Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Is statistician involvement as co-author associated with reduced time to publication of quantitative research in general medical journals? A bibliometric study

Is statistician involvement as co-author associated with reduced time to publication of... Abstract Objective We aimed to compare the number of submissions until acceptance and the time to publication between articles co-authored and articles not co-authored by statisticians. Methods We randomly selected 781 articles published in 2016 in 18 high impact factor journals of general internal medicine and primary care. For each article, we retrieved its date of submission to the journal and its first publication; we also contacted its corresponding author and asked about the number of submissions necessary from the first submission to a journal until acceptance and whether the article was co-authored by a statistician. After having excluded qualitative studies, we compared the articles co-authored with those not co-authored by statisticians in terms of number of submissions and submission-to-publication time, using negative binomial and Cox regressions, adjusted for intracluster correlations. Results One hundred fifty-eight authors completed the questionnaire (20%); 136 articles with quantitative design were included in the study. Overall, 63 articles (46%) were co-authored by statisticians. There was no statistically significant difference in the number of submissions (statistician group: mean 2.1 (SD 1.1) versus 2.2 (SD 1.2), P value 0.87). By contrast, we found a statistically significant difference in the submission-to-publication time (statistician group: median 211 days [interquartile range (IQR) 171] versus 260 (IQR 144); hazard ratio 1.44 (95% CI 1.01–2.03), adjusted P value 0.04). Conclusions Papers co-authored by statisticians have a shorter time to publication. We encourage researchers to closely involve statisticians in the design, conduct and statistical analysis of research, not only to ensure high standards of quality but also to speed up its publication. Delay of publication, general medical journals, number of submissions, publication speed, retrospective study, statistician, time to acceptance, time to publication Introduction Despite a continuous rise in the number of journals publishing research, the increase in scientific productivity is expected to negatively impact the workload of journals and consequently the acceptance rate of submitted manuscripts and the publication speed of accepted manuscripts (1). This trend is problematic because high rejection rates and long delays of publication could discourage scientific researchers from publishing their research in the traditional publishing system; it could also delay the progression of academic careers, as well as postpone the dissemination of new knowledge or the implementation of new interventions. In order to avoid misuse of statistics in medical research, it was suggested that active participation of statisticians should be required at all stages of a study project (2–5). It has been shown that an effective and close collaboration between statisticians and clinical investigators may improve the methodological quality of randomized trials (6–9) and observational studies (10). Yet, despite the large number of bibliometric studies from a wide range of disciplines aimed to quantify the publication speed of scientific research (1,11–14); there are currently no data on the influence of statistician involvement on the publication speed of medical research. The objective of our study was therefore to assess whether papers co-authored by statisticians and submitted to general medical journals (i.e. journals of general internal medicine and primary care) require lower numbers of submissions and are associated with shorter times to publication. For this project, we planned to restrict the analysis to quantitative studies, where the role of the statistician may be particularly important; indeed, studies with qualitative design use non-statistical methods of analysis. Methods Search strategy and identification of studies This is the second part of a bibliometric study designed to assess the publication speed in journals of general internal medicine and primary care and to explore the link with various author, paper and journal characteristics. We obtained the 2015 impact factor list of general internal medicine and primary care journals through ‘ISI Web of Knowledge’. Then, two investigators (CR and PHG) randomly selected 810 original articles published between the 1st of January and the 31th of December 2016 using simple randomization based on computer-generated random numbers (45 articles from each of the 9 highest impact factors journals of general internal medicine and 45 articles from each of the 9 highest impact factors journals of primary care). Journals which did not report submission and publication dates, as well as commentaries, editorials, brief reports, correspondence, case reports and non-systematic reviews were excluded. Data collection All corresponding authors of the articles included in the study were contacted by e-mail by a research assistant located in Geneva, Switzerland; they were informed about the aim of the study and the practical procedures to complete a short web-based anonymous questionnaire. The questionnaire included sociodemographic questions about the corresponding authors [gender, age, academic rank of professor (Y/N)] and about the articles [total number of submissions necessary from the first submission to a journal until acceptance by the journal that published the article, statistician involvement as co-author (Y/N)]. Non-responders were sent a maximum of four reminder messages. In addition, using a standardized data extraction form, the two investigators extracted, ‘from the articles’, the place of residence of the corresponding author, the study design, the form of first publication (online or paper form) and the date of submission, of acceptance and of first publication, and ‘from the journals’, the discipline (general internal medicine or primary care) and the impact factor. We assessed the investigator inter-rater variability over a random sample of 15% of articles included in the study (data extraction of the remaining articles was therefore undertaken by only one investigator); the concordance was >95%, except for study design where the concordance was only 80%. Therefore, the design of all studies was assessed with the support of the main investigators (PS, JPF, HM), except when the design was clearly mentioned in the paper. Disagreements were resolved through discussion and consensus within the study team. Sample size justification and statistical analyses The minimal required sample size (n = 100) was calculated in order to measure a relative hazard of 1.75 between two groups of observations of equal size (proportion of articles co-authored by statistician: 0.5), taking a Type I error rate of 5% and a Type II error rate of 20%. Without reported intraclass correlation coefficients in similar studies, we decided to increase the sample size to 150 to take into account the cluster design (since observations coming from the same journal are likely to be more similar to each other than to observations in other journals) and also the presence of incomplete questionnaires. As we assumed a participation rate of 20%, we had to select at least 750 articles. First, according to our plan, we excluded all studies with qualitative design. Then, we compared the articles co-authored with those not co-authored by statisticians in terms of corresponding author, paper and journal characteristics, using frequency tables and univariate logistic regressions adjusted for intracluster correlations (15). In addition, we computed for each article the time to acceptance from the submission to the journal that accepted the paper (i.e. the submission-to-acceptance time) and the time to publication from the submission to the journal that accepted the paper (i.e. the submission-to-publication time). We summarized these data using medians and interquartile ranges (IQRs) because the distribution of these variables was clearly asymmetric. We compared the articles co-authored with those not co-authored by statisticians in terms of number of submissions, time to acceptance and time to publication, using univariate negative binomial and univariate Cox regressions adjusted for intracluster correlations (15–17). There was no censoring in the Cox regression because all these articles were published; we tested the proportional hazards assumption on the basis of Schoenfeld residuals (16,18). For statistically significant associations in univariate analysis, we performed multivariable regressions, using a non-automatic backward stepwise procedure so as to remove any covariates associated with a P value of >0.20. Statistical significance was set at a two-sided P value of ≤0.05. The sample size was estimated with ‘sample-size.net’ (version 9 October 2017, University of California, San Francisco). All other analyses were carried out with STATA version 12.0. Results The two investigators selected 781 articles published in 2016 (398 in general internal medicine and 383 in primary