Technical Web Appendix: Search Strategies for Interventions to Reduce Racial and Ethnic Disparities in Healthcare

Debra Werner, Biomedical Reference Librarian, University of Chicago

This document provides additional details about the search strategies used by authors of papers included in the Finding Answer Symposium to identify intervention studies with relevance to reducing disparities in health care.

Choosing Search Terms

The Finding Answers project began the selection of search terms with the MEDLINE database, identifying appropriate Medical Subject Headings, or MeSH terms. For each term, we reviewed the scope note to understand its definition, discover previous terms under which a concept had been indexed, and learn the date the term was introduced in MeSH. From the review of scope notes we found the MeSH term ‘Healthcare Disparities’ was introduced in 2008. Since this term had no Previous Indexing terms, meaning there were no older terms under which articles were indexed for this precise concept, we needed to determine which, if any, MeSH terms were used to represent the health care disparities concept in articles prior to 2008. We found that the manner in which literature is indexed and described varies, even within a single database, depending upon the needs and conventions of a particular discipline. For example, the asthma team found that using health care disparities terms to identify intervention studies relevant to reducing disparities in asthma excluded many relevant articles. The team discovered that combining terms for minority or ethnic groups with asthma terms produced a more comprehensive collection of articles. But while minority and ethnic group terms served as a proxy for health disparities in the asthma search, the same was not true of the other diseases or conditions, in which health disparities terms were used in their searches. In another example, the HIV team found the MeSH terms for minority and population groups too limiting. Many studies in the HIV literature did not explicitly state the inclusion of minority populations, yet the studies included greater than 50% minority participants, which met the inclusion criteria. Future researchers should bear these discipline-specific concerns in mind, and explore the terms used to describe the literature particular to their field, and tailor their search accordingly. Researchers should review the terms used to index known, relevant articles and compare the results of different search strategies—an iterative, time-intensive process—noting which searches exclude relevant articles.

Applying the Search Across Databases

We found it was necessary to select a unique set of search terms for each database or information resource searched. While there is considerable overlap between the controlled vocabularies found in MEDINE and CINAHL (Cumulative Index to Nursing and Allied Health Literature), MeSH and CINAHL Headings, respectively, there are enough differences that a search strategy used in MEDLINE usually needs to be modified to take full advantage of CINAHL’s indexing. While an unedited MEDLINE search may produce results when run in CINAHL, articles may be missed due to differences in index terms. A quick review of the terms used to represent racial and ethnic groups in each database, available at the end of each paper, illustrates the differences.

A limitation that did impact the teams’ ability to streamline its searches was PsycINFO’s explode feature in its Thesaurus of Psychological Index Terms on the PsycNET platform. PsycINFO includes the index term “Racial and Ethnic Groups,” which has narrower, or more specific, terms such as “African Cultural Groups,” “Asians,” “Latinos/Latinas,” and more. A few of these narrower terms themselves contain narrower terms still, for example “Latinos/Latinas” has a narrower term of “Mexican Americans.” The explode feature, according to the Help file on PsycNET, “allows you to search for an index term and its narrower terms.” The problem was that it only searched for one tier of narrower terms but not deeper. For example, when exploding the term “Racial and Ethnic Groups,” the terms one level below, such as “Latinos/Latinas,” are searched but not the terms that are below this tier, such as “Mexican Americans.” In order to include all of the racial and ethnic groups in a search, we found it necessary to add them individually. Future researchers should keep this limitation in mind, and when using PsycINFO via PsycNET only rely on the explode feature when there is just one level of narrower terms.