Are you looking into publishing or getting involved in research?
This might be the high-level overview you need to simplify the whole mysterious process.
I was recently invited to be a conference writer at HACK Aotearoa 2020, a conference held at the University of Auckland regarding healthcare, tech and artificial intelligence (AI).
As a published research author myself, I found this workshop delivered by Professor David Pilcher to be a very simple overview of the scientific writing process that beginners looking into publishing or getting involved in research should find helpful.
4 Steps to Turn Data Into Publications
“There’s often people that understand the clinical bits. There’s often people that understand how to analyse data. Things fall down because it’s not constructed in a way that makes sense and puts out a clear message in a manuscript that gets published.”
That’s Melbourne-based intensive care specialist, Professor David Pilcher, who has given talks around the world on how to craft your data into a publishable scientific story.
The highly interactive, limited-seating workshop delivered at HACK Aotearoa 2020 brought data scientists, statisticians and clinicians together to turn data from an intensive care registry into a series of compelling results and discussion points.
He presents a simple four-step system to seamlessly bridge the gap between raw data and meaningful results.
Step 1: Defining your data
Identify your datasets, clarify exclusion and inclusion criteria, and define primary and secondary outcomes.
Step 2: Allocating groups
Define the exposure and comparison group. Most of the time, unless you are hunting for correlations and groups, the exposure and comparison groups will already be established. This step is for you to specify the criteria for these groups based on the type of data you have.
For example: In the workshop, we were looking at outcomes for those with infection vs. no infection. Because the dataset had diagnoses much more specific than that, we had to identify which particular diagnoses we would classify as “infection”, and which were not.
This can have important ramifications based on step 1. For example, approximately half of intensive care patients admitted for asthma, chronic obstructive pulmonary disease (COPD) and mechanical airway obstruction, are a result of infection. Do we include these diagnoses or not? If not, do we exclude them from our analyses entirely?
Step 3: Analysis and confounding
Identify the analyses required and confounding factors. There are often four or five tables, in the following order:
- Table 1: Basic descriptives on the cohort in exposure group vs control
- Table 2: Total non-adjusted outcomes
- Table 3: Primary non-adjusted outcomes in exposure vs comparison
- Table 4: Adjusted* primary outcomes in exposure vs comparison
- Table 5: Further sensitivity analyses or secondary outcomes reporting
*Adjustments for outcomes will be based on the confounding factors that you have identified. This can be a lengthy process with important ramifications on the results. A systemic bias affecting only one group may completely change the outcome! This is especially important when using retrospective datasets where you have no knowledge of the rigour of data collection.
Step 4: Discussion
“There is a fairly standard format for discussions… every good discussion generally follows this format”Professor David Pilcher
His format for building a discussion is as follows:
- Summarise the findings.
- Give us background information on the subject.
- Confirm or reject the hypothesis.
- What are the implications of the study?
- What are the strengths and limitations?
- How does this impact future research or care?
- What is the conclusion?
“You can take this template on how to publish a scientific article and apply it to anything else… This has been put together through… years of trying to do this, but actually once you do it over and over again, you realise you can apply this same system.”
What are your thoughts?
Did you have any questions? If you do, comment below – we’d love to discuss!