care), which represents 50% of the total number of articles published in 2016 by the 18 journals included in the study (1561 articles published). We expected 810 articles (45 articles × 18 journals), but this number was not obtained because of a low number of articles published in 2016 for some journals. The questionnaire, which was sent to the corresponding author of these 781 articles, was completed by 158 of them (20%). After having excluded qualitative studies (N = 22; only qualitative: 15; mixed quantitative-qualitative: 7), 136 articles were included in our study (Figure 1). Overall, 63 articles (46%) were co-authored by statisticians. Figure 1. View largeDownload slide Flowchart of the study Figure 1. View largeDownload slide Flowchart of the study Table 1 lists the 18 journals, stratified by discipline (general internal medicine or primary care) and sorted by 2015 impact factor; 80 articles (58.8%) were published in general internal medicine journals and 56 (41.2%) in primary care journals. Their study design was as follows: cross-sectional (39), cohort (23), systematic review (21), trial (20), case-control (10) and other (23). Table 1. Number of articles published in 2016 that were included in our study on the association between statistician involvement and number of submissions and time to publication, stratified by discipline and journal (N = 136) Journal Country housing the journal 2015 Impact factor Number of articles, N (%) General internal medicine 80 (58.8)  British Medical Journal UK 19.697 12 (8.8)  BMC Medicine UK 8.005 13 (9.6)  European Journal of Clinical Investigation Netherlands 2.687 6 (4.4)  International Journal of Medical Sciences USA 2.232 5 (3.7)  International Journal of Clinical Practice USA 2.226 9 (6.6)  Journal of the Formosan Medical Association Taiwan 2.018 7 (5.1)  Archives of Medical Science Poland 1.812 16 (11.8)  Korean Journal of Internal Medicine South Korea 1.679 5 (3.7)  American Journal of the Medical Sciences USA 1.575 7 (5.1) Primary care 56 (41.2)  Annals of Family Medicine USA 5.087 9 (6.6)  British Journal of General Practice UK 2.741 7 (5.1)  Journal of the American Board of Family Medicine USA 1.989 5 (3.7)  BMC Family Practice UK 1.641 8 (5.9)  Scandinavian Journal of Primary Health Care UK 1.556 9 (6.6)  European Journal of General Practice UK 1.364 9 (6.6)  Australian Journal of Primary Health Australia 1.152 2 (1.5)  Atencion Primaria Spain 1.098 5 (3.7)  Primary Health Care Research and Development UK 1.090 2 (1.5) Journal Country housing the journal 2015 Impact factor Number of articles, N (%) General internal medicine 80 (58.8)  British Medical Journal UK 19.697 12 (8.8)  BMC Medicine UK 8.005 13 (9.6)  European Journal of Clinical Investigation Netherlands 2.687 6 (4.4)  International Journal of Medical Sciences USA 2.232 5 (3.7)  International Journal of Clinical Practice USA 2.226 9 (6.6)  Journal of the Formosan Medical Association Taiwan 2.018 7 (5.1)  Archives of Medical Science Poland 1.812 16 (11.8)  Korean Journal of Internal Medicine South Korea 1.679 5 (3.7)  American Journal of the Medical Sciences USA 1.575 7 (5.1) Primary care 56 (41.2)  Annals of Family Medicine USA 5.087 9 (6.6)  British Journal of General Practice UK 2.741 7 (5.1)  Journal of the American Board of Family Medicine USA 1.989 5 (3.7)  BMC Family Practice UK 1.641 8 (5.9)  Scandinavian Journal of Primary Health Care UK 1.556 9 (6.6)  European Journal of General Practice UK 1.364 9 (6.6)  Australian Journal of Primary Health Australia 1.152 2 (1.5)  Atencion Primaria Spain 1.098 5 (3.7)  Primary Health Care Research and Development UK 1.090 2 (1.5) View Large Table 1. Number of articles published in 2016 that were included in our study on the association between statistician involvement and number of submissions and time to publication, stratified by discipline and journal (N = 136) Journal Country housing the journal 2015 Impact factor Number of articles, N (%) General internal medicine 80 (58.8)  British Medical Journal UK 19.697 12 (8.8)  BMC Medicine UK 8.005 13 (9.6)  European Journal of Clinical Investigation Netherlands 2.687 6 (4.4)  International Journal of Medical Sciences USA 2.232 5 (3.7)  International Journal of Clinical Practice USA 2.226 9 (6.6)  Journal of the Formosan Medical Association Taiwan 2.018 7 (5.1)  Archives of Medical Science Poland 1.812 16 (11.8)  Korean Journal of Internal Medicine South Korea 1.679 5 (3.7)  American Journal of the Medical Sciences USA 1.575 7 (5.1) Primary care 56 (41.2)  Annals of Family Medicine USA 5.087 9 (6.6)  British Journal of General Practice UK 2.741 7 (5.1)  Journal of the American Board of Family Medicine USA 1.989 5 (3.7)  BMC Family Practice UK 1.641 8 (5.9)  Scandinavian Journal of Primary Health Care UK 1.556 9 (6.6)  European Journal of General Practice UK 1.364 9 (6.6)  Australian Journal of Primary Health Australia 1.152 2 (1.5)  Atencion Primaria Spain 1.098 5 (3.7)  Primary Health Care Research and Development UK 1.090 2 (1.5) Journal Country housing the journal 2015 Impact factor Number of articles, N (%) General internal medicine 80 (58.8)  British Medical Journal UK 19.697 12 (8.8)  BMC Medicine UK 8.005 13 (9.6)  European Journal of Clinical Investigation Netherlands 2.687 6 (4.4)  International Journal of Medical Sciences USA 2.232 5 (3.7)  International Journal of Clinical Practice USA 2.226 9 (6.6)  Journal of the Formosan Medical Association Taiwan 2.018 7 (5.1)  Archives of Medical Science Poland 1.812 16 (11.8)  Korean Journal of Internal Medicine South Korea 1.679 5 (3.7)  American Journal of the Medical Sciences USA 1.575 7 (5.1) Primary care 56 (41.2)  Annals of Family Medicine USA 5.087 9 (6.6)  British Journal of General Practice UK 2.741 7 (5.1)  Journal of the American Board of Family Medicine USA 1.989 5 (3.7)  BMC Family Practice UK 1.641 8 (5.9)  Scandinavian Journal of Primary Health Care UK 1.556 9 (6.6)  European Journal of General Practice UK 1.364 9 (6.6)  Australian Journal of Primary Health Australia 1.152 2 (1.5)  Atencion Primaria Spain 1.098 5 (3.7)  Primary Health Care Research and Development UK 1.090 2 (1.5) View Large Table 2 presents the author, paper and journal characteristics of the 136 articles included in our study, stratified by statistician involvement as co-author. Only the journal impact factor was statistically associated with statistician involvement (56% of the articles published in journals with impact factor ≥2 were co-authored by statisticians, versus 37% in journals with impact factor <2, P value of 0.03). Table 2. Author, paper and journal characteristics of the 136 articles published in 2016 that were included in our study on the association between statistician involvement and number of submissions and time to publication, stratified by statistician involvement Characteristics Papers co-authors by statisticians (n = 63) Papers not co-authored by statisticians (n = 73) P valuea N (%) N (%) Author characteristics  Gender (male) 43 (68.3) 51 (69.9) 0.87  Age group)b 0.70   <45 years 30 (47.6) 31 (43.1)   ≥45 years 33 (52.4) 41 (56.9)  Academic rank of professor 34 (54.0) 34 (46.6) 0.31  Place of residence 0.09   Europe 39 (61.9) 37 (50.7)   North America 13 (20.6) 11 (15.1)   Other 11 (17.5) 25 (34.2) Paper characteristics  Study designc 0.38   Cross-sectional, cohort or case-control study 37 (60.7) 35 (52.2)   Systematic review or experiment 16 (26.2) 25 (37.3)   Other 8 (13.1) 7 (10.5)  Online first publication 55 (87.3) 64 (87.7) 0.91 Journal characteristics  Journal discipline 0.78   General internal medicine 38 (60.3) 42 (57.5)   Primary care 25 (39.7) 31 (42.5)  2015 impact factor 0.03   <2 25 (39.7) 43 (58.9)   ≥2 38 (60.3) 30 (41.1) Characteristics Papers co-authors by statisticians (n = 63) Papers not co-authored by statisticians (n = 73) P valuea N (%) N (%) Author characteristics  Gender (male) 43 (68.3) 51 (69.9) 0.87  Age group)b 0.70   <45 years 30 (47.6) 31 (43.1)   ≥45 years 33 (52.4) 41 (56.9)  Academic rank of professor 34 (54.0) 34 (46.6) 0.31  Place of residence 0.09   Europe 39 (61.9) 37 (50.7)   North America 13 (20.6) 11 (15.1)   Other 11 (17.5) 25 (34.2) Paper characteristics  Study designc 0.38   Cross-sectional, cohort or case-control study 37 (60.7) 35 (52.2)   Systematic review or experiment 16 (26.2) 25 (37.3)   Other 8 (13.1) 7 (10.5)  Online first publication 55 (87.3) 64 (87.7) 0.91 Journal characteristics  Journal discipline 0.78   General internal medicine 38 (60.3) 42 (57.5)   Primary care 25 (39.7) 31 (42.5)  2015 impact factor 0.03   <2 25 (39.7) 43 (58.9)   ≥2 38 (60.3) 30 (41.1) The threshold value of statistical significance is P = 0.05.aUnivariate logistic regressions adjusted for intracluster correlations. bNumber of missing data: 1. cNumber of missing data: 8. View Large Table 2. Author, paper and journal characteristics of the 136 articles published in 2016 that were included in our study on the association between statistician involvement and number of submissions and time to publication, stratified by statistician involvement Characteristics Papers co-authors by statisticians (n = 63) Papers not co-authored by statisticians (n = 73) P valuea N (%) N (%) Author characteristics  Gender (male) 43 (68.3) 51 (69.9) 0.87  Age group)b 0.70   <45 years 30 (47.6) 31 (43.1)   ≥45 years 33 (52.4) 41 (56.9)  Academic rank of professor 34 (54.0) 34 (46.6) 0.31  Place of residence 0.09   Europe 39 (61.9) 37 (50.7)   North America 13 (20.6) 11 (15.1)   Other 11 (17.5) 25 (34.2) Paper characteristics  Study designc 0.38   Cross-sectional, cohort or case-control study 37 (60.7) 35 (52.2)   Systematic review or experiment 16 (26.2) 25 (37.3)   Other 8 (13.1) 7 (10.5)  Online first publication 55 (87.3) 64 (87.7) 0.91 Journal characteristics  Journal discipline 0.78   General internal medicine 38 (60.3) 42 (57.5)   Primary care 25 (39.7) 31 (42.5)  2015 impact factor 0.03   <2 25 (39.7) 43 (58.9)   ≥2 38 (60.3) 30 (41.1) Characteristics Papers co-authors by statisticians (n = 63) Papers not co-authored by statisticians (n = 73) P valuea N (%) N (%) Author characteristics  Gender (male) 43 (68.3) 51 (69.9) 0.87  Age group)b 0.70   <45 years 30 (47.6) 31 (43.1)   ≥45 years 33 (52.4) 41 (56.9)  Academic rank of professor 34 (54.0) 34 (46.6) 0.31  Place of residence 0.09   Europe 39 (61.9) 37 (50.7)   North America 13 (20.6) 11 (15.1)   Other 11 (17.5) 25 (34.2) Paper characteristics  Study designc 0.38   Cross-sectional, cohort or case-control study 37 (60.7) 35 (52.2)   Systematic review or experiment 16 (26.2) 25 (37.3)   Other 8 (13.1) 7 (10.5)  Online first publication 55 (87.3) 64 (87.7) 0.91 Journal characteristics  Journal discipline 0.78   General internal medicine 38 (60.3) 42 (57.5)   Primary care 25 (39.7) 31 (42.5)  2015 impact factor 0.03   <2 25 (39.7) 43 (58.9)   ≥2 38 (60.3) 30 (41.1) The threshold value of statistical significance is P = 0.05.aUnivariate logistic regressions adjusted for intracluster correlations. bNumber of missing data: 1. cNumber of missing data: 8. View Large Finally, Table 3 shows the association between statistician involvement and the number of submissions necessary from the first submission to a journal, as well as the time to acceptance and the time to publication from the first submission to the journal that accepted the paper. Overall, the mean and median number of submissions were 2.2 (SD 1.1) and 2 (IQR 2), whereas the median time to acceptance and to publication were respectively 117.5 (IQR 125.5) and 242.5 days (IQR164). There were no statistically significant differences in the number of submissions and the time to acceptance between the articles co-authored and those not co-authored by statisticians. By contrast, the difference in the time to publication [statistician group: median 211 days (IQR 171) versus 260 days (IQR 144)] achieved statistical significance [survival analysis: hazard ratio 1.30 (95% CI 1.00–1.70), P value of 0.05]. The Kaplan–Meier curves (Figure 2) confirm in graphic form that articles co-authored by statisticians were published quicker than others. In multivariate analysis, the hazard ratio was 1.44 (95% CI 1.01–2.03, P value of 0.04). Tests of the Schoenfeld residuals did not indicate that the proportional hazards assumption was violated. Table 3. Association between statistician involvement of the 136 articles published in 2016 that were included in our study and number of submissions, time to acceptance and time to publication Mean number of submissions (SD) P valuea Median time to acceptance (IQR), days Time to acceptance, hazard ratio (95% CI) P valueb Median time to publication (IQR), days Time to publication, hazard ratio (95% CI) P valueb Time to publication, hazard ratio (95% CI) Adj. P valuec Statistician involvement 0.87 0.18 0.05 0.04  Yes 2.1 (1.1) 108 (102) 1.29 (0.89–1.86) 211 (171) 1.30 (1.00–1.70) 1.44 (1.01–2.03)  No 2.2 (1.2) 135 (130) 1 260 (144) 1 1 Mean number of submissions (SD) P valuea Median time to acceptance (IQR), days Time to acceptance, hazard ratio (95% CI) P valueb Median time to publication (IQR), days Time to publication, hazard ratio (95% CI) P valueb Time to publication, hazard ratio (95% CI) Adj. P valuec Statistician involvement 0.87 0.18 0.05 0.04  Yes 2.1 (1.1) 108 (102) 1.29 (0.89–1.86) 211 (171) 1.30 (1.00–1.70) 1.44 (1.01–2.03)  No 2.2 (1.2) 135 (130) 1 260 (144) 1 1 The threshold value of statistical significance is P = 0.05.aUnivariate analysis (negative binomial regression adjusted for intracluster correlations). bUnivariate analysis (Cox proportional hazard regression adjusted for intracluster correlations). cMultivariate analysis (Cox proportional hazard regression adjusted for intracluster correlations); the following variables: statistician involvement, gender, age group, academic rank (professor), place of residence, study design, form of first publication, journal discipline and impact factor were included in the initial model; the variables statistician involvement and study design were included in the final model. View Large Table 3. Association between statistician involvement of the 136 articles published in 2016 that were included in our study and number of submissions, time to acceptance and time to publication Mean number of submissions (SD) P valuea Median time to acceptance (IQR), days Time to acceptance, hazard ratio (95% CI) P valueb Median time to publication (IQR), days Time to publication, hazard ratio (95% CI) P valueb Time to publication, hazard ratio (95% CI) Adj. P valuec Statistician involvement 0.87 0.18 0.05 0.04  Yes 2.1 (1.1) 108 (102) 1.29 (0.89–1.86) 211 (171) 1.30 (1.00–1.70) 1.44 (1.01–2.03)  No 2.2 (1.2) 135 (130) 1 260 (144) 1 1 Mean number of submissions (SD) P valuea Median time to acceptance (IQR), days Time to acceptance, hazard ratio (95% CI) P valueb Median time to publication (IQR), days Time to publication, hazard ratio (95% CI) P valueb Time to publication, hazard ratio (95% CI) Adj. P valuec Statistician involvement 0.87 0.18 0.05 0.04  Yes 2.1 (1.1) 108 (102) 1.29 (0.89–1.86) 211 (171) 1.30 (1.00–1.70) 1.44 (1.01–2.03)  No 2.2 (1.2) 135 (130) 1 260 (144) 1 1 The threshold value of statistical significance is P = 0.05.aUnivariate analysis (negative binomial regression adjusted for intracluster correlations). bUnivariate analysis (Cox proportional hazard regression adjusted for intracluster correlations). cMultivariate analysis (Cox proportional hazard regression adjusted for intracluster correlations); the following variables: statistician involvement, gender, age group, academic rank (professor), place of residence, study design, form of first publication, journal discipline and impact factor were included in the initial model; the variables statistician involvement and study design were included in the final model. View Large Figure 2. View largeDownload slide Time to publication of the 136 articles published in 2016 that were included in our study, stratified by statistician involvement as co-author (Kaplan–Meier survival curves) Figure 2. View largeDownload slide Time to publication of the 136 articles published in 2016 that were included in our study, stratified by statistician involvement as co-author (Kaplan–Meier survival curves) Discussion Main findings In this bibliometric study of original articles published in high impact factor journals of general internal medicine and primary care, we found that the mean number of submissions was 2 and the median times to acceptance and to publication were respectively 118 and 243 days (i.e. ~4 and 8 months). The differences in the number of submissions and the time to acceptance between the articles co-authored and those not co-authored by statisticians were not statistically significant. By contrast, the difference in the time to publication (49 days) was statistically significant. Comparison with existing literature The association that we found between statistician involvement and time to publication might be explained by the influence of research quality on publication speed; indeed, some authors showed that a close collaboration between statisticians and clinical investigators may improve the methodological quality of randomized trials (6–9) and observational studies (10). As we did not assess the methodological quality of the articles included in our study, we cannot however verify this assertion. We also found a difference in the time to submission (~1 month less in articles co-authored by statisticians), but the association did not achieve statistical significance, perhaps because statisticians may also have contributed to studies that they did not co-author. In an Indian survey evaluating the reporting of statistical parameters in clinical trials, statisticians were involved in 20% of trials according to the acknowledgement section (19). Alternatively, our study was insufficiently powered to detect small differences between the subgroups; only 20% of corresponding authors participated in our study. However, these participation rates are usual in similar web-based studies (20,21). In addition, several authors showed that low participation rates do not necessarily introduce selection bias and that non-response bias may be of less concern in physician surveys than in surveys of the general public (22,23). The number of submissions needed from the first submission to a journal until acceptance by the journal that published the article was in general low (mean 2.1, median 2). This finding is exactly in line with data coming from a large study having assessed publication speed in