Ongoing digitalization offers significant potential for healthcare. From electronic patient records (ePA) to telemedicine solutions, digital technologies are changing how healthcare services are organized and delivered. An important part of this transformation is Natural Language Processing (NLP) in combination with Large Language Models (LLMs). Below, we outline how these technologies contribute to healthcare digitalization and which opportunities—and challenges—are associated with their use.

1. Automated documentation

One of the most relevant application areas for NLP and LLMs in healthcare is automated documentation. Medical professionals often spend substantial time creating and maintaining patient records and reports. NLP-based approaches can support these tasks by converting spoken information into text and identifying relevant medical terms and codes. This can help streamline documentation processes and support consistency in recorded information.
 

2. Diagnostic support

LLMs such as GPT-4 demonstrate advanced capabilities in text processing and language understanding. In the future, these capabilities may be used to support physicians in diagnostic workflows and the preparation of treatment plans. With access to large volumes of medical literature and clinical documentation, LLMs can surface relevant information and highlight alternative considerations as part of structured decision-support processes. