Posters Presented at North American Skull Base Society 2016–2018: What Factors Influence Their Publication?

Abstract

Objective  Research productivity impacts an individual's academic credentials and serves to advance the field of neurosurgery at large. Poster presentations allow researchers to share preliminary results with respected colleagues; however, more critical is the ability to publish peer-reviewed articles. Key factors that lead posters to journal publication are not well understood and difficult to quantify. This study investigates the association between bibliometrics of authors who presented posters at the North American Skull Base Society (NASBS) meeting and odds of journal publication.

Methods  Posters from the 2016 to 2018 NASBS archive were reviewed. Hirsch-index (h-index) of first (FH) and senior (SH) authors, research type, research topic, and number of poster authors (nAuthPost) were collected. For posters published as journal articles, number of days from poster presentation to publication (nDays), number of authors in published articles (nAuthArt), and journal impact factor (JIF) were recorded.

Results  One-hundred sixty-nine of 481 posters (35.1%) were published as articles. Median FH and SH for published versus unpublished posters were 7 versus 5 ( p  = 0.01) and 29 versus 19 ( p  < 0.001), respectively. When adjusted with multivariate regression, only SH ( p  < 0.001) and nAuthPost ( p  = 0.001) were significantly associated with odds of publication. Median (interquartile range [IQR]) nDays was 361 (394). Increased authors from poster to article ( p  = 0.017) and lower FH ( p  = 0.08) were correlated with increased time to publication. Median (IQR) JIF for all publications was 1.723 (1.068).

Conclusions  Bibliometrics such as h-index and number of authors from posters can help objectively characterize and predict future success in research productivity.

Introduction

The discipline of modern neurosurgery has a relatively brief yet immensely rich history of clinical innovation through academic productivity. Since the birth of neurosurgery as its own distinct specialty in the early 20th century, professional organizations like the American Association of Neurologic Surgeons (AANS) and Congress of Neurologic Surgeons (CNS), as well as specialty-specific journals such as The Journal of Neurosurgery and Neurosurgery , were formed out of a desire for neurosurgeons to connect and educate one another more effectively. ¹ Research output is now valued at every level of training: from prospective residents to fellows and attendings. According to the National Resident Match Program, neurosurgery ranked the highest in mean number of abstracts, presentations, and publications among allopathic seniors who matched into the specialty in 2018. ² From 2005 to 2015, the average graduate from an accredited neurosurgery residency in the United States had 8.3 publications just during residency. ³ More surprisingly, Rallo et al uncovered a trend of increased average research productivity among neurosurgery residents with the absence of a formally designated “research requirement.” ⁴

At present, an impressive volume of literature has been produced by and is available to neurosurgeons. The “core” journals exerting the most influence on the neurosurgical community have doubled since the mid-1970s, now encompassing The Journal of Neurosurgery, Neurosurgery, Spine, Acta Neurochirurgica, Stroke, and Journal of Neurotrauma . ^{1 5} From just 1996 to 2009, a total of 53,425 PubMed-indexed articles specific to neurosurgery were published worldwide. ⁶ Consequently, there is an ever-growing need to critically evaluate these research contributions and to characterize the academic caliber of this prolific group of surgeon-scientists. Consensus reporting guidelines and an extensive network of peer reviewers provide much-needed quality control for these articles. ^{7 8 9} Meanwhile, total citation count, number of papers published, and the more recent h-index are among the various bibliometrics used to objectively compare academic productivity among the authors. ^{10 11 12}

There has been growing interest within the neurosurgery community to better illustrate emerging trends in neurosurgical training and academic promotion. Author bibliometrics have been especially scrutinized due to their increasing role as an objective measure of academic achievement outside of the status quo of standardized exam scores in residency admissions. ¹³ It has been shown that a resident's total number of publications, number of first-author publications, and program size were predictive of residents choosing and succeeding in an academic career path. ¹⁴ Further along in training, peer-reviewed metrics like h-index are used for the academic promotion and tenure evaluations. ^{15 16} Thus, bibliometrics may hold the key to elucidating the otherwise nebulous and often lengthy process of publishing papers.

