MLA 2026 Annual Meeting Research Awards and JMLA Biannual Research Award
Submitted by: Lindsay Blake
Congratulations to the award-winning research papers and posters from MLA ’26 in Milwaukee, WI, and to the Biannual JMLA Research Article Award winner! On a special note, our Biannual JMLA Research Paper Winner is a previous MLA Annual Meeting Research Award Winner!
The MLA Research Caucus is pleased to announce the winners for best research papers and posters presented at the MLA 2026 Milwaukee annual meeting. Thank you to all the judges who volunteered their expertise to help select these deserving awardees both in the pre-judging phase and during the conference. To learn more about the awards and selection process, visit the Research Caucus on MLAnet.org.
Look for the Research Caucus announcement on a webinar with all research winners presenting their projects in the Fall!
JMLA 2026 Biannual Research Paper Award
Sometimes the Apple Does Fall Far from the Tree: A Case Study on Automatic Indexing Precision Errors in PubMed
Paije Wilson, University of Wisconsin-Madison, School of Medicine and Public Health
Objective: This case study identifies the presence and prevalence of precision indexing errors in a subset of automatically indexed MEDLINE records in PubMed (specifically, all MEDLINE records automatically indexed with the MeSH term Malus, the genus name for apple trees). In short, how well does automatic indexing compare [figurative] apples to [literal] apples?
Methods: 1,705 MEDLINE records automatically indexed with the MeSH term Malus underwent title/abstract and full text screening to determine whether they were correctly indexed (i.e., the records were about Malus, meaning they discussed the literal fruit or tree) or incorrectly indexed (i.e., they were not about Malus, meaning they did not discuss the literal fruit or tree). The context and type of indexing error were documented for each erroneously indexed record.
Results: 135 (7.9%) records were incorrectly indexed with the MeSH term Malus. The most common indexing error was due to the word “apple” being used in similes, metaphors, and idioms (80, or 59.2%), with the next most common error being due to “apple” being present in a name or term (50, or 37%). Additional indexing errors were attributed to the use of “apple” in acronyms, and, in one case, a reference to Sir Isaac Newton.
Conclusion: As indicated by this study’s findings, automatic indexing can commit errors when indexing records that have words with non-literal or alternative meanings in their titles or abstracts. Librarians should be mindful of the existence of automatic indexing errors and instruct authors on how best to ameliorate the effects of them within their own manuscripts.
MLA 2026 Annual Meeting Contributed Paper Awards
1st Place – Mind vs. Machine: A Quantitative Analysis of Human vs. AI Critical Appraisal Quality
Authors: Laura Menard, Librarian, Cincinnati Children’s Hospital, Amy Blevins, Molly Montgomery, Heather Abshire, Dan Trujillo, Molly Morckel, Danette Lopp, Karen Vonderhaar
Objectives: We aim to test the hypothesis that human expertise will be superior to artificial intelligence in conducting critical appraisal of standard study designs as assessed by questions 6 and 7 of the validated FRESNO rubric. Our study will identify and report any statistically significant differences between humans and AI in critical appraisal performance.
Methods: Each human expert will conduct a critical appraisal of three pre-selected studies: one RCT, one systematic review, and one diagnosis study. They will then prompt their AI system to perform the same task using a validated prompt. Participants will submit their critical appraisal samples as well as the AI outputs to the PI, who will de-identify each sample. Once all samples are obtained and de-identified, they will be assigned to a separate and independent group of graders, who will be trained and normed in the application of Fresno questions 6 and 7 to grade each sample. Once all data are collected and scored by trained raters, we will analyze the information for group differences. Tentatively, we will use the Friedman Test (non-parametric equivalent of repeated measures ANOVA) followed by a post-hoc Wilcoxon signed-rank test with Bonferroni correction to analyze the data.
Results: A preliminary analysis used a Welch t-test to determine whether there were differences between AI and human graders in overall scores according to the FRESNO rubric. AI-graded samples had a higher overall score (M = 31.4, SD = 7.1) than human scores (M = 21.3, SD = 11.2), a statistically significant difference of M = 10.1, 95% CI [5.4 to 14.7], t(54.068) = 4.363, p <.001.
Conclusions: As AI summaries and clinical decision-making tools become more prevalent, understanding the quality and rigor of the AI output compared to human expertise will become increasingly important. Our validated prompt and prompt architecture will be useful to clinical librarians, health sciences professionals, and vendors as we aim to make informed decisions regarding responsible use of AI in research and patient care.
