Article Text

Evolution of methodology and reporting of emergency medicine quantitative research over a 20-year period
  1. Jesse Smith1,
  2. Patrick Date1,
  3. William Spencer1,
  4. Erik de Tonnerre1,
  5. David McDonald Taylor1,2
  1. 1 Emergency Department, Austin Hospital, Heidelberg, Victoria, Australia
  2. 2 Department of Medicine, University of Melbourne, Melbourne, Victoria, Australia
  1. Correspondence to Professor David McDonald Taylor, Emergency, Austin Health, Heidelberg, VIC 3084, Australia; david.taylor{at}austin.org.au

Abstract

Objective We aimed to determine trends over time in article origin, and article and methodology characteristics.

Method We examined original research articles published every fifth year over a 20-year period (1997–2017) in six emergency medicine (EM) journals (Ann Emerg Med, Acad Emerg Med, Eur J Emerg Med, Emerg Med J, Am J Emerg Med, Emerg Med Australas). Explicit data extraction of 21 article characteristics was undertaken. These included regional contributions, specific article items and research methodology.

Results 2152 articles were included. Over the study period, the proportional contributions from the USA and the UK steadily fell while those from Australasia, Europe and ‘other’ countries increased (p<0.001). All specific article items increased (p<0.01). Institutional Review Board/Ethics Committee approval and conflicts of interest were almost universal by 2017. There were substantial increases in the reporting of keywords and authorship contributions. The median (IQR) number of authors increased from 4 (2) in 1997 to 6 (3) in 2017 (p<0.001) and the proportion of female first authors increased from 24.3% to 34.2% (p<0.01). Multicentre and international collaborations, consecutive sampling, sample size calculations, inferential biostatistics and the reporting of CIs and p values all increased (p<0.001). There were decreases in the use of convenience sampling and blinding (p<0.001). The median (IQR) study sample size increased from 148 (470) to 349 (2225) (p<0.001).

Conclusion Trends over time are apparent within the EM research literature. The dominance in contributions from the US and UK is being challenged. There is more reporting of research accountability and greater rigour in both research methodology and results presentation.

  • emergency department
  • research
  • clinical

Statistics from Altmetric.com

Key messages

What is already known on this subject

  • Although the emergency medicine specialty is relatively young, a substantial output of research literature has been achieved.

  • Increasing amounts of research literature originate from countries as the specialty becomes more established.

What this study adds

  • We analysed the literature published in six emergency medicine journals over a 20-year period.

  • The proportional contributions of research literature from the USA and the UK steadily fell while those from Australasia, Europe and other countries increased.

  • Author numbers and the proportion of female first authors have increased.

  • There is more reporting of research accountability including Institutional Review Board/Ethics Committee approval, conflicts of interest and author contribution.

  • There is also a greater rigour in both research methodology and results presentation.

Introduction

Many countries have only recognised emergency medicine (EM) as an independent specialty in the last 30 years.1 The development of academic EM soon followed the specialty’s foundation and substantial research literature has been published.2 3 The value of research in EM is well recognised.4 It provides an evidence base for improvements in care, informs the development of practice guidelines and research-based interventions have improved health outcomes.4 Also, research active sites have been associated with improved clinical outcomes.5

Given the importance of research, it is necessary to understand the evolution of this growing field. EM research output is increasing worldwide.3 This has been associated with increasing numbers of international EM journals and journal impact factors.2 Despite this, a paucity of articles has reported on the state of EM research.3 Furthermore, many are outdated and neither reflect the current status nor detail its progress. Another difficulty is the wide range of journals in which it can be published makes it challenging to identify this work.6

We aimed to determine the evolution of EM research, with a particular interest in the reporting of a range of article items and methodological rigour. The findings will demonstrate its progression and inform the development of strategies aimed at improving further the quality of both the research and its reporting.

Methods

Study design and setting

We undertook a retrospective, observational study of a sample of EM journals, over a 20-year period (1997–2017), in the ED of a metropolitan hospital in Melbourne, Australia. As the study examined articles in the public domain, approval by a human research ethics committee was not required. The study was unfunded.

Journal selection

Journals were included if their content primarily focused on the EM setting. They were excluded if they focused on other settings or on specific patient categories, for example, prehospital medicine, critical care. General high-impact journals (eg, The Lancet, NEJM) were also excluded due to the difficulties and time involved in identifying their EM publications. From those journals meeting the eligibility criteria, we selected the top six based on their impact factors reported in the ‘Emergency Medicine’ category of Journal Citation Reports.7

The journals selected were Annals of Emergency Medicine, Academic Emergency Medicine, European Journal of Emergency Medicine, Emergency Medicine Journal, American Journal of Emergency Medicine and Emergency Medicine Australasia. All original research articles published in these journals in 1997, 2002, 2007, 2012 and 2017 were retrieved. The inclusion of all articles in these years provided a very large sample size and avoided random sampling with the potential for selection bias. It also provided a comprehensive snapshot of article status at well-separated time points, allowing meaningful comparisons between those time points. Original research was defined as having an identifiable research aim or hypothesis and measurable outcomes.

