Palabras clave: Evidence-based medicine, Bibliographies as topic, Databases, Bibliographic
The currently abundant bibliography on healthcare can make the search process an exhausting and frustrating experience. For this reason, it is essential to learn the basic concepts of research question formulation, information sources, and search strategies to make this process more efficient and user-friendly. The search strategy is an iterative process that allows the incorporation of tools and terms in the strategy design to optimize evidence retrieval. Each strategy varies according to the questions, the language used, the source of information accessed, and the available tools. This article is part of a methodological series of narrative reviews on biostatistics and clinical epidemiology. This narrative review describes the essential elements for developing a literature search strategy and identifying the relevant evidence concerning a clinical question through familiar and accessible sources (such as Google and Google Scholar), as well as search interfaces and technical-scientific databases focused on biomedical knowledge (PubMed and The Cochrane Library).
Evidence-based medicine (EBM) is a tool that seeks to converge the best available evidence with clinical expertise, in addition to values and patient preferences [1]. Traditionally, five steps are described for a practical implementation, known as the five As. These are: ask (formulate the question), acquire (search for the best evidence), appraise (evaluate findings), apply (apply the results), and assess (evaluate the results) [2].
Nowadays, the available information on a given topic can become very broad, resulting in an overwhelming and burdensome process if the necessary tools are unavailable or unknown. It is estimated that only half of healthcare professional questions are searched, and only about two-thirds of these are answered [3]. The main challenges encountered by physicians are the lack of training in bibliographic search, the lack of knowledge of the multiple digital sources of information, and poor search strategies [3].
This article is part of a methodological series of narrative reviews on general biostatistics and clinical epidemiology topics, which explore and summarize published articles available in the main databases and specialized reference texts in a user friendly language. The series is oriented to the training of undergraduate and graduate students. It is carried out by the Chair of Evidence-Based Medicine of the School of Medicine of the 'Universidad de Valparaiso', Chile, in collaboration with the Research Department of the University Institute of the 'Hospital Italiano de Buenos Aires', Argentina, and the Catholic University Evidence Center. This narrative review describes the essential elements for developing a literature search strategy and identifying relevant evidence concerning a clinical question.
The main objective of a literature search is to find scientific articles and other sources of information that can answer a focused clinical question while making a practical, structured, and reproducible search.
The search should begin once the critical elements of the clinical question have been defined. If the clinical question is not formulated correctly, the search will probably not yield good results. Consequently, it is essential to formulate a well-defined question that allows us to establish criteria for selecting the evidence and approach the results of our interest. For practical purposes, we can classify the questions into:
a) Fundamental or global deficit (background) questions are general questions on basic academic concepts about a given disease or clinical concept [4]. For example: What is the pathophysiology of heart failure? What are the possible differential diagnoses for a given symptomatology? The ideal sources of information to answer this type of question can be textbooks, narrative reviews, or bibliographic abstracts. These are the so-called secondary sources of information [4].
b) Advanced or foreground questions are questions concerning specific knowledge applied to a particular patient or problem [4]. The PICO format is the most widely used to answer foreground questions. The letter "P" refers to the population or patient we are interested in and its characteristics. The "I" corresponds to the intervention we seek to carry out, the "C" is the comparison (therapeutic alternative or placebo), and the "O" is the objective or outcome we wish to evaluate [4]. For example: "In infants with bronchiolitis (P), does treatment with salbutamol (I) compared with placebo (C) decrease overall mortality (O)?". This type of question focuses on treatment, but there are also modifications to the PICO approach depending on whether the question is prognostic, etiologic, diagnostic, prevalence, or harm-focused (adverse events). For didactic purposes, we take therapeutic questions as a model in this article.
Knowing and understanding the available resources to answer a specific clinical question allows us to search more effectively and efficiently. This search is intended to lead to the most appropriate answer for the clinical context in which it is developed. A clear example can be seen in the abovementioned questions. Background questions can be answered by searching secondary information sources (textbooks, encyclopedias, or bibliographic summaries). On the other hand, foreground questions require an exhaustive search from many sources (scientific journals and medical bibliography databases). In the latter, the amount of information can be substantial, and the quality can be varied. Considering this aspect, Haynes proposed the "6S pyramid", a model in which he organized the different types of evidence (Table 1) [5],[6].
