Abstract
Introduction
Depression and physical conditions such as gestational diabetes and hypertension frequently co-occur during pregnancy. However, the combined effect of these conditions on postnatal maternal health outcomes remains unclear. This study systematically synthesises evidence on the impact of prenatal depression comorbid with diabetes and/or hypertension on adverse postnatal maternal health outcomes.
Methods
A systematic review was conducted across PubMed/Medline, Scopus, Web of Science, Embase, PsycINFO, and Google Scholar to identify relevant studies. The study protocol is registered with the International Prospective Register of Systematic Reviews (PROSPERO; CRD42024573322). The Joanna Briggs Institute (JBI) quality appraisal tool was used to assess study quality. An inverse variance-weighted random-effects meta-analysis was performed to pool effect estimates. Subgroup analyses, publication bias assessment, trim-and-fill analysis, and sensitivity analyses were conducted.
Results
Eleven studies, including over one million participants, were analysed. Prenatal depression comorbid with diabetes was associated with an increased risk of postpartum depression (Odds Ratio (OR)= 1.80, 95% CI: 1.26–2.59). Additionally, depression comorbid with hypertension during pregnancy was linked to a higher risk of cardiovascular disease (OR = 1.28, 95% CI: 1.14–1.43). Our subgroup and sensitivity analyses confirmed the robustness of the main findings.
Conclusion
Prenatal depression, when co-occurring with gestational diabetes or hypertension, is associated with an elevated risk of postpartum depression (PPD) and cardiovascular disease (CVD). Early screening and integrated care for prenatal depression and pregnancy-related medical conditions may reduce postnatal complications and long-term maternal health risks.

Keywords: Prenatal Depression, Comorbidity, Postnatal Health Outcomes, Systematic Review

Background
People with disability experience profound inequalities in health and wellbeing and are increasingly being recognised as a priority population in public health research. However, since disability information is not routinely collected in Australian administrative data collections, research has been limited to population surveys. Thus, there is urgent need to develop methodologies that accurately identify people with disability in administrative data to advance understanding of disability-related inequalities and progress towards equity.

Methods
We used data from the Person Level Integrated Data Asset (PLIDA), which brings together whole population linked data from various administrative sources, to identify disability items within available linked data collections that align with the functional conceptualisation of disability of the International Classification of Functioning, Disability and Health. We used three algorithms (broad, intermediate, narrow) to derive disability flags and evaluated the sensitivity and specificity of each flag against the Survey of Disability, Ageing and Carers (SDAC), the gold standard for measuring disability in Australia. We demonstrate the utility of the disability flags for examining disability-related inequalities in COVID-19 vaccination.

Results
We used data items from social services payments/benefits, National Disability Insurance Scheme participants, the Medicare Benefit Schedule and the Pharmaceutical Benefit Scheme that met the inclusion criteria. We found that the disability flags were not fully representative of people with disability in Australia. Compared to people with disability in SDAC, people identified using the disability flags were more likely to be of working age, male, have low income, and have severe disability.

Conclusion
Though the disability flags are not appropriate for estimating disability prevalence, linked administrative data can provide insights into disability-related inequalities that cannot be obtained from unlinked administrative or survey data. However, it is important to understand the disability cohort identified (and who is missed) and interpret findings in light of these limitations.

Background: This study aims to systematically synthesise the literature on the associations between longitudinal changes in multiple modifiable risk factors and cardiometabolic outcomes in adulthood. It addresses inconsistencies in defining, measuring, and reporting risk components and assessment frequency that hinder meaningful comparisons across studies and the development of clear guidelines for prevention and management.
Method: The protocol was registered in PROSPERO (ID: CRD42023455203). Ovid MEDLINE, Ovid EMBASE, and Web of Science databases were searched on October 12, 2023. Population-based longitudinal studies that included composite scores of multiple modifiable factors in both childhood and adulthood and adulthood cardiometabolic outcomes were included. Two independent reviewers conducted screening, data extraction, and methodological quality assessment.
Results: Out of 4693 articles identified, 11 met the inclusion criteria; consisting of nine cohorts from four countries. The composite scores mostly included diet (n=5), smoking (n=8), physical activity (n=5), body mass index (n=9), blood glucose (n=7), blood pressure (n=6), and total cholesterol (n=7). An increase in continuous composite scores from childhood to adulthood was inversely associated with adverse vascular structure and function measures (n=7), type 2 diabetes (n=3), cardiac structure and function measures (n=1), as well as fatal and nonfatal cardiovascular events (n=1) in adulthood. Compared to individuals with consistently low risk from childhood to adulthood, both persistently high-risk and worsening groups had a higher risk of adulthood cardiometabolic outcomes. The resolution group showed a similar cardiometabolic risk to the consistently low-risk group, except for type 2 diabetes.
Conclusion: Worsening or persistently poor composite scores of modifiable risk factors from childhood to adulthood were associated with a high risk of adulthood cardiometabolic outcomes. Resolution of poor composite scores may mitigate many of the cardiometabolic risks in adulthood. Future research should prioritize expanding to diverse populations and including new established behavioural factors (e.g., sleep).

Background
Epidemiological research groups are typically comprised of epidemiologists at different levels of career, as well as statisticians, data managers and students. Regardless of preferred statistical analysis package for data management and statistical analysis, users tend to utilize these packages in slightly different manners. Inconsistencies in the use of a particular package might compromise the quality of research conducted. To ensure the accuracy and validity of data management processes and statistical analyses, consistent training and shared tacit knowledge among teams is essential.

Methods
An online Stata eLearning tool was developed to standardise our research organisation’s approach to data management and statistical analysis. Training modules were developed to clarify basic Stata concepts and to promote best-practice methods. Stata-specific material included how to configure settings; windows; project manager; file types, syntax; finding and loading user-written software; writing simple Stata code; best practice methods e.g. labelling variables, creating log files do-file set up. General material covered privacy and data sharing. Content was transformed into an eLearning tool in collaboration with our organisation’s Learning Experience Design team.

Results
A 4-hour eLearning tool was developed which included 11 modules. Several do-files and datasets were provided to be used for a range of practical activities set throughout the course. For the module on writing simple Stata code, users could run the provided do files, step-by-step, with explanations of the output included throughout. Questions were included so users would write their own Stata code which could be compared with the sample solutions. Users could pause the course at any time and pick up from where they left off.

Conclusion
New and experienced users have expanded their Stata knowledge and have been made aware of the importance of best-practice methods. This should improve consistency, reduce errors and enhance transparency of research processes at our organisation.

Introduction: The incidence of early-onset bowel cancer (EOBC) is increasing in Australia and globally. However, the burden of EOBC among First Nations Australians is rarely determined. This study aimed to quantify the diagnosis and survival rates of EOBC among First Nations Peoples in Queensland, Australia.

Methods: CancerCostMod, a linked administrative dataset of patients diagnosed with cancer in Queensland from 1st July 2011 to 30th June 2015, was used. EOBC was defined as a diagnosis of bowel cancer (i.e. colon, rectosigmoid, or rectal cancer) at 18–49 years of age. A multivariable logistic regression analysis was employed to determine the association of Indigenous status and other factors with a diagnosis of EOBC. Five-year survival rates were used to estimate the survival rate.

Results: Of 11,702 bowel cancer cases, 9.2% (95% CI: 8.7%–9.7%) were EOBC, with 19% among First Nations peoples and 9% among Non-First Nations. First Nations Australians had 2.6 times the odds of EOBC diagnosis (95% CI: 1.7–4.0) compared with Non-First Nations Australians. Overall EOBC patients showed a significantly higher 5-year survival rate of 77% compared with 60% for late-onset bowel cancer patients. However, First Nations EOBC patients showed a lower 5-year survival rate (73%) than Non-First Nations EOBC patients (77%).

Conclusion: First Nations Australians have more than double the diagnosis rates and lower 5-year survival for EOBC compared to Non-First Nations. Whilst the recent lowering of the age eligibility for the National Bowel Cancer Screening Program is a beneficial strategy to address the increasing incidence of EOBC, special consideration should be given to addressing the higher diagnosis and lower survival among First Nations Australians. This study raises the potential of further lowering the age eligibility for First Nations Australians to ensure younger First Nations Australians can access screening for the earlier detection thereby improving their survival from bowel cancer.

Abstract
Background: Traditional birth attendance (TBA) remains common in Sub-Saharan Africa (SSA), impacting maternal and neonatal mortality rates. This study aimed at producing high-resolution geospatial estimates and identifying predictors of TBA-assisted childbirth in SSA.
Methods: We used the latest Demographic and Health Survey (DHS) data (2012-2023) from 32 SSA countries. Our sample included 231,189 reproductive-aged women who had given live birth to a child within the past five years. A multilevel binary logistic regression model was employed to identify the predictors of TBA-assisted childbirth, accounting for individual, household, and community-level factors. Geospatial analysis identified geographic hotspot areas where TBA is most prevalent.
Result: The proportion of TBA-assisted childbirth among reproductive-aged women in SSA was 12.43% (95% CI: 10.02%, 14.84%), ranging from 0.3% (South Africa) to 49.4 % (Chad). Hotspot clusters of TBA-assisted childbirth were found in Chad, Ethiopia, Madagascar, Guinea, and Niger. TBA-assisted childbirth was associated with women with community low women literacy (AOR=2.82; 95% CI; 2.57, 3.09), low household wealth status (AOR=1.42; 95% CI; 1.34, 1.49), and residing in rural areas (AOR=2.95; 95% CI; 2.68, 3.24) or had major problems with distance from the health facilities (AOR=1.22; 95% CI; 1.17, 1.26).
Conclusion: Significant geographic variation in TBA-assisted childbirth among women in SSA indicates the need for targeted health interventions to improve access to skill delivery services and empower women through financial and literacy initiatives.

