We found a significant and exponential increase in the number of SDM publications in the 15 highest impact factor journals in general internal medicine between 1996 and 2011. Nevertheless, the number of SDM publications across the journals varied widely. When focusing on research publications, the growth was not exponential but remained mostly linear since 1996. We observed few publications assessing decision aids, however their growth pattern was similar to the general trend.
Several studies have explored the frequency of concepts in medical literature [5, 21] or focused on research productivity [18, 32]. Through a PubMed search of the term SDM, a previous study showed a sharp growth of the number of publications in the overall literature from 1996 to 2003 . Our results are concordant with these findings. We also obtained relative numbers to make sure that the increase was not only due to the growing number of overall publications in the studied journals.
More recently, a bibliometric analysis assessed the quality as well as the volume of research in primary care between countries . Our review in full-text, however, identified not only research publications but other publication types as well, such as editorials or narrative reviews, which could also have an impact on the readers. Moreover, we showed that the exponential increase of SDM observed in the last 16 years was mainly due to these types of publications, while research publications followed a linear trend during the same period.
Another study measured the distribution of topics in two major medical journals by disease categories and domains over a calendar year . They found that the topic prevention was underrepresented in the journals in comparison to the importance of this topic to patients and public health. In our study, we observed a large variation in the number of SDM publications across the 15 journals studied. In relative numbers, the Journal of General Internal Medicine released fifty times more publications than the last journal in the top 15. This variation may be partially explained by differences in editorial choices, which were not studied here. Nevertheless, the trends were analysed according to various components. Breaking down the increase in SDM publications into bibliometric data brought some keys to better understand it.
To our knowledge combining techniques of review through full-text search and bibliometric analysis has not been reported before. Full-text search techniques have been used in biomedicine,  especially genomics,  but not yet in clinical medicine. This new combined method offers both search precision and simplicity of use. The sensitivity of detecting phrases in full-text publications is much higher than when limited to a classical PubMed search [13, 34]. We validated the method of full-text search on journal websites through a good concordance rate when comparing with the validation dataset. Moreover, this method did not present the difficulties that we met in collecting the validation dataset from some uneasy publishers, a known issue in the scientific field . Validity of our abstraction form was also gained through high kappa values for the different categories of bibliometric data.
Our study has some limitations. First, our study was limited to the 15 major medical journals chosen according to their impact factor. Despite the widespread use of the impact factor metric,  it has inherent limitations [37–39]. We considered an alternative journal selection process based on journal circulation as a reflection of readership . Nevertheless, circulation counts are affected by the current increase in readers accessing journals online.
Second, categorising publication types from 15 different journals was limited by our a priori criteria as each journal has its own indexing method. Our category list was developed internally, inspired by PubMed and Embase classifications, yet we reached a high degree of agreement between both reviewers (agreement rate 95.4%, Cohen kappa 0.89). The discrepancy between agreement rate (93.6%) and Cohen kappa (0.24) on publications assessing decision aids could be explained by the imbalance of the 2×2 table and the low number of publications reporting on decision aid tools [41, 42].
Third, we did not divide papers into those where the concept of SDM was the primary topic and those where the concept was briefly mentioned. Our aim was to perform a scoping review of the concept of SDM in the selected medical journals. Future, more detailed studies could aim at better understanding if the concept was used as the primary topic or if authors just mentioned it.
Fourth, two journals did not exist before 1996 and were therefore excluded from our analyses. These were two open access journals (PLOS Medicine and the Annals of Family Medicine). We cannot determine if the trends in SDM would have been different in open access journals. Future studies might aim at determining if the trends in SDM are different between open access and other journals.
Finally, if we showed an increase in the number of SDM publications, we did not assess to which extent the medical community reads them. Therefore, the impact of this increase remains unknown. A future study could alternatively focus on the number of citations per SDM publication compared to citations of other publications .
The growth in SDM publications found in our study nevertheless supports the call of experts in the field for the medical community to implement SDM in practice [2, 3, 44, 45]. Some of these calls have been published in major medical journals. These journals are thought to have a large impact on a vast population of physicians [15, 46]. More than three years after the enactment of the Patient Protection and Affordable Care Act, there are concerns that the SDM model has not been promoted enough [7, 8]. Addressing SDM in major journals is therefore more important than ever, as it could be an efficient way to disseminate it among the medical community. Moreover, our study may capture the transition of interest and advocacy for SDM from experts to clinicians and policy makers [8, 47].
SDM has been called to ensure evidence-based patient choice, especially for equipoise situation . Even though some issues have been raised threatening the compatibility between EBM and SDM, such as practice or financial incentives to achieve quality standards, both approaches are now clearly justified to promote efficient care by integrating patient autonomy .