Table 1. Types of information sources.
Once a clinical question has been formulated, and the vital source of information has been identified, we can move on to the next step. At this stage, we need to select the necessary databases, evaluate the availability of search interfaces, design the search strategy and adapt it to the interface(s) that allow us to retrieve the information relevant to our question.
In the following sections, we describe the methodological process of a bibliographic search, considering the types and characteristics of the information sources. Specifically, we address the search tools that are most used in Latin America. In Box 1, we present a glossary of the terms used in the article. Some specialized resources (EMBASE, PsycINFO, and CINAHL) will not be addressed because they are relevant to specific areas of knowledge, require institutional subscriptions, and are not the most frequently used by health professionals.
Box 1. Glossary of terms. Database Thesaurus/controlled language Natural language Boolean operator Search interface Indexing Repository Gray literature |
Google’s search interface is the most widely used because it is intuitive and accessible. Google uses artificial intelligence to
search, index, and retrieve the most relevant results [7]. We likely find synthesized and simplified information intended for patients or the general population using natural or colloquial language. Instead, technical language, preferably in English, should be used to find scientific information (e.g., types of research design such as clinical trials or case-control studies). Some strategies to obtain better results include:
a) The use of scientific language.
b) Describe the methods sought (e.g., study design).
c) Add terms related to reliable databases/interfaces (PubMed, SciELO, Cochrane, or others).
Figure 1. Example of Google searches.
Suggestions usually appear from the Google Scholar interface (https://scholar.google.com/). The algorithms used by the conventional Google search engine are different from those of Google Scholar since the latter prioritizes searching and indexing complete and original articles based on relevance, with less advertising and commercial priority. It has the advantage of having a wide variety of bibliographic material, including books and articles.
Google Scholar uses more frequently non-English literature (up to 40%) for its citation count, unlike other databases (Web of Science and Scopus), whose frequency of English literature reaches up to 90% [8]. In addition, it indicates how many times and by whom an article has been cited. It is a source that allows access to information resources from different areas of knowledge, retrieving results not found in area-specific sources [9]. It has limited search support from Boolean operators and other search operators as a disadvantage. In addition, the ranking of search results considers the availability of full text, place and site of publication, author(s), and citation frequency. However, the relative weight of these factors is unclear [10].
Finally, the search options are inflexible (it is not possible to filter by document type, search by field, or refine by subject, among others), and its instructions are not very accessible. There is also no rigorous quality control of the sources processed, and duplicates may be found. For these reasons, it may be necessary to resort to the advanced search option, which filters the results by date, author, and article content. Another option this platform provides is linking the search to university libraries [11] (Figure 2).
Figure 2. Advanced search in Google Scholar.
PubMed (https://pubmed.ncbi.nlm.nih.gov) is the search interface of the US National Library of Medicine that allows free access to more than 30 million bibliographic citations, including the MEDLINE database – one of the most important databases in health sciences [12]. The results are more
appropriate for a healthcare-related search than a basic Google search as this engine focuses on biomedical knowledge. Its tools allow advanced, reproducible, and more controlled searches, meaning the retrieved results will be similar using the same search strategy. However, the interface is not intuitive for people unfamiliar with Boolean operators. PubMed retrieves
bibliographic data to access the full text, including the digital object identifier (DOI). The DOI allows searching the document in different information resources to access its contents and add the bibliographic data in managers such as Paperpile and Zotero [13].
PubMed has two types of searches: basic and advanced. Enter the term or phrase of interest and press the Search button to start a basic search. PubMed uses the "AND" operator between word spaces in a search phrase by default. Therefore, we must check that there are no extra spaces between terms and correct logical operators to perform the search.
PubMed: basic search
When searching in PubMed, we must consider the language used. Natural language is preferred if we want a more sensitive search. If we want the search to be more specific, we must use the controlled language of the MeSH thesaurus. Another feature of the controlled language is that it includes all the synonyms and variants of the search term, considerably reducing the risk of losing articles that do not include the specific term used in the search strategy. For example, "Renal Insufficiency" [MeSH] retrieves articles that cover "renal insufficiency", "kidney insufficiency", "renal failure", "kidney failure", among others. It is important to note that these languages (natural and controlled) are complementary, and it is advisable to use both in the same search (Table 2).
