Saving time and money in biomedical publishing: the case for free-format submissions with minimal requirements

In our analysis, we identified significant diversity in initial formatting requirements. For each category, a substantial proportion of journals did not specify requirements (or we were unable to locate them). In particular, the following percentage of journals did not indicate information related to running titles and their length (65%), length of the title (64%), maximum number of references (58%), number of figures and tables (48%), main text length (34%), manuscript structure (27%), abstract structure (21%), and abstract length (14%). The lack of information regarding these parameters did not necessarily promote free-format submission; for example, we differentiated between journals explicitly stating no word limit for the main text (3%) and those for which this information was not available (or we were unable to locate it) (34%). This high percentage of missing information suggests that submission guidelines are incomplete and/or difficult to navigate, adding an additional burden to the already cumbersome process of (re)formatting a manuscript.

We categorized the requirements for maximum length and overall structure; where these were specified, running-title lengths tended to be between 50 and 59 characters (15%), title lengths were between 150 and 200 characters (13%), abstract word limits were between 200 and 300 words (65%), main text word limits were between 2500 and 4000 words (32%), the number of items were 6–7 (19%), and the number of references were between 45 and 60 (17%) (Fig. 1). With regard to abstracts, 21% did not specify a structure (or we did not find it), 22% required unstructured abstracts, and the remaining 57% required a structured abstract, with Background, Methods, Results, and Conclusions being the most frequently specified sections (Fig. 2). In terms of manuscript structure, the Introduction, Methods, Results, and Discussion format (known as IMRaD) was the most frequently specified (43%) (Fig. 3). While journals belonging to the same family (e.g., Journal of the American Medical Association [JAMA] family, Science family) usually follow the same formatting conventions, the vast majority of journals we analyzed had a unique set of requirements. We acknowledge that the selection process of leading biomedical journals (i.e., those with the highest CiteScore values) might bias our findings; however, we find it unlikely that selecting journals with lower average scores or selecting journals according to other criteria would significantly impact our conclusions regarding either the diversity or the variation in submission requirements.

In our calculation of the number of possible formats, we obtained a staggering number of ~ 6.3 billion possible unique sets of formatting requirements. In the next section, we review and estimate the financial cost of reformatting articles for a new submission.

We obtained average annual salaries in the European Union and the United States for PhD students (33,000 USD), postdoctoral researchers (43,000 USD), assistant professors (56,000 USD), associate professors (67,000 USD), and full professors (86,000 USD), and then calculated an average researcher salary of 57,000 USD. With an estimate of ~ 1750 work hours per year, we arrived at an average hourly salary of 32.6 USD. Using four hours of reformatting time, we estimated the cost of reformatting one article to be approximately 130 USD. In 2000, the number of published entries on PubMed was ~ 0.5 million, while in 2021, a record number of ~ 1.8 million were published. Assuming one rejection per paper (on average), we calculated that ~ 230 million USD were lost in 2021 alone. Should the current practice remain unchanged within this decade, we estimate ~ 2.5 billion USD could be lost between 2022 and 2030—solely due to reformatting articles after a first editorial desk rejection. These estimates are depicted in Fig. 4, which also shows projected losses with an error bound based on a previously reported minimum estimate of one hour [14, 16] and a maximum estimate of 14 h per article [12]. The extrapolated financial loss for the period 2022–2030 roughly equates to 75 million hours of lost time. Calculating based on 1750 work hours per year, this means that there could be a cumulative delay of approximately 43,000 years in the public dissemination of results—solely due to the reformatting of papers from 2022 to 2030.

For many authors, the first step in publishing a scientific article is identifying the most appropriate journal(s) to disseminate their research, and then formatting the manuscript according to the target journal’s submission guidelines. The percentage of papers rejected without peer-review can vary across journals, disciplines, and publishing models, with some journals rejecting more than two-thirds of all first submissions at the editorial level [36]; this is known as a ‘desk reject’. When an article is returned to the corresponding author, they often want to resubmit to another journal quickly to avoid delays in dissemination. Once the author identifies another appropriate journal, they must then understand its specific submission requirements. It is at this stage that reformatting can be particularly cumbersome and time-consuming; numerous tasks are involved in reformatting manuscripts to adhere to the second journal’s specific requirements, as well as additional tasks related to the submission process itself.

