	7998	9364
Mean age in years (SE)	14.1 (0.02)	14.4 (0.02)
Male	3622 (45.3%)	4492 (48.0%)
English spoken at home	6500 (81.3%)	6942 (74.1%)
School located in metropolitan area	5811 (72.7%)	7878 (84.1%)
Neighbourhood disadvantage
Most disadvantaged	2007 (25.1%)	2378 (25.4%)
Disadvantaged	2004 (25.1%)	2466 (26.3%)
Advantaged	1982 (24.8%)	2164 (23.1%)
Most advantaged	2005 (25.1%)	2356 (25.2%)

* In both 1999 and 2009, special schools were not included in the sampling frame.

2 Rates of carrying a weapon and attacking someone with intent to harm in 1999 and 2009

	Carrying a weapon				Attacking someone with intent to harm
	1999		2009		1999		2009

	No.	% (95% CI)	No.	% (95% CI)	No.	% (95% CI)	No.	% (95% CI)
Population	7998		9364		7998		9364
Overall sample*	1208	15.1% (14.3%–15.9%)	1396	14.9% (14.2%–15.6%)	598	7.5% (6.9%–8.1%)	685	7.3% (6.8%–7.8%)
Girls	334	7.6% (6.9%–8.4%)	346	7.1% (6.4%–7.8%)	192	4.4% (3.8%–5.0%)	215	4.4% (3.8%–5.0%)
Boys	874	24.1% (22.7%–25.5%)	1050	23.4% (22.1%–24.6%)	406	11.2% (10.2%–12.2%)	470	10.5% (9.9%–11.4%)

* For the overall sample, we found no difference in rates between 1999 and 2009 surveys whether findings were or were not adjusted for demographic characteristics. This was the case for both carrying a weapon (unadjusted odds ratio [OR], 1.0; 95% CI, 0.9–1.1 and adjusted OR, 0.9; 95% CI, 0.8–1.1) and attacking someone with intent to harm (unadjusted OR, 1.0, 95% CI, 0.8–1.1 and adjusted OR, 0.9; 95% CI, 0.8–1.0).

Amanita phalloides poisoning and treatment with silibinin in the Australian Capital Territory and New South Wales

To the Editor: Roberts and colleagues recently reviewed the frequency and clinical outcomes of poisoning with Amanita phalloides (“deathcap”) mushrooms in the Australian Capital Territory and New South Wales.1 Two widely publicised cases of fatal ingestion occurred on 31 December 2011 after a chef had prepared a meal containing wild mushrooms for his colleagues in the kitchen of an ACT restaurant. The link with the restaurant was only discovered after emergency department staff notified public health authorities, who then interviewed unaffected associates of the index case. It was fortunate that no wild mushrooms were used to prepare food for the public. Material at the restaurant was inspected by the public health unit and destroyed to ensure no Amanita mushrooms entered the food supply.

However, the potential for a cluster of poisonings to occur, for people to be in the early stages of toxicity, and for uneaten mushrooms to pose an ongoing risk to food safety should be considered when a sentinel clinical case of poisoning occurs. Because acutely unwell patients may not provide a thorough account of who consumed wild mushrooms, public health units must seek out contacts to clarify the extent of exposure. The ultimate goal of public health programs is to prevent people collecting and eating the deadly A. phalloides mushroom. But, should poisoning occur, a potential public health emergency may arise. ACT Health has now formalised reporting of cases to the relevant public health unit in the protocol for managing cases of poisoning. Public health messaging has also been strengthened and targeted at high-risk groups.

Reinforcing the iodine message for pregnant women in Australia

To the Editor: The recently released clinical practice guidelines
on antenatal care,1 which have been endorsed by the National Health
and Medical Research Council, recommend nutritional supplementation with 500 μg/day
of folic acid, from 12 weeks before conception and for the first trimester, and 150 μg/day of iodine throughout pregnancy. This recommendation recognises that, despite the introduction in 2009 of mandatory fortification of bread with both iodine and folic acid, fortification does not meet the increased needs of pregnant and lactating women. Urinary iodine concentrations of pregnant women improved after the introduction of the iodine fortification program; however, a study in regional New South Wales found that only those women who were taking iodine-containing supplements had urinary concentrations indicating sufficiency (≥ 150 μg/L).2

