The Latin America and the Caribbean search strategy proposal

Introduction

It is important to understand health promotion in a developing region such as Latin America and the Caribbean (LAC) because the medical literature therein is under-recognized by scientists around the world (1), who tend to conduct searches only in English and in English databases (2,3). Overall, Latin America contributes only 2% of global scientific production, mostly due to the contributions of Argentina, Brazil, Chile and Mexico (4). As a result, South American countries’ science has been compared to a night sky with just a few specks of light (5). We believe it is important to make those brightening stars as reachable as possible to acknowledge the importance of current health promotion innovations in LAC; we will accomplish this by improving the way we search for best practices and evidence.

Even though LAC remains as one of the most imbalanced regions in socio-economic, demographic and health characteristics (6), the development of its research has been growing, most of it through several country-based initiatives and ‘North-to-South’ collaborations (7,8). For example, Peruvian articles are usually co-authored with scientists outside LAC that allow a greater visibility (9). The diffusion of the research conducted in these countries is not only essential to promote consciousness about a striking region with remarkable investigations, but also to encourage partnerships in public health between the LAC countries and the international community (1,7).

We want to provide readers with a new method to exhaust the possibilities of their search results about LAC within MEDLINE/PubMed, an idea similar to that recommended by the Cochrane Highly Sensitive Search Strategy for maximizing sensitivity in systematic reviews (10) and an easier approach than that proposed by Perel et al. (11). With the intention of validating our query, we compared it with the Medical Subject Headings (MeSH) terms most frequently used for searching studies regarding the LAC region in MEDLINE/PubMed.

Methods

In order to standardize our search, we took as a reference the way the MeSH term ‘Latin America’ is defined (12). Usually, this MeSH term is broken down into the following subterms: ‘Caribbean Region’, ‘Central America’, ‘Mexico’, and ‘South America’. With these results, we created the ‘Classic Query’ as the comparator for our proposed system.

Our proposal takes each subterm, and does an iterative search on each single country within the region. We call this ‘The Latin America and the Caribbean search strategy’ (LACSS). The first part of the search query consists of all terms that conform to the LAC region according to its MeSH, combined with ‘AND’ and followed by a Keyword. Both queries are shown in Table 1.

OPEN IN VIEWER

Table 1.

Classic query and the LACSS.

Classic query

(“Latin America”[Mesh] OR “South America”[Mesh] OR “Central America”[Mesh] OR “Caribbean Region”[Mesh] OR “Mexico”[Mesh]) AND Keyword

LACSS

(“Anguilla” OR “Antigua and Barbuda” OR “Argentina” OR “Aruba” OR “Bahamas” OR “Barbados” OR “Bolivia” OR “Belice” OR “Belize” OR “Brasil” OR “Brazil” OR “British Virgin Islands” OR “Cayman Islands” OR “Chile” OR “Colombia” OR “Costa Rica” OR “Cuba” OR “Dominica” OR “Dominican Republic” OR “Republica Dominicana” OR “Ecuador” OR “El Salvador” OR “French Guiana” OR “Grenada” OR “Guadalupe” OR “Guatemala” OR “Guiana” OR “Guyana” OR “Haiti” OR “Honduras” OR “Leeward Islands” OR “Jamaica” OR “Martinique” OR “Mexico” OR “Montserrat” OR “Netherlands Antilles” OR “Nicaragua” OR “Panama” OR “Paraguay” OR “Peru” OR “Puerto Rico” OR “Saint Kitts and Nevis” OR “Saint Lucia” OR “Saint Vincent and the Grenadines” OR “Suriname” OR “Surinam” OR “Trinidad and Tobago” OR “Turks and Caicos Islands” OR “Uruguay” OR “Venezuela” OR “Virgin Islands of the United States” OR “Windward Islands” OR “Caribbean” OR “Central America” OR “Latin America” OR “South America” OR “West Indies”) AND Keyword

We conducted a search in MEDLINE/PubMed on 22 February 2016, using several keywords categorized as non-communicable diseases (NCD), neglected tropical diseases (NTD), injuries and other important public health relevant topics in the region. We chose these subjects because the LAC countries have experienced a fast epidemiological transition since the 1990s, where a combination of NCDs and injuries, on top of a well-known infectious scenario, has generated a unique setting (6). Keywords are displayed in Table 2.

OPEN IN VIEWER

Table 2.

Keywords and results according to each query.