biomedical journals (median number 2) (20) and suggests that only few submissions are generally required prior to journal acceptance. It was pointed out that researchers prefer to submit their research to lower impact journals to ensure that manuscripts do not require multiple submissions prior to acceptance (20). However, our results show that the number of submissions required is also reasonable when publishing in high impact journals (we included only high impact journals in our study). Limitations We need to point out some limitations. First, the size of our sample may be too small to detect some associations, and results have to be interpreted accordingly. As already mentioned, only 20% of authors completed our questionnaire; however, we assumed a total participation rate of 20% in the estimation of our sample size. Second, we included only journals with high impact factors and journals of general internal medicine and primary care; our results are therefore not necessarily generalizable to lower impact journals and to journals in other disciplines. Third, we did not include several high impact journals because of missing data on outcomes (dates of submission and/or publication), which also limits the generalizability of our results. Fourth, corresponding authors were not asked about author sequence (i.e. where the statisticians appear in the author list), as well as contributions and qualification of the statisticians involved in these studies; for example, though it is likely that some of them had a PhD or a master degree in statistics, others might have a qualification in epidemiology or public health. Fifth, some but not all submitted papers are generally reviewed by a statistical referee; this could have influenced the findings. Sixth, the two investigators carried out the double extraction of the data only for 15% of the articles; however, we believe that the risk of information bias is low because the inter-rater concordance was >95% for all variables, except for study design and the latter was assessed with the support of the main investigators. Finally, in this study, we assessed the association between statistician involvement as co-author and (i) the total number of submissions needed from the first submission to a journal and (ii) the time between submission to the journal that published the paper and acceptance, respectively publication. It would also be interesting to assess whether statistician involvement is associated with the total time to publication (i.e. from the first submission to a journal), ideally by using prospective observational design to decrease the risk of measurement error. Conclusion Close collaboration between researchers and statisticians has the potential to significantly reduce the time to publication of articles published in journals of general internal medicine and primary care. These data should encourage researchers to involve statisticians in the design, conduct, statistical analysis and reporting of research, not only to ensure high standards of quality but also to speed up its publication. Further studies using higher sample size and targeting other scientific journals are however necessary to confirm our results. Declaration Funding: this project was supported by institutional funding from the Faculty of medicine, University of Geneva. Ethical approval: not required (under Swiss law, ethical approval is not required when collecting non-personal health data). Conflicts of interest: none. Acknowledgements We would like to warmly thank Amir Moussa, administrative assistant, and Claire Ragot and Pierre-Henri Gorioux, MD students, for their precious contribution to the study. We would also like to thank Bernard Cerutti and François Herrmann for methodological support and Melania Bembea, Leonardo Silvestri and Dagmar Haller for their support and assistance throughout the study. References 1. Kalcioglu MT , Ileri Y , Karaca S , Egilmez OK , Kokten N . Research on the Submission, Acceptance and Publication Times of Articles Submitted to International Otorhinolaryngology Journals . Acta Inform Med 2015 ; 23 : 379 – 84 . Google Scholar Crossref Search ADS PubMed WorldCat 2. Moses L , Louis TA . Statistical consulting in clinical research: the two-way street . Stat Med 1984 ; 3 : 1 – 5 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Dobson AJ . The role of the statistician . Int J Epidemiol 1983 ; 12 : 274 – 5 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Simon R . The role of statisticians in intervention trials . Stat Methods Med Res 1999 ; 8 : 281 – 6 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Gore SM , Jones IG , Rytter EC . Misuse of statistical methods: critical assessment of articles in BMJ from January to March 1976 . Br Med J 1977 ; 1 : 85 – 7 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Delgado-Rodriguez M , Ruiz-Canela M , De Irala-Estevez J , Llorca J , Martinez-Gonzalez A . Participation of epidemiologists and/or biostatisticians and methodological quality of published controlled clinical trials . J Epidemiol Community Health 2001 ; 55 : 569 – 72 . Google Scholar Crossref Search ADS PubMed WorldCat 7. Diaz-Ordaz K , Froud R , Sheehan B , Eldridge S . A systematic review of cluster randomised trials in residential facilities for older people suggests how to improve quality . BMC Med Res Methodol 2013 ; 13 : 127 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Kloukos D , Papageorgiou SN , Doulis I , Petridis H , Pandis N . Reporting quality of randomised controlled trials published in prosthodontic and implantology journals . J Oral Rehabil 2015 ; 42 : 914 – 25 . Google Scholar Crossref Search ADS PubMed WorldCat 9. Tokolahi E , Hocking C , Kersten P , Vandal AC . Quality and reporting of cluster randomized controlled trials evaluating occupational therapy interventions: a systematic review . OTJR (Thorofare N J) 2016 ; 36 : 14 – 24 . Google Scholar PubMed WorldCat 10. Kumar R , Indrayan A , Chhabra P . Evaluation of quality of multivariable logistic regression in Indian medical journals using multilevel modeling approach . Indian J Public Health 2016 ; 60 : 99 – 106 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Amat CB . Editorial and publication delay of papers submitted to 14 selected Food Research journals. Influence of online posting . Scientometrics 2008 ; 74 : 379 – 89 . Google Scholar Crossref Search ADS WorldCat 12. Björk B-C , Solomon D . The publishing delay in scholarly peer-reviewed journals . J Informetr 2013 ; 7 : 914 – 23 . Google Scholar Crossref Search ADS WorldCat 13. Chen H , Chen CH , Jhanji V . Publication times, impact factors, and advance online publication in ophthalmology journals . Ophthalmology 2013 ; 120 : 1697 – 701 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Palese A , Coletti S , Dante A . Publication efficiency among the higher impact factor nursing journals in 2009: a retrospective analysis . Int J Nurs Stud 2013 ; 50 : 543 – 51 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Katz MH. Multivariable analysis: a practical guide for clinicians . 2 nd ed. Cambridge, NY : Cambridge University Press , 2006 . Google Preview WorldCat COPAC 16. UCLA: Statistical Consulting Group. Survival Analysis with Stata [Internet]. [cited 2018 Oct 14] . https://stats.idre.ucla.edu/stata/seminars/stata-survival/ (accessed on 15 March 2018). 17. UCLA: Statistical Consulting Group. Regression Models with Count Data [Internet] . [cited 2018 Oct 14]. https://stats.idre.ucla.edu/stata/seminars/regression-models-with-count-data/ (accessed on 15 March 2018). 18. UCLA: Statistical Consulting Group. Testing the proportional hazard assumption in Cox models [Internet]. [cited 2018 Oct 14] . https://stats.idre.ucla.edu/other/examples/asa2/testing-the-proportional-hazard-assumption-in-cox-models/ (accessed on 15 March 2018). 19. Jaykaran C , Chavda N , Yadav P . Proper reporting of statistical parameters in clinical trials published in Indian medical journals. Is inclusion of statistician play any significant role ? J Young Pharm 2011 ; 3 : 167 – 8 . Google Scholar Crossref Search ADS PubMed WorldCat 20. Wallach JD , Egilman AC , Gopal AD et al. Biomedical journal speed and efficiency: a cross-sectional pilot survey of author experiences . Res Integr Peer Rev 2018 ; 3 : 1 . Google Scholar Crossref Search ADS PubMed WorldCat 21. Sebo P , Maisonneuve H , Cerutti B et al. Rates, delays, and completeness of general practitioners’ responses to a postal versus web-based survey: a randomized trial . J Med Internet Res 2017 ; 19 : e83 . Google Scholar Crossref Search ADS PubMed WorldCat 22. Kellerman SE , Herold J . Physician response to surveys. A review of the literature . Am J Prev Med 2001 ; 20 : 61 – 7 . Google Scholar Crossref Search ADS PubMed WorldCat 23. Asch DA , Jedrziewski MK , Christakis NA . Response rates to mail surveys published in medical journals . J Clin Epidemiol 1997 ; 50 : 1129 – 36 . Google Scholar Crossref Search ADS PubMed WorldCat © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Family Practice Oxford University Press

Is statistician involvement as co-author associated with reduced time to publication of quantitative research in general medical journals? A bibliometric study

Download PDF
6 pages

Loading next page...