A series of cross-sectional studies
of women attending a major public antenatal clinic in a regional area of NSW has consistently shown suboptimal dietary practices and knowledge, particularly with regard to the iodine requirements for optimal neurocognitive development of the fetus.2,3 While supplement use has increased from 59% in 2008 to 71%–77% in 2011–2012, only 20% of women in 2008 and 60%–66% in 2011–2012 were taking supplements containing iodine. We do not know why up to 40% of women are still
not taking recommended iodine supplements, but we surmise that women are not being informed about the need for supplementation by their antenatal care providers, despite there being written advice on supplements in the publication from NSW Health, Having a baby, which is provided in public health facilities statewide.4

We acknowledge that data on nutritional supplementation in pregnancy may differ in other states, and that nationally representative data are required, but we are not aware of any state or national campaign to promote iodine supplementation during pregnancy and breastfeeding.

Clearly, better service delivery models are required to ensure consistent messages about nutritional supplements are provided by all health care professionials. New data from Tasmania have shown for the first time in Australia that mild iodine deficiency during pregnancy is associated with reduced educational outcomes in children born to mothers with this deficiency (decreased performance in the NAPLAN [National Assessment Program — Literacy and Numeracy] of up to 10%).5 Similarly, in the United Kingdom, mild maternal iodine deficiency has been shown to result in decreased IQ in the offspring.6 It is indefensible that women of reproductive age, and especially pregnant women, are not being adequately informed about the need for iodine supplementation to prevent irreversible neurodevelopmental effects in their children. The time has come for a national public health education and supplementation program to ensure iodine requirements are met for all, not just
a few, women of childbearing age.

Removing the GST exemption for fresh fruits and vegetables could cost lives

To the Editor: There have been rumours that the current exemption from the goods and services tax (GST) for fresh foods, such as fruits and vegetables, might be removed.1 Changes in government are often accompanied by a review of policies, and broadening the tax base is appealing because it would increase efficiency and raise revenue. But what implications would a rise in the price of fruits and vegetables have for the Australian diet and health?

From estimates of the price elasticity of demand for fruits and vegetables in the United States, we estimate that fruit consumption would decline by 4.9% (95% CI, 2.6%–8.1%) and vegetable consumption by 4.8% (95% CI, 2.6%–7.2%) with removal of the current 10% GST exemption.2 To model lifetime health effects, we used methods we developed for the Assessing Cost-Effectiveness in Prevention (ACE Prevention) project.3 The model mimics the 2003 Australian population in terms of the health-adjusted life-years lived and compares current practice with a scenario in which fruit and vegetable consumption is reduced. Based on available evidence, the model assumes a reduction in fruit and vegetable consumption is associated with an increase in the incidence of ischaemic heart disease (IHD), ischaemic stroke, and cancer of the lung, oesophagus, stomach and colon, leading to increased prevalence and mortality in later years. In the model, health care costs are associated with prevalence for IHD and stroke and with incidence for cancer. Future costs and health loss are discounted at 3% per year.

Using the model, we calculate that adding GST to fruits and vegetables could cost about 100 000 healthy life-years over the lifetime of the 2003 Australian adult population, due to an additional 90 000 cases of IHD, stroke and cancer (Box). This extra disease burden could add a billion dollars in health care costs over the same period.

This suggests that abolishing the GST exemption for fruits and vegetables could have a large detrimental impact on health and health care budgets. Of course, reality is more complex than we have modelled here. For example, other food categories would also be affected by a removal of the GST exemption for fresh foods, leading to complex shifts in diet. But it is very clear that governments should explicitly consider the potential health consequences before making changes to Australia’s tax system.

Results of model estimating the lifetime health effects of adding the goods and services tax to fruits and vegetables

Mean (95% CI)

Healthy life-years*

− 101 379 (− 60 294 to − 145 690)

Health care costs (2003 A$ billion)

1.04 (0.55 to 1.77)

Ischaemic heart disease incident cases

64 495 (23 814 to 108 385)

Stroke incident cases

13 572 (4485 to 24 695)

Cancer incident cases

12 077 (5368 to 20 333)

* Over the lifetime of the 2003 Australian adult population.