Background: Air pollution is associated with an increased risk of cardiovascular morbidity and mortality. This study aims to systematically synthesise scientific evidence on the effectiveness of pharmacological interventions and dietary supplements (fish oil and vitamin C) in mitigating the cardiovascular adverse effects of air pollution exposure.
Methods: We searched electronic databases including PubMed, MEDLINE, EMBASE, Scopus, Web of Science and reference lists of relevant studies. Population-based observational and interventional studies published in English from 2002 to September 2024 were included. Quality was assessed using the Newcastle-Ottawa Scale for observational studies and the Cochrane risk of bias tool for interventional studies.
Results: The database search identified 1,521 articles, with 28 studies included after full text screening. Among the included studies, 10 were interventional and 18 were observational, with a total sample size of 2,533,434 participants. People with heart failure not taking beta-blockers had a 1.23 (95% CI; 1.05, 1.46) times higher risk of PM2.5 related readmission compared to those taking beta-blockers. Additionally, beta-blockers have beneficial effects in mitigating heart rate variability (HRV) associated with air pollution exposure. Similarly, statin use has shown benefits in reducing inflammation and cardiovascular mortality associated with air pollution exposure. Vitamin C supplementation reduced particulate matter-induced levels of Interleukin-6 (IL-6) by 22% (95% CI: -29%, -14%). Moreover, fish oil supplementation was associated with a protective effect against inflammation and alterations in HRV induced by air pollutants.
Conclusion: Pharmacological interventions, such as statins and beta-blockers, and dietary supplements, including fish oil and vitamin C, have the potential to offer protection against the harmful cardiovascular effects of air pollution.

Background: Most Australian general practice registrars seeking to become specialist GPs train within a profession-led, community-based model organised and administered by the Royal Australian College of General Practitioners. Registrars consult with patients in accredited training facilities under the guidance of dedicated supervisors. While registrars attend regular away-from-practice education, most of their learning occurs in-practice, within this apprenticeship-like model. Over the course of their training, registrars gain competency in managing common clinical conditions and scenarios through exposure to a range of conditions that characterise general practice. Prior to the Registrar Clinical Encounters in Training (ReCEnT) project there was little research into the content of registrars’ consultations despite its importance to quality of training. ReCEnT aims to establish the consultation-based clinical and educational experiences of individual registrars.
Methods: ReCEnT is an ongoing inception cohort study, conducted since 2010, and comprises interrelated research and educational components. ReCEnT collects demographic information about registrars, information about their current training practices, and, once in each of the three training terms, registrars record details of 60 consecutive patient consultations (data on the patient, the consultation, and actions taken (directly linked to the problem/diagnosis prompting the action)) via an online portal. Analysis of data uses standard techniques including linear, logistic, and mixed-effects regression modelling.
Results: Analyses answer research questions that can be categorised as: mapping analyses of clinical exposure; exploratory analyses of associations of clinical exposure; mapping and exploratory analyses of educational actions; mapping and exploratory analyses of more complex constructs; longitudinal ‘within-registrar’ analyses; longitudinal ‘within-program’ analyses; testing efficacy of educational interventions; and analyses of ReCEnT data together with data from other sources (examples included).
Conclusion: Educationally, this epidemiological project generates a formative reflective tool for each participating registrar and informs the GP training program via identification of training needs and translation of subsequent evidence-based educational innovations.

Traditional healing complementary effect on mental health care in Sub-Saharan Africa, A systematic review
Background: In areas with limited and unaffordable biomedical mental health services, such as sub-Saharan Africa (SSA), traditional healers are an incredibly well-used source of mental health care. This systematic review synthesises the available evidence on traditional healing practices, factors to access it, and its effectiveness in improving people's mental health in Sub-Saharan Africa.
Methods: Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) approach was used for the systematic review. Qualitative and quantitative studies published before December 1, 2022 on traditional healing practices to treat mental health problems in Sub-Saharan African countries were included from Pub Med, Medline, CINAHL, and Scopus. Data was extracted using Covidence software and thematically analyzed. The methodological quality of the papers included was evaluated using Joanna Briggs Institute quality appraisal tools.
Results: In total, 51 studies were included for analysis. Traditional healing practices included faith based (spiritual or religious) healing, diviner healing practices, and herbal therapies as complementary to other traditional healing types. Objectively measured studies stated that people's mental health improved through collaborative care of traditional healing and biomedical care services. In addition, other subjectively measured studies revealed the effect of traditional healing in improving the mental health status of people. Human rights abuses occur as a result of some traditional practices, including physical abuse, chaining of the patient, and restriction of food or fasting or starving patients. Individual, social, traditional healers, biomedical health care providers, and health system-related factors were identified to access traditional healing services.
Conclusion: Although there is no conclusive solid evidence to support the effectiveness of traditional healing alone in improving mental health status, studies included in this review indicated that traditional healing and biomedical services collaborative care improves people's mental health.

Administrative health data have revolutionised contemporary epidemiology. Administrative health data allow whole populations to be followed-up from the cradle to the grave. These data are increasingly being used in cutting edge epidemiological research, drug discovery, and to train machine learning models. However, the accuracy of diagnoses in administrative health data is rarely measured or discussed in studies using these data. This is particularly alarming as administrative health data are not collected for research purposes, so its accuracy may be affected by policies affecting the coding and billing practices of health systems in which the data are collected.
We conducted a systematic search of peer-reviewed validation studies which measured the diagnostic accuracy of any health condition in administrative data compared to one or more reference standards.
We identified 277 unique reports met our inclusion criteria. Studies were predominantly conducted using data from the United States and Canada. Most studies data in included studies were collected between 1999 and 2015. Most studies focussed on the accuracy of diagnostic codes for cancers, heart disease, and other non-communicable chronic health conditions. Most studies used hospital data that were validated through medical chart review followed by insurance claims data validated using medical chart reviews. Studies varied in their use of combinations of sensitivity, specificity, positive predictive value, and negative predictive value to assess the accuracy of diagnostic codes.
The heterogeneity of health conditions, measures of accuracy, and data sources precluded meta-analysis, so we narratively synthesised our results.
Our study provides an important overview of the current evidence measuring the accuracy of diagnoses in administrative health data. This will help to inform the quality of evidence generated using administrative health data.

Background and Aims: Historical guidelines endorsed beta-blocker use following cardiovascular events based on clinical trials primarily involving patients with significant left ventricular dysfunction. However, its role in reducing mortality in contemporary populations, particularly those with preserved left ventricular ejection fraction (LVEF), is being re-examined. This study evaluated the effect of beta-blockers on all-cause mortality following cardiovascular events and percutaneous coronary intervention, stratified by LVEF.

Methods: A statewide nested case-control study was conducted using data from the Victorian Cardiac Outcomes Registry. After excluding individuals with in-hospital mortality or missing data, 71,053 patients were analysed. Patients were categorised by LVEF: preserved (≥50%), mildly reduced (45-49%), moderately reduced (35-44%), and severely reduced (<35%). Propensity score matching was used to compare beta-blocker users and non-users, assessing average treatment effects on the treated (ATT).

Results: Among patients with preserved LVEF, beta-blockers were associated with a 1% increased probability of all-cause mortality compared to non-users (ATT: 0.01, 95% CI: 0.002 to 0.02). No significant differences were observed between groups for mildly reduced (ATT: 0.01, 95% CI: -0.02 to 0.04), moderately reduced (ATT: -0.04, 95% CI: -0.09 to 0.01), or severely reduced LVEF (ATT: -0.04, 95% CI: -0.15 to 0.08). Mortality risk was higher with beta-blockers in preserved LVEF patients with prior revascularization, non-acute events, unstable angina, and long-term mortality. Similarly, the mortality risk was higher with beta-blockers in mildly reduced LVEF with unstable angina. On the contrary, beta-blockers reduced mortality in patients with moderately reduced LVEF without prior revascularization, those who had multiple events, or STEMI. No benefit was observed in severely reduced LVEF.

Conclusions: Beta-blockers were associated with increased mortality risk in preserved LVEF and showed no survival benefit in reduced LVEF. These findings highlight the need to re-evaluate beta-blocker therapy in the context of contemporary acute coronary syndrome and heart failure management.

Background:
There is increasing interest in the role of skin microbiome in the development of eczema, which may inform targeted preventive approaches. Results from previous studies have been inconsistent, possibly due to failure to consider different eczema phenotypes and ages of skin microbiome measurement.