2nd Place – From Brie-f Comments to Lasting Revisions: Librarian Peer Review in Action
Authors: Melissa L. Rethlefsen, Health Sciences Library and Informatics Center, University of New Mexico, Tara J. Brigham, Carrie Price, John M. Reynolds
Objectives: Three biomedical journal submission systems supported a randomized controlled trial (RCT) in which librarians and information specialists (LIS) were invited to be peer reviewers for systematic reviews (SRs). Though this trial found that LIS peer reviewers did not lessen risk of bias or improve reporting quality of SR searches, a follow-up study found that the LIS peer reviewers involved perceived that they had impacted editorial decision-making and authors’ revisions. To better quantify the impact of LIS peer reviewers’ contributions, we sought to assess differences between LIS and non-LIS peer reviewer comments and to analyze their comments’ comparative impact on authors’ revisions.
Methods: We identified BMJ, BMJ Open, and BMJ Medicine manuscripts from the RCT that had at least one peer review report submitted by an LIS and was ultimately published. For each manuscript, we downloaded the original submitted manuscript; the first revision manuscript, including authors’ responses to peer reviewers; and the first round of peer reviewer comments. All three journals use open peer review and publish all manuscript versions publicly. We first analysed all peer reviewers’ comments using content analysis. We split comments between methodological and non-methodological comments. For methodological comments, we categorized each comment by topic and whether a revision was suggested. For comments where a revision was suggested, we analysed authors’ responses to reviewers and the first revision manuscript for changes. We consulted the original submission to validate changes where needed. The primary outcome is the number of changes made in authors’ revised manuscripts in response to methodological comments made by LIS compared to non-LIS peer reviewers. Secondary outcomes include the quantification of the scope and content of LIS peer reviewer feedback compared to other peer reviewers’ reports and differences between the types of content authors revised versus did not revise.
Results: Of the 68 manuscripts initially identified for potential analysis, only 57 were ultimately available for analysis, as eleven had been rejected in subsequent rounds of review and were therefore not publicly available. 194 total first-round peer review reports were available for these 57 manuscripts, ranging from 2 to 6 reviewer reports per manuscript. 59 (30%) of the review reports were by LIS; for all but two manuscripts where there were two LIS peer reviewers, there was one LIS peer reviewer per manuscript. Preliminary results show that approximately one-third (1115/3819) of the excerpts coded were considered “Constructive Feedback.” Peer reviewers commented on all sections of manuscripts, though LIS focused more on methodological components, including the search, than non-LIS peer reviewers. Authors changed their manuscript in response to 52% (n=196) of LIS versus 58% (n=162) non-LIS peer review comments.
Conclusions: Preliminary results highlight the significant, unique contributions of LIS peer reviewers to the peer review process. LIS peer reviewers are more likely to contribute to requests for revisions on methodological topics, especially search-related topics, than non-LIS. LIS and non-LIS peer reviewers impacted authors’ manuscript changes similarly. LIS peer reviewers should be invited to peer review more evidence syntheses to guide editorial decision-making.
3rd Place – A Statewide Survey to Determine the Potential Impact of Federal Funding Changes
Author: Deirdre Caparoso, Health Sciences Library and Informatics Center, University of New Mexico
Objectives: To determine the potential and realized impact of federal funding changes across the library landscape of a single state via the administration of state-wide surveys.
Methods: There have recently been multiple attempted changes to federal library funding sources in the United States. In particular, rural and tribal libraries have been heavily impacted by changes to Institute of Museum and Library Services funding. In late Spring 2025, an academic health sciences librarian developed a survey to administer to all members of a state’s library association. This survey, reviewed by an institutional review board, utilized 15 multiple-choice and open-ended questions to collect quantitative and qualitative data on the potential impact of reduced federal funding on a wide variety of libraries and archives. Questions were designed to determine library type, the percentage of libraries utilizing federal resources, the types of funds utilized, how funds were utilized, and potential short- and long-term impacts of funding changes, including possible changes to library association participation. Due to embedded survey logic, the number of questions provided to each participant varied. The survey, created and administered via Qualtrics and sent to the library association’s membership (250 members in April 2025) by email, was open between April 21 and May 9, 2025. A follow-up survey, based on and administered like the previous year’s survey but designed to measure the realized impact of federal funding changes since the previous year, was sent to the library association’s membership (268 members in April 2026) and open between April 20 and May 8, 2026.