 
/lp/oxford-university-press/is-statistician-involvement-as-co-author-associated-with-reduced-time-CordF3d03s

References (22)

Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0263-2136
eISSN
1460-2229
DOI
10.1093/fampra/cmy115
Publisher site
See Article on Publisher Site

Abstract

Abstract Objective We aimed to compare the number of submissions until acceptance and the time to publication between articles co-authored and articles not co-authored by statisticians. Methods We randomly selected 781 articles published in 2016 in 18 high impact factor journals of general internal medicine and primary care. For each article, we retrieved its date of submission to the journal and its first publication; we also contacted its corresponding author and asked about the number of submissions necessary from the first submission to a journal until acceptance and whether the article was co-authored by a statistician. After having excluded qualitative studies, we compared the articles co-authored with those not co-authored by statisticians in terms of number of submissions and submission-to-publication time, using negative binomial and Cox regressions, adjusted for intracluster correlations. Results One hundred fifty-eight authors completed the questionnaire (20%); 136 articles with quantitative design were included in the study. Overall, 63 articles (46%) were co-authored by statisticians. There was no statistically significant difference in the number of submissions (statistician group: mean 2.1 (SD 1.1) versus 2.2 (SD 1.2), P value 0.87). By contrast, we found a statistically significant difference in the submission-to-publication time (statistician group: median 211 days [interquartile range (IQR) 171] versus 260 (IQR 144); hazard ratio 1.44 (95% CI 1.01–2.03), adjusted P value 0.04). Conclusions Papers co-authored by statisticians have a shorter time to publication. We encourage researchers to closely involve statisticians in the design, conduct and statistical analysis of research, not only to ensure high standards of quality but also to speed up its publication. Delay of publication, general medical journals, number of submissions, publication speed, retrospective study, statistician, time to acceptance, time to publication Introduction Despite a continuous rise in the number of journals publishing research, the increase in scientific productivity is expected to negatively impact the workload of journals and consequently the acceptance rate of submitted manuscripts and the publication speed of accepted manuscripts (1). This trend is problematic because high rejection rates and long delays of publication could discourage scientific researchers from publishing their research in the traditional publishing system; it could also delay the progression of academic careers, as well as postpone the dissemination of new knowledge or the implementation of new interventions. In order to avoid misuse of statistics in medical research, it was suggested that active participation of statisticians should be required at all stages of a study project (2–5). It has been shown that an effective and close collaboration between statisticians and clinical investigators may improve the methodological quality of randomized trials (6–9) and observational studies (10). Yet, despite the large number of bibliometric studies from a wide range of disciplines aimed to quantify the publication speed of scientific research (1,11–14); there are currently no data on the influence of statistician involvement on the publication speed of medical research. The objective of our study was therefore to assess whether papers co-authored by statisticians and submitted to general medical journals (i.e. journals of general internal medicine and primary care) require lower numbers of submissions and are associated with shorter times to publication. For this project, we planned to restrict the analysis to quantitative studies, where the role of the statistician may be particularly important; indeed, studies with qualitative design use non-statistical methods of analysis. Methods Search strategy and identification of studies This is the second part of a bibliometric study designed to assess the publication speed in journals of general internal medicine and primary care and to explore the link with various author, paper and journal characteristics. We obtained the 2015 impact factor list of general internal medicine and primary care journals through ‘ISI Web of Knowledge’. Then, two investigators (CR and PHG) randomly selected 810 original articles published between the 1st of January and the 31th of December 2016 using simple randomization based on computer-generated random numbers (45 articles from each of the 9 highest impact factors journals of general internal medicine and 45 articles from each of the 9 highest impact factors journals of primary care). Journals which did not report submission and publication dates, as well as commentaries, editorials, brief reports, correspondence, case reports and non-systematic reviews were excluded. Data collection All corresponding authors of the articles included in the study were contacted by e-mail by a research assistant located in Geneva, Switzerland; they were informed about the aim of the study and the practical procedures to complete a short web-based anonymous questionnaire. The questionnaire included sociodemographic questions about the corresponding authors [gender, age, academic rank of professor (Y/N)] and about the articles [total number of submissions necessary from the first submission to a journal until acceptance by the journal that published the article, statistician involvement as co-author (Y/N)]. Non-responders were sent a maximum of four reminder messages. In addition, using a standardized data extraction form, the two investigators extracted, ‘from the articles’, the place of residence of the corresponding author, the study design, the form of first publication (online or paper form) and the date of submission, of acceptance and of first publication, and ‘from the journals’, the discipline (general internal medicine or primary care) and the impact factor. We assessed the investigator inter-rater variability over a random sample of 15% of articles included in the study (data extraction of the remaining articles was therefore undertaken by only one investigator); the concordance was >95%, except for study design where the concordance was only 80%. Therefore, the design of all studies was assessed with the support of the main investigators (PS, JPF, HM), except when the design was clearly mentioned in the paper. Disagreements were resolved through discussion and consensus within the study team. Sample size justification and statistical analyses The minimal required sample size (n = 100) was calculated in order to measure a relative hazard of 1.75 between two groups of observations of equal size (proportion of articles co-authored by statistician: 0.5), taking a Type I error rate of 5% and a Type II error rate of 20%. Without reported intraclass correlation coefficients in similar studies, we decided to increase the sample size to 150 to take into account the cluster design (since observations coming from the same journal are likely to be more similar to each other than to observations in other journals) and also the presence of incomplete questionnaires. As we assumed a participation rate of 20%, we had to select at least 750 articles. First, according to our plan, we excluded all studies with qualitative design. Then, we compared the articles co-authored with those not co-authored by statisticians in terms of corresponding author, paper and journal characteristics, using frequency tables and univariate logistic regressions adjusted for intracluster correlations (15). In addition, we computed for each article the time to acceptance from the submission to the journal that accepted the paper (i.e. the submission-to-acceptance time) and the time to publication from the submission to the journal that accepted the paper (i.e. the submission-to-publication time). We summarized these data using medians and interquartile ranges (IQRs) because the distribution of these variables was clearly asymmetric. We compared the articles co-authored with those not co-authored by statisticians in terms of number of submissions, time to acceptance and time to publication, using univariate negative binomial and univariate Cox regressions adjusted for intracluster correlations (15–17). There was no censoring in the Cox regression because all these articles were published; we tested the proportional hazards assumption on the basis of Schoenfeld residuals (16,18). For statistically significant associations in univariate analysis, we performed multivariable regressions, using a non-automatic backward stepwise procedure so as to remove any covariates associated with a P value of >0.20. Statistical significance was set at a two-sided P value of ≤0.05. The sample size was estimated with ‘sample-size.net’ (version 9 October 2017, University of California, San Francisco). All other analyses were carried out with STATA version 12.0. Results The two investigators selected 781 articles published in 2016 (398 in general internal