Academic journal article Journal of the Medical Library Association

Nuggets: Findings Shared in Multiple Clinical Case Reports

Academic journal article Journal of the Medical Library Association

Nuggets: Findings Shared in Multiple Clinical Case Reports

Article excerpt

INTRODUCTION

The problem of identifying subsets of documents that "say (more or less) the same thing" is related to, but distinct from, existing text-mining techniques that seek to classify or cluster documents according to overall similarity of the topics that they discuss, to identify predominant themes within the literature, or to extract or summarize knowledge from an entire body of documents [1]. Clinical case reports are an ideal type of biomedical literature for undertaking this type of analysis, for medical as well as methodological reasons. Roughly 1.7 million articles are indexed in PubMed as case reports, about 7% of all biomedical articles; yet case reports rank at the bottom in the hierarchy of evidence-based medicine, far below randomized controlled trials. Case reports are generally not included in the assessment of clinical evidence carried out by systematic reviews and meta-analyses. Case reports are generally poorly cited relative to other research articles [2, 3], leading some journals either to stop accepting case reports or to classify them as "letters" or "comments" that will not affect the journal's impact factor. Like any other type of eyewitness reports, case reports may be prone to observer bias, may reflect wrong assumptions or premature interpretation of findings, and generally lack controlled conditions so that, at best, findings are correlative instead of conclusive.

On the other hand, eyewitness reports often provide the first observations of new phenomena or new innovations. For example, three case reports appearing in 1981 provided the first delineation of AIDS [4-6]. Randomized clinical trials generally follow patients for short periods and may be subject to sponsor bias, so case reports may be a more reliable, unbiased venue for reporting long-term adverse effects. Recent editorials have argued that case reports provide essential evidence in medicine [7]. In fact, there has been a recent renaissance of the case report literature. Leading journals such as Lancet continue to publish case reports regularly, and scores of new journals devoted specifically to case reports have emerged from publishers such as BMJ, Elsevier, Wiley, Oxford University Press, and Hindawi. Case reports may not be entirely independent of each other (since publishing one report may spur interest in reporting additional patients), but except in rare situations [8], each article may be taken as an independent source of evidence.

Methodologically, case reports are a favorable test bed for exploring discovery of main findings because each case report tends to be short and concise. Case reports tend to state a single main finding that is often directly stated in its title. This situation is much simpler than occurs for some publication types, such as clinical trials or genetic association studies [9].

In summary, case reports are a valuable, unique, yet noisy and underutilized type of evidence. The authors believe that there is considerable value in identifying findings that have been independently published in multiple case reports, since that would alert readers to evidence that has particularly high reliability and potential impact. In turn, this might encourage wider judicious use of case reports in evidence-based medicine and other tasks such as surveillance of drug side-effects [10]. Despite the fact that many case report articles are accessible and indexed in MEDLINE, no automated tool exists that can find subsets of articles that report the same or very similar main findings.

METHODS

Search strategies

Overall, about forty-five PubMed queries were performed and analyzed to learn whether there were, indeed, multiple case reports that stated the same or almost the same main finding. We also attempted to validate our expectation that the main finding of a case report could generally be discerned by simply looking at the title.

First, in an exploratory strategy, fifteen ad hoc PubMed queries were performed in July 2011 by entering the name of a common drug and the name of a common disease (limited to case reports [Publication Type]) into Anne O'Tate, a value-added PubMed search engine that is publicly accessible [11, 12]. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.