Insights from ESCMID 2025: trends and future in diagnostic testing

AMR and Antibiotic Use: Access to Testing Is Critical

Antimicrobial resistance remains a central challenge. With an estimated 40–50 billion antibiotic doses taken daily and use projected to rise 50% by 2030, intervention is urgent. The vast majority of prescriptions occur in primary care, often without diagnostic support.

Improved access to diagnostics is key. Many current platforms focus on performance where demand already exists, but the biggest opportunity lies in reaching settings where no testing is currently available. Systems must support appropriate treatment decisions, balancing speed, accuracy, and pathogen identification with usability, affordability, and integration into clinical workflows.

Syndromic Testing: Performance vs Cost

Multiplexed syndromic panels are established in hospitals and acute care, but uptake in primary care remains low. Their value in guiding antibiotic use is clear yet cost and complexity are barriers to broader use.

Developers must reduce system cost and complexity to reach more healthcare environments and fit into reimbursement frameworks. Technical capability alone is no longer enough. Integration, ease of use, and clinical decision support are now central to adoption.

Sequencing-based diagnostics are also beginning to compete with syndromic approaches, raising the bar for accessibility and performance in multiplexed testing.

Sequencing: Moving Toward Clinical Routine

Genetic sequencing is rapidly approaching routine use in clinical diagnostics. With falling costs and expanding platform availability, it holds the potential to reshape infectious disease diagnostics, particularly in syndromic or multiplexed contexts.

At ESCMID, multiple case studies demonstrated the use of same-day metagenomic sequencing for pathogen identification in respiratory and bloodstream infections. Amplicon-based approaches were also discussed, particularly for rapid variant detection and resistance gene profiling. Broader sequencing methods, including microbiome analysis, may also play a role in future clinical applications.

However, sequencing workflows are not yet plug-and-play. Upstream processes, such as sample collection, DNA extraction, host depletion, and library preparation, remain technically demanding. Downstream, interpretation, bioinformatics pipelines, and clinically actionable reporting are major hurdles. While sequencing speed and cost are improving, the challenge now lies in integrating these steps into streamlined, automated, and interpretable systems that deliver value at the point of care.

Developers must ask strategic questions: should initial diagnostic applications focus on relatively simple, targeted sequencing, for example, amplicon sequencing for variant detection, or on broader, hypothesis-free metagenomic approaches? Will sequencing be decentralized or remain concentrated in specialized hubs? How can systems present complex data in ways that support decision-making by non-specialist clinicians? The opportunity is significant, but realising it will require systems that balance performance, usability, and clinical relevance.

ML, AI, and Automation: From Hype to Implementation

Artificial intelligence has already made an impact in diagnostic imaging, including MRI, CT, and ultrasound. At ESCMID, attention turned to the broader use of AI and machine learning in in-vitro diagnostics, with exciting potential across three main areas:

• Image analysis: AI is enabling rapid interpretation of high-resolution optical data in fields like digital pathology, haematology, and spectral imaging. These tools can increase diagnostic accuracy while reducing the burden on human reviewers.
• Multi-marker data interpretation: AI models are increasingly used to integrate complex biomarker datasets, where the combined signal across markers yields diagnostic insight. Some platforms already use deterministic models; the shift toward machine learning promises greater flexibility and performance, particularly as training datasets grow.
• Workflow optimization and automation: AI is being applied to streamline laboratory operations, reducing hands-on time, standardizing results, and minimizing error rates. This is especially valuable in high-throughput or resource-constrained settings.

Several ESCMID sessions showed AI being used to support antimicrobial resistance prediction, improve taxonomic resolution, and aid diagnostic decision-making. Importantly, the field is moving from exploratory development to real-world deployment. The emphasis is now on clinical validation, regulatory clarity, reproducibility, and integration with existing systems.

For developers, success will depend on more than accuracy. Trust, interpretability, and usability are emerging as key differentiators. Tools that embed smoothly into clinical workflows, minimize training requirements, and deliver repeatable, high-confidence outputs will define the next wave of AI in diagnostics.