Journal of Physiotherapy 63 (2017) 129–130

Journal of
PHYSIOTHERAPY
journal homepage: www.elsevier.com/locate/jphys

Editorial

Publishing code: an initiative to enhance transparency of data

analyses reported in Journal of Physiotherapy
Rob Herbert a,b, Mark Elkins c,d
a
Editorial Board Member, Journal of Physiotherapy; b Neuroscience Research Australia (NeuRA), Sydney; c Editor, Journal of Physiotherapy; d Sydney Local Health District,
Sydney, Australia

In a recent interview,1 Professor Michael Gazzaniga, one of the Journal Editors (ICMJE) has recently mandated that, for clinical
founders of cognitive neuroscience, provided a wonderful insight trials reported in ICMJE member journals, a data sharing plan must
into how good science works: ‘You know, when you’re giving a be provided at the time the trial is registered and submitted with
scientific presentation, everybody in the audience is sitting there the trial report.8
saying what’s another explanation for this?’ That’s exactly how it Now the Journal of Physiotherapy is taking the next step.
should be. External scrutiny is a key feature of the scientific Henceforth the authors of accepted reports of quantitative research
process. Sometimes scientists get it wrong, and when they do it is will be strongly encouraged to publish the statistical analysis plan
the responsibility of other scientists to find the mistakes. So, and code used to analyse their data. The policy applies to primary
ideally, all research findings are perpetually open to scrutiny. Then quantitative research (not to secondary research such as systematic
differing interpretations of the data can compete for acceptance. reviews of summary data). The mechanism will be much the same
The hope is that eventually the best interpretation wins. as that used for publishing raw data. The statistical analysis plan and
However, scrutiny is only possible when there is transparency. code will be provided as supplementary files on the journal’s
Readers of a research report must be able, if they wish, to look website and linked to the primary research report.
deeply into how the research was conducted to decide if the Some readers may wonder what we mean when we refer to the
authors’ conclusions are or are not supported by the data. ‘statistical analysis plan’ and ‘code’ used to analyse the data, so we
Many readers are prepared to trust the published results of explain those terms below.
research without scrutinising the data, perhaps because they trust A statistical analysis plan is a detailed description of intentions
the peer review process. However, even when there has been about how the investigators intend to analyse the data. Impor-
rigorous peer review, published manuscripts may still contain tantly, the statistical analysis plan is written before the data are
errors: there may be errors in the analysis of data, errors in available for analysis, and is dated and ratified by the investigators.
typesetting, or (hopefully rarely) deliberate misrepresentation of The plan may be made publicly available as part of the registered
the data. Transparency of analysis procedures allows enthusiastic protocol, as in these examples,9,10 or as part of a published protocol,
readers to detect errors in analyses. For this reason Munafò and as in this example.11,12 Some statistical analysis plans, particularly
colleagues have claimed that ‘transparency is superior to trust’.2 for large and complex studies, are published as stand-alone papers,
The Editorial Board has tried to maximise the transparency of as in these examples.13,14 The value of a statistical analysis plan is
research reports published in the Journal of Physiotherapy. That is that it provides a public declaration of the investigators’ intentions
why, for example, the Journal of Physiotherapy was amongst the for the analysis. This makes it clear, for both the investigators and
first physiotherapy journals to mandate prospective registration of the readers of the resulting research report, which of the analyses
clinical trials3 and encourage prospective registration of systematic were decided upon a priori and which were decided upon after
reviews,4,5 and why authors are encouraged to conform to looking at the data. Analyses that were pre-specified in a statistical
reporting guidelines such as those generated and made freely analysis plan are more credible than analyses that were not pre-
available online by the Equator Network.6 specified. Analyses that were only conceived after looking at the
In 2008, the Journal of Physiotherapy led the way in an initiative data should be considered exploratory (‘hypothesis generating’)
designed to improve research transparency. The Editorial Board rather than confirmatory.
resolved that authors of research reports accepted for publication Almost all analysis of quantitative research published in the
in the Journal of Physiotherapy would be strongly encouraged to Journal of Physiotherapy (and, these days, of some qualitative
publish raw data (eg, data from individual participants in a research too) involves statistical analyses conducted using
randomised trial) on the web alongside the primary research programs such as SPSS, Stata, and SAS. Most programs allow the
report. Publication of raw data enables scrutiny, and even re- user to conduct the analysis interactively by pulling down menus,
analysis of the data, and makes it possible for other researchers to specifying variables, and selecting various options. The program
incorporate the data in conventional or individual-patient-data then processes those instructions, runs the appropriate analyses,
meta-analyses.7 This initiative has been tremendously successful: and displays the output on a screen or writes the output to a file. A
since 2008, 92% of primary reports of randomised trials published problem with this approach is that the output might not provide a
in the Journal of Physiotherapy have provided raw data as a web- complete or permanent record of exactly how the analysis was
based supplement to the primary research report. Nearly a decade conducted.
later, other clinical journals have begun to implement policies Fortunately, most high-end software has the facility to
promoting data sharing. The International Committee of Medical automatically generate a permanent record of exactly how the