Methods:
We undertook a nested case-control study (from control arm of an ongoing PEBBLES clinical trial) of 30 infants with, and 44 infants without eczema, by 12 months of age. Skin microbiome samples were collected from volar forearm at 3 timepoints (1-21 days, 6 weeks, and 12 months of age) and analysed using shotgun metagenomic sequencing. Eczema was assessed at 12 months of age using UK Working Party criteria. Sensitisation was measured using skin prick tests (SPT). Eczema phenotypes were defined as: Eczema Only, Sensitization Only, Eczema with Sensitization, Neither Eczema nor Sensitization [reference group]. Associations were examined by logistic and multinomial regression models, using the exposures of α-diversity and two specific species (S.aureus and S.hominis, a common commensal skin bacterium), with adjustments for possible confounding factors.

Results:
The α-diversity of skin microbiome increased with increasing age. Decreased α-diversity at 6 weeks of age was associated with subsequent increased risk of eczema, particularly Eczema with Sensitization. Presence of S.aureus at 1-21 days, but not at other timepoints, predicted increased risk of eczema (aOR=9.60, 95%CI=1.59-57.9), particularly Eczema Only. While presence of S.hominis at 1-21 days was associated with increased risk of eczema (aOR=3.35, 95%CI=0.87-12.90), its presence at 6 weeks was associated with a reduced risk of eczema (aOR=0.20, 95%CI=0.04-1.12).

Conclusions:
We observed S.aureus colonisation and limited skin microbial diversity in the first 6 weeks of life predicted eczema development. There was some evidence that these associations were both age and eczema phenotype specific. Interventions that increase skin microbial diversity and suppress S.aureus colonisation may help prevent eczema.

Cancer incidence and fatality vary within and between countries. Social disparities, defined as systematic differences between social groups, are a crucial component of the observed differences. Quantifying these disparities in cancer outcomes has been identified as a research priority, however this is challenging as death from other causes is a competing event. In a seminal paper, Young et al (Stat Med. 2020;39(8):1199-236) used a formal counterfactual framework to clarify the interpretation, identifiability, and estimation of common causal estimands in the survival analysis literature in the presence of competing events. In this work, we consider descriptive estimands with competing events using a similar framework. We review the practical value of possible estimands for describing disparities in cancer incidence and fatality in terms of their interpretation, policy relevance, comparability, and identifiability. Specifically, we consider three typical descriptive estimands: 1) cancer incidence or fatality disparities, for example by socioeconomic position, under a hypothetical intervention that would eliminate the competing event; 2) disparities in the presence of the competing event; and 3) disparities among those who would never experience the competing event. We also propose a new estimand based on conceptualising the competing event as a mediator and using the “interventional effects” mediation framework. The proposed estimand represents cancer incidence or fatality disparities, by socioeconomic position, under a hypothetical intervention that would close the socioeconomic gap in the risk of the competing event. We illustrate the concepts in the context of an investigation using data from the Melbourne Collaborative Cohort Study to describe socioeconomic disparities in colorectal cancer fatality.

Objectives
This study quantifies the impact of accelerated non-communicable disease (NCD) mortality decline from 2021 on health system expenditure in OECD countries, compared to business-as-usual (BAU) trends, to achieve the Sustainable Development Goal (SDG 3.4) of a 1/3rd reduction in NCD mortality risk in 30-70-year-olds by 2030.

Methods
Global burden of disease data was used to estimate average annual percentage changes in disease incidence, remission and case fatality rates (CFR) from 1990 to 2021, and extended to 2030 as a BAU scenario in a proportional multistate lifetable (PMSLT). For countries not achieving SDG 3.4 under BAU, three intervention scenarios were specified: (1) prevention (acceleration of incidence rate reductions) (2) treatment (acceleration of CFR reduction and remission rate increase) (3) blended. The PMSLT incorporated country-specific health expenditure data and accounted for competing risks and delayed disease onset to estimate changes in health expenditure for intervention scenarios compared to BAU.

Results
Under BAU, only five countries achieved SDG 3.4. The prevention scenario required an average 3.45 percentage point (pp) acceleration in annual incidence rate declines, while the treatment scenario needed a 2.0 pp acceleration in remission rate increases and CFR decreases.
The prevention scenario produced the largest savings in disease expenditure from 2022 to 2030 (median 0.70%), while the treatment scenario produced the least (median 0.08%) for 30-70-year-olds. When considering all ages (30+ years), the treatment scenario incurred some costs due to increased longevity and competing risks. Extending the time horizon to 2050, savings approximately doubled for treatment and tripled for prevention scenarios, despite potential increases in other age-related conditions.

Conclusion
Reductions in NCD mortality, through either preventive or treatment means, modestly reduce expenditure in the short-term. These savings do not appear to be over-run by deferred and competing diseases in the medium term, highlighting the long-term economic benefits of NCD interventions.

Background: Type 1 diabetes and its related complications significant impact on individuals and society across a wide spectrum. Our objective was to utilise machine learning techniques to predict diabetic ketoacidosis (DKA) and HbA1c>7%.
Methods: Nine different models were implemented and model performance evaluated via the Area under the Curve (AUC). These models were applied to a large multi-centre dataset of 13761 type 1 diabetes individuals prospectively recruited from Australia and New Zealand. Predictive features included a number of clinical demographic and socio-economic measures collected at previous visits.
Results: In our study, 2.9% reported at least one episode of DKA since their last clinic visit. A number of features were significantly associated with DKA. Our results showed that Deep Learning (DL) model performed well in predicting DKA with an AUC of 0.887. The DL also provided the lowest classification error rate of 0.9%, highest sensitivity of 99.9% and F-measure of 99.6%. As for HbA1c >7%, the optimal Support Vector Machine provided a good AUC of 0.884.
Conclusion:
Machine learning models can be effectively implemented on real-life large clinical datasets and they perform well in terms of identifying individuals with type 1 diabetes at risk of adverse outcomes.

Background: Non-communicable disease screening programs are a key public health activity to improve primary prevention. However, there are few examples of successful screening programs at scale and participation in existing screening programs broadly, is known to be suboptimal. This systematic review aimed to identify interventions associated with increased participation in non-communicable disease risk factor screening among adults.

Methods: Using three online databases, a systematic search of English language peer-reviewed journal articles was performed. Articles quantitatively examining person-centric interventions to increase non-communicable disease screening among adults without existing disease diagnosis were eligible.

Results: Sixty-four studies spanning 23 countries and over 569,741 observations were included. Non-communicable disease screening outcomes of breast cancer (n=14), bowel/colorectal cancer (n=31), cardiovascular disease (n=1), cervical cancer (n=14), heart health check [diabetes/hypertension] (n=1), health check (n=2), lung cancer (n=2), and melanoma (n=1) were observed. Six categories of intervention, comprised of 23 differing approaches, were determined for the promotion of non-communicable disease screening: invitation, education, nudge, navigation, and self-affirmation interventions. Of these interventions, the provision of patient navigator support, telephone-based promotion, written invitations to screen, and face-to-face/workplace education were the most consistently associated with greater screening engagement.

Conclusion: This systematic review is the first to detail screening behaviour interventions spanning multiple research disciplines and a range of non-communicable diseases. Four intervention methods were identified to be consistently associated with greater engagement non-communicable disease screening among adults.
Many studies presented stand-alone interventional techniques; this siloed approach may limit interventional effect compared to a multi-pronged approach. To maximise likelihood of effectiveness, future interventions to increase non-communicable disease screening should consider combined approaches utilising the consistently effective interventions identified in this review.

The introduction of generative AI tools such as ChatGPT, DALL-E and Microsoft Copilot changed the way we teach and learn. With functionality to draft programming code, undertake complex calculations and to write paragraphs, genAI has revolutionised the way students learn, and prompted new approaches to teaching and assessment.

The Undergraduate course "Translating Health Information" requires our students to communicate health data to different audiences in different ways. One assessment relates to the visual representation of published data, in which students use software (e.g Piktochart, Canva, Tableau) to convert results from nationally representative surveys into infographics, billboards, or interactive data stories.

In 2024, this assessment was re-framed, giving students license to use two different generative AI tools to create a billboard or infographic using data from nationally representative surveys. Students were required to critique their output from the generative AIs in the context of visual design theories taught in class.

The genAI outputs varied considerably in terms of quality, accuracy, and the representation of data and ethnic diversity. At a distance, genAI visualisations looked broadly consistent with infographics from previous years, however on closer inspection the graphs or words were incomplete or incoherent. Some packages had an inherent bias in their portrayal of deprived communities or ethnic groups.

In this presentation, I provide an overview of the workflow, an assessment of the genAI tools chosen by the students, with example infographics and billboards from various genAI tools students used, and describe the marking rubric developed. This successful pilot will be extended to obtain more detailed feedback on the learning experience of the students, and their understanding of the strengths and limitations of genAI tools in this setting.

Introduction/Background
COVID-19 exacerbated many existing health inequities. It is important to understand if long COVID, the most enduring impact of the pandemic, will have similar inequitable effects. This study explores the influence area deprivation has on patient-reported symptom scales and health-related quality-of-life (HRQoL) of individuals with long COVID.