medicine and 383 in primary care), which represents 50% of the total number of articles published in 2016 by the 18 journals included in the study (1561 articles published). We expected 810 articles (45 articles × 18 journals), but this number was not obtained because of a low number of articles published in 2016 for some journals. The questionnaire, which was sent to the corresponding author of these 781 articles, was completed by 158 of them (20%). After having excluded qualitative studies (N = 22; only qualitative: 15; mixed quantitative-qualitative: 7), 136 articles were included in our study (Figure 1). Overall, 63 articles (46%) were co-authored by statisticians. Figure 1. View largeDownload slide Flowchart of the study Figure 1. View largeDownload slide Flowchart of the study Table 1 lists the 18 journals, stratified by discipline (general internal medicine or primary care) and sorted by 2015 impact factor; 80 articles (58.8%) were published in general internal medicine journals and 56 (41.2%) in primary care journals. Their study design was as follows: cross-sectional (39), cohort (23), systematic review (21), trial (20), case-control (10) and other (23). Table 1. Number of articles published in 2016 that were included in our study on the association between statistician involvement and number of submissions and time to publication, stratified by discipline and journal (N = 136) Journal Country housing the journal 2015 Impact factor Number of articles, N (%) General internal medicine 80 (58.8)  British Medical Journal UK 19.697 12 (8.8)  BMC Medicine UK 8.005 13 (9.6)  European Journal of Clinical Investigation Netherlands 2.687 6 (4.4)  International Journal of Medical Sciences USA 2.232 5 (3.7)  International Journal of Clinical Practice USA 2.226 9 (6.6)  Journal of the Formosan Medical Association Taiwan 2.018 7 (5.1)  Archives of Medical Science Poland 1.812 16 (11.8)  Korean Journal of Internal Medicine South Korea 1.679 5 (3.7)  American Journal of the Medical Sciences USA 1.575 7 (5.1) Primary care 56 (41.2)  Annals of Family Medicine USA 5.087 9 (6.6)  British Journal of General Practice UK 2.741 7 (5.1)  Journal of the American Board of Family Medicine USA 1.989 5 (3.7)  BMC Family Practice UK 1.641 8 (5.9)  Scandinavian Journal of Primary Health Care UK 1.556 9 (6.6)  European Journal of General Practice UK 1.364 9 (6.6)  Australian Journal of Primary Health Australia 1.152 2 (1.5)  Atencion Primaria Spain 1.098 5 (3.7)  Primary Health Care Research and Development UK 1.090 2 (1.5) Journal Country housing the journal 2015 Impact factor Number of articles, N (%) General internal medicine 80 (58.8)  British Medical Journal UK 19.697 12 (8.8)  BMC Medicine UK 8.005 13 (9.6)  European Journal of Clinical Investigation Netherlands 2.687 6 (4.4)  International Journal of Medical Sciences USA 2.232 5 (3.7)  International Journal of Clinical Practice USA 2.226 9 (6.6)  Journal of the Formosan Medical Association Taiwan 2.018 7 (5.1)  Archives of Medical Science Poland 1.812 16 (11.8)  Korean Journal of Internal Medicine South Korea 1.679 5 (3.7)  American Journal of the Medical Sciences USA 1.575 7 (5.1) Primary care 56 (41.2)  Annals of Family Medicine USA 5.087 9 (6.6)  British Journal of General Practice UK 2.741 7 (5.1)  Journal of the American Board of Family Medicine USA 1.989 5 (3.7)  BMC Family Practice UK 1.641 8 (5.9)  Scandinavian Journal of Primary Health Care UK 1.556 9 (6.6)  European Journal of General Practice UK 1.364 9 (6.6)  Australian Journal of Primary Health Australia 1.152 2 (1.5)  Atencion Primaria Spain 1.098 5 (3.7)  Primary Health Care Research and Development UK 1.090 2 (1.5) View Large Table 1. Number of articles published in 2016 that were included in our study on the association between statistician involvement and number of submissions and time to publication, stratified by discipline and journal (N = 136) Journal Country housing the journal 2015 Impact factor Number of articles, N (%) General internal medicine 80 (58.8)  British Medical Journal UK 19.697 12 (8.8)  BMC Medicine UK 8.005 13 (9.6)  European Journal of Clinical Investigation Netherlands 2.687 6 (4.4)  International Journal of Medical Sciences USA 2.232 5 (3.7)  International Journal of Clinical Practice USA 2.226 9 (6.6)  Journal of the Formosan Medical Association Taiwan 2.018 7 (5.1)  Archives of Medical Science Poland 1.812 16 (11.8)  Korean Journal of Internal Medicine South Korea 1.679 5 (3.7)  American Journal of the Medical Sciences USA 1.575 7 (5.1) Primary care 56 (41.2)  Annals of Family Medicine USA 5.087 9 (6.6)  British Journal of General Practice UK 2.741 7 (5.1)  Journal of the American Board of Family Medicine USA 1.989 5 (3.7)  BMC Family Practice UK 1.641 8 (5.9)  Scandinavian Journal of Primary Health Care UK 1.556 9 (6.6)  European Journal of General Practice UK 1.364 9 (6.6)  Australian Journal of Primary Health Australia 1.152 2 (1.5)  Atencion Primaria Spain 1.098 5 (3.7)  Primary Health Care Research and Development UK 1.090 2 (1.5) Journal Country housing the journal 2015 Impact factor Number of articles, N (%) General internal medicine 80 (58.8)  British Medical Journal UK 19.697 12 (8.8)  BMC Medicine UK 8.005 13 (9.6)  European Journal of Clinical Investigation Netherlands 2.687 6 (4.4)  International Journal of Medical Sciences USA 2.232 5 (3.7)  International Journal of Clinical Practice USA 2.226 9 (6.6)  Journal of the Formosan Medical Association Taiwan 2.018 7 (5.1)  Archives of Medical Science Poland 1.812 16 (11.8)  Korean Journal of Internal Medicine South Korea 1.679 5 (3.7)  American Journal of the Medical Sciences USA 1.575 7 (5.1) Primary care 56 (41.2)  Annals of Family Medicine USA 5.087 9 (6.6)  British Journal of General Practice UK 2.741 7 (5.1)  Journal of the American Board of Family Medicine USA 1.989 5 (3.7)  BMC Family Practice UK 1.641 8 (5.9)  Scandinavian Journal of Primary Health Care UK 1.556 9 (6.6)  European Journal of General Practice UK 1.364 9 (6.6)  Australian Journal of Primary Health Australia 1.152 2 (1.5)  Atencion Primaria Spain 1.098 5 (3.7)  Primary Health Care Research and Development UK 1.090 2 (1.5) View Large Table 2 presents the author, paper and journal characteristics of the 136 articles included in our study, stratified by statistician involvement as co-author. Only the journal impact factor was statistically associated with statistician involvement (56% of the articles published in journals with impact factor ≥2 were co-authored by statisticians, versus 37% in journals with impact factor <2, P value of 0.03). Table 2. Author, paper and journal characteristics of the 136 articles published in 2016 that were included in our study on the association between statistician involvement and number of submissions and time to publication, stratified by statistician involvement Characteristics Papers co-authors by statisticians (n = 63) Papers not co-authored by statisticians (n = 73) P valuea N (%) N (%) Author characteristics  Gender (male) 43 (68.3) 51 (69.9) 0.87  Age group)b 0.70   <45 years 30 (47.6) 31 (43.1)   ≥45 years 33 (52.4) 41 (56.9)  Academic rank of professor 34 (54.0) 34 (46.6) 0.31  Place of residence 0.09   Europe 39 (61.9) 37 (50.7)   North America 13 (20.6) 11 (15.1)   Other 11 (17.5) 25 (34.2) Paper characteristics  Study designc 0.38   Cross-sectional, cohort or case-control study 37 (60.7) 35 (52.2)   Systematic review or experiment 16 (26.2) 25 (37.3)   Other 8 (13.1) 7 (10.5)  Online first publication 55 (87.3) 64 (87.7) 0.91 Journal characteristics  Journal discipline 0.78   General internal medicine 38 (60.3) 42 (57.5)   Primary care 25 (39.7) 31 (42.5)  2015 impact factor 0.03   <2 25 (39.7) 43 (58.9)   ≥2 38 (60.3) 30 (41.1) Characteristics Papers co-authors by statisticians (n = 63) Papers not co-authored by statisticians (n = 73) P valuea N (%) N (%) Author characteristics  Gender (male) 43 (68.3) 51 (69.9) 0.87  Age group)b 0.70   <45 years 30 (47.6) 31 (43.1)   ≥45 years 33 (52.4) 41 (56.9)  Academic rank of professor 34 (54.0) 34 (46.6) 0.31  Place of residence 0.09   Europe 39 (61.9) 37 (50.7)   North America 13 (20.6) 11 (15.1)   Other 11 (17.5) 25 (34.2) Paper characteristics  Study designc 0.38   Cross-sectional, cohort or case-control study 37 (60.7) 35 (52.2)   Systematic review or experiment 16 (26.2) 25 (37.3)   Other 8 (13.1) 7 (10.5)  Online first publication 55 (87.3) 64 (87.7) 0.91 Journal characteristics  Journal discipline 0.78   General internal medicine 38 (60.3) 42 (57.5)   Primary care 25 (39.7) 31 (42.5)  2015 impact factor 0.03   <2 25 (39.7) 43 (58.9)   ≥2 38 (60.3) 30 (41.1) The threshold value of statistical significance is P = 0.05.aUnivariate logistic regressions adjusted for intracluster correlations. bNumber of missing data: 1. cNumber of missing data: 8. View Large Table 2. Author, paper and journal characteristics of the 136 articles published in 2016 that were included in our study on the association between statistician involvement and number of submissions and time to publication, stratified by statistician involvement Characteristics Papers co-authors by statisticians (n = 63) Papers not co-authored by statisticians (n = 73) P valuea N (%) N (%) Author characteristics  