Anyone involved in long-term research is familiar with the trials and rigmarole of transforming a poster presentation into a full-fledged journal article. Research conferences and regional meetings provide opportunities for investigators, clinicians, and students to disseminate their original research and exchange ideas. Comments and constructive criticisms from conference attendants can then be used to formulate salient points of discussion within a manuscript. Despite widespread anecdotal understanding of the publishing experience among those in academia, there is a paucity of actual research characterizing the objective factors associated with progression of poster to publication. In this study, we analyzed the bibliometric data of poster presentations from the North American Skull Base Society (NASBS) from 2016 to 2018 to determine the factors that impact (1) odds of publication, (2) time from presentation date to publication date, and (3) journal impact factor (JIF) of the destination journal.

Methods

Data Collection

In total, 547 poster presentations from 2016 to 2018 were available on the NASBS online archive. The following variables were collected based on each poster presentation: research type (clinical, basic, or case report) research topic (surgical approach, oncology, vascular, trauma, imaging, or all other topics that do not fall into the previous categories), and date of presentation. Additionally, the following author bibliometrics were collected: h-index of first author (FH), h-index of senior author (SH), and total number of authors listed on the poster (nAuthPost). Scopus was used as the primary search engine to identify FH and SH.

To identify posters that were ultimately published as a journal article, the main keywords from the poster title with and without author last names were searched on Google Scholar , PubMed , and Google. If a similar article was identified, the content of the article was reviewed to confirm that the same topic and data from the original NASBS poster were being discussed. If a published journal article was found, the following variables were then collected: date of publication (E-publication date if listed), total number of authors listed on the article (nAuthArt), journal name, and JIF. The time needed to publish (nDays) was calculated as the difference between the number of days from poster presentation date to article publication date ( Fig. 1 ) .

Fig. 1.

Data collection method of North American Skull Base Society (NASBS) posters from 2016 to 2018.

Statistical Analysis

Pairwise association studies were conducted between all recorded variables with primary focus on variables of main interest: publication status (yes vs. no), nDays, and JIF. Nonparametric methods were used due to skewed distributions of variables. Series of chi-squared test, Wilcoxon rank sum test (Mann–Whitney U test), and Kruskal–Wallis test were applied to test for associations. For two continuous variables, Spearman's correlation coefficients (ρ) along with the 95% confidence intervals and p -values were computed.

To investigate the effect of a variable when adjusting for potential confounding variables, multivariate regression models were applied. Multivariate logistic regression was applied to publication status. Multivariate linear regressions were applied to nDays and JIF. Variables that showed significant pairwise associations with response variables were included as explanatory variables in the multivariate regression models.

All data processing and analysis were performed using R v.4.0.3 ( http://r-project.org ) via Rstudio v. 1.3.1093 ( http://rstudio.com ).

Results

Summary Statistics

Overall, 481 out of 547 posters presented at NASBS during 2016 to 2018 were analyzed. Sixty-six articles were excluded due to withdrawal of the poster from the meeting, incomplete data, unretrievable FH or SH on Scopus , or unidentifiable JIF for published posters.

In each year from 2016 to 2018, the total numbers of posters listed in the NASBS archive were 143 (29.7%), 150 (31.2%), and 188 (39.1%), respectively. Median (IQR = interquartile range) of FH was 6 (6), while that of SH was 22 (14). Median (IQR) of nAuthPost was 6 (3). With respect to research type, there were 227 (47.2%) posters classified as clinical research, followed by 178 (37.0%) case reports and 76 (15.5%) classified as basic science research. With respect to research topic, the three most common neurosurgery topics were surgical approach, oncology, and vascular with 187 (38.9%), 139 (28.9%), and 50 (10.4%) posters, respectively ( Table 1 ).

Table 1. Summary statistics for categorical measures.

Variable		Publication				nDays		JIF
		Yes	No	Total	p χ	Median (IQR)	p κ	Median (IQR)	p κ
Year	2016	57 (39.9%)	86 (60.1%)	143	0.3213	438.5 (423.2)	0.552	1.723 (1.068)	0.283
	2017	52 (34.7%)	98 (65.3%)	150		405 (396)		1.723 (0.574)
	2018	60 (31.9%)	128 (68.1%)	188		311 (353)		1.886 (0.771)
Research type	Basic	34 (44.7%)	42 (55.3%)	76	0.0018*	489.5 (404.7)	0.1657	1.886 (1.413)	0.176
	Clinical	90 (39.6%)	137 (60.4%)	227		325 (392.2)		1.723 (1.029)
	Case report	45 (25.3%)	133 (74.7%)	178		393 (383)		1.723 (0.72)
Research topic	Surgical approach	64 (34.2%)	123 (65.8%)	187	0.5573	352 (338.5)	0.6029	1.723 (0.946)	0.8119
	Oncology	54 (38.8%)	85 (61.2%)	139		430 (376)		1.723 (1.058)
	Vascular	16 (32%)	34 (68%)	50		414 (756)		1.886 (1.203)
	Imaging	14 (36.8%)	24 (63.2%)	38		607 (431.5)		1.886 (1.118
	Trauma	1 (10%)	9 (90%)	10		282 (NA)		NA (NA)
	Others	20 (35.1%)	37 (64.9%)	57		323.5 (354.6)		1.723 (1.579)

