The birth of a Medicare item: evidence, politics and cardiovascular disease

Dr Jason Agostino of NCEPH gave this RSPH seminar on Thursday 18 April.

His was a story of killer statistics, early morning press releases, an election and a news organisation – a heady mix in anyone’s book!

Jason took us through the story of the development of MBS item 715, from the slow slow building of relationships between researchers and Departmental officials to the announcement of MBS funding from two political parties, ten months apart, all before 8am on a Sunday!

Jason cleverly left the audience to fill in his last slide, “Lessons learnt”, with him. I think one of the keys was being a member of a great team, and being able to play both the long game and the short one at the same time.

A great talk to bring on a long weekend!

Advertisements

Multiple imputation in three level models

I was thrilled to be able to present this talk via Skype to the Multilevel Congress in Utrecht on Tuesday 9 April. A 3.50 pm time slot turned into 11.50 pm for me, so armed with a strong cup of tea and the laptop, I presented to about 40 people on the other side of the world.

This work is joint with my PhD student Nidhi Menon. Inspired by the 4th National Family and Health Survey conducted in India in 2015-16, she constructed a random intercept model simulated data set for poking holes in to the rate of 20% and 50% missingness at various combinations of levels in the model. Two methods are being explored, multiple imputation using chained equations (mice) and joint modelling (jomo). Passive imputation via “impute then transform” is being used to impute for derived variables such as higher-level group means.

Results are starting to flow, now that Nidhi has access to the ANU supercomputer! Performance is being assessed through bis, mean square error, coverage and alignment of distributions of observed and imputed data. Patterns are hard to discern as yet, so watch this space for more presentations of this work as the results continue to roll in!

 

Psychological distress and CVD: an investigation into the epidemiological evidence

Jenny Welsh presented her PhD exit seminar to a full house in RSPH on Thursday 4 April. Her RSPH supervisors for this research comprised Rosemary Korda, Grace Joshy, Emily Banks and Lyndall Strazdins. Psychological distress is a tem use to capture depression and anxiety. It is biologically plausible that distress causes cardio-vscular disease (CVD) – strong statements made by the Heart Foundation to that effect are out there in the community. But there is reason to be sceptical and consider a confounded pathway, which was the major focus of Jenny’s PhD work. Measurement error is also a thing that she spent much time thinking about.

After a fantastic presentation, compellng, coherent and measured, jenny concludeed that distress is best characterised as a marker of risk rather than an independent risk factor. And in the end, distress should be treated whether or not it is related to CVD. As Jenny poined out, anxiety and dpression are “miserable” conditions that deserve proper treatment in and of themselves, not just for their association with CVD.

Preparing for death: what is important to people with dementia at the end of life?

Professor Helen Mason of Glasgow Caledonian University is visiting the School this week, and she gave this seminar on Thursday 28 March.

She reported on aspects of the Supporting Excellence in End of life care for people with Dementia (SEED) project in the UK. Right now about 1.3% of the UK population has dementia, and this is predicted to rise to 2.6% by 2050. It’s a growing issue (and similar in size to the ME/CFS issue that occupies much of my research time).

First Helen reported on the Q methodology her team used to identify the features of end of lifecare that are important. Fufty-eight participants were asked to rank their reactions to 24 statements e.g. strongly disagree to strongly agree. A factor analysis was then used to identify clusters of people who responded in similar ways.

An intervention was then proposed, namely a Dementia Nurse Specialist. In the second part of the project that helen reported, participants were asked to come up with willingness to pay numbers in a contingent valuation study. Basically the willingness to pay was the amount of extra taxation participants were prepared to imagine paying pre month in order to have different levels of service from Dementia Nurse Specialists available. It turned out that amongst the general public, amounts of £20 were typical, with a highly right skewed distribtion around that.

These willingness to pay figures, though subtle and not easy to elicit from participants, will be very useful in the next phase ofthe study, a cost benefit analysis.

R Ladies: Bayesian Modelling in R

R Ladies in Canberra met again on Monday 25 March. Jiali Wang of CSIRO Data61 spoke on Bayesian Modelling in R. She began in a most modest manner, quite unneccessary – she wasn’t presenting because no one else would, she was presenting because she has a lot to offer on the topic of Bayesian modelling! I loved her idea that you if you don’t have a pen (Read: don’t know how!) to do an integral, you need a simulation based approach to integration such as Gibbs or Metropolis-Hastings. She had three main examples after an introduction – a normal model, hierarchical model and semiparametric regression.

