Skip to main content
Open Access Emergency Medicine : OAEM logoLink to Open Access Emergency Medicine : OAEM
editorial
. 2026 Sep 29;18:657506. doi: 10.2147/OAEM.S657506

Evidence in the Age of Algorithms: Redefining Emergency Medicine Research

Amado Alejandro Baez 1,✉
PMCID: PMC13633729  PMID: 42829634

I have spent more than three decades dedicated to emergency care. During that time, I have seen initiatives arrive with enormous confidence and disappear without leaving a lasting mark. I have also seen quieter, more deliberate efforts fundamentally change how we practice. The difference has rarely been the strength of the idea itself. More often, it has been whether we built the structures needed to sustain it.

Those same three decades have been extraordinary for our specialty. Emergency medicine has gone from a discipline that had to argue for its own legitimacy to one that shapes resuscitation science, trauma systems, prehospital care, disaster response and the design of health systems. In many countries, we have become the place where the health system is seen most honestly, because it is where people arrive when everything else has failed them.

What I have not yet seen — and what I believe this moment demands — is a scientific literature that genuinely represents the world it claims to serve.

Emergency Medicine is entering a new scientific era. Artificial intelligence, precision medicine, genomics and large-scale clinical datasets are changing not simply how we conduct research, but what we may come to define as evidence. Millions of clinical encounters can now be analyzed across health systems. Continuous physiologic information is captured in real time. Genomic, phenotypic, environmental and clinical data are increasingly integrated with outcomes. Algorithms can identify relationships that conventional analytical approaches could never realistically interrogate.

Evidence-based medicine remains a foundation of modern clinical practice. Yet the hierarchy of evidence we inherited was constructed for a scientific environment very different from the one now emerging, and it is beginning to strain.

These capabilities create extraordinary opportunity. They also create a paradox I want to state plainly: the age of more data creates a greater need for rigorous evidence, not a lesser one.

Big data is not the same thing as good evidence. An algorithm that predicts well has not thereby earned clinical authority. Scale cannot compensate for poor methodology, biased data, or inadequate validation. The next generation of artificial intelligence in emergency medicine must itself become evidence-based — externally validated, reproducible, transparent about its data and methods, examined for bias and algorithmic drift, and ultimately judged by whether it improves outcomes that matter to patients and health systems.

The AI black box cannot become a substitute for the scientific method. Instead, the scientific method must make the black box clear.

Transparency does not require reducing every computational process to a simple explanation. It requires rigorous evidence: knowing where the data came from, how a model was developed and validated, in whom it performs well or poorly, and where its limitations lie. Through independent validation, reproducibility, external testing, and continuous evaluation, science can progressively illuminate what initially appears opaque. Trust in clinical artificial intelligence should rest not on our ability to understand every line of an algorithm, but on our ability to objectively test its performance, limitations, safety and clinical value.

The COVID-19 pandemic provided an extraordinary test of how evidence is generated and translated into practice under profound uncertainty. I experienced this both as a practicing physician and as lead presidential public health adviser and Executive Director of the Pandemic Task Force in the Dominican Republic, where consequential decisions often had to be made rapidly and with limited resources.

Early in the pandemic, the challenge was a paucity of reliable evidence. Preliminary observations, small studies, expert opinion, and preprints sometimes shaped practice before adequate validation, allowing plausible hypotheses, pseudoscience, and unsupported therapies to compete with legitimate scientific inquiry. Within months, the problem reversed: an unprecedented volume of publications made it increasingly difficult to distinguish reliable evidence from noise.

Yet the pandemic ultimately demonstrated the power of rigorous science. Better observational studies, randomized trials, genomic surveillance, and international collaboration progressively replaced uncertainty with knowledge. Effective interventions were identified, ineffective ones abandoned, and vaccines developed and evaluated at remarkable speed.

Scientific conclusions changed because the evidence changed. That was not a failure of science, but the scientific method working as intended. Science is neither ideology nor certainty; it is a disciplined, self-correcting process whose authority rests on rigor, transparency, reproducibility, and the willingness to change when better evidence emerges.