Abbreviations: IQR, interquartile range; JIF, journal impact factor; NA, not applicable due to small sample size; nDays, number of days from presentation to publication for days > 0 only; p ^κ , p -value from Kruskal–Wallis test; p ^χ , p -value from chi-squared test; *, significant for p -value <0.05.

Overall, 169 of 481 posters (35.1%) were published as articles. For published articles, median (IQR) time to publication as represented by nDays was 361 (394) and JIF was 1.723 (1.068) ( Fig. 2 ). Based on the calculation of nDays for published articles, it was noted that 28 (16.6%) posters were published as articles before the date of poster presentation ( Fig. 2 ). nDays for these posters ranged from −465 to −2 with a median value of −72 (negative number of days implying publication date preceding the presentation date). We excluded these posters from final statistical analysis of nDays and JIF.

Fig. 2.

Spread of number of days from poster presentation date to publication date. Left of the red bar notes the frequency of posters that were published prior to poster presentation.

It was also observed that 88 (62.4%) posters had different numbers of authors in poster and journal articles. Since the difference in numbers of authors between article and poster (nAuthDiff = nAuthArt-nAuthPost; ranging from −4 to 10) could imply substantial changes in the study during the period between poster presentation and publication, nAuthDiff was used as an additional explanatory variable in analysis of nDays and JIF.

Association Studies

The proportion of posters that reach publication increases each year of the study but did not reach statistical significance ( p  = 0.321); however, there is a significant difference between the publication rates of posters among research types ( p  = 0.01). The publication rates were 44.7, 39.6, and 25.3% across basic science, clinical research, and case reports, respectively ( Fig. 3 ). There is no significant correlation between the publication rate and research topic ( p  = 0.557).

Fig. 3.

Number and percentage of posters that were published as articles, organized by research type.

Publication rate, nDays, and JIF were analyzed through single variable analysis. FH and SH had significant positive correlations with publication rate. The median (IQR) FH and SH for published versus unpublished posters was 7 (7) versus 5 (6) ( p  = 0.01) and 29 (30) versus 19 (13) ( p  < 0.001), respectively. The median nAuthPost for published versus unpublished posters was 5 (3) versus 5 (3) ( p  < 0.0001) with averages being 5.82 (SD ± 2.59) versus 4.77 (SD ± 1.96) ( Fig. 4 ). Only nAuthDiff had a significant positive correlation ( p  = 0.017) with nDays. SH and nAuthArt both had significant positive correlations with JIF ( p  < 0.001 and p  = 0.031, respectively ( Table 2 ).

Fig. 4.

Comparison of author bibliometrics for published versus unpublished posters.

Table 2. Summary statistics for numerical measures.

Variable	Publication			Days until publication		Journal impact factor
	Yes Median (IQR)	No Median (IQR)	p w	Spearman's ρ (95% CI)	p	Spearman's ρ (95% CI)	p
FH	7 (7)	5 (6)	0.0146*	−0.147 (−0.315, 0.021)	0.0817	0.145 (−0.031, 0.320)	0.1020
SH	29 (30)	19 (13)	<0.0001**	−0.052 (−0.227, 0.123)	0.5424	0.404 (0.242, 0.565)	<0.0001**
nAuthPost	5 (3)	5 (3)	<0.0001**	−0.132 (−0.305, 0.042)	0.1201	0.135 (−0.043, 0.312)	0.1283
nAuthArt	5 (3)			0.027 (−0.141, 0.195)	0.7533	0.19 (0.020, 0.360)	0.0311*
nAuthDiff	0 (2)			0.201 (0.030, 0.372)	0.0168*	0.042 (−0.132, 0.217)	0.6344
JIF	1.723 (1.068)			0.038 (−0.220, 0.144)	0.6678
nDays	361 (394)					0.038 (−0.220, 0.144)	0.6678

Abbreviations: CI, confidence interval; FH, first author h-index; IQR, interquartile range; JIF, journal impact factor; nAuthArt, total number of authors on article; nAuthDiff, nAuthArt-nAuthPost; nAuthPost, total number of authors on poster; nDays, number of days from presentation to publication for days > 0 only; p , p-value from Spearman's correlation; p ^w , p-value from Wilcoxon rank sum test for independent samples; SH, senior author h-index; *, significant for p -value <0.05; **, significant for p -value <0.01.