Method
Mātauranga Raranga | Long COVID Registry Aotearoa was designed in partnership with those with lived experience of long COVID. Comprehensive data are collected from participants in the registry, including the Index of Multiple Deprivation (IMD2018) area deprivation score and domain ranks, and information on symptoms, their severity using symptom scales, self-reported health (SRH) and HRQoL using the EQ-5D-5L. Participants have been followed up monthly using the EQ-5D-5L and 6-monthly for the symptom scales. Analyses describe the symptoms/self-reports and explore differences by deprivation Quintile. We additionally explore variation using the Range and Extremal Quotient to measure the absolute and relative levels of inequities respectively.

Results/Findings
The registry remains open and data collection is ongoing, analyses of collected data as of the end of March 2024 have formed the bases of this presentation (8.6% Māori; 91.4% non-Māori). Nationally, each IMD2018 Quintile comprises 20% of Aotearoa’s population. 45% of non-Māori respondents lived in less deprived Quintile 1(Q1) and Q2 areas. The deprivation pattern for Māori respondents is variable, with fewer Q1 (15%) participants, but an overrepresentation in both Q2 and Q5 (25% each). Compared to participants living in Q1 areas, participants in more deprived areas (Q4+Q5) were statistically more likely to have poorer symptoms, particularly for breathlessness, depression, anxiety, psychological distress, fatigue, and stigma. However, there was little or no difference in HRQoL.

Implications/Key Message
Area deprivation reveals variations in the long COVID burden of our participants. Further research is needed to understand this variation, in order to inform service provision and policy.

Abstract:
Background:
Pacific peoples in New Zealand (NZ) have the highest rates of type 2 diabetes (T2DM) compared to other population groups. There is a paucity in the design and implementation of culturally tailored approaches in NZ focused on targeting major long-term conditions, and particularly in reducing the T2DM epidemic among high risk population groups. This study implements a culturally adapted diabetes prevention programme (Sia Tō’onga Mo’ui Diabetes Prevention Programme (STM-DPP)) for use amongst Tongans living in NZ.


Methods: The STM-DPP was run for 6 months (June 2021 to November 2021), where participants attended weekly health educational zoom sessions, and were encouraged to do moderate exercise for 30min or more every day. Data from participants collected at baseline and post-survey, include the clinical markers such as blood pressure, anthropometric measurements, physical functioning based on the six-minute walk test, pre-diabetes risk test, and participants diet over a seven day period.

Results: Forty one Tongan participants (males = 25, females =16) were recruited to the STM-DPP, of which 95.2% had BMI≥30.0 and 75.6% with Prediabetes Risk Test (PRT) score of baseline PRT≥5, indicating a high risk. At the conclusion of the intervention, post-programme data revealed significant (p<0.0001) changes across all areas for participants in weight; waist; waist to height ratio; physical activity capability; and systolic blood pressure. Using the exploratory factor analysis, it found participants mainly ate two foods; first factor from meat, poultry and fish, fruits, and vegetables; and second factor was junk foods, sweets, and sweet snacks, savory, and take away foods.

Conclusion: These results provide evidenced-based research affirming that the culturally adapted T2DM prevention programme, STM-DPP, has a positive impact on reducing the risk of diabetes among the Tongan population in NZ, and could be made available and adapted to other Pacific populations across NZ.

Background: There is a need to understand better the link between ‘culture and food systems’, and ‘diet-related conditions’ (e.g., diabetes), particularly from an Indigenous Pacific worldview. In this study, we explored the perspectives of Samoan traditional healers about their insights on how Indigenous culture and food systems have impacted on modern-day diet related conditions.

Methods: Fourteen traditional healers from New Zealand and Samoa participated in semi-structured narrative interviews, from June 2022 to October 2023. Transcribed interviews were coded and analysed using thematic analysis.

Results: Three major themes were identified: (1) highlighting the generational knowledge and practices of the healers within the health, (2) treatment using holistic health approach versus disease pathology, and (3) the modernity of the societal shifts that have impacted on health and wellbeing.

Conclusion: These insights have delineated the role of Indigenous Pacific healers and why Pacific peoples continue to opt for services in modern day society as the first port of call to address their health needs. Further, Indigenous healers knowledge and practices can provide a potentially alternative model of care in modern-day community context that may be better suited for Indigenous peoples.

Background: Meta-analyses are fundamental in synthesising evidence across studies in epidemiology. However, determining when enough reliable evidence has been gathered is challenging. The aim of this study is to present an R Shiny app that offers a complementary and user-friendly tool to assess the sufficiency (whether enough evidence exists to establish an association) and stability (whether new evidence is likely to alter the conclusion) of evidence in meta-analyses.
Methods: The app uses cumulative meta-analysis to compute the relative risk (RR) convergence over time and identifies the stability threshold— the minimum or maximum RR that a new study would need to report to shift the pooled relative risk (RRp) of the meta-analysis to a null effect (RRp = 1). Users only upload data including Author/Year, effect sizes (e.g., RR, OR), and confidence intervals in CSV format. The app then automatically displays forest plot, cumulative meta-analysis (both random and fixed models) and the stability threshold, displaying results through interactive plots.
Results:
Applying the app to the PM2.5 and preterm birth case study, an association with notable heterogeneity, showed that the cumulative RR converged over time, and the stability threshold was identified. By 2015, the pooled RR for PM2.5 exposure and preterm birth stabilised, indicating that additional studies were unlikely to change the existing evidence. The app provided specific thresholds for future studies to shift the conclusion to a null association (RR 0.93, 95% CI limit 1.02) underlining its utility in preventing redundant research efforts.
Conclusion: This R Shiny app offers a powerful and complementary tool for systematically assessing when meta-analyses have reached sufficiency and stability, thereby reducing research waste. By focusing future efforts on understanding risk, addressing biases, and developing interventions, the app has the potential to streamline the meta-analysis process, improve evidence synthesis, and guide research agendas.

Background
New pharmacological treatment options are increasingly introduced to the market for patients with existing disease. However, patients who switch to new treatment options from existing treatment regimens are traditionally excluded from observational studies to avoid ‘prevalent user bias’.
The prevalent new user design (PNUD) has been developed as a novel pharmacoepidemiologic method that allows for the inclusion of prevalent patients by employing time-conditional propensity scores to match prevalent new users (treatment switchers) to comparator patients (those with similar prior treatment that did not switch to the new medicine). To date, simple proxies of disease severity such as time on prior treatment or number of previous prescriptions dispensed have been used to match patients, however, it is unknown which strategy is most appropriate for different and potentially more complex clinical scenarios.
Aim: To validate the use of the PNUD for medicines safety surveillance in complex clinical scenarios.

Methods
Dispensing claims were simulated for a patient cohort where a new medicine is introduced as an alternative to existing treatments. Different confounding scenarios were induced based on prior treatment; for example, those who switch to the new medicine are sicker than those who do not and therefore have an increased risk of experiencing an outcome.
The impact of different matching strategies for selecting comparators for each prevalent new user on treatment effect estimates was assessed using measures of bias, precision (empirical and model SE), and coverage.

Results
Matching strategies considering both time on prior treatment and number of previous prescriptions outperformed those focusing only on time in terms of bias reduction, however precision decreased due to a smaller sample size from more restrictive matching.

Conclusion
New approaches to post-market surveillance of new medicines as they enter the market are critical to ensure that medicines safety issues are identified earlier, and harms avoided.

Abstract:
Background:
Pacific peoples have traditionally relied on fresh, local foods deeply connected to cultural and spiritual practices. Over time, globalisation, colonisation, and environmental changes have significantly altered dietary patterns across the South Pacific, contributing to rising rates of diet-related diseases such as diabetes and cardiovascular conditions. This research explores how global, political, cultural, spiritual, and environmental factors have shaped Pacific dietary habits and population health from the 1800s to the present day.

Methods:
Using a historical timeline, we focused on colonial policies, global food systems, migration, environmental disruptions, and the marginalisation of traditional food knowledge. Historical photographs are included in the presentation to visually illustrate these shifts.

Results:
The findings reveal how colonial rule, trade policies, and global market integration led to the widespread adoption of imported, processed foods. Environmental disruptions, including cyclones and land-use changes, further affected local food production. Community efforts to revive traditional food systems show promise in addressing these health challenges.

Conclusion:
This research highlights the importance of contextualising epidemiological methods to better address the health needs of Pacific peoples.

Dengue fever remains a pressing public health concern in Indonesia, with substantial spatial and temporal variability across its 34 provinces. This study aims to deepen the understanding of dengue transmission dynamics by distinguishing between fixed and random effects using an advanced hierarchical modeling framework. Leveraging 11 years of Indonesian surveillance data, we applied a Bayesian hierarchical model and employed Markov Chain Monte Carlo (MCMC) methods to estimate both case incidence and mortality associated with dengue.

To enhance model robustness, we explored several prior distribution scenarios, adjusting for different assumptions regarding mean and variance. This approach allowed us to assess the sensitivity of the model outcomes to prior selection and refine our understanding of parameter behavior. Posterior trace analysis was conducted to evaluate model convergence and stability, ensuring reliable parameter estimation. Additionally, Bayesian credible intervals were calculated to quantify the uncertainty of estimates, providing a probabilistic range for case and mortality predictions.

