gms | German Medical Science

Introduction: Lack of time and information-searching skills are the two most frequent barriers to searching for information amongst physicians. In contrast, easily searching for and finding synthesized evidence-based information would facilitate information searching [1].

In 80 per cent of all searches on PubMed, only the first 20 references are browsed [2]. NLM has addressed such findings by making a ‘Best match’ relevance sorting of their results to meet searchers’ needs [3]. However, this solution is based on automatic algorithms and not information specialist expertise that may provide even more relevant results.

We aimed to develop a search tool where lack of time and search skills are no longer barriers to searching for relevant health information.

Methods: The tool was developed by a team of knowledge brokers and information specialists working on information retrieval and management at a specialised diabetes hospital.

We developed a web application based on the front-end JavaScript framework Vue.js and NLM E-utilities to make queries to the PubMed database based on pre-defined search strings. All search strings were created by information specialists.

Results: The tool is implemented on a publicly accessible website (https://danishdiabetesknowledgecenter.dk/quickpubmed) and requires only short guiding to use (Figure 1 [Fig. 1]). Search results, including abstracts, are displayed below the search form.

In default mode, the tool is available in a simple version suitable for quick searches that are expected to yield highly relevant results. The user composes a query by combining available options in dropdown menus and checkboxes that contain topics relevant to diabetes, such as complications, medication, lifestyle and technology, and limits such as study type, geography, language, and age group. Because each option and limit activates a predefined search string, no knowledge of search string development is required from the user. However, the actual combined search strings can be found by clicking a ‘Details’ link.

An advanced mode of the tool is also available. Here, the user can choose between three search string versions to compose searches with higher specificity or sensitivity. Further, free-text can be entered manually.

We created the three search strings for each topic with different sensitivity to meet different user needs. A ‘narrow’ version with very high specificity, a ‘broad’ version with lower specificity, and a ‘normal’ version balancing the two.

Preliminary feedback from clinicians at a Danish specialised diabetes hospital has been positive.

Conclusion: We have developed a tool to conduct quick PubMed searches in a clinical area applying options relevant to clinicians and searchers scoping the literature. The tool uses predefined search strings and can easily be implemented and adapted to other health-specific areas. The tool will be evaluated in 2022.

Keywords: PubMed, searching, development, implementation, E-utilities