Multivariate Regression Analysis

Multivariate regression was applied to odds of publication, nDays, and JIF to adjust for confounding variables. When we controlled for all relevant variables from associations studies on odds of publication, only SH ( p  < 0.001) and nAuthPost ( p  = 0.001) remained statistically significant. FH ( p  = 0.72) and research type ( p  = 0.10) that showed significant correlation from association studies were no longer significant. The estimated regression coefficient for SH and nAuthPost with respect to publication rate were 0.0193 and 0.1524, respectively. Because of the logarithmic nature of the analysis, this implies that as SH increases by 10, we expect a 21.3% increase in chance of the respective poster reaching publication on average. Similarly, as nAuthPost increases by 1, we expect a 16.5% increase in chance of the respective poster reaching publication on average.

When we controlled for all relevant variables from association studies on nDays, both FH ( p  = 0.039) and nAuthDiff ( p  < 0.001) remained significant with multivariate linear regression analysis, consistent with the results from pairwise association studies. The estimated coefficients for FH and nAuthDiff were −9.671 and 49.283, respectively. As FH increases by 1, nDays decreases by 9.671 on average. As nAuthDiff increases by 1, nDays increases by 48.283 on average.

When we controlled for all relevant variables from association studies on JIF, both SH ( p  = 0.004) and nAuthArt ( p  = 0.016) remained statistically significant with multivariate linear regression analysis. The estimated coefficients for SH and nAuthArt were 0.0117 and 0.0753, respectively. As SH increases by 10, JIF increases by 0.1169 on average. As the nAuthArt increases by 1, JIF increases by 0.0753 on average. Table 3 displays the full results of the multivariate analysis for odds of publication, nDays, and JIF.

Table 3. Summary statistics of multivariate regression analysis.

	Odds of publication			Days until publication			Journal impact factor
	Regression coefficient	p *		Regression coefficient	p		Regression coefficient	p
SH	0.0193	<0.001	FH	−9.671	0.039	SH	0.0117	0.004
nAuthPost	0.1524	0.001	nAuthDiff	49.28	<0.001	nAuthArt	0.0753	0.016

Abbreviations: FH, first author h-index; nAuthArt, total number of authors on article; nAuthDiff, nAuthArt-nAuthPost; nAuthPost, total number of authors on poster; p *, p -value from multivariate logistic regression; p , p -value from multivariate linear regression; SH, senior author h-index.

Discussion

The purpose of this study was to investigate the impact of author bibliometrics and poster characteristics that effected the odds of publication, time until publication (nDays), and JIF of posters presented at NASBS. Briefly, it was found that SH and nAuthPost were each positively related to the odds of publication. It was also found that a higher nAuthDiff and lower FH were both positively related to nDays. SH and nAuthArt had significant positive correlation on JIF.

It is well known that publication of novel research often requires extensive rounds of edits and feedback among the researchers. There are also the uniquely human factors, such as delays in communication and technical issues in editing, which impact the publication process. There are multiple factors that affect journal acceptance that may prove difficult to isolate and quantify. To our knowledge, this is one of the few papers in neurosurgical literature that has explored factors that could be linked to publication potential and even speed. Our study showed that specific bibliometrics just from poster presentations have a significant role in affecting publication outcomes.

People often join a research team based on the reputation of the senior author, but there are few studies that objectively characterize an author's reputation and verify whether it can truly predict future publication productivity. Interestingly, in line with findings from analogous studies in other specialties like otorhinolaryngology, orthopaedic surgery, psychiatry, and emergency medicine, we found that higher bibliometrics positively correspond with odds of publication. ^{17 18 19 20} For example, higher SH (i.e., greater research productivity and impact of senior author) was found to yield higher odds of publication for a given poster.