Sobani et al. estimated that the total time used to resubmit one article is ~ four hours for submitting authors [15]. Others have estimated that reformatting a single manuscript can take from one to 14 h or more [12, 14, 16]. In practice, the time needed to reformat can be substantially longer, especially when the resubmission requires substantive reductions to the manuscript’s word count. For instance, an article rejected by Science (max. 4500 words) cannot be immediately resubmitted to The Lancet (max. 3500 words) or The New England Journal of Medicine (max. 2700 words). Such substantive differences in length also exist among the specialized journals, a specific barrier that was mentioned in our interviews. One researcher said, “For a typical manuscript, [reformatting takes] a couple of days … If you have to rewrite a lot because you have to cut 2,000 words, then it probably takes more time”.

When significant rewriting takes place, the corresponding author will typically circulate a revised manuscript to any co-authors in order to obtain their approval for the new submission. In these cases, one of the researchers said, “It could take a week [because] if we need to add an extra section, then we need to talk again between the different authors”. Reducing the word count is not the only challenge. Often, additional changes need to be implemented throughout the manuscript; e.g., the number of figures and tables, adjustments to the abstract text/structure, reference style, or the title. As one researcher put it, “When you need to reformat, it takes forever, because sometimes [journals] are really picky with some detail that you really have to follow, and they won’t send it out to review if it doesn’t fulfil their formatting criteria”. However, we did find that approximately 11% of the journals in our study provided a template on their website to guide authors with formatting.

Our estimate of the financial loss per article was lower than others found in the literature. A report by LeBlanc et al. surveyed ~ 400 individuals on the topic of resubmitting papers and arrived at an estimate of 477 USD lost per paper. This estimate took into account 14 h of work per manuscript and a median number of two resubmissions per paper based on survey responses [12]. Using this estimate, they calculated a loss of 1908 USD per year in research funding for each author. Sobani et al. estimated 2.6 million hours of lost academic work per year based on half a million biomedical articles annually, three hours of resubmission time, and one rejection per submitted article [15]. Another report by Khan et al. arrived at approximately 1.5 million hours lost in reformatting rejected articles per year, based on an average rejection rate of 62% for 2.5 million published papers every year, and one hour spent on reformatting time per manuscript [14]. Single author Budd estimated the time lost due to resubmission after editorial rejection by the journal Nature alone (desk-rejection rate: 92%). They concluded ~ 10,000 h of work time were lost per year, based on one hour of work per article to be resubmitted elsewhere [16]. Finally, Jiang et al. estimated the annual cost of reformatting manuscripts to be ~ 1.1 billion USD, a fivefold increase on our calculations [37].

We believe that our calculations are conservative estimates; the real cost in lost time and money is likely to be higher. Research is a collaborative process, and reformatting articles results in wasted time for everyone in the finite academic ecosystem, including all authors, peer-reviewers, editors, publishers, and ultimately, the public.

Thus far, we have shown that reformatting manuscripts takes an enormous toll on the research ecosystem in terms of both time and money. Next, we propose a simple solution that could help to prevent these projections from being realized.

Based on our investigation, we have identified two possible solutions (Fig. 5) that could significantly decrease the time and money lost due to heterogeneous initial-submission guidelines. The first suggestion is to adopt the same universal guidelines across all biomedical journals. The second is to remove the requirement for specific formatting at first submission; in this case, journals could request specific formatting guidelines but only after acceptance [14].

Both solutions have benefits and drawbacks. While universal guidelines would drastically reduce the burden on authors, this approach could force journals to lose their distinctive ‘look’ (house style) as well as remove flexibility specific to their subject matter or print format. But, in practice, it is unlikely to impact the editorial process. For example, one editor told us: “There are journals that in principle have formats but don’t care. […] The editor hardly ever cares about how long [the text] is or how many figures it has on first instance”. The introduction of universal submission guidelines could also be a costly endeavor for journals/publishers, which could lead to resistance, as another editor noted: “It would be a very high lift to get all publishers together and agree on universal submission guidelines. And people use different submission systems that have different dependencies. […] If you’re an editor at one journal, you’ll know that in a way we’re at the tyranny of very clunky submission systems that are not very good. […] Changing that is very costly for individual publishers.”