Our findings reveal significant spatial heterogeneity in both incidence and mortality rates, highlighting the influence of province-specific factors alongside broader temporal trends. This framework offers a nuanced understanding of dengue transmission patterns and serves as a valuable tool for informing region-specific public health interventions and policy decisions aimed at controlling dengue in Indonesia.

Background: Many people with MS (PwMS) experience multiple symptoms, often simultaneously. This makes understanding how symptoms present as a ‘bigger picture’ challenging. We aimed to categorise PwMS into profiles using intensity scores for 14 MS symptoms.
Methods: 13 common MS symptoms were assessed using MS Symptom Scores (range 0-10, higher=worse); sleep quality was assessed using the Pittsburgh Sleep Quality Index (1 symptom, rescaled 0-10). Cross-sectional data were analysed using latent profile analysis. Symptom profiles were validated against AQoL-8D health-related quality of life (HRQoL) health state utility, and super- and individual-dimensional scores (range 0-1) using linear regression.
Results: Latent profile analyses suggested four symptom profiles: ‘low’ (12%), ‘moderate’ (20%), ‘high’ (52%), and ‘mixed’ (16%). The latter profile constituted PwMS with moderate-high symptom intensity, but minimal anxiety/depression. Profiles had different demographic and clinical characteristics, indicating underlying heterogeneity. Greater symptom intensity was associated with lower HRQoL in a dose response manner. However, PwMS without depression and anxiety reported relatively high HRQoL despite high physical symptom intensity. Sleep may be an independent silent symptom of MS.
Conclusion: Symptom profiles had complex dose-response relationships with HRQoL with sleep having important independent effects. Reducing anxiety and depression and improving sleep may improve HRQoL even in those with high symptom intensity. Profiles have distinct demographic and clinical characteristics, suggesting drivers of symptom profiles are complex and include factors such as relapses and obesity. This suggests that using symptom profiles to categorise symptoms in PwMS. In particular, understanding why some individuals have minimal anxiety / depression in spite of moderate-high physical symptoms will inform future approaches to managing MS symptoms.

Background
Artificial intelligence (AI) tools in epidemiology and health data science have the potential to change how we work. However, the extent of AI adoption and the need for AI literacy to be incorporated into university curricula is unclear. Hence, the aim of this study was to investigate (1) AI utilisation in these fields and (2) employer demand for AI literate graduates.

Methods
Industry partners participated in 1-on-1 interviews focusing on the use and acceptability of AI tools in their workplaces. Recent University of Melbourne graduates with degrees in epidemiology and/or biostatistics were invited to complete an online survey on AI tools in their workplace.

Results
As of January 2025, six of the 20 invited industry partners had participated in interviews. Preliminary themes that emerged include (1) enthusiasm for the use of AI to enhance efficiency and (2) the need for safe and ethical guidelines for AI usage. All reported that AI usage was in its early stages and that they expected demand for AI literacy to increase.

The survey was sent to 412 recent graduates, and at a preliminary point 27 individuals had completed the survey. Of these, 11 reported commonly using AI tools, primarily for writing assistance, coding, and data visualisation. Fourteen of the remaining 16 respondents who did not commonly use AI tools were open to its adoption but expressed uncertainty on implementation. All respondents emphasised the importance of responsible AI use.

Conclusion
Initial analyses indicate that epidemiological and health data science employers and graduates viewed the use of AI positively; however, a lack of guidelines on its use was a barrier. Incorporating AI tools into teaching may better prepare graduates for future workplace demands but should be accompanied by guidance on responsible use.

Background
C-reactive protein (CRP) is a widely used marker of chronic low-grade inflammation. Several methylation-based markers of CRP (mCRP) have been proposed as alternatives to measured CRP and used for disease risk stratification and ageing marker development. We aimed to i) compare the validity of four mCRP markers, ii) assess their associations with three health-related traits, iii) compare their stability over a decade.
Methods
Blood samples were collected from 947 participants (34% females) in the Melbourne Collaborative Cohort Study at baseline (1990-1994) and follow-up (2003-2007). CRP was assayed from plasma samples and log-transformed for analyses. Genome-wide DNA methylation data (HM450k) were used to calculate mCRP-Wielscher (68 CpGs), mCRP-EpiScore (100 CpGs), mCRP-GrimAge2 (132 CpGs) and mCRP-Hillary (1468 CpGs). Explained variance by mCRP was estimated. Associations of follow-up CRP markers with mortality (17-year follow-up, Ndeaths=319) were assessed using Cox models, and linear regression for age-adjusted PCGrimAge, and body mass index (BMI). Intraclass correlation coefficients (ICC) for CRP markers at baseline and follow-up were calculated to assess temporal stability.
Results
mCRP explained 5.8% (mCRP-Wielscher) to 15.2% (mCRP-Hillary) of the variance in plasma CRP. Compared with plasma CRP, mCRP had stronger associations with PCGrimAge, whereas for mortality and BMI, the associations were similar or weaker for mCRP. Associations between mCRP and mortality greatly attenuated after accounting for PCGrimAge, which was not observed for CRP. The association of plasma CRP with mortality remained strong after adjusting for mCRP markers. ICCs were greater for mCRP-Wielscher (0.67) and mCRP-EpiScore (0.64) than for plasma CRP (0.49) or the other mCRP markers.
Conclusion
The results were largely marker-dependent and mCRP did not consistently outperform plasma CRP in their associations with health-related traits or temporal stability as claimed in previous studies. Clarification of mCRP marker composition will allow a better understanding of the interplay/overlap between age-related inflammation and methylation processes.

Background:
Causal inference in observational epidemiology is challenged by confounding and selection biases. Directed Acyclic Graphs (DAGs) help identify adjustment strategies, but are non-parametric and do not distinguish effect size or direction (risk vs protective). This project extends the traditional DAG through developing a web-based application that allows users to parameterise a DAG with assumed direction and magnitude of effects. Aligned with the conference theme "embracing change", the tool bridges traditional epidemiology with a convenient digital interface, enabling researchers to test assumptions and quantify biases.

Methods:
We developed an interactive web application named DAGknife, using the R Shiny Golem framework. Users can choose from various pre-specified DAGs representing common scenarios (e.g. basic confounding, mediation, collider bias etc). Users parameterize the selected DAG by specifying the magnitude and direction of each causal arrow, then DAGknife simulates data and enables bias comparisons under different assumptions or adjustment strategies. The tool reports standard metrics (e.g., percent bias, MSE) and provides the underlying R code for reproducibility and extension.

Results:
DAGknife allows researchers to (1) simulate data from user-defined DAGs, (2) quantify bias across analytical approaches, and (3) assess unmeasured confounding or misclassification. Testing with case studies (e.g., collider bias in survival analyses) shows how the DAGknife reveals the scale and direction of biases under different assumptions. Initial validation shows robust performance for complex DAGs, with iterative refinements improving usability.

Conclusion:
DAGknife advances epidemiological methods by integrating traditional DAG theory with the flexibility to quickly parameterise DAGs and run DAG-based quantitative bias analyses, empowering researchers to critically evaluate the potential magnitude and direction of biases in their observational analyses. DAGknife supports capacity building by democratizing complex bias analysis, reducing reliance on custom coding, which aligns with evolving trends in digital transformation and methodological innovation for more credible epidemiological evidence.

Key learnings from Tokelauan youth on the Health and Wellbeing Benefits of Fagatua the indigenous wrestling of Tokelau


Ilai Elekana Manū (PhD Candidate)

Abstract:
Background:
Individuals of multi-ethnic Pacific/Other descent (i.e., belonging to two or more ethnic groups) have a high incidence of mental illness, as well as a markedly elevated estimated lifetime prevalence, attributed to a considerably diminished sense of cultural connection.
Tokelauans possess a significant multi-ethnic and New Zealand-born demographic. This study examined the viewpoints of Tokelauan youth to provide insights into the comprehensive health and wellbeing advantages. Fagatua, the indigenous wrestling of Tokelau

Methods: In a mixed methods approach, 10 Tokelauan elders from around NZ, Australia and Tokelau participated in semi-structured narrative interviews, 100 Tokelauans from participated in a Qualitative and Quantitative survey, 3 groups made up of Community Leaders, Health Teams and Youth Co-design a health programme based around fagatua and 10 youth participated in that programme as well as a Talanoa.
Transcribed Talanoa were coded and analysed using thematic analysis.

Results: Three major themes were identified: (1) The physical nature of fagatua is unique and offers a unique way to connect to the culture, (2) engaging in the fagatua programme allowed for social interaction and an opportunity to discuss mental health (3) the spiritual health benefits of incorporating prayer, learning about Tokelau mythology and interacting with elders in the community were beneficial

Conclusion: Youth shared that they believed fagatua had holistic benefits, emphasising mental, physical, spiritual and social health and well-being benefits.