1836-9553/© 2017 Australian Physiotherapy Association. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.
org/licenses/by-nc-nd/4.0/).
130 Editorial

analysis was conducted. In SPSS this is done with ‘syntax’ files, in Ethics approval: N/A.
Stata it is done with ‘do’ files, and in SAS it is done with ‘SAS Competing interests: The authors declare that they have no
program’ files. This is what we refer to as ‘code’. Other programs competing interests.
may store code in other formats. The code provides a record of Source of support: Nil.
exactly how the data were analysed. It is these files that authors Acknowledgements: Nil.
will be encouraged to make publicly available as supplementary Provenance: Invited. Not peer reviewed.
files on the Journal of Physiotherapy website. Correspondence: Mark Elkins, Centre for Education & Work-
Providing public access to individual participant data, the force Development, Sydney Local Health District, Australia. Email:
statistical analysis plan and code make it possible for other [email protected]
researchers to examine and reproduce the analyses exactly as they
were originally conducted. Obviously, this increases transparency;
References
but the potential benefits of making the code publicly available
extend beyond enhancing transparency. Another advantage of 1. Campbell V. Interview with Dr Michael Gazzaniga, author of tales from both sides of
making the code public is the potential for education by allowing the brain: A life in neuroscience. Brain Science Podcast 2015. http://
brainsciencepodcast.com [Accessed 5th May 2017].
inexperienced researchers to see rigorous ways to analyse data.
2. Munafò MR, et al. Nat Hum Behav. 2017;1:0021.
Two members of the Editorial Board are each publishing an 3. Askie L, et al. Aust J Physiother. 2006;52:237–239.
original research paper in the current issue.15,16 These Editorial 4. Elkins M. J Physiother. 2011;57:67–68.
Board members and their respective co-authors have kindly agreed 5. Elkins M, Ada L. J Physiother. 2010;56:69.
6. Elkins M. J Physiother. 2015;61:103–105.
to lead the way with this new initiative by providing the data file, 7. Herbert RD. Aust J Physiother. 2008;54:3.
the code used to analyse the data, and the protocol showing the 8. Taichman DB, et al. New Engl J Med. 2017;376:2277–2279.
original statistical analysis plan. All these files are freely available 9. https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?
id=364402&isReview=true.
as electronic appendices to the published papers.15,16 10. https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?
While the Editor and the Editorial Board of the Journal of id=370896&isReview=true.
Physiotherapy are committed to this new policy, we understand 11. Harvey LA, et al. J Physiother. 2016;62:88–95.
12. Harvey LA, et al. Trials. 2011;12:14.
that some researchers will not have followed pre-specified 13. Williams CM, et al. Trials. 2013;14:248.
statistical analysis plans or may not have coded their analyses, 14. Mathieson S, et al. Trials. 2016;17:53.
and therefore may not be able to comply with requests to supply 15. Lambert T, et al. J Physiother. 2017;63:161–167.
16. Oosterhuis T, et al. J Physiother. 2017;63:144–153.
the statistical analysis plan and code. For now, the absence of a
statistical analysis plan and code will not preclude publication in
the Journal of Physiotherapy. However, provision of a statistical Websites
analysis plan and code at the time a manuscript is submitted will www.equator-network.org
strengthen the prospects of publication. Authors are strongly
encouraged to provide statistical analysis plans and code when
they submit their papers to the Journal of Physiotherapy. http://dx.doi.org/10.1016/j.jphys.2017.05.011

Readers’ Choice Award

The Editorial Board is pleased to announce the annual Readers’ Choice Award, which recognises the paper published in Journal of Physiotherapy
that generates the most interest by readers of the journal. The winning paper is chosen based on the number of times that each paper published
in a given year is downloaded in the six months after its day of publication.

The winning paper from among those published in 2016 is ‘Physiotherapy rehabilitation for people with spinal cord injuries’ by Professor Lisa
Harvey from the University of Sydney.1 The winning paper is one of the journal’s new Invited Topical Reviews. It deftly summarises the results of
a large amount of research into the assessment and rehabilitation management of people with spinal cord injury. The section on assessment
includes clear and concise guidance on how to convert the assessment into goal setting for individual patients. The physiotherapy interventions
considered by the paper include interventions to increase strength, interventions to improve the performance of motor tasks, and interventions
to prevent and treat contracture. The paper also calls for spinal cord injury to be the topic of further high-quality clinical studies, several of which
have subsequently been published in Journal of Physiotherapy.2–4

The only other Invited Topical Review5 published in the same year was the second ranked paper, indicating the popularity of this relatively new
category of paper in the journal.

The Editorial Board of Journal of Physiotherapy congratulates Professor Harvey on her success.

References

1. Harvey LA. Physiotherapy rehabilitation for people with spinal cord injuries. J Physiother. 2016;62:4–11.
2. Nooijen CFJ, et al. A behavioural intervention increases physical activity in people with subacute spinal cord injury: a randomised trial. J Physiother. 2016;62:4–11.
3. Harvey LA, et al. Early intensive hand rehabilitation is not more effective than usual care plus one-to-one hand therapy in people with sub-acute spinal cord injury (‘Hands On’):
a randomised trial. J Physiother. 2016;62:88–95.
4. Jørgensen V, et al. Falls and fear of falling predict future falls and related injuries in ambulatory individuals with spinal cord injury: a longitudinal observational study. J Physiother.
2016;63:108–113.
5. Granger C. Physiotherapy management of lung cancer. J Physiother. 2016;62:60–67.

http://dx.doi.org/10.1016/j.jphys.2017.05.012