Introduction: One still hears from critics of the chiropractic profession there is no research about chiropractic, or by chiropractors, or that chiropractic is unscientific. One possible reason why this myth might continue is the difficulty in searching for chiropractic research. The Journal of Manipulative and Physiological Therapeutics (JMPT) has been the preeminent chiropractic scientific journal and was the first chiropractic journal indexed in MEDLINE. Given its status within the chiropractic profession, the likelihood is that the majority of the papers published in it are by and about the chiropractic profession. MEDLINE currently contains over 13 million citations. However, the ability to search exclusively within the contents of JMPT offers a unique opportunity to determine the best methods to find research by or about chiropractors, due to the manageable size of the data set.
Objective:The purpose of this study was to investigate accuracy of indexing of chiropractic generated or related research in MEDLINE using papers published in JMPT as the sample.
Methods:The MEDLINE database of biomedical citations and abstracts was searched for citations from JMPT with limits that were set to each of the publication types used by MEDLINE (clinical trials, randomized controlled trial, reviews, meta-analysis, practice guideline, letters, editorials, and studies involving human subjects). Additionally, the search was limited to those papers classified as systematic reviews using the search term “systematic[sb]”. Each of the citations identified within a publication type were searched using the Boolean operator NOT and the MeSH terms “chiropractic” or “manipulation, chiropractic” to find those papers within each publication type that were not indexed by NLM with either of these terms. JMPT citations identified as being from the highest forms of clinical evidence (reviews, meta-analyses, randomized clinical trials and clinical trials) were analyzed by reviewing the complete MEDLINE record to determine if each study was associated with chiropractic. From these data, the following were calculated: sensitivity; specificity; and positive and negative likelihood ratios with 95% confidence intervals of using the MeSH terms chiropractic and manipulation, chiropractic to identify chiropractic literature in the JMPT MEDLINE dataset.
Results: In August of 2005, there were 1899 total citations from JMPT in MEDLINE. A total of 702 citations from JMPT were not indexed using the MeSH term “chiropractic” or “manipulation, chiropractic”. The accuracy of the indexing the four publication types from JMPT within MEDLINE. was 62.98%. With 95% confidence intervals, the specificity was 82.93% (73.36 - 89.55), sensitivity was 58.45% (53.3 – 63.42), positive likelihood ratio was 3.423 (2.108 - 5.56), and negative likelihood ratio was 0.501 (0.428 - 0.586).
Discussion: The aggregate poor sensitivity and negative likelihood ratios of using the MeSH terms “chiropractic” or “manipulation, chiropractic” to search the content of JMPT in MEDLINE demonstrates that these terms are not very useful in screening for papers about or by chiropractors within a chiropractic journal. One would infer then that these terms would be no better or in fact much worse at finding chiropractic literature within the complete MEDLINE dataset.
Conclusion: Searching the MEDLINE database for papers published in JMPT using the MeSH terms “chiropractic” or “manipulation, chiropractic” had a poor ability to identify appropriate chiropractic literature. Thus, the presence of the chiropractic contribution to biomedical scholarship may be hidden to those searching for that contribution. This finding reinforces the importance of developing a best practices database related to chiropractic that is available for scholars, clinicians and policy makers and easy to search.
This abstract is reproduced with the permission of the publisher.