This study found that having a larger number of authors on a poster yielded a significant increase in odds of publication, while also increasing the time it took to ultimately publish the work. This suggests that inviting a greater diversity of input from various researchers may be helpful in increasing the chance of publishing the findings of a project, albeit at the expense of timeliness, which is in line with the latest studies in neurosurgery focusing on faculty productivity. ^{12 21} The implied importance of collaboration by the number of authors is further evident in other studies that have shown multicenter collaborations are more likely to be published as journal articles than that of single-centers. ²² Another finding that suggests the positive effect of having a larger team was our measurement of the difference in authors from poster to journal article, as reflected by nAuthDiff. The effect of having more authors on publications has been studied before in journals such as Journal of Neurosurgery and Neurosurgery . In one study, Harsh et al showed that in both Journal of Neurosurgery and Neurosurgery, increased citation scores were directly correlated with increased author count. ²³

This study also found that for published articles, FH was significantly negatively associated with nDays, while SH was not. This demonstrates that the senior author does not affect the speed of publication, while the first author may play a relatively greater role in maturing the poster into a publication. Furthermore, this finding suggests that while SH may be important in understanding one's potential to publish a project in a macroscopic sense, the onus of taking a project “to the finish line” may fall more on the first author. However, it's also likely that quantifying every single variable that influences speed to publication may be an inherently difficult, if not impossible task. Some research works could be in their nascent stages of development at the time of poster presentation, thus requiring considerable long-term effort to withstand a more rigorous review by the journal editors. Conversely, some projects could be ostensibly complete by the time of poster presentation, so much so that they may have already submitted their work for consideration by a journal for publication while waiting a favorable review of a pending abstract submission. This may explain why we found a surprisingly large number of abstracts that had a “negative” time to publication. This may simply indicate that the speed with which a publication is achieved depends on too many factors to name.

Finally, our study found that SH and nAuthArt had a positive significant effect on JIF. This suggests that increased level of expertise and pedigree from an increased number of authors may qualify the article for consideration by more prestigious journals, as portrayed by JIF. However, such a suggestion is only valid in the context of JIF being a reliable proxy for the quality of published research. For example, one could argue that the median JIF of published authors in this study was only 1.723, appearing especially low when compared against internationally reputable journals like Science which had a JIF around 41 in 2019. ²⁴ However, JIF has been found to be an inconsistent indicator for the “quality” of research, especially among highly specialized research subjects like neurosurgery, where the “citability” of a project may be intrinsically constrained by the subject of the project itself. ²⁵

Our study has a few notable limitations. First, the posters included in this study reflect the work presented from 2016 to 2018 at just one conference. While the raw number of posters analyzed is quite large, there may be concern as to the generalizability of our findings given the timespan and single meeting data this analysis relied on. However, given the highly respected reputation of the NASBS meeting within the neurosurgical community, as well as the wide array of research topics and subspecialties represented at this meeting, we do not believe that our data misrepresents the neurosurgical community in any significant way. In fact, our summative statistics in publication rates fall largely in line with that of similar studies that collected data from the AANS, the CNS, and the International Society for Pediatric Neurosurgery meetings. ^{26 27 28} Moreover, our study on publication rates can be expanded further in future studies by encompassing other presentation types such as electronic posters to podium presentations as the publication rates could vary among these subtypes. ²⁹

Second, the cross-sectional study design limits any causative conclusions to be drawn. For example, as mentioned previously, inconsistencies between JIF and quality of research would weaken such relationship. Thus, we found that while there is preliminary benefit finding correlations between the variables studied, further work to better characterize the relationship between these variables is required to draw meaningful causative conclusions.

Lastly, the distinction between the research topics represented in the NASBS meeting was determined by only the first author of this study. The exact differences in topic were not clearly delineated by the NASBS committee itself, thus raising the question as to whether the research at this meeting was properly characterized by this study. However, the first author used a system of categorization depending largely on their own thorough review of the existing neurosurgical literature and specific research topics delineated by similar meetings. Thus, it is unlikely that specific subspecialties of research were misidentified in any significant way during the process of data collection.

Conclusion

We investigated various factors associated with publishing a body of research within the field of neurosurgery. We found that higher h-index values of first and senior authors may portend a favorable outcome in terms of publication. The findings of our study suggest that the guidance of an experienced senior author with a robust publication history, as well as collaborative effort between various researchers, may significantly increase the likelihood of publishing a project presented at a research meeting. These implications underline the importance of bibliometric data in objectively predicting and characterizing academic productivity at various levels of training.