		Classic Query	LACSS	Ratio
NCD Keyword	‘Alzheimer Disease’[Mesh]	259	1284	4.96
	‘Asthma’[Mesh]	1035	2385	2.30
	‘Atherosclerosis’[Mesh]	121	653	5.40
	‘Cataract’[Mesh]	124	313	2.52
	‘Diabetes Mellitus’[Mesh]	2844	8128	2.86
	‘Epilepsy’[Mesh]	509	3369	6.62
	‘Fibromyalgia’[Mesh]	35	200	5.71
	‘Hypertension’[Mesh]	2009	5610	2.79
	‘Kidney Failure, Chronic’[Mesh]	529	1699	3.21
	‘Myocardial Infarction’[Mesh]	496	1882	3.79
	‘Obesity’[Mesh]	2623	5944	2.27
	‘Osteoporosis’[Mesh]	224	872	3.89
	‘Peripheral Arterial Disease’[Mesh]	17	68	4.00
	‘Pulmonary Disease, Chronic Obstructive’[Mesh]	211	822	3.90
	‘Stroke’[Mesh]	438	1573	3.59
NTD Keyword	‘Chagas Disease’[Mesh]	2723	6424	2.36
	‘Chikungunya Fever’[Mesh]	41	68	1.66
	‘Cysticercosis’[Mesh]	410	1209	2.95
	‘Dengue’[Mesh]	1572	2319	1.48
	‘Helminthiasis’[Mesh]	5437	9638	1.77
	‘Leishmaniasis’[Mesh]	2287	4356	1.90
	‘Leprosy’[Mesh]	1055	1795	1.70
	‘Taeniasis’[Mesh]	452	1341	2.97
Injuries Keyword	‘Accidents, Traffic’[Mesh]	465	728	1.57
	‘Domestic Violence’[Mesh]	704	1035	1.47
	‘Drowning’[Mesh]	25	44	1.76
Other Keywords	‘Bartonella Infections’[Mesh]	117	200	1.71
	‘HIV’[Mesh]	1246	2564	2.06
	‘Malaria’[Mesh]	2404	3635	1.51
	‘Helicobacter pylori’[Mesh]	541	1270	2.35
	‘Stomach Neoplasms’[Mesh]	542	1277	2.36
	‘Tuberculosis’[Mesh]	2878	4840	1.68
	‘Zika Virus’[Mesh]	3	5	1.67
Total		34376	77550	2.26

Finally, we compared the performance of our search strategies (number of articles retrieved) and calculated the ratio between them by dividing the total results in the LACSS by the total results in the Classic Query.

Results

We found around 40,000 more results with our proposed method than with the classic MeSH query (Table 2). Interestingly enough, we can see up to a 6-fold increase by using LACSS. When we calculated the ratio differences between the results, the biggest ratios were seen for ‘Epilepsy’ and ‘Fibromyalgia’, whilst the smaller ratio differences were observed in ‘Domestic Violence’ and ‘Accidents, Traffic’. If we separate the keywords by groups (NCD, NTD, Injuries) we noticed that the group with the biggest change in results is in the NCD group (median of 3.79); followed by NTD group (median of 1.84) and finally the Injuries group (median of 1.57).

Discussion

As we had previously described, LAC is a region where chronic conditions overlap with infectious diseases. Because both are so relevant in our area, we focused on articles on those topics. When comparing the number of items found using the Classic Query versus the LACSS, our results demonstrated that our structured approach allowed us to obtain as much as six times more information than the Classic Query. The maximization of the results with this new LACSS hopefully prevents the loss of important elements that could contribute to the inquiry readers are seeking.

The increase in the number of items found is because the Classic Query presents some drawbacks and pitfalls. First, it requires the article to be indexed by the MeSH term of the country or the region. A second important factor to consider is that the National Library of Medicine (NLM) indexing process may fail when assigning the MeSH terms, retrieving fewer results (13). The example that Hooijmans et al. use in their paper is that the NLM may miss MeSH terms for a specific topic, such as only assigning the MeSH term ‘mice’ instead of ‘animals’ and ‘mice’ for mice-related experiments (13). Furthermore, Fielding and Powell stated that MeSH terms may be poorly indexed (14), and this was later demonstrated by Minguet et al. when they proved the low quality of assignment of the pharmacy-specific MeSH terms (15). This means that if the MeSH term is not set as a keyword in the article, it would not be found in the search. Third, there is a delay when MeSH-indexing newly submitted articles (13), which makes them impossible to find. Finally, many scientists do not know how to use MEDLINE/PubMed to its full extent, especially when applying the MeSH terms (13). Liu demonstrated that MeSH terms differ in usefulness according to the user’s level of expertise (16). This means that the lack of familiarity with a specific topic and its MeSH term may lead to the loss of crucial information, even when using separate terms for each country, and could lead our brightening stars of LAC to fade away in the archive. This problem arises with new epidemics, such as Zika Virus, in which LACSS proved to be 1.67 times as efficient when retrieving results. Getting more information about the virus in our region of the world would help others know what is happening and help innovate new public health ideas to fight it and promote good health practices.

There are four limitations to our LACSS. A first criticism of our proposed system is that we limited ourselves to only creating a method to perform literature searches in MEDLINE/PubMed and avoided using LAC databases which perhaps could be more representative of the quantity of articles in the region (17). Undesirably, the LAC database search engines are limited and do not allow complex searches. Although some initiatives have been developed to maximize results when using them (18), they have not had a noteworthy impact. This is the reason why we only chose MEDLINE/PubMed to create our LACSS. The second limitation refers to its cumbersome syntax which can make it harder for researchers and scientists to replicate. The third limitation is that our query uses MeSH keywords to look for specific topics, which may limit the number of results (as stated before). However, this was done mostly in order to compare with the classical method of searching. The fourth limitation is if, for example, a country’s name is not listed in the publication but only the nationality (Brazil vs. Brazilian), our query will not find it. LACSS was created to help increase sensitivity of the search results; however, it could be argued that some articles that will come up are not relevant to the researcher. This can occur in two scenarios; the first one is when an author’s affiliation is associated with a specific country but the research was not done there. The second scenario where this could occur is when a country’s name is just mentioned in the abstract without having actual relevance.

We acknowledge that aside from being an innovative search strategy, our system is also a powerful tool for authors, because if we consider that these publications are failing to be read, they will not be cited, and thus the dissemination of their findings will be hindered.

Conclusion

We encourage readers and researchers interested in obtaining the largest amount of public health information about the LAC region in MEDLINE/PubMed to use our LACSS. It will brighten a little bit our dark night sky of science, and will help others build on prior knowledge to help promote and improve public health. If we are lucky enough this might even help our region have a night sky as bright as day.