Novel Population Search Filter Development Methodology - the Black Persons Living in the United States Project

Article type
Authors
Datko M1, Davies M2, Glover S3, Lubker I4, Paynter R5, Vieira D6, Voisin C7
1AHRQ Effective Health Care Program ECRI-Penn Medicine Evidence-based Practice Center
2AHRQ Effective Health Care Program Kaiser Permanente Evidence-based Practice Center
3Northeast Georgia Health System
4Medical University of South Carolina
5AHRQ Effective Health Care Program Scientific Resource Center
6NYU Health Sciences Library
7AHRQ Effective Health Care Program Research Triangle Institute-University of North Carolina Evidence-based Practice Center
Abstract
Background: Increasingly, systematic review searchers need to limit searches to specific populations. This is a difficult task given the wide array of synonyms, semantic drift over time, and poor population reporting in abstracts. This is especially evident in the terms associated with Black Persons Living in the United States (BPLiUS), i.e., US-born Black Americans and US immigrants of Sub-Saharan African ancestry.

Objective: To increase the efficiency, reliability, and inclusivity of population searches, we will develop a novel method to create and validate search filters to retrieve biomedical and health sciences–related information on BPLiUS persons in Ovid MEDLINE and PubMed.

Methods: We will use a handsearch method to create our gold standard citation set. To gather the most representative citation sample, we will draw from Black-focused journals along with journals in the following categories: race/ethnicity-focused, cultural anthropology/psychosocial, representative high-impact journals, clinical/health systems, and, lastly, regional/state medical journals. To capture the breadth and depth of Black ethnonym terms over time, we will sample citations in the above categories in 20-year increments—i.e., 1955 (Jim Crow era), 1975 (Civil Rights era), 1995 (African American era), 2015 (People of Color era), and 2021 (BIPOC era)—to capture more recent language. We will include explicit population terms (e.g., Blacks), implicit terms (e.g., minorities), and contextual terms occurring with explicit/implicit terms (e.g., immigrant). To identify filter terms, we will text-mine our test set and Black-focused journal citations using the AntConc software; consult the National Library of Medicine’s MeSH database; identify example search filters; review published systematic review search strategies; and analyze potential implicit words/phrases for inclusion. The gold standard set will be divided into test and validation sets. The filters will be internally and externally validated. We will create a highly sensitive (recall)-maximizing filter (prespecified performance minimum ≥ 75%), a precision-maximizing filter, and a best balance of sensitivity and precision filter.

Relevance and importance to patients: This article contributes to methods resulting in more robust evidence production.