Can we avert a diabetes catastrophe in Australia?

Diabetes is likely to cement its place as the fastest growing epidemic in history

At the present time, one person is dying of diabetes every seven seconds, but the news can only talk about victims of hurricanes with houses flying in the air.

Nassim Taleb, Antifragile: things that gain from disorder1

During the past three decades, the number of people with diabetes has more than doubled globally, making it one of the most important public health challenges for all nations.2 During this time, the Medical Journal of Australia (MJA) has had a consistent history of highlighting the “rise and rise” of diabetes as a major public health threat.

In a 1985 MJA editorial, the sparse information on diabetes epidemiology, and its socioeconomic and public health effects on Indigenous and non-Indigenous communities, was highlighted.3 This was in spite of evidence on overseas trends that predicted a future epidemic. Ten years later, in another MJA editorial, it was noted that the predictions were being largely ignored by public health planners.4

In 1999–2000, the first national study of diabetes, the Australian Diabetes, Obesity and Lifestyle Study (AusDiab) was conducted.5 It revealed that about 1 million Australians were affected by diabetes and 60% of adults were overweight or obese.6

In a 2006 MJA editorial, my coauthor and I pointed out that Australia was already in the throes of an unprecedented epidemic of diabetes and obesity.7 In 2007, the federal government announced a 4-year $103 million program for prevention of type 2 diabetes. However, it was poorly designed and the funding stopped in 2011. In 2008, in an address to the Sydney Institute, Prime Minister Kevin Rudd had stated:

If current trends continue, by 2020, diabetes will be the leading cause of disease for men and the second leading cause for women.8

The global diabetes epidemic can be compared with the cholera and typhoid epidemics of the 19th century and the HIV/AIDS epidemic of the 20th century. There are now 370 million people with diabetes worldwide, and this number is predicted to reach 500 million by 2030.9

Testament to the global recognition of this scenario is the landmark and unanimous 2006 United Nations General Assembly resolution which declared diabetes an international public health issue.10 This has been followed by a recent call from the World Health Assembly to reduce avoidable mortality from non-communicable diseases, including diabetes, by 25% by 2025.11

Regrettably, Australia has not seen coordinated action on this epidemic.12 We have not had a national diabetes strategy and action plan for a decade.13 Government attempts at preventing type 2 diabetes have had minimal success, and since the Prime Minister’s warning on diabetes was made, Australia has moved even closer to a “diabetes apocalypse”.

At least 1.5 million people in Australia now have diabetes, and 2 million have prediabetes.12 Consequently, about 30% of Australia’s adult population is directly affected by this important cardiovascular risk factor.

Diabetes strikes at the heart of our Indigenous population — the remote Northern Territory town of Alice Springs is close to becoming the world’s capital of diabetes in terms of prevalence.13 Diabetes is ripping through our Indigenous communities at a frightening pace, causing one of the highest rates of diabetic kidney disease in the world.

We have a frightening national scenario, from the personal individual costs of diabetes through to the costs of medical care and the adverse effects on national productivity.12,14 A recently released report from the Australian Institute of Health and Welfare15 notes that in the 2008–09 financial year, estimated expenditure for diabetes allocated by health care sector totalled over $1507 million, which was 2.3% of all allocated health care expenditure. An additional $153 million was spent on government programs and subsidies, research and gestational diabetes programs. Notably, while expenditure for all diseases increased by 60% between 2000–01 and 2008–09, that for diabetes increased by 86% in the same period. These costs are likely to be an underestimate and would have increased substantially since then.

The June 2013 launch of a new National Diabetes Strategy and Action Plan by Diabetes Australia at the National Press Club was designed to raise awareness of the urgency of addressing the diabetes agenda before the forthcoming federal election. The plan highlights the major issues requiring action and sets out five key goals, all with an emphasis on prevention:

prevent complications — through optimal management and earlier diagnosis
prevent more people from developing type 2 diabetes
reduce the impact of diabetes in pregnancy for women and children
reduce the impact of diabetes on Aboriginal and Torres Strait Islander people
strengthen prevention through knowledge and evidence.14

From a preventive aspect, there is now a focus on early life developmental events, particularly the early life impact on the fetus and the epigenetic effects of risk factors during pregnancy, including hyperglycaemia.2 It is clear that these epigenetic changes can result in intergenerational changes to risk of diabetes, creating a vicious cycle. This is a new target for prevention of type 2 diabetes.