PubMed: advanced search
The advanced search is helpful for a structured and reproducible approach. In these cases, natural and controlled language terms are selected a priori, corresponding to each concept that composes our query. By identifying the search interface and its possibilities, the terms can be added and combined in specific bibliographic records fields, combining them with the Boolean operators that best suit our search. In addition, there are additional tools such as filters, wildcards, and truncations, which allow refining the search even further (Box 2).
Box 2. Advanced search tools. Field labels Filters Truncation (*) Wildcardsa NEAR or NEXTa proximity operators Source: Prepared by the authors of this study. |
Targeted or sensitive search
These multiple tools available in PubMed allow a more flexible search. If we want an even more specific search (i.e., as few results as possible and more relevant to our query), we should add more components of the predefined PICO query linked with the Boolean operator "AND". It is preferred not to use the Boolean operator "NOT" because relevant references may be excluded. In addition, filters available in the interface (e.g., for systematic reviews or clinical trials) can be used to narrow the search and further decrease results based on the study design (T) component of the PICO question [14]. Some
generally inadequate methods used to increase search specificity include:
a) Limiting by date: relevant articles may be published before the imposed limit. However, it would be reasonable to limit by date to identify systematic reviews since older ones may be outdated.
b) Limit by article availability ("free full text" in PubMed): the relevance of an article is not determined by its accessibility. In these cases, it is suggested to consult with your institution to obtain the full text of the reference.
Suppose we want a more sensitive search (with the largest possible number of results, but with many results that may not be relevant to our question). In that case, we can remove components of the PICO question and apply the least number of filters (or omit them). Other options to broaden the search are the application of truncations and adding additional terms with the Boolean operator “OR”.
It is essential to understand that search is an iterative practice: the process does not end with a single search. In contrast, we need multiple searches to refine the quantity and quality of results obtained from analysis and the aforementioned tools. Ultimately, the sequence of searches, the evaluation of the results, and the appropriate use of the available tools determine the final accuracy.
Steps of an advanced search design
Once the advanced search structure has been created, the first step is conceptualizing the PICO question’s elements with controlled language (MeSH) and natural language. We need to use natural language terms in the title and abstract fields ([title/ abstract] or [tiab]) to avoid irrelevant results. Terms that relate to the same element of the PICO question are joined by an "OR" and closed in parentheses, thus forming the search line (e.g., line number 1 or #1). We should repeat this process with all the elements until the lines join the final strategy with an "AND". Finally, the results obtained by our search are displayed and analyzed. In case the retrieved results are not as expected, it is essential to review the literature on the topic we are trying to address and apply the tools described in the previous section to refine the sensitivity and specificity of the search (Table 3).
Table 3. Advanced search on PubMed.
The Cochrane Library (https://www.cochranelibrary.com) is a collection of databases containing high-quality evidence for decision-making in healthcare. This platform has more than 8300 reviews, 2400 protocols, 1 600 000 clinical trials in the Cochrane Central Register of Controlled Trials (CENTRAL), 2400 Cochrane Clinical Answers (brief summaries of Cochrane systematic reviews), 300 000 Epistemonikos reviews, 130 editorials, and 30 special collections. In addition, it has the advantage of being available in multiple languages, including English and Spanish [15].
Like PubMed, it has two search options: basic and advanced. Type the term of interest in the search box and select the
desired search field to start a basic search. The Cochrane Library also allows searching by topic or Cochrane Review Group. The logic is similar to an advanced search in PubMed. However, it is worth mentioning that The Cochrane Library has a search system based on the PICO format, which is currently in the testing phase and is limited to Cochrane systematic reviews. In addition, the advanced search adds other search tools, such as the proximity operators (Box 2).
In brief, the Cochrane Library has a search system similar to PubMed and a wide range of information content. Mainly, it is a source that can be accessed when our objective is to obtain a complete summary of the available evidence on our clinical question.
