Meta-Analysis Studies on Artificial Intelligence in Healthcare: A Bibliometric Analysis (2010-2025)

Artificial intelligence has moved from the realm of science fiction into the daily machinery of modern medicine. It is no longer just a theoretical concept but a set of tools that help doctors recognize patterns in X-rays, predict which patients might get sick, and manage hospital schedules. As these tools have become more common, researchers have begun to ask a critical question: do they actually work? To answer this, scientists have been writing "meta-analyses," which are special types of studies that gather and combine the results of many smaller studies to find a clearer picture. Think of a meta-analysis as a way to look at a forest instead of just individual trees, allowing researchers to see the overall health of the evidence rather than getting lost in the details of a single experiment. However, while the field of artificial intelligence in healthcare is exploding with new discoveries, no one had ever stepped back to look at the collection of these summary studies themselves to see how the field is growing, who is doing the work, and where the focus lies.

A team of researchers from Istanbul Medeniyet University decided to fill this gap by conducting a massive review of the literature on artificial intelligence meta-analyses published between 2010 and 2025. They did not test new medical devices or run new clinical trials; instead, they acted as cartographers, mapping the existing landscape of research to understand its structure. Using powerful computer software, they gathered 646 research articles from two major global databases, filtering out anything that was not a formal summary of other studies. Their goal was to see the big picture: how fast the field is expanding, which countries and universities are leading the way, and what specific medical topics are capturing the most attention.

The results reveal a field that has been quietly building for over a decade before suddenly taking off. For the first nine years of their study, from 2010 to 2018, very few such summary papers were published, with only one or two appearing in a single year. But starting in 2019, the number of publications began to climb, and the growth has been explosive in recent years. In 2024, researchers published 153 of these studies, and in 2025, that number jumped to 280. These last two years alone account for nearly two-thirds of all the research in the entire dataset. This surge mirrors the broader adoption of artificial intelligence in medicine, which only became possible once hospitals had enough digital data to train the computer systems. However, the researchers note that while the volume of papers is growing rapidly, the time it takes for these new studies to be read and cited by others means their full impact is still being measured.

When the team looked at who is writing these papers, a clear pattern of concentration emerged. While the research is global, the most productive authors are not spread evenly across the world. A small group of researchers, particularly a team based in Iran, has produced a significant portion of the most cited work. The University of Tehran in Iran stands out as the single most productive institution, followed by other medical universities in the same country. China leads the world in the sheer number of publications, producing more than a third of all the studies, followed closely by the United States. Yet, when it comes to influence, measured by how often these papers are cited by other scientists, the United Kingdom takes the top spot, even though it publishes fewer papers than China or the US. This suggests that a smaller number of high-quality studies from the UK are having a disproportionately large impact on the field.

The topics covered in these studies have also shifted over time. In the early years, the research focused on broad applications and specific diagnostic procedures like endoscopy. As the years passed, the focus narrowed and became more technical. By 2024 and 2025, the most common themes were specific to diagnostic accuracy, predicting patient outcomes like mortality, and using advanced imaging techniques. The researchers found that the most popular journals for publishing this work are those dedicated to digital health and medical informatics, rather than journals focused on a single disease like cancer or heart disease. This indicates that the evidence is being gathered across many different areas of medicine, from mental health to surgery, rather than being confined to one specialty.

Despite the rapid growth and the international nature of the collaboration, the researchers found a surprising lack of deep connection between the major institutions. While countries are working together across borders, the universities and hospitals themselves are often working in isolation. The network of collaboration between institutions is sparse, with most papers coming from a few tightly knit teams rather than a broad web of global partners. Furthermore, the study highlights that while the technical performance of these tools is being studied intensely, other important areas are being overlooked. There is very little research in these summary papers about how artificial intelligence affects the mental well-being of healthcare workers or how it changes the management of hospitals. The focus remains heavily on whether the machines can diagnose diseases accurately, leaving questions about the human side of the technology less explored.

The authors conclude that while the field is undeniably young and expanding at a breathtaking pace, the sheer number of studies does not automatically mean the evidence is settled. The rapid increase in publications suggests that researchers are eager to summarize what they know, but many of these summaries still point to gaps in the data, such as small sample sizes or a lack of long-term testing. The growth in research activity is real, but it has not yet caught up with the need for methodological maturity. The study serves as a reminder that while we are building a vast library of knowledge about artificial intelligence in healthcare, we must be careful not to confuse the speed of publication with the certainty of the results. The map is being drawn, but the territory is still being explored.