In December 2013, the World Diabetes Congress returns to Australia after 25 years. Over 10 000 delegates are expected from developed and developing nations and they will be looking for answers on how to address this epidemic. Much has changed: we have an enviable record in diabetes research, education and care, despite the failure of governments to face the rolling epidemic head on.

The rapid growth of numbers of people with diabetes is placing an increasing burden on our society and is outpacing the resources that we have to deal with the challenge. Our governments and health authorities must address the epidemic now — the only alternative is a disease burden that could overwhelm the national health budget and damage national productivity. The options are: act now or face the consequences.

In Australia, the alarm bells continue to ring loud, yet authorities continue to ignore the warnings. Almost 20 years ago, my coauthor and I called for the establishment of a national diabetes commission to address the epidemic,4 a recommendation again made in the new National Diabetes Strategy and Action Plan.14 The delay in establishing such a commission is a national disgrace. Again I ask, is anyone in Canberra listening?

Comprehensive primary health care and social determinants as top priorities

Make everyone more equal and good health will follow

Compared with other countries, Australia does very well in terms of health. We rank second in terms of life expectancy, reflecting our level of education, housing and living standards. We have a health system many envy — based on universal public health insurance. And yet, there is room for improvement.

Evidence suggests that more equal societies are healthier.1 An explicit goal of the next Australian federal government, therefore, should be reducing health inequities. To achieve this, vastly improved coordination of primary health care (PHC) services is needed. In addition, the new government’s agenda should prioritise action to improve the social determinants of health.2 By improving community health overall, our continually increasing and economically unsustainable health care costs that are excessively focused on hospital care will be reduced.

The World Health Organization has stated that health systems should be built on the basis of a strong, comprehensive PHC system.3,4 Such a system needs to cure disease, provide rehabilitation, prevent disease and promote health. For the past 20 years, Australian governments have been instituting reforms to achieve this. But the task is made difficult by lack of coordination between fee-for-service general practice, which provides episodic care, and parallel services offered by state and territory governments, such as community health services that provide multidisciplinary curative (eg, diabetes management), rehabilitative (diabetes self-care), preventive (exercise groups and promoting healthy eating) and promotive (lobbying to restrict the sale of high-fat and high-sugar foods) services. The origins of the latter services go back to the Whitlam government’s Community Health Program,5 and, despite uneven development, these services have offered comprehensive PHC services as recommended by the WHO.3,4 The Rudd–Gillard health reforms aimed to produce improved coordination of PHC services through Medicare Locals, created to assume responsibility for regional coordination and planning. It is too early to comment on their effectiveness. However, a new government would do well to build on these structures and to use the federal–state health care agreements to ensure that all states and territories do not abandon disease prevention and health promotion, as Queensland and South Australia appear to have done. The new government should also institute formal trials of community-managed community health centres that enrol general practice patients (rather than offering fee-for-service), offer a range of allied health services, run support groups (for people with chronic disease, mental illness and people coping with issues such as domestic violence) and promote health, including through community development. The existing network of Aboriginal community-controlled health services should be strengthened and new services established with increased funding and more supportive governance models. A well coordinated PHC sector would contribute to reducing health inequities by increasing access to health care for Indigenous people living in remote areas and people living on low incomes.

The second priority of an incoming government should be responding to the social determinants of health. The Senate Community Affairs References Committee has emphasised the importance to Australia of prioritising social determinants.6 The response should be systematic, whole-of-government and based on health in all policies (HiAP).7 An HiAP approach has been trialled in a number of countries and in South Australia.8 Its aim is to hold all government sectors accountable for their health impact and for determining how negative impacts can be minimised and positive ones maximised; for example, urban planning would be done in a way that encourages physical activity and social connectivity, and food supply would be regulated to reduce fat and sugar content in diets. An incoming government should ask the Department of the Prime Minister and Cabinet to take responsibility by ensuring that all departments are conscious of their health impact and take measures to maximise health. This will have a far greater impact on health than health care services are able to have.