Data collection

Twenty-one article items of interest (see online supplementary appendix 1) were explicitly extracted from electronic copies of each article. These items related to the region(s) of origin (based on regions where EM was most established in 1997), specific article items (acronyms, keywords, study approval, funding, conflicts of interest, references and aspects of authorship) and research methodology (collaborations, study design, sampling techniques, blinding, biostatistics and results presentation). This technique differed from that of a systemic review in that data on reporting items and methodology were extracted, not data on outcome measures.

Supplemental material

A single investigator (JS) extracted and entered all data directly onto the study spreadsheet. A second investigator (DMT) reviewed 10% of articles to confirm accuracy of data extraction. In addition, he reviewed other articles to extract missing data items and reconcile concerns regarding data ambiguity. No systemic errors in the original data extraction were found.

Outcome measures

The primary outcome was article publication rate per country/region over the 20-year study period. The secondary outcomes were changes over time in the range of reported article items and methodology.

Patient and public involvement

No patient was involved.

Data analysis

Data are reported descriptively with levels of uncertainty, for example, frequency and percentage (95% CIs). Median (IQR) values are reported as figures. The χ2 and Kruskal-Wallis tests were employed for categorical and continuous data comparison, respectively. SPSS for Windows statistical software (V.24.0, SPSS) was used for all analyses. The level of significance was 0.05.

This report of this study’s findings has been prepared using the Strengthening the Reporting of Observational Studies in Epidemiology statement for the reporting of observational studies.

Results

A total of 2152 articles were analysed. There were 2303 individual country contributions as some articles originated from international collaborations. The numbers and proportions of articles published by region are described in table 1 and figure 1, respectively. More than twice as many articles were published in 2017 compared with 1997. Also, the contributions from the six regions changed significantly (p<0.001). While the proportional contributions of the USA and the UK steadily fell, those from Australasia, Europe and ‘other’ countries increased. Contributions from Canada remained unchanged. By 2017, however, the USA remained the region with the greatest contribution.

Table 1

Articles published by international region over the study period

Supplemental material

Figure 1

Publication contribution (percentage) by region.

Over the study period, each of the prespecified article items was reported with increasing frequency (p<0.01, table 2). The reporting of Institutional Review Board (IRB)/Ethics Committee approval and author conflicts of interest had become almost universal by 2017. There were also substantial increases in the reporting of keywords and author contributions. Modest increases were observed in the use of acronyms and the number of funded studies.

Table 2

Reported article items over the study period

The median (IQR) number of authors increased significantly from 4 (2) in 1997 to 6 (3) in 2017 (p<0.001, figure 2). Concurrently, the proportion of female first authors increased significantly. Also, the median (IQR) number of article references increased from 17 (15) in 1997 to 25 (13) in 2017 (p<0.001, figure 3).

Figure 2

Author number per article, median (IQR).

Figure 3

Reference number per article, median (IQR).

For all but three research methodology items, there were significant increases in items of methodological technique and reporting (table 3). Notable increases were in the proportions of international studies, the use of inferential statistics and the reporting of CIs. Modest increases were observed in the proportion of multicentre studies, and the use of consecutive sampling and p values. The proportion of articles that reported experimental studies (randomised controlled (parallel or crossover), pre/postinterventional, quasiexperimental, cluster randomised, other experimental) did not change and there were decreases in the use of both convenience sampling and blinding. Although there was an increase in the reporting of a sample size calculation, the absolute increase was small. However, the median (IQR) study sample size increased significantly from 148 (470) in 1997 to 349 (2225) in 2017 (p<0.001, figure 4).

Table 3

Reported items of research methodology over the study period

Figure 4

Sample size per study, median (IQR).