Gender (male) 43 (68.3) 51 (69.9) 0.87  Age group)b 0.70   <45 years 30 (47.6) 31 (43.1)   ≥45 years 33 (52.4) 41 (56.9)  Academic rank of professor 34 (54.0) 34 (46.6) 0.31  Place of residence 0.09   Europe 39 (61.9) 37 (50.7)   North America 13 (20.6) 11 (15.1)   Other 11 (17.5) 25 (34.2) Paper characteristics  Study designc 0.38   Cross-sectional, cohort or case-control study 37 (60.7) 35 (52.2)   Systematic review or experiment 16 (26.2) 25 (37.3)   Other 8 (13.1) 7 (10.5)  Online first publication 55 (87.3) 64 (87.7) 0.91 Journal characteristics  Journal discipline 0.78   General internal medicine 38 (60.3) 42 (57.5)   Primary care 25 (39.7) 31 (42.5)  2015 impact factor 0.03   <2 25 (39.7) 43 (58.9)   ≥2 38 (60.3) 30 (41.1) Characteristics Papers co-authors by statisticians (n = 63) Papers not co-authored by statisticians (n = 73) P valuea N (%) N (%) Author characteristics  Gender (male) 43 (68.3) 51 (69.9) 0.87  Age group)b 0.70   <45 years 30 (47.6) 31 (43.1)   ≥45 years 33 (52.4) 41 (56.9)  Academic rank of professor 34 (54.0) 34 (46.6) 0.31  Place of residence 0.09   Europe 39 (61.9) 37 (50.7)   North America 13 (20.6) 11 (15.1)   Other 11 (17.5) 25 (34.2) Paper characteristics  Study designc 0.38   Cross-sectional, cohort or case-control study 37 (60.7) 35 (52.2)   Systematic review or experiment 16 (26.2) 25 (37.3)   Other 8 (13.1) 7 (10.5)  Online first publication 55 (87.3) 64 (87.7) 0.91 Journal characteristics  Journal discipline 0.78   General internal medicine 38 (60.3) 42 (57.5)   Primary care 25 (39.7) 31 (42.5)  2015 impact factor 0.03   <2 25 (39.7) 43 (58.9)   ≥2 38 (60.3) 30 (41.1) The threshold value of statistical significance is P = 0.05.aUnivariate logistic regressions adjusted for intracluster correlations. bNumber of missing data: 1. cNumber of missing data: 8. View Large Finally, Table 3 shows the association between statistician involvement and the number of submissions necessary from the first submission to a journal, as well as the time to acceptance and the time to publication from the first submission to the journal that accepted the paper. Overall, the mean and median number of submissions were 2.2 (SD 1.1) and 2 (IQR 2), whereas the median time to acceptance and to publication were respectively 117.5 (IQR 125.5) and 242.5 days (IQR164). There were no statistically significant differences in the number of submissions and the time to acceptance between the articles co-authored and those not co-authored by statisticians. By contrast, the difference in the time to publication [statistician group: median 211 days (IQR 171) versus 260 days (IQR 144)] achieved statistical significance [survival analysis: hazard ratio 1.30 (95% CI 1.00–1.70), P value of 0.05]. The Kaplan–Meier curves (Figure 2) confirm in graphic form that articles co-authored by statisticians were published quicker than others. In multivariate analysis, the hazard ratio was 1.44 (95% CI 1.01–2.03, P value of 0.04). Tests of the Schoenfeld residuals did not indicate that the proportional hazards assumption was violated. Table 3. Association between statistician involvement of the 136 articles published in 2016 that were included in our study and number of submissions, time to acceptance and time to publication Mean number of submissions (SD) P valuea Median time to acceptance (IQR), days Time to acceptance, hazard ratio (95% CI) P valueb Median time to publication (IQR), days Time to publication, hazard ratio (95% CI) P valueb Time to publication, hazard ratio (95% CI) Adj. P valuec Statistician involvement 0.87 0.18 0.05 0.04  Yes 2.1 (1.1) 108 (102) 1.29 (0.89–1.86) 211 (171) 1.30 (1.00–1.70) 1.44 (1.01–2.03)  No 2.2 (1.2) 135 (130) 1 260 (144) 1 1 Mean number of submissions (SD) P valuea Median time to acceptance (IQR), days Time to acceptance, hazard ratio (95% CI) P valueb Median time to publication (IQR), days Time to publication, hazard ratio (95% CI) P valueb Time to publication, hazard ratio (95% CI) Adj. P valuec Statistician involvement 0.87 0.18 0.05 0.04  Yes 2.1 (1.1) 108 (102) 1.29 (0.89–1.86) 211 (171) 1.30 (1.00–1.70) 1.44 (1.01–2.03)  No 2.2 (1.2) 135 (130) 1 260 (144) 1 1 The threshold value of statistical significance is P = 0.05.aUnivariate analysis (negative binomial regression adjusted for intracluster correlations). bUnivariate analysis (Cox proportional hazard regression adjusted for intracluster correlations). cMultivariate analysis (Cox proportional hazard regression adjusted for intracluster correlations); the following variables: statistician involvement, gender, age group, academic rank (professor), place of residence, study design, form of first publication, journal discipline and impact factor were included in the initial model; the variables statistician involvement and study design were included in the final model. View Large Table 3. Association between statistician involvement of the 136 articles published in 2016 that were included in our study and number of submissions, time to acceptance and time to publication Mean number of submissions (SD) P valuea Median time to acceptance (IQR), days Time to acceptance, hazard ratio (95% CI) P valueb Median time to publication (IQR), days Time to publication, hazard ratio (95% CI) P valueb Time to publication, hazard ratio (95% CI) Adj. P valuec Statistician involvement 0.87 0.18 0.05 0.04  Yes 2.1 (1.1) 108 (102) 1.29 (0.89–1.86) 211 (171) 1.30 (1.00–1.70) 1.44 (1.01–2.03)  No 2.2 (1.2) 135 (130) 1 260 (144) 1 1 Mean number of submissions (SD) P valuea Median time to acceptance (IQR), days Time to acceptance, hazard ratio (95% CI) P valueb Median time to publication (IQR), days Time to publication, hazard ratio (95% CI) P valueb Time to publication, hazard ratio (95% CI) Adj. P valuec Statistician involvement 0.87 0.18 0.05 0.04  Yes 2.1 (1.1) 108 (102) 1.29 (0.89–1.86) 211 (171) 1.30 (1.00–1.70) 1.44 (1.01–2.03)  No 2.2 (1.2) 135 (130) 1 260 (144) 1 1 The threshold value of statistical significance is P = 0.05.aUnivariate analysis (negative binomial regression adjusted for intracluster correlations). bUnivariate analysis (Cox proportional hazard regression adjusted for intracluster correlations). cMultivariate analysis (Cox proportional hazard regression adjusted for intracluster correlations); the following variables: statistician involvement, gender, age group, academic rank (professor), place of residence, study design, form of first publication, journal discipline and impact factor were included in the initial model; the variables statistician involvement and study design were included in the final model. View Large Figure 2. View largeDownload slide Time to publication of the 136 articles published in 2016 that were included in our study, stratified by statistician involvement as co-author (Kaplan–Meier survival curves) Figure 2. View largeDownload slide Time to publication of the 136 articles published in 2016 that were included in our study, stratified by statistician involvement as co-author (Kaplan–Meier survival curves) Discussion Main findings In this bibliometric study of original articles published in high impact factor journals of general internal medicine and primary care, we found that the mean number of submissions was 2 and the median times to acceptance and to publication were respectively 118 and 243 days (i.e. ~4 and 8 months). The differences in the number of submissions and the time to acceptance between the articles co-authored and those not co-authored by statisticians were not statistically significant. By contrast, the difference in the time to publication (49 days) was statistically significant. Comparison with existing literature The association that we found between statistician involvement and time to publication might be explained by the influence of research quality on publication speed; indeed, some authors showed that a close collaboration between statisticians and clinical investigators may improve the methodological quality of randomized trials (6–9) and observational studies (10). As we did not assess the methodological quality of the articles included in our study, we cannot however verify this assertion. We also found a difference in the time to submission (~1 month less in articles co-authored by statisticians), but the association did not achieve statistical significance, perhaps because statisticians may also have contributed to studies that they did not co-author. In an Indian survey evaluating the reporting of statistical parameters in clinical trials, statisticians were involved in 20% of trials according to the acknowledgement section (19). Alternatively, our study was insufficiently powered to detect small differences between the subgroups; only 20% of corresponding authors participated in our study. However, these participation rates are usual in similar web-based studies (20,21). In addition, several authors showed that low participation rates do not necessarily introduce selection bias and that non-response bias may be of less concern in physician surveys than in surveys of the general public (22,23). The number of submissions needed from the first submission to a journal until acceptance by the journal that published the article was in general low (mean 2.1, median 2). This finding