That lesson is directly relevant to artificial intelligence. We should neither reject new technologies because we do not yet fully understand them nor embrace them simply because they appear powerful. As during the pandemic, the path from uncertainty to trustworthy practice must run through rigorous evidence.

Our concept of evidence must also extend outward — from evidence-based medicine toward evidence-based policy and evidence-based health systems. Emergency departments sit at the intersection of clinical medicine, public health and society. The volume and diversity of information generated in emergency care can inform not only individual patient care, but emergency medical services, public health surveillance, resource allocation, disaster preparedness, workforce planning and national health policy. Emergency departments are among the most information-rich yet underutilized observation points in health systems. That should not remain true.

Evidence Must Also Be Global

The next scientific transformation cannot occur within geographic or institutional silos.

Algorithms developed within a limited number of countries, populations and health systems cannot be exported to the rest of the world on an assumption of universal validity. Differences in genetics, epidemiology, disease prevalence, resources, clinical practice, language, culture and health system architecture are not noise to be adjusted away. They are the signal. Representation matters. Data sovereignty matters.

This is not merely an ethical concern. It is a scientific one.

Our scientific effort has to represent north and south, rich and poor, and every ethnicity that walks through an emergency department door. A model deployed in populations it has never encountered should not be presumed validated; its performance in those populations remains an untested hypothesis. An algorithm that performs extraordinarily well in one institution and fails across different populations or resource environments is not globally validated science.

International research partnerships must therefore stop reducing institutions and countries to data sources for models built elsewhere. Researchers and communities should participate meaningfully in determining how their data are governed, analyzed, interpreted, and translated into scientific and clinical value. Data diversity becomes a methodological strength, allowing us to establish not merely whether something works, but where, for whom, and under what conditions it works.

This pursuit of global external validity is the single idea that will shape my tenure as Editor-in-Chief.

These changes also require us to reconsider how we organize evidence itself. Randomized clinical trials will remain indispensable for questions of causality and therapeutic efficacy, and systematic reviews will continue to provide critical synthesis. But enormous, diverse and longitudinal datasets can answer questions conventional trials often cannot: rare outcomes, real-world effectiveness, population heterogeneity, surveillance, prediction, treatment-response phenotypes and variation across health systems.

Big data should not automatically sit at the top of the evidence hierarchy. Neither should it remain artificially constrained near the bottom simply because it is observational.

The future of evidence is less about placing one methodology permanently above another and more about identifying the methodology best suited to the question. The traditional pyramid will increasingly need to coexist with a multidimensional framework — one that weighs data quality, methodological rigor, reproducibility, representativeness, external validity, transparency and clinical relevance alongside study design. The question is not which design ranks highest. It is which evidence best answers the question before us.

The Next Chapter for Open Access Emergency Medicine

Our ambition is to make Open Access Emergency Medicine more representative of emergency medicine as it is practiced around the world. The future of our specialty cannot be defined only by evidence generated in highly resourced health systems. Patients, challenges, resources and models of emergency care differ enormously across regions, and our scientific literature should reflect that diversity.

We will expand the journal’s global footprint, strengthen representation across our Editorial Board, and develop dedicated scientific and regional areas of focus. We will also work more closely with international and national professional societies to understand how the journal can better serve their scientific communities, support emerging investigators, and create new opportunities for collaboration and innovation.

Global representation is more than geographic diversity. It strengthens external validity, brings different questions and perspectives into the scientific conversation, and ultimately produces evidence that better reflects the world in which emergency medicine is practiced.

Thirty years in emergency medicine have taught me which ideas endure. The ones that endure belong to more than one country, more than one institution, and more than one generation.

So, if you are doing work of this kind — in a university hospital or a district emergency department, with a national dataset or a single cohort, in any country and in any language — send it to us. We are building the place where that conversation happens, and we would rather build it with you than describe it to you.

Disclosure

The author reports no conflicts of interest in this work.


Articles from Open Access Emergency Medicine : OAEM are provided here courtesy of Dove Press

RESOURCES