Jiali recommended two books that inspired her presentation: Hoff “A First Course in Bayesian Statistics” and Harezlak, Ruppert & Wand “Semiparametric regression with R”.

An examination of factors that challenge or facilitate implementation of the Baby Friendly Hospital Initiative in Australia and the Ten Steps to Successful Breastfeeding in Indonesia

Andini Pramono of the Department of Health Services Research and Policy here at NRSPH gave her thesis proposal eview seminar on Thursday 21 March. She’s got a strong bid for seminar-with-the-longest-title there!

Her supervisors of this PhD research are Jane Desborough, Siobhan Bourke and Julie Smith. Andini did a great job of establishig the backgrund and motivation of her research, with a wide array of inforgraphisc on the advantages of breastfeeding. She walked us through all her proposed research questions in this mixed methods research. This included a Social Return on Investment analysis of breastfeeding, and use of the Nonadoption-Abandonment-Scaleup-Spread-Sustainability framework of change implementation.

Questions at the end targeted specific aspects of the data collection, and also issuea around expression of breast milk which is an interesting extra strand to the whole “breast is best” debate.

ME/CFS: the biomedical basis, diagnosis, treatment and management

This international research symposium was organised in Geelong by Emerge Australia, the peak body For ME/CFS advocacy. It was a pleasure to chair the first session, Research Innovation, Big Data and Bioinformatics, at this conference on Thursday 14 March.

Dr Ronald Davis, Director of the Stanford Genome Technology Center spoke first, on Establishing new mechanistic and diagnostic paradigms for ME/CFS. His interest in ME/CFS is personal, as he described his experience caring for an adult son with a severe form of the condition. I firmly believe that a personal passion can be very helpful when the research going gets tough, however a more dispassionate passion is also valuable in terms of evaluating the evidence and reacting with caution in the face of uncertainty.

Ron has been using the Stanford supercomputer to compare masses of gene sequences to masses of other gene sequences in his search for signals of ME/CFS. He’s been looking at viruses too, and parasites (think trypanosomiasis or sleeping sickness, and Leishmaniasis). Even metallomics is a thing now.

Dr Wenzhong Xiao of Harvard Medical Center spoke next, on The Severely Ill Patient Study of ME/CFS. I like his use of heat maps to visualise everything from correlation matrices to the usual gene expression data, but I look forward to measure of variability being included in these plots such as is now packaged up in R by my colleague Petra Kuhnert. I was also keen to know more about the 60-odd metabolites out of 600 that were found to be significant. Would these be hypothesis-testing, needing adjustment for multiple testing, or more in the way of hypothesis-generating, ready for further research.

Dr Robert Phair of Integrative Bioinformatics Inc, completed the session with a presentation on Metabolic Traps in ME/CFS. He firmly placed ME/CFS at the intersection of three concepts – bistability, biochemistry and genetics. Bistability refers to the multiple stable points of metabolite curves. Robert’s main research focus is common gene variants, wondering if ME/CFS is actually a common condition with common variants, rather than the more widely studies rare conditions with rare variants.

The second session was on Metabolomics and Transcriptomics. Dr Chris Armstrong of the Open Medicine Foundation spoke on Longitudinal ME/CFS MMR Metabolomics. He briefly introduced results from studies of B cells then moved on to the part that interested me much more, a longitudinal study involving 25 blood samples in 25 days from ten patients. The analysis of this data is work-in-progress and I see it as a huge opportunity to apply techniques such as functional data analysis to a problem that so clearly involves autocorrelated data.

Ruud Raijmakers of Radboud University Medical Centre in the Netherlands came at things from a completely different direction, to speak on Transcriptome analysis of QFS and CFS in the Netherlands. His jumping off point was an “outbreak” over several years in the early 2010s of Q fever in the south of the Netherlands, traced to infected goats on farms in the region. A proportion of people who contract Q fever go on to experience CFS-lime symptoms and Ruud was interested in seeing if the experience in treating Q fever fatigue syndrome had any parallels in the chronic fatigue space.