The Safety Net of the Future: How Learning Systems Detect Risks Before People Are Harmed
September 7, 2026
In the emergency department, a missing piece of information can become a risk within minutes. A test result reaches the team too late, a handoff remains incomplete, or a change in a patient’s condition is overlooked. Patient safety is therefore a fundamental aspect of the entire healthcare system. Sensors, connected data, and artificial intelligence can help identify critical developments earlier. The key is whether this leads to a better decision at the right moment.
Safety does not begin with an error, but with the system
Modern safety research focuses not on individual errors, but on conditions that either contribute to or mitigate them. Medical expertise, staff, communication, technology, and processes must all work together under time pressure. Safety is achieved through robust structures that take human limitations into account. Max Pongratz, team leader of the Central Emergency Department at Nuremberg Hospital, experiences this reality every day. “The key is whether a system is designed to detect and catch errors as early as possible.” Technology thus becomes an integral part of a safety net that identifies risks, organizes information, and supports people at the critical moment.
This safety net must be established right from the development phase, emphasizes Herna Munoz-Galeano, Managing Director of PGXperts: “Safety must be part of the development process from the very beginning: the scientific basis, clinical requirements, software development, and the systematic assessment of potential risks must all go hand in hand.”
From Reaction to Prevention
Digital health approaches shift the focus from reaction to prevention. While medical decisions are often based on one-time examinations, sensors can continuously monitor vital signs, movements, or behavioral patterns. Algorithms identify relevant changes in this data that might otherwise go unnoticed in everyday care. The added value lies not in the continuous data collection itself, but in reliable warning signals that enable timely action.
For Dr. Christian Münzenmayer of Fraunhofer IIS, this is a central component of future patient safety. In the early detection of fall risks, even small changes in gait or stability can indicate a developing health issue before an incident occurs. This creates a window of opportunity for healthcare providers during which preventive measures can still be effective. The real progress, therefore, lies not solely in a new sensor or a more powerful algorithm, but in the time gained between the first warning sign and potential harm.
Why Data Alone Is Not Enough
However, health data alone does not ensure safety. “More data does not automatically mean more safety,” Münzenmayer emphasizes. Its value stems from quality, clinical interpretability, and processing that provides guidance rather than adding to the burden. Systems must distinguish relevant information from irrelevant information, make uncertainties transparent, and present results in a way that fits into existing decision-making processes. Otherwise, the hoped-for overview becomes yet another source of cognitive overload.
Drug therapy illustrates this need for translation particularly clearly. In addition to the disease and concomitant medications, drug interactions, organ function, demographic characteristics, and genetic differences can be relevant to the selection and dosing of an active ingredient. PGXperts therefore brings together pharmacogenetic, pharmacological, clinical, and demographic information. “The crucial question today is how to turn this information into a reliable, patient-specific decision-making tool,” says Munoz-Galeano. Digital clinical decision support systems can make this knowledge available at the point of care without replacing medical judgment.
That is why interoperability is more than just a technical interface. Information must not only be transferable but must retain the same meaning across different institutions, professions, and systems. Semantic clarity, structured data, and reliable integration into clinical workflows determine whether connectivity actually leads to safety. Technical functionality alone is not enough either. “Software can function technically and still be unsuitable for a medical purpose,” warns Munoz-Galeano. Regulation establishes the framework within which safety and performance for the intended purpose are systematically evaluated and demonstrated.
The biggest challenge lies between the systems
Risks arise particularly frequently at transition points. When a patient is transferred from emergency medical services to the hospital, from the emergency room to the ward, or between different healthcare professionals, information and responsibilities are transferred simultaneously. Under time pressure, incomplete findings, differing documentation practices, or unclear responsibilities can become safety concerns. “A good handoff is therefore not merely a transfer of information, but a safety-critical process,” says Pongratz. Digital systems can structure handoffs and reduce information loss. The key factor remains whether the relevant information is received in full, is understandable, and arrives in a timely manner.
The practical benefit of artificial intelligence lies not in autonomous decision-making, but in targeted relief of the workload. It can search through large amounts of data, highlight patterns, and identify priorities. Especially in highly dynamic care situations, it can direct attention to areas where a rapid professional assessment is required. “AI should help us identify relevant information more quickly and reduce cognitive load. The decision must remain with humans,” says Pongratz. The quality of an AI system is thus measured not only by its analytical performance but also by how transparent and actionable its recommendations are.
People Remain the Most Important Technology
Cooperation remains the decisive safety factor. Teams must reliably share information, address uncertainties, and learn from critical situations. Human factors and safety culture are therefore prerequisites for technological innovation. A technically sound solution loses its effectiveness if it complicates processes, conveys warnings in an incomprehensible manner, or leads to alarm fatigue. Equally important are clear responsibilities and the ability to critically evaluate technical recommendations. Learning organizations therefore treat errors and near misses as indicators of conditions that can be improved.
For manufacturers, this means that quality and risk management should not be treated as a mere after-the-fact documentation requirement. Potential hazards must be identified early, assessed, and mitigated through both technical and organizational measures. In the case of software, this task does not end with the market launch, as changes, new data, and experience gained from use must be continuously taken into account. “Quality and safety are not retroactively built into a finished product. They must be integrated throughout the development process and continued even after the product is launched,” explains Munoz-Galeano.
The future is not a single product
Patient safety cannot be reduced to a single product. Nor are regulation and innovation opposites. High standards for evidence, technical documentation, and clinical evaluation are necessary, but they tie up considerable personnel, time, and financial resources—especially in smaller medtech companies. Munoz-Galeano therefore does not advocate for lower standards, but rather for a regulatory framework that better supports high-quality clinical trials and regulatory processes. “Good regulation can help ensure that robust innovations safely make their way into clinical practice.”
The patient safety of the future emerges from the interplay of robust processes, appropriate technology, and human responsibility. If this is achieved, data becomes guidance, warning signs become prevention, and experiences become opportunities for shared learning. This is where Bayern Innovativ Gesundheit comes in. We connect companies, research institutions, healthcare providers, and public partners; we share knowledge and support collaborations that bridge technological possibilities with concrete needs and bring safe healthcare innovations into practice.