Keywords: Indigenous Health, Mental Health, Indigenous Practices

Background
Recent epidemiological studies have found associations between air pollution exposure during early childhood and neurological conditions in children. Our study aims to investigate the possible link between the exposure to annual concentrations of particulate matter less than 2.5 microns in diameter (PM2.5) while in utero and early childhood and attention-deficit/hyperactivity disorder (ADHD) medication prescribing as a proxy for ADHD diagnosis in Victoria, Australia.
Methods
Our study included children born in Victoria from March 2012 to December 2015. PM2.5 concentrations were estimated using satellite-based land-use regression models to assign in utero and early childhood exposure at the Australian Bureau of Statistics (ABS) Statistical Area 2 (SA2) level based on the residential address at birth. Data linkage was conducted by the Centre for Victorian Data Linkage (CVDL) with support from the Australian Institute of Health and Welfare (AIHW) to match birth records with the PBS data. Lifetime prescription of ADHD medication was defined by PBS prescribing section N06B, which is psychostimulants agents used for ADHD and nootropics.
We are currently analyzing the data on exposure to annual PM2.5 concentrations and lifetime ADHD prescription status by sociodemographic characteristics using chi-squared for categorical factors and t-tests for continuous factors. Survival analysis will be used to control for confounding factors. All statistical analysis is being conducted using R 4.4.1.
Results
While the analysis is currently ongoing, we plan to present sociodemographic characteristics, such as biological sex and socioeconomic status, by lifetime ADHD prescription status. Tables and figures will be presented to show our study sample and findings. We expect to have results available in a few months pending approvals from the CVDL and AIHW.
Conclusion
Our study will add to the growing body of research on particulate matter exposure and neurological conditions.

Background: Rent regulation (e.g. 'rent caps') is a topic of policy debate in many countries, as a possible lever for addressing a lack of affordable housing. In Australia, the amount by which rental providers can increase rent is only very lightly regulated, putting pressure on household budgets and often forcing households to relocate. Unaffordable and insecure housing has been linked to poor mental health, but the mental health cost of imposed rent increases specifically has not been established. Further, any such mental health effects may be distributed unevenly.

Methods: For low-income households in the private rental sector, we investigated the relationship between rent increases (none, <10%, or ≥10% within a year) and mental health (MHI-5 score) in 20-64-year-old participants in the Household Income & Labour Dynamics in Australia survey, between 2001 and 2019. Analyses were restricted to non-movers to isolate landlord-imposed rent increases. We used longitudinal fixed effects regression models and tested for potential effect modification by age, pre-existing rental affordability stress and markers of potential marginalisation in the rental market.

Results: In any given year 12% of our sample was subjected to a rent increase of ≥10%. Compared to years when their rent remained stable, people’s mental health was worse following a rent increase, though only among renters aged 20-44 years (MHI-5 score 1.6 point points lower after a rent increase of at least 10% (95%CI:-0.3,-2.9)). As well as age, we observed evidence of effect modification by pre-existing rental affordability stress and some markers of marginalisation in the rental market.

Conclusions: When subjected to a substantial rent increase, younger private renters in low-income households experience a decline in mental health. Debate about rent caps should consider the potential protective impact on mental health.

Introduction
Pneumococcal nasopharyngeal carriage studies provide critical insights into serotype dynamics, especially in settings where disease surveillance is limited. Pneumococcal conjugate vaccines (PCVs), including the 10-valent (PCV10) and 13-valent (PCV13) formulations, target common disease-causing serotypes. This global meta-analysis evaluates changes in vaccine serotype (VT) and non-vaccine serotype (NVT) carriage post PCV10 and PCV13 introduction.
Methods
We conducted a systematic review and requested individual participant data from authors of included studies. Meta-analysis included pre- and/or post-PCV10/13 introduction carriage prevalence in children <2 years. Studies published until 31/12/2020 were included. Using mixed-effects logistic regression with restricted cubic splines, we modelled log-odds of carriage as a function of time since PCV10/13 introduction compared to pre-PCV rate, incorporating study-level random effects. Analyses were stratified by WHO region.
Results
We identified 381 eligible studies; 63 studies provided individual-level data from 59,230 children <2 years. Data from PCV10-using countries were limited primarily to WHO Africa region. Post-PCV10, the odds of VT carriage declined (OR 0.08, 95%CI: 0.05–0.14) and odds of NVT carriage increased (OR 6.01, 95%CI: 3.95–9.15). Total carriage remained stable for up to four years. Post-PCV13 introduction carriage data was obtained from all WHO regions. Seven years post-PCV13 introduction, reduction in VT carriage was 71% (OR 0.29 (0.21-0.39)) in the African region, 86% (OR 0.14 (0.03-0.57)) in the Americas and 53% (OR 0.47 (0.32-0.69) in the European region. Despite only three years data in the Western Pacific, reduction in VT carriage was 75% (OR 0.25 (0.17-0.36)). NVT carriage slightly declined in all regions, except for the African region (increase).
Conclusion
VT carriage declined substantially after PCV10/13 introduction across all regions. Total carriage stayed stable in all regions except the African region, where it increased. This underscores the need for ongoing surveillance, with changing vaccination schedules and considering extended valency vaccine formulations.

Background: With Australia's ageing population, fall incidence is expected to rise, posing significant public health challenges. The proportion of Australians aged ≥65 years is projected to increase from 15% in 2017 to 22% by 2057. Understanding fall trends and their determinants is critical for developing targeted prevention strategies. This study examines longitudinal trends in falls and associated risk factors using repeated follow-up data from a population-based cohort.
Methods: Participants were drawn from the Geelong Osteoporosis Study (GOS). Falls were self-reported for the previous 12 months and age-standardized to the Australian population. Additional data included self-reported information on fractures, medications, comorbidities, alcohol consumption, and smoking, alongside measured anthropometrics, muscle strength, biochemical markers, and imaging results. A multivariable Generalized Estimating Equation (GEE) model was used to identify fall determinants, reporting adjusted odds ratios (AORs) with 95% confidence intervals (CIs).
Results: Among men, age-adjusted fall prevalence declined over 15 years. In contrast, women initially had a 4.2% decrease, followed by a slight 0.6% rise. Each additional year of age increased the likelihood of falling by 1% (AOR = 1.01, 95% CI: 1.00–1.02). Women had a 52% higher risk of falls compared to men (AOR = 1.52, 95% CI: 1.22–1.88). Diabetes was associated with a 69% increased risk (AOR = 1.69, 95% CI: 1.23–2.31). Greater hip flexion strength was protective, with each 1 N/kg increase reducing fall risk by 3% (AOR = 0.97, 95% CI: 0.95–0.99).
Conclusion: Men showed a steady decline in falls over time, while women exhibited a nonlinear pattern, with an initial decline followed by a slight rise. Key predictors included age, sex, diabetes, and hip flexion strength, highlighting the need for sex-specific fall prevention strategies. Future policies should integrate age-adapted strength training, diabetes management, and continuous fall surveillance, particularly among older adults and women, to mitigate the evolving fall burden.

Backrgound: Tongue cancer is a growing health issue in Australia, with the tongue being the most common site for oral cancer worldwide, particularly in men. This study aims to explore the trends of tongue cancer in Australia, focusing on gender differences across age, period, and cohort factors.

Methods: The data used comes from the 2017 version of the Australian Cancer Database (ACD), published by the Australian Institute of Health and Welfare (AIHW). It focuses on tongue cancer cases from 1982 to 2017, using repeated cross-sectional data for individuals aged 25 to over 85. We performed an age-period-cohort analysis with canonical parameterization to address identification issues in age, period, and cohort, and modeled it using Poisson regression. The analysis includes the incidence and mortality counts and rates of tongue cancer.

Results:
The total sample included 1,064 participants, evenly divided between males and females. The highest incidence rate for males was in the 65-69 age group, with an IR of 1.47 (95% CI 1.42, 1.53), and for females, it was in the >80 age group, with an IR of 0.94 (95% CI 0.77, 1.11). Mortality rates were highest in the >90 age group for both males (MR 0.66, 95% CI 0.53, 0.80) and females (MR 0.71, 95% CI 0.61, 0.79). Based on the APC model, we found that the main non-linear factors affecting tongue cancer are related to age and cohort in the general population. For incidence, age and period-related effects were observed among females, while for mortality, age and cohort-related effects were found among females.

Conclusion: This suggests that the overall trend of tongue cancer in Australia shows an increasing incidence with older age groups. However, in males, it can occur in any age, period or cohort group. This emphasizes the need for gender- and age-specific targeted screening interventions.