Discussion

In the USA, EM was officially recognised in 19798 and is now becoming an established specialty worldwide. EM incorporates a unique body of knowledge and research is essential for its development.9 EM-specific journals were established in the 1980s and 1990s1 and the escalation in publications in the early 1990s reflected the emergence of its academic discipline.8 Lee et al 2 reported an overall increase in EM-related journals of 58% between 2000 and 2009. Initially, the journals were predominantly from USA and Europe and were published in English. Subsequently, other regions became involved and journals were more frequently published in other languages.2

Wilson and Itagaki3 reported that, prior to 2007, no investigation had characterised the type and quantity of worldwide EM research publications. It is clear from the present study that, over the 20-year period, EM research and its reporting has evolved considerably. Our finding of an increase in publication numbers is consistent with the conclusions of Wilson and Itagaki.3

Regions of article origin

In the early 2000s, the USA was the largest contributor to the EM literature. Between 1996 and 2005, 58.5% of original articles were from the USA, followed by the UK (8.4%) and Japan (4.5%).3 Also, Rosenzweig et al 9 reported that, between 1994 and 2003, article locations of origin were the USA (63%), non-USA (15%) and combined (22%). Furthermore, between 2008 and 2011, 53% of EM randomised controlled trials originated in the USA.10

Our findings indicate that the overall proportion of US publications is decreasing. Others have also reported this trend. Singer et al 11 reported that, between 1974 and 1997, there were increases in the proportions of non-US lead authors. Furthermore, Bounes et al 12 reported that original articles from the USA had decreased to 47.9%. Our finding of substantial increases in publications from ‘other’ countries is encouraging, especially the emergence of certain countries which each contributed zero to two articles in 1997 to more than 20 in 2017, for example, Turkey, South Korea, China and Japan. In many of these countries, EM has only recently been recognised as a specialty and the publication trends may indicate an emerging research infrastructure. This bodes well for ongoing development of the specialty worldwide.

It is notable that the UK was the only region where absolute and proportionate numbers of publications decreased. It is not known if the low 2017 numbers were an aberration, a true reflection of UK research trends, due to the upsurge of other countries or consequent on the potential selection bias associated with journal selection.

Reported article items

Over the study period, there was a significant increase in the median number of authors per article. This phenomenon has also been observed in EM scientific meeting abstracts11 and published articles.9 13 14 In an early EM study, Podolsky et al 14 found that the mean number of authors was small but increased from 1.1 in 1972 to 2.6 in 1981. More recently, Rosenzweig et al 9 reported that the percentage of articles having six or more authors increased from 12% in 1995 to 18% in 2003.

Most investigators went from having almost no coauthors at the start of the 20th century to having between two and seven by the end of it.13 Furthermore, the number of authors has increased since 1980 at a faster rate than the number of publications.13 Our findings are consistent with increases observed across all medical disciplines.15 In a Scopus database survey, the mean number of authors increased from 3.5 in 2003 to 4.2 in 2013.15 The reasons for these increases are unclear but may warrant further investigation.

In our study, the proportion of female first authors increased significantly over the study period. This is consistent with the report by Li et al.16 They examined four US EM journals and found that the proportion of female first authors increased from 9% in 1985 to 24% in 2005. The rising number of female first authors is likely related to the increasing numbers of female graduates.17 In our study, the proportion of female first authors was well short of 50%. The reasons for this are not known and further research of this issue is indicated.

The proportion of funded studies increased significantly although it has plateaued in recent years. Others have reported similar increases. In the early years, small proportions of studies were funded.9 14 18 19 Rosenzweig et al 9 reported that 22% of EM articles published between 1994 and 2003 had been funded, a proportion similar to ours around that period. Ernst et al 18 reported that 27% of studies published in 1994 had been funded. However, Birkhahn et al 19 reported that the proportions of published funded studies increased from 28% in 1994 to 36% in 2003.

Unfunded research remains the major source of the EM literature. However, a lack of funding is a barrier to quality of EM research6 and there are recommendations for grant activity to increase.12 Absolute proportions of studies being funded can be misleading as it takes into account neither its adequacy nor the efficiency with which the funds are used.

Early in our study period, the proportion of articles that reported IRB/Ethics Committee approval was moderate. Bounes et al 12 reported that, in 2010, only 68.5% of studies mentioned an IRB review. In our study, the reporting of IRB/Ethics Committee approval was almost universal by 2017. However, for some articles, it was unclear if IRB/Ethics Committee approval had not been obtained, was not necessary or whether it was simply not reported.

The greatest increase in any variable was the reporting of conflicts of interest. While no study reported these in 1997, the large majority did so in 2017. Ivanov et al 20 concluded that improved reporting of conflicts of interest and clarity around financial sponsorship represent a positive step forward. Journals now require authors to report conflicts of interests.

Study methodology

Multicentre studies can facilitate the recruitment of large sample sizes and increase external validity.21 It is encouraging, therefore, that the proportion of multicentre studies increased significantly. However, other reports are mixed. Wilson and Itagaki3 reported that the proportion of multicentre EM trials did not increase between 1995 and 2005. Conversely, Rosenzweig et al 9 reported that multicentre studies increased from 16% in 1994 to 26% in 2003. In a more recent 2010 study, Bounes et al 12 reported a proportion of 27.9%, a proportion similar to ours (31.7%) in 2012. Overall, the available evidence does suggest that multicentre EM studies are becoming more frequent. This is consistent with the global increase in multicentre trials.20 21 It is also encouraging that international collaborations have increased.