The second option of a free-format first submission is now easier than ever. An increasing number of journals are published online-only, so articles in these journals require less post-processing upon acceptance, and there are fewer restrictions on length. Indeed, the editors we interviewed indicated that there is an increasing trend towards free-format initial submission. One said, “My understanding [is that] less and less journals […] have rigid format restrictions. I think most journals are […] moving towards initial-submission agnostic views”. Another said, “Some journals have introduced what they call ‘hassle-free’ submission [and] our journal is moving towards it”. Despite this development, the number of journals with free-format initial submissions is still estimated at only around ~ 5% of all biomedical journals [37].

With regard to a free-format approach, several editors raised concerns about the burden of having to review papers that were unreasonably long, poorly structured, or just badly written. For example, one editor noted, “Sometimes you get submissions […] which are, basically, somebody’s whole PhD thesis just dumped there”. Another editor said: “Every so often somebody sends me an article that […] looks like […] somebody had a dictaphone in their hand and just spoke to it for 20 minutes and got someone to transcribe it”. For a format-free approach to be successful, this same editor reflected that, “It would depend on everybody submitting having a vague sense of what a scientific article looks like”. Thus, the key point in favor of a free-format approach was nicely summarized as such: “I like the articles to be standardly structured. But there is flexibility around that. It doesn’t always have to be the same. It should be recognizable as a scientific article.”

Our proposed minimum set of requirements, presented in Table 1, was developed based on the current submission requirements of the 302 biomedical journals we reviewed as well as on our interviews with the editors. We propose no requirements regarding titles, running titles, title pages, abstract structure, item formatting, maximum number of items, maximum number of references, or reference styles at first submission—we believe that these details can be addressed later during the ‘revise and resubmit’ stage, or even after acceptance. As we learned from our interviews with the editors, it is at the resubmission stage when most journals tend to commit to shepherding manuscripts towards acceptance and publication: “Where we do a lot of checking is when the revised version comes back in. Because we have a strong commitment to revisions”. Similarly, another editor said, “We don’t bother with any format until we are inviting for revision. When we invite for revision, we are pretty confident. […] After peer-review, that’s when we make a commitment.”

As outlined in Table 1, even the two requirements that we propose allow for flexibility. In cases where a journal may have additional specific requirements/sections (e.g., clinical relevance, key messages, graphical abstract), we encourage them to ask authors to upload these parts in separate files to keep the main manuscript easy to resubmit to another journal. If print journals with space constraints need to enforce word limits that are stricter than those defined in our minimum set of requirements, then we propose that these journals still adopt the same basic requirements and also that they should also be willing to review longer papers. However, they should clearly explain to the submitting author(s) that word count will need to be reduced upon acceptance in order to adhere to their pre-specified maximum lengths.

Furthermore, we recommend that online content-management systems incorporate auto-fill tools that require an initial upload of the manuscript file and then use the available information to populate fields in the system. Manually copying titles, abstracts, and key words as well as filling in authorship and funding information can substantially prolong the submission process. For example, one researcher explained, “In the submission system, there are a lot of steps that we need to go through. […] You have to write the name, the address—things that you already have on the first page of the article. […] And the names are not recognized by the system. These are unnecessary things that really waste time.” These manual processes can also be prone to error, especially when manuscripts have many authors and/or funders. Oftentimes, following a submission of a manuscript, all co-authors need to register their profile on the journal-submission site, provide information related to expertise, keywords, and fill in other fields; e.g., the newly introduced mandatory diversity and inclusion surveys (which we fully support—after acceptance), and then proceed to approve the corresponding author’s submission. When a manuscript has multiple co-authors, such processes can cause significant delays. One of the editors we interviewed had had similar experiences on both sides of the editorial desk, explaining, “I’ve published articles […] where it is a bloody pain trying to work it out how to format it for [a journal], and jump through the ludicrous hoops of their content-management system. […] I’ve used most of the content-management systems or editorial management systems, and found that all of them tripped me up at various points”. Artificial-intelligence tools already implemented by some journals can automatically recognize and process these sections, which can save significant time for the corresponding author. If journals do not have the means to incorporate such tools, then we suggest requesting this detailed information only at the resubmission/acceptance stage.