Background
In late 2012, Fiji introduced the ten-valent pneumococcal conjugate vaccine (PCV10) as a three‐dose (3+0) schedule. Three doses of pneumococcal conjugate vaccine (PCV) is effective, but very costly. To inform recommendations of reduced-dose PCV schedules, the WHO is reviewing evidence comparing the efficacy and effectiveness of two (1+1) and three‐dose schedules. The Fijian government is considering a 1+1 PCV schedule. We modelled the projected impact of switching the PCV schedule in Fiji from 3+0 to 1+1 on vaccine-type carriage and invasive pneumococcal disease (IPD).
Methods
We developed and calibrated a deterministic transmission model, fitted using Fijian pre- and post-PCV10 pneumococcal carriage (2012–2015) and IPD (2012–2023) data. We assumed the 3+0 schedule, given at 6, 10, and 14 weeks, had 95% coverage, and vaccine efficacy of 56% against vaccine-type carriage and 80% against vaccine-type IPD. We assumed the 1+1 schedule, given at 14 weeks and 9 months, had 83% coverage, and vaccine efficacy of 49% against vaccine-type carriage and 70% against vaccine-type IPD. We modelled the 3+0 schedule from 2012–2034 and compared vaccine-type carriage and IPD with 3+0 (2012–2024) switching to 1+1 (mid-2024–2034).
Results
Model projections suggest a continued 3+0 schedule in Fiji would suppress vaccine-type carriage in children under five to 0.003% (95% credible interval (CrI) 0.000–0.016%) by 2028; switching to 1+1 would achieve similar levels (0.005% [95% CrI 0.000–0.026%]) by 2034. Compared with 3+0, a 1+1 schedule is projected to add 32 extra IPD cases by 2034 (577 [95% CrI 475–684] vs. 545 [95% CrI 444–649]), most in children aged 2–10 years.
Conclusions
By 2034, switching from 3+0 to 1+1 in Fiji achieves similar, though delayed, vaccine-type carriage suppression, and 32 extra IPD cases. For any schedule changes, continued surveillance is needed as there is inherent uncertainty around model outputs.

Background: Attention-deficit/hyperactivity disorder (ADHD) may increase the risk of criminal behaviour due to deficits in attention and impulse control. This study examines the risk of first and repeat criminal offending among children and young adults with ADHD and whether the risk is reduced by treatment with stimulant medications.
Methods: A population-based data-linkage cohort study of individuals born in New South Wales, Australia between January 1990 and December 2005 and followed until May 2016. Individuals with ADHD were frequency matched on age, sex, and postcode at birth to controls (1:10) without ADHD. Treatment status was identified from stimulant authorizations for ADHD and proven criminal offenses were identified using court records. Analyses were stratified by sex and age period (10-17 years and 18-25 years) and first and repeat offenses were examined separately using modified Cox regression and Prentice-Williams-Peterson models, respectively.
Results: The study cohort included 75,650 individuals with ADHD (147,855 treated and 525,595 untreated person years) and 745,634 non-ADHD controls. The risk of a first offense was increased among cases compared with controls and reduced by treatment (Males 10-17 years, untreated Hazard Ratio (HR) 2.02 95%CI 1.95-2.10; treated HR 1.52 95%CI 1.41-1.62). The effect estimates for males were reduced for repeat offenses, though treatment remained protective (10-17 years: untreated HR 1.09 95%CI 1.05-1.13; treated HR 0.97 95%CI 0.90-1.04). No association in reoffending was found in females, except treated individuals aged 10-17 years had increased risk (HR 1.26 95%CI 1.02-1.56).
Conclusion: Individuals with ADHD were at increased risk of criminal offending, but stimulant treatment reduced the risk of a first offense. This association was reduced among repeat offenders, with only modest benefit of treatment among males. Adequate treatment resources may help keep young people with ADHD out of the criminal justice system.

Background: Hyperemesis gravidarum (HG), characterised by severe and constant nausea and vomiting in pregnancy, can lead to nutritional deficiencies and other pregnancy complications. HG has also been linked with adverse long-term outcomes for children. However, previous studies have not accounted for potential confounding due to shared family-level factors. This study aimed to determine whether maternal HG was associated with long-term educational, health and neurodevelopmental outcomes for children, and whether these findings were substantiated using a sibling-comparison design.
Methods: This population-based record linkage study featured livebirths in New South Wales, Australia from 2002-2010. Maternal HG was classified using hospital admissions and emergency department presentations during pregnancy. Outcomes included standardised educational testing in Grade 3 (age 7–9 years), age-specific hospitalisations up to 7 years, and neurodevelopmental disorders (NDD). Robust Poisson and Cox Proportional Hazards models with generalised estimating equations were used to estimate the risk of outcomes with inverse probability of treatment weighting used to account for potential confounding. Analyses were also restricted to an exposure-discordant sibling cohort to account for unmeasured genetic and familial factors.
Results: Of 700,082 livebirths, 10,282 (1.5%) were born to mothers with HG during their pregnancy. Maternal HG was associated with a higher risk of their offspring scoring below the national testing standard in reading (adjusted Risk Ratio [aRR] 1.19, 95% CI 1.09–1.29) and numeracy (aRR 1.14, 95% CI 1.03–1.23), increased age-specific hospitalisations (e.g. <1 year, aRR 1.35, 95% CI 1.30–1.39) and NDD (aHR 1.38, 95% CI 1.22–1.56). For most outcomes, these associations were nullified in the sibling cohort analyses.
Conclusion: We found no association between HG exposure and long-term educational, health and NDD outcomes for children when using a sibling-comparison design, suggesting no direct impact of maternal HG.

Background:
Antibiotics may cause gut microbiota dysbiosis, potentially leading to autoimmune diseases such as childhood Type 1 diabetes (T1D) and neurodevelopmental disorders like attention-deficit/hyperactivity disorder (ADHD).
Methods:
Using linked data, we identified antibiotic-use during pregnancy and the first five years of life for all children born between 2005 and 2010 in New Zealand (n=315,786) and followed them until 2021. Cox proportional hazards regression, adjusted for parental and perinatal factors, was used to assess associations with T1D risk. For prenatal exposure, associations were assessed with T1D diagnosis from birth; for early-childhood exposure, associations were assessed with T1D diagnosis from age five. In parallel, we are examining the association between early-life antibiotic exposure and ADHD diagnosis in the same cohort.
Results:
Antibiotic use (≥1 course) was 29.2% for pregnant mothers and 96.3% for children in the first five years. T1D was diagnosed in 0.36% (n=1,132), with 0.28% (n=909) diagnosed after age five. Antibiotic use during pregnancy was associated with T1D when expressed as a continuous variable (Hazard ratio (HR) 1.05, 95% CI 1.03–1.08) and as a dichotomous variable (0 vs. ≥1 course; HR 1.23, 95% CI 1.07–1.40). Categorisation into 0, ≤1, ≤2, and ≥3 courses revealed a significant positive trend (p=0.0008). For postnatal exposure, significant dose-response trends were observed (HR 1.02, 95% CI 1.02–1.03; HR 1.62, 95% CI 1.00–2.63; p=0.0017). The strongest associations were observed for moderate and broad-spectrum antibiotics. Prenatal exposure to penicillin and macrolides, and postnatal exposure to all antibiotic classes except tetracyclines, were associated with higher T1D risk. Analyses for ADHD are ongoing.
Conclusion:
Antibiotic use during both prenatal and early childhood was associated with an increased risk of T1D. Ongoing analyses will further elucidate the relationship between early-life antibiotic exposure and ADHD risk.
Keywords: antibiotic use, microbiome dysbiosis, Type 1 diabetes, ADHD, early-life, data linkage
(See Disclaimer).

Background: Cancer is notifiable under Tasmanian legislation. The Tasmanian Cancer Registry is a population-based registry responsible for collecting and reporting cancer incidence and mortality among Tasmanian residents. We describe the number of cases and deaths from cancer in the state in 2022 and survival in 2013-17, describe the main users of registry data, and identify data gaps and future directions.

Methods: Data were entered into a database and used to produce publicly-available incidence, mortality and survival reports. Summary and individual level data were available by request under relevant approvals. Data requests were summarised by requestor and nature of request. We referred to the National Cancer Data Framework to identify data gaps.

Results: In 2022 in Tasmania, 4,094 new cancers were diagnosed and 1,357 people died from cancer. The most common cancer diagnoses were prostate (n=858), breast (n=447) and colorectal (n=406). The most common cancer deaths were from lung (n=289), pancreatic (n=96) and prostate (n=84) cancers. Five-year survival for all cancers improved between 1988-92 and 2013-17 from 48% to 68% but remained low for lung (20%) and stomach (28%) cancers. The registry received 43 data requests in 2022 (23% from Tasmanian sources). Data users included academic researchers (47%), government (19%), population-based and clinical quality registries (12%), industry (9%), health services (7%), and non-governmental organizations (7%). The National Cancer Data Framework identified cancer stage at diagnosis as a critical data gap; however, the registry does not have capacity to record cancer stage routinely due to lack of resources and ambiguous and/or incomplete information to ascertain this information.

Conclusion: Tasmanian cancer data show improving survival in some, but not all, cancers. There is broad interest in the data, primarily from academic researchers. Coordinated efforts and adequate resource allocation will be required to meet the demand for cancer stage at diagnosis in Tasmania.

Aim: The New Zealand population defined as ‘Pacific’ is ethnically diverse, but this diversity is seldom examined in health research. This research applies novel methods for describing health outcomes for specific Pacific populations in New Zealand using all cancer and stomach cancer incidence and mortality as examples. Effects of loss to follow-up from leaving the country are also assessed.
Method: The New Zealand Cancer Registry was linked to administrative datasets and analysed within Stats NZ’s Integrated Data Infrastructure (IDI). All cancer and gastric cancer incidence was examined over the 1995 – 2022 period, and one-to-five-year mortality among 1998 – 2017 diagnoses.
Results: There was variability in age-standardised all-cancer incidence and gastric cancer incidence for different Pacific groups. Less variation in mortality was identified between groups, and these rates increased only modestly when adjusting for those who left the country. Lower all-cancer mortality was observed in 2008 – 2017 compared to 1998 – 2007.
Conclusion: Variation in health outcomes among specific Pacific ethnicities is masked when examining Pacific peoples as an aggregated ethnic group. However, small counts among small ethnicities create challenges for producing detailed, reliable data when using the IDI.