Results: The initial 2025 survey received 140 usable responses and the 2026 survey received 53 usable responses. Analysis of the 2025 survey showed that most libraries and archives in the state reported receiving federal funds in some way and expected to experience short- and/or long-term impacts to programs, services, and library association participation, including possibly canceling projects, scaling back operations, ceasing functions or closing facilities, eliminating or reducing internet access, and losing professional development opportunities. Analysis of the second survey showed that since the previous survey, about half of libraries and archives reported receiving federal funds in some ways, the average amount of funds received was less than previously reported, federal funds supported less library programs and services, and a third reported experiencing no short- or long-term impacts. Of note; in 2025 less than 5% of institutions considered staff cuts to replace or supplement lost or reduced funding but in 2026 almost 10% reported actual staff cuts.
Conclusions: Last year, multiple libraries and archives in the state reported that they expected to experience significant impacts to library operations if funding continued to be different from previous years. This year, libraries and archives in the state reported fewer actual impacts to library operations than expected. However, less libraries and archives reported receiving federal funds in some ways and the libraries and archives receiving federal funds in the past year reported receiving significantly less funds overall. It must be noted that survey analysis is limited by the dissimilar participation rates between the two surveys and inherent imprecision of self-reported data.
MLA 2026 Annual Meeting Contributed Poster Awards
1st Place – Beyond the Desk: Using Targeted Email Campaigns to Boost Student Engagement and Success in Clinical Rotations
Authors: Laura Barkema, Harold S. Bright IV, Dot Winslow, A.T. Still University
Objectives: To evaluate the effectiveness of a proactive library intervention designed to maintain Physician Assistant (PA) student engagement with essential clinical resources during off-campus clinical rotations.
Methods: Email templates were created by the liaison librarian from each rotation LibGuide subject page and sent out to a list of physician associate students per rotation that was provided by their program. This occurred each Friday before the start of every rotation. Engagement with the emails and resources was measured through clicks on both emails and links, through LibGuide usage comparison, through ebook usage, and through a 5 and 10-point Likert scale survey distributed to the students post-clinical year.
Results: Data showed that delivering rotation-specific resources directly to students increased library outreach and resource utilization and strengthened the library’s role in supporting student clinical outcomes. Click data showed proactive outreach increased engagement with library communication and resources. The survey results highlighted: emails are an effective reminder of library resources, students refer back to the email multiple times, and the content of the emails is useful. The students also agreed that the rotation emails contributed to the success of their clinical year. LibGuide usage statistics showed an increase of visits to clinical year pages.
Conclusions: Student survey results favored the continuation of library resource emails for clinical rotations. We plan to expand this method to the DO schools rotations and continue with our current PA emails. Email clicks, LibGuide usage, and survey responses all support the success of this project. Overall, this proactive communication strategy provides a scalable option.
2nd Place – Does the Library Offer Research Data Services and Where do I Click to Find out? A Content Analysis of North American Health Sciences Library Websites
Author: Danielle Maurici-Pollock, Health Sciences Library and Informatics Center, University of New Mexico
Objectives: In the wake of increasing mandates for data management and sharing, more health sciences libraries are offering research data services (RDS). This content analysis examines the websites of academic health sciences libraries in the U.S. and Canada to answer the following research questions:
1. What research data services (RDS) do these libraries offer, as evidenced by the information on their websites?
2. Where in the research data lifecycle do these libraries offer support?
3. Where is information about RDS located on libraries’ websites?
Results can inform the provision of library RDS as well as improvements in library website design.
Methods: The study analyzes the public-facing websites of 203 academic health sciences libraries in the United States and Canada, including: a) Association of Academic Health Sciences Libraries (AAHSL) members, b) libraries serving Association of American Medical Colleges (AAMC)-accredited medical schools, and/or c) libraries serving colleges accredited by the American Association of Colleges of Osteopathic Medicine (AACOM).
Each identified site is hand searched for mentions of the phrases “research data,” “data services,” and “data management.” If mentions of RDS are found on the library’s homepage, this is recorded. If RDS are mentioned on a subpage or LibGuide, its name and URL is recorded, as is the shortest path to it from the homepage.