Experimental study designs, especially randomised controlled trials, are widely seen as a desirable study design. In this study, the number of experimental studies increased over time. Singer et al 11 also reported an increase in EM randomised trials between 1974 and 1997. However, our proportion of experimental studies remained unchanged at approximately 20%. Bounes et al 12 reported that only 9% of EM articles published in 2010 were randomised trials. It is difficult to compare these proportions as ours represents all experimental studies (not just randomised trials) and different journals were examined. It is unclear why the proportion of experimental studies did not increase over our study period. This is especially so as there was a significant increase in the proportion of prospective randomised trials published in The Lancet, New England Journal of Medicine and Journal of American Medical Association between 1988 and 2008.20

Over the study period, there were improvements in a range of aspects of study design, a finding also reported by others.11 Convenience and consecutive sampling decreased and increased, respectively, and may have helped minimise selection bias. Sample size increased significantly which may have afforded more study power. Although the reporting of a sample size calculation also increased, the magnitude was small. Our proportions were, however, substantially greater than those reported in early studies (range 0%–3%).11 22 The use of inferential statistics, CIs and p values also increased. Although the value of the p value is debatable, its increase may reflect the increase in inferential statistics observed in this study and one other.11

The proportion of studies that employed blinding decreased. Blinding is employed to minimise measurement bias and can increase methodological rigour. The reason for the observed decrease is not known, especially as many other aspects of the studies improved.

Limitations

The use of impact factors for journal selection was chosen as it is associated with higher methodological quality and broader article dissemination.23 However, impact factors are affected by differing article types in the denominator and numerator, self-citation and negative citation and by editorial citation suggestions.24

The findings may not truly reflect trends in EM research and do not include any information on qualitative studies. They are more likely to reflect publication trends which may be affected by journal publication bias. Also, the findings are subject to selection bias as only the top six EM journals were examined. The nature and quality of articles published in non-predominantly EM journals (eg, Resuscitation) and other EM journals is not known. It is notable that the majority of research published by emergency researchers is published in non-EM journals.3 Specific countries may have been under-represented if their national journal was not included, for example, Canadian Journal of Emergency Medicine. Non-random sampling of articles for examination was undertaken.

The reasons for the apparent changes over time in reporting items and methodology cannot be determined from our data. It is possible that the quality of these variables improved in line with the maturation of academic EM. However, revision of journal requirements and author guidelines are likely to have driven some changes, for example, the requirement for IRB/Ethics Committee approval. It is notable, however, that the quality of reporting in abstracts of EM trials is not improved by the use of the Consolidated Standards of Reporting Trials extension for abstracts.25

Data extraction was undertaken by a single author (JS) and measurement bias may have been introduced in the interpretation of some data items. However, the data validation exercise revealed no systematic errors. If a data item was not mentioned in an article, it was assumed it was not done, for example, sample size calculation. This may have led to an under-reporting of some items. Occasionally, the first author’s given name did not clearly indicate their sex, leaving this item data set incomplete.

Recommendations

Ongoing surveillance of the EM literature is recommended to monitor contemporary trends. Rather than absolute numbers of publications per region, a better measure of productivity may be the number of articles per 100 emergency physicians. However, such a productivity measure would be difficult to determine. Similar studies examining different journals are recommended. These could include a broader range of EM-specific journals as well as EM articles published in general, high-impact journals. Further research should examine the involvement of women in academic EM, the potential for increasing international collaborations, the apparent decline in UK output and the increasing contributions of other countries.

Conclusion

Clear trends over time are apparent within the EM literature. The range of contributing countries has increased and the previous dominance of the USA and the UK is being challenged. The proportion of female first authors has increased and there is more reporting of research accountability including IRB/Ethics Committee approval, conflicts of interest and author contribution. There is a greater rigour in both research methodology and results presentation with increases in multicentre and international studies, improved patient sampling techniques, increased sample sizes and improved statistical techniques and reporting. However, the proportions of experimental studies have not changed and the use of blinding has decreased.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Related Data

Footnotes

  • Contributors All authors contributed to the design of the study and its protocol. JS and DMT undertook data collection. All authors participated in collation of the data and its interpretation. DMT undertook the statistical analysis and supervised the study overall. All authors contributed to development of the manuscript and all authors have approved the final version.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement All study data are available upon reasonable request from the corresponding author and may be reused as required.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles