Stephen Morrell

BACKGROUND: Significant variation exists in published Aboriginal mortality and life expectancy (LE) estimates due to differing and evolving methodologies required to correct for inadequate recording of Aboriginality in death data, under-counting of Aboriginal people in population censuses, and unexplained growth in the Aboriginal population attributed to changes in the propensity of individuals to identify as Aboriginal at population censuses. The objective of this paper is to analyse variation in reported Australian Aboriginal mortality in terms of LE and infant mortality rates (IMR), compared with all Australians. METHODS: Published data for Aboriginal LE and IMR were obtained and analysed for data quality and method of estimation. Trends in reported LE and IMR estimates were assessed and compared with those in the entire Australian population. RESULTS: LE estimates derived from different methodologies vary by as much as 7.2 years for the same comparison period. Indirect methods for estimating Aboriginal LE have produced LE estimates sensitive to small changes in underlying assumptions, some of which are subject to circular reasoning. Most indirect methods appear to under-estimate Aboriginal LE. Estimated LE gaps between Aboriginal people and the overall Australian population have varied between 11 and 20 years. Latest mortality estimates, based on linking census and death data, are likely to over-estimate Aboriginal LE. Temporal LE changes by each methodology indicate that Aboriginal LE has improved at rates similar to the Australian population overall. Consequently the gap in LE between Aboriginal people and the total Australian population appears to be unchanged since the early 1980s, and at the end of the first decade of the 21st century remains at least 11-12 years. In contrast, focussing on the 1990-2010 period Aboriginal IMR declined steeply over 2001-08, from more than 12 to around 8 deaths per 1,000 live births, the same level as Australia overall in 1993-95. The IMR gap between Aboriginal people and the total Australian population, while still unacceptable, has declined considerably, from over 8 before 2000 to around 4 per 1,000 live births by 2008. CONCLUSIONS: Regardless of estimation method used, mortality and LE gaps between Aboriginal and non-Aboriginal people are substantial, but remain difficult to estimate accurately.

Importance: Mammography screening currently relies on subjective human interpretation. Artificial intelligence (AI) advances could be used to increase mammography screening accuracy by reducing missed cancers and false positives. Objective: To evaluate whether AI can overcome human mammography interpretation limitations with a rigorous, unbiased evaluation of machine learning algorithms. Design, Setting, and Participants: In this diagnostic accuracy study conducted between September 2016 and November 2017, an international, crowdsourced challenge was hosted to foster AI algorithm development focused on interpreting screening mammography. More than 1100 participants comprising 126 teams from 44 countries participated. Analysis began November 18, 2016. Main Outcomes and Measurements: Algorithms used images alone (challenge 1) or combined images, previous examinations (if available), and clinical and demographic risk factor data (challenge 2) and output a score that translated to cancer yes/no within 12 months. Algorithm accuracy for breast cancer detection was evaluated using area under the curve and algorithm specificity compared with radiologists' specificity with radiologists' sensitivity set at 85.9% (United States) and 83.9% (Sweden). An ensemble method aggregating top-performing AI algorithms and radiologists' recall assessment was developed and evaluated. Results: Overall, 144 231 screening mammograms from 85 580 US women (952 cancer positive ≤12 months from screening) were used for algorithm training and validation. A second independent validation cohort included 166 578 examinations from 68 008 Swedish women (780 cancer positive). The top-performing algorithm achieved an area under the curve of 0.858 (United States) and 0.903 (Sweden) and 66.2% (United States) and 81.2% (Sweden) specificity at the radiologists' sensitivity, lower than community-practice radiologists' specificity of 90.5% (United States) and 98.5% (Sweden). Combining top-performing algorithms and US radiologist assessments resulted in a higher area under the curve of 0.942 and achieved a significantly improved specificity (92.0%) at the same sensitivity. Conclusions and Relevance: While no single AI algorithm outperformed radiologists, an ensemble of AI algorithms combined with radiologist assessment in a single-reader screening environment improved overall accuracy. This study underscores the potential of using machine learning methods for enhancing mammography screening interpretation.

BACKGROUND: The clinical effectiveness of intensive lifestyle interventions in preventing or delaying diabetes in people at high risk has been established from randomised trials of structured, intensive interventions conducted in several countries over the past two decades. The challenge is to translate them into routine clinical settings. The objective of this review is to determine whether lifestyle interventions delivered to high-risk adult patients in routine clinical care settings are feasible and effective in achieving reductions in risk factors for diabetes. DATA SOURCES: MEDLINE (PubMed), EMBASE, CINAHL, The Cochrane Library, Google Scholar, and grey literature were searched for English-language articles published from January 1990 to August 2009. The reference lists of all articles collected were checked to ensure that no relevant suitable studies were missed. STUDY SELECTION: We included RCTs, before/after evaluations, cohort studies with or without a control group and interrupted time series analyses of lifestyle interventions with the stated aim of diabetes risk reduction or diabetes prevention, conducted in routine clinical settings and delivered by healthcare providers such as family physicians, practice nurses, allied health personnel, or other healthcare staff associated with a health service. Outcomes of interest were weight loss, reduction in waist circumference, improvement of impaired fasting glucose or oral glucose tolerance test (OGTT) results, improvements in fat and fibre intakes, increased level of engagement in physical activity and reduction in diabetes incidence. RESULTS: Twelve from 41 potentially relevant studies were included in the review. Four studies were suitable for meta-analysis. A significant positive effect of the interventions on weight was reported by all study types. The meta-analysis showed that lifestyle interventions achieved weight and waist circumference reductions after one year. However, no clear effects on biochemical or clinical parameters were observed, possibly due to short follow-up periods or lack of power of the studies meta-analysed. Changes in dietary parameters or physical activity were generally not reported. Most studies assessing feasibility were supportive of implementation of lifestyle interventions in routine clinical care. CONCLUSION: Lifestyle interventions for patients at high risk of diabetes, delivered by a variety of healthcare providers in routine clinical settings, are feasible but appear to be of limited clinical benefit one year after intervention. Despite convincing evidence from structured intensive trials, this systematic review showed that translation into routine practice has less effect on diabetes risk reduction.