LILACS (https://lilacs.bvsalud.org) is a database that includes technical-scientific papers produced by authors from Latin America and the Caribbean related to health sciences. This database has its thesaurus called DeCS, which is based on MeSH terms and is available in English, Spanish, and Portuguese [16],[17]. The methodological process is similar to the other information sources seen above, although advanced searches are limited. LILACS has an interface with tools and filters specific to the information source. Generally, we suggest using a few search terms and natural language (Table 4).
Table 4. Examples of searches in other databases.
To find the best available evidence for a focused (foreground) clinical question, we need to determine the type of question (therapeutic, prognostic, etiological, diagnostic, prevalence, or harm), identify its components, and apply them to the PICO format or its variants. Once our question has been formulated, it is necessary to determine which source of information (6S pyramid) is the most relevant for obtaining the results. And finally, after selecting search locations, we must always keep in mind that a search process is iterative. It is necessary to try different variants in the search strategy – using natural and controlled language and tools such as truncation, Boolean operators, filters, among others – until we obtain satisfactory results.
Contributor roles
LFT: conceptualization, research, writing original draft, reviewing and editing, visualization. CEL: conceptualization, methodology, research, writing original draft, reviewing and editing, visualization, supervision. LVM: conceptualization, research, writing original draft, reviewing and editing, visualization. JPB: conceptualization, methodology, research, writing original draft, reviewing and editing, visualization, supervision. JVAF: conceptualization, methodology, research, writing original draft, reviewing and editing, visualization, supervision.
Acknowledgments
We thank the Chair of Evidence-Based Medicine of the School of Medicine of the 'Universidad de Valparaíso', Chile, for promoting this series and the collaboration of the Research Department of the 'Instituto Universitario' from the 'Hospital Italiano de Buenos Aires', Argentina.
Competing interest
The authors completed the ICMJE conflict of interest statement and declared that they received no funding for the completion of this article; have no financial relationships with organizations that may have an interest in the published article in the last three years; and have no other relationships or activities that may influence the publication of the article.
Funding
The authors declare that they have no external sources of funding associated with this article.
Ethics
Due to the nature of this study, approval by an ethics committee was not required.
Language of submission
Spanish.
The currently abundant bibliography on healthcare can make the search process an exhausting and frustrating experience. For this reason, it is essential to learn the basic concepts of research question formulation, information sources, and search strategies to make this process more efficient and user-friendly. The search strategy is an iterative process that allows the incorporation of tools and terms in the strategy design to optimize evidence retrieval. Each strategy varies according to the questions, the language used, the source of information accessed, and the available tools. This article is part of a methodological series of narrative reviews on biostatistics and clinical epidemiology. This narrative review describes the essential elements for developing a literature search strategy and identifying the relevant evidence concerning a clinical question through familiar and accessible sources (such as Google and Google Scholar), as well as search interfaces and technical-scientific databases focused on biomedical knowledge (PubMed and The Cochrane Library).
Citación: Trivisonno LF, Escobar Liquitay C, Vergara-Merino L, Pérez-Bracchiglione J, Franco JVA. Key concepts for searching evidence: an introduction for healthcare professionals. Medwave 2021;21(11):e002512 doi: 10.5867/medwave.2022.01.002512
Fecha de envío: 20/7/2021
Fecha de aceptación: 15/11/2021
Fecha de publicación: 7/1/2022
Origen: No solicitado
Tipo de revisión: Con revisión por pares externa, por dos árbitros a doble ciego
Nos complace que usted tenga interés en comentar uno de nuestros artículos. Su comentario será publicado inmediatamente. No obstante, Medwave se reserva el derecho a eliminarlo posteriormente si la dirección editorial considera que su comentario es: ofensivo en algún sentido, irrelevante, trivial, contiene errores de lenguaje, contiene arengas políticas, obedece a fines comerciales, contiene datos de alguna persona en particular, o sugiere cambios en el manejo de pacientes que no hayan sido publicados previamente en alguna revista con revisión por pares.
Aún no hay comentarios en este artículo.
Para comentar debe iniciar sesión