Also recorded is evidence of library RDS beyond the existence of the web page or LibGuide itself, including, but not limited to, assistance with data management planning, locating data sets and data repositories, and data visualization. Using DataONE’s Data Lifecycle model, services are categorized by the lifecycle stage researchers would most likely seek them out. Results will be analyzed to determine trends in RDS offerings, phases in the data lifecycle most supported, average path length from homepage to RDS information, and differences by institution location and type.
Results: Preliminary descriptive results are presented on the poster. While evidence was found that a majority (64.53%) of health sciences libraries with public-facing websites are offering RDS, only 33.59% of these mention RDS on their homepage. Further, where to click to find information about library RDS is not always obvious. The average number of click or hover actions needed to access information about RDS from the library homepage is 2.39 and often the first link in these paths does not contain the word “data.” Health sciences libraries vary in the prominence and location of RDS-related information on their websites and in the names given to pages and guides containing this information. This variance presents challenges, suggesting that even when libraries do offer RDS, finding information about these services may not be easy for researchers, particularly if they are unfamiliar with the library’s website and the types of support available.
Conclusions: While a majority of North American health sciences libraries appear to offer at least some RDS, the existence of these services is not always obvious from the library homepage. Finding information to answer the question “Does the library offer research data services?” is often challenging, and this is likely to be particularly true for researchers who do not have previous knowledge that a library offers RDS or who are unfamiliar with the specific vocabulary a library uses to describe its services.
This study is still in progress. Future phases will further explore the specific types of RDS health sciences libraries are offering, where in the data lifecycle libraries provide support, and what differences exist in RDS provided by institutional location and type. It is expected to provide valuable data on health sciences libraries’ current RDS offerings in the wake of increasing funder, publisher, and institutional mandates for data management and sharing, and to inform the design and location of RDS-related web content.
3rd Place – Brewed Differently: Comparing Subheading use in Embase and MEDLINE
Authors: Danielle Westmark, Cindy Schmidt, University of Nebraska Medical Center
Objectives: Our long experience with MEDLINE searching and attendance at NNLM-sponsored PubMed training sessions taught us to depend on specific patterns of subheading use by MEDLINE’s indexers. For instance, we learned to expect that drug trials would be indexed with a drug heading combined with a therapeutic use subheading. Our library licensed EMBASE in 2012. We soon began to suspect that subheading application was not as predictable in EMBASE as it was in MEDLINE. We wanted the recommendations we provide novice EMBASE searchers to be based on evidence, so we decided to compare use of subheadings in EMBASE and MEDLINE.
Methods: On December 5, 2024, we searched MEDLINE (via EBSCOhost) and EMBASE (via embase.com) for articles with 1) indexed “trial” publication type/heading plus 2) a “chronic-myelogenous leukemia” title phrase or a “chronic myelogenous leukemia” (CML) heading that 3) did not have a “CML” heading combined with a “therapeutic use/therapy”-related subheading and that 4) were within the MEDLINE-subset. No language or publication type filters were applied. Both authors independently reviewed all records retrieved. Articles were included in our analysis if they met the NIH’s definition of a clinical trial, reported tests of a therapeutic intervention aimed at CML treatment, and focused on that intervention and its outcomes. Any disagreements concerning article inclusion were settled by discussion. The authors then looked at both the MEDLINE and EMBASE records (M-record and E-record) for each included article to document any issues in the areas of CML-related heading and subheading use and clinical trial indexing.
Results: 306 records for 271 articles were retrieved. Records corresponding to 90 articles met our study’s inclusion criteria. 7 of the E-records and two of the M-records used an incorrect heading for CML, and one M-record included no disease heading. Of those with a “CML” heading, 65 E-records and 7 M-records failed to combine the “CML” heading with one of the “therapy” subheadings. All of the 22 E-records lacking clinical trial indexing post-dated the introduction of clinical trial indexing in EMBASE (1974). 4 of the 18 M-records lacking clinical trial indexing post-dated the introduction of clinical trial indexing in MEDLINE (1991).
Conclusions: Our study provides evidence that EMBASE searches depending on subheadings will miss more relevant articles than similar searches in MEDLINE. It’s important to consider this inconsistent use of subheadings when developing routine mediated literature searches, creating searches for evidence synthesis projects, or testing or developing search hedges/filters. An approach to EMBASE search training that advises against routine subheading use seems justified.