is exactly in line with data coming from a large study having assessed publication speed in biomedical journals (median number 2) (20) and suggests that only few submissions are generally required prior to journal acceptance. It was pointed out that researchers prefer to submit their research to lower impact journals to ensure that manuscripts do not require multiple submissions prior to acceptance (20). However, our results show that the number of submissions required is also reasonable when publishing in high impact journals (we included only high impact journals in our study). Limitations We need to point out some limitations. First, the size of our sample may be too small to detect some associations, and results have to be interpreted accordingly. As already mentioned, only 20% of authors completed our questionnaire; however, we assumed a total participation rate of 20% in the estimation of our sample size. Second, we included only journals with high impact factors and journals of general internal medicine and primary care; our results are therefore not necessarily generalizable to lower impact journals and to journals in other disciplines. Third, we did not include several high impact journals because of missing data on outcomes (dates of submission and/or publication), which also limits the generalizability of our results. Fourth, corresponding authors were not asked about author sequence (i.e. where the statisticians appear in the author list), as well as contributions and qualification of the statisticians involved in these studies; for example, though it is likely that some of them had a PhD or a master degree in statistics, others might have a qualification in epidemiology or public health. Fifth, some but not all submitted papers are generally reviewed by a statistical referee; this could have influenced the findings. Sixth, the two investigators carried out the double extraction of the data only for 15% of the articles; however, we believe that the risk of information bias is low because the inter-rater concordance was >95% for all variables, except for study design and the latter was assessed with the support of the main investigators. Finally, in this study, we assessed the association between statistician involvement as co-author and (i) the total number of submissions needed from the first submission to a journal and (ii) the time between submission to the journal that published the paper and acceptance, respectively publication. It would also be interesting to assess whether statistician involvement is associated with the total time to publication (i.e. from the first submission to a journal), ideally by using prospective observational design to decrease the risk of measurement error. Conclusion Close collaboration between researchers and statisticians has the potential to significantly reduce the time to publication of articles published in journals of general internal medicine and primary care. These data should encourage researchers to involve statisticians in the design, conduct, statistical analysis and reporting of research, not only to ensure high standards of quality but also to speed up its publication. Further studies using higher sample size and targeting other scientific journals are however necessary to confirm our results. Declaration Funding: this project was supported by institutional funding from the Faculty of medicine, University of Geneva. Ethical approval: not required (under Swiss law, ethical approval is not required when collecting non-personal health data). Conflicts of interest: none. Acknowledgements We would like to warmly thank Amir Moussa, administrative assistant, and Claire Ragot and Pierre-Henri Gorioux, MD students, for their precious contribution to the study. We would also like to thank Bernard Cerutti and François Herrmann for methodological support and Melania Bembea, Leonardo Silvestri and Dagmar Haller for their support and assistance throughout the study. References 1. Kalcioglu MT , Ileri Y , Karaca S , Egilmez OK , Kokten N . Research on the Submission, Acceptance and Publication Times of Articles Submitted to International Otorhinolaryngology Journals . Acta Inform Med 2015 ; 23 : 379 – 84 . Google Scholar Crossref Search ADS PubMed WorldCat 2. Moses L , Louis TA . Statistical consulting in clinical research: the two-way street . Stat Med 1984 ; 3 : 1 – 5 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Dobson AJ . The role of the statistician . Int J Epidemiol 1983 ; 12 : 274 – 5 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Simon R . The role of statisticians in intervention trials . Stat Methods Med Res 1999 ; 8 : 281 – 6 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Gore SM , Jones IG , Rytter EC . Misuse of statistical methods: critical assessment of articles in BMJ from January to March 1976 . Br Med J 1977 ; 1 : 85 – 7 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Delgado-Rodriguez M , Ruiz-Canela M , De Irala-Estevez J , Llorca J , Martinez-Gonzalez A . Participation of epidemiologists and/or biostatisticians and methodological quality of published controlled clinical trials . J Epidemiol Community Health 2001 ; 55 : 569 – 72 . Google Scholar Crossref Search ADS PubMed WorldCat 7. Diaz-Ordaz K , Froud R , Sheehan B , Eldridge S . A systematic review of cluster randomised trials in residential facilities for older people suggests how to improve quality . BMC Med Res Methodol 2013 ; 13 : 127 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Kloukos D , Papageorgiou SN , Doulis I , Petridis H , Pandis N . Reporting quality of randomised controlled trials published in prosthodontic and implantology journals . J Oral Rehabil 2015 ; 42 : 914 – 25 . Google Scholar Crossref Search ADS PubMed WorldCat 9. Tokolahi E , Hocking C , Kersten P , Vandal AC . Quality and reporting of cluster randomized controlled trials evaluating occupational therapy interventions: a systematic review . OTJR (Thorofare N J) 2016 ; 36 : 14 – 24 . Google Scholar PubMed WorldCat 10. Kumar R , Indrayan A , Chhabra P . Evaluation of quality of multivariable logistic regression in Indian medical journals using multilevel modeling approach . Indian J Public Health 2016 ; 60 : 99 – 106 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Amat CB . Editorial and publication delay of papers submitted to 14 selected Food Research journals. Influence of online posting . Scientometrics 2008 ; 74 : 379 – 89 . Google Scholar Crossref Search ADS WorldCat 12. Björk B-C , Solomon D . The publishing delay in scholarly peer-reviewed journals . J Informetr 2013 ; 7 : 914 – 23 . Google Scholar Crossref Search ADS WorldCat 13. Chen H , Chen CH , Jhanji V . Publication times, impact factors, and advance online publication in ophthalmology journals . Ophthalmology 2013 ; 120 : 1697 – 701 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Palese A , Coletti S , Dante A . Publication efficiency among the higher impact factor nursing journals in 2009: a retrospective analysis . Int J Nurs Stud 2013 ; 50 : 543 – 51 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Katz MH. Multivariable analysis: a practical guide for clinicians . 2 nd ed. Cambridge, NY : Cambridge University Press , 2006 . Google Preview WorldCat COPAC 16. UCLA: Statistical Consulting Group. Survival Analysis with Stata [Internet]. [cited 2018 Oct 14] . https://stats.idre.ucla.edu/stata/seminars/stata-survival/ (accessed on 15 March 2018). 17. UCLA: Statistical Consulting Group. Regression Models with Count Data [Internet] . [cited 2018 Oct 14]. https://stats.idre.ucla.edu/stata/seminars/regression-models-with-count-data/ (accessed on 15 March 2018). 18. UCLA: Statistical Consulting Group. Testing the proportional hazard assumption in Cox models [Internet]. [cited 2018 Oct 14] . https://stats.idre.ucla.edu/other/examples/asa2/testing-the-proportional-hazard-assumption-in-cox-models/ (accessed on 15 March 2018). 19. Jaykaran C , Chavda N , Yadav P . Proper reporting of statistical parameters in clinical trials published in Indian medical journals. Is inclusion of statistician play any significant role ? J Young Pharm 2011 ; 3 : 167 – 8 . Google Scholar Crossref Search ADS PubMed WorldCat 20. Wallach JD , Egilman AC , Gopal AD et al. Biomedical journal speed and efficiency: a cross-sectional pilot survey of author experiences . Res Integr Peer Rev 2018 ; 3 : 1 . Google Scholar Crossref Search ADS PubMed WorldCat 21. Sebo P , Maisonneuve H , Cerutti B et al. Rates, delays, and completeness of general practitioners’ responses to a postal versus web-based survey: a randomized trial . J Med Internet Res 2017 ; 19 : e83 . Google Scholar Crossref Search ADS PubMed WorldCat 22. Kellerman SE , Herold J . Physician response to surveys. A review of the literature . Am J Prev Med 2001 ; 20 : 61 – 7 . Google Scholar Crossref Search ADS PubMed WorldCat 23. Asch DA , Jedrziewski MK , Christakis NA . Response rates to mail surveys published in medical journals . J Clin Epidemiol 1997 ; 50 : 1129 – 36 . Google Scholar Crossref Search ADS PubMed WorldCat © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

Family Practice – Oxford University Press

Published: Jul 31, 2019

There are no references for this article.