Background: Access to healthcare services plays a crucial role in patient outcomes, particularly for those requiring specialised treatments like kidney replacement therapy (KRT). This study investigates the accessibility of kidney healthcare centres for individuals starting KRT.
Methods: Data were sourced from the Australian and New Zealand Dialysis and Transplant registry for the period 2000-2020. The Statistical Area Level 3 (SA3) region is used as the primary geographical unit for the investigation, with 326 regions considered where KRT patients are located. The catchment area of each kidney healthcare centre was assessed using multiple-ring buffer analysis with 10, 20, 50, 100 and 200 km radii. Access to services from major cities to remote areas was considered using the Accessibility/Remoteness Index of Australia Plus to capture variations in kidney healthcare.
Results: A total of 51,715 patients starting KRT in 300 kidney healthcare centres across Australia were considered, with nearby areas showing similar incidence (Moran’s I=0.4). In the major cities, 132 kidney health care centers covered one to eleven SA3 regions within a 10 km radius. Whereas in the very remote areas, 9 kidney health care centers covered one to three SA3 regions within a 200 km radius. About 1016 kidney failure cases reside within a 10-km radius of a kidney healthcare center in major cities (out of an average population of 506,885). In very remote areas, 407 kidney failure cases reside within a 200-km radius of a kidney healthcare center (out of an average population of 24,151).
Conclusion: Kidney healthcare center accessibility was unevenly distributed, with higher kidney failure needs in remote areas compared to major cities, which had a larger at-risk population but lower kidney failure incidence. This suggests that poor access may contribute to kidney failure progression, as distance barriers may prevent patients from making sufficient visits for their health management.

Background: Chronic low-grade inflammation has been found to be associated with a range of neurodegenerative diseases, including dementia. Previous research on the association between the inflammatory biomarker, C-reactive protein (CRP), and dementia have yielded mixed results. We aimed to investigate the association between high sensitivity CRP (hs-CRP) levels and dementia risk in a group of initially healthy older people aged 70 years and over.
Methods: This post-hoc analysis included 11,909 participants from the ASPirin in Reducing Events in the Elderly (ASPREE) study. At recruitment, all participants were without diagnosed dementia and had a score of over 77 on the Modified Mini-Mental State exam. Cox proportional hazards regression models adjusted for potential confounders were used to determine the association of baseline hs-CRP and incident dementia. Hs-CRP was analysed in quintiles (Q) where Q1 represented the lowest levels. Dementia was diagnosed according to the Diagnostic and Statical Manual of Mental Disorders – version IV criteria.
Results: The median (IQR) age at baseline was 73.9 (71.9-77.4) years and 53.6% (n=6377) of participants were female. Over a median follow-up period of 8.4 (7.3-9.5) years, 747 developed incident dementia. Baseline hs-CRP was higher in those who did not develop incident dementia compared to those who did (1.72 vs 1.36 g/mL). Results from the Cox-regression models showed that elevated hs-CRP levels were associated with a decreasing risk of dementia. Compared to Q1, the adjusted hazard ratio for Q5 was 0.59 (95%CI 0.47-0.75).
Conclusion: In a group of initially healthy older adults, elevated levels of hs-CRP showed to be protective of incident dementia.

Background
Endometrial cancer (EC) is the 5th most diagnosed cancer in Australian women. Risk factors include obesity and hormonal factors, and there is evidence that gynaecological conditions (e.g. pelvic inflammatory disease, PID) may also increase risk. Little is known about whether different treatment options for these conditions may modify future risk.

Methods
Using data from two large population-based cohort studies (IMPROVE and HOLD), we initially conducted a nested case-control study of women from New South Wales (NSW) (born 1966-1970), Victoria (VIC) (born 1958-1965) and Western Australia (WA) (all women aged 18+) diagnosed with EC from 2004 to 2022 (2017 for WA) (n=4,580). We randomly selected 10 controls for each case, matched by year of birth, state, and eligibility at the index date (n=45,800). We used the Australian Cancer Database to identify EC cases and hospital morbidity data to identify diagnoses of gynaecological conditions and EC where ACD records were not available (WA 2015-2017, NSW/VIC 2020-2022). We investigated the association between endometriosis, fibroids and PID and risk of EC, adjusting for Socio-Economic Indexes for Areas quintile (all states) and parity (WA).
We have also commenced an emulated trial analysis for women diagnosed with gynaecological conditions. We aim to explore whether different treatments (e.g. myomectomy, hormonal intrauterine device) mediate the risk of EC.

Results
Fibroids and PID diagnosed at least 1 year prior to the index date were associated increased risk of EC (Fibroids: NSW/VIC OR=1.6, 95%CI:1.3-2.0; WA OR=1.7, 95%CI:1.3-2.1; PID NSW/VIC OR=1.3, 95%CI:1.1-1.6; WA OR=1.4, 95%CI:1.1-1.6). Endometriosis was not associated with EC risk. Results from the emulated trial will be presented.

Conclusion
We found that women with fibroids and PID have an increased risk of EC. Our study may provide important insights for EC prevention for women with gynaecological conditions.

Background
Limited research has focused on risk factors for community acquired pneumonia (CAP) in adults in low-middle-income countries. Mongolia has high rates of CAP hospitalisations with extreme winter temperatures and significant air pollution. We aimed to examine risk factors for severe CAP among hospitalised adults aged ≥18 years in Mongolia.

Methods
Adults hospitalised with clinical pneumonia were enrolled over 3 years (2019-2022) into a prospective pneumonia surveillance program in Ulaanbaatar. Participants had clinical information and risk factors collected using a case report form and nasopharyngeal swabs and urine were collected. Severe pneumonia was defined as clinical pneumonia with ICU admission or ≥2 severity signs (confusion, hypotension, tachypnoea or hypoxaemia). Multivariable logistic regression included risk factors for severe pneumonia chosen based on previous literature, applicability to respiratory infections and those deemed significant by odds ratios on univariate regression if relevant. An interaction term was included between age and underlying medical conditions.

Results
Overall, 3178 participants met the CAP study case definition including 10% (n=322) with severe CAP. Underlying medical conditions (aOR 2.00; 95%CI 1.41-2.80), cigarette smoking (aOR 1.50; 95%CI 1.07-2.11), alcohol intake (aOR 2.16; 95%CI 1.48-3.17), previous hospital admission within the last year (aOR 1.72; 95% CI:1.25-2.35), pneumococcal carriage (aOR 2.68; 95%CI 1.90-3.80) and serotype specific urine antigen detection positivity (aOR 1.73; 95% CI 1.14-2.62) were identified as risk factors for severe CAP on multivariable analysis. Socioeconomic factors and older age did not increase the odds of severe CAP.

Conclusion
We identified key risk factors for severe CAP in hospitalised adults in urban Mongolia. Our findings suggest to decrease severe CAP the Mongolian government should consider implementation of a targeted adult-based PCV program for older adults and all adults with co-morbidities as well as support additional health policies aimed at decreasing alcohol use and cigarette smoking.

Background
Extreme bushfires pose significant occupational health risks. Research to date has mostly focused on emergency responders, while less is known about the impact on the general workforce. This study examined the association between extreme bushfires and workers’ compensation claims in the general workforce in Victoria, Australia, identifying affected industries, occupation groups, and injury/disease types.
Methods
Workers' compensation claims of the general workforce (10% random sample) were obtained from WorkSafe Victoria (January 2005–June 2023). Claims made by emergency responders (firefighters, police officers, ambulance officers) were excluded from the study. Claims for injury/illness that originated during extreme bushfire time periods in Victoria (Black Saturday 2009; Black Summer 2019–2020) were compared to all other claims using logistic regression modelling, examining the effects of sociodemographic, injury/disease, occupational, and seasonality variables.
Results
Of 78,057 claims, 1,646 (2.1%) occurred during extreme bushfire time periods. Extreme bushfire period claims were associated with age groups 25–34 years (OR=1.21, 95%CI=1.04–1.42) and ≥55 years (OR=1.30, 95%CI=1.12–1.51) compared to 35–44 years. Working in construction (OR=1.25, 95%CI=1.03–1.51), financial services (OR=2.01, 95%CI=1.35-3.00) and accommodation and food services industries (OR=1.53, 95%CI=1.18-1.99) were associated with extreme bushfire period claims, relative to working in manufacturing. Respiratory conditions (OR=3.81, 95%CI=2.58–5.62) and fractures (OR=1.31, 95%CI=1.09–1.56) were associated with extreme bushfire periods when compared to musculoskeletal conditions.
Conclusion
This is the first study to examine the occupational health effects of extreme bushfires on the general workforce, excluding those directly involved as emergency responders. The overrepresentation of respiratory disease claims suggests that air quality may have been affected. The overrepresentation of specific industry claims and sociodemographic patterns uncovered in this study provides the impetus for further research to establish causal pathways and ultimately inform targeted occupational injury/disease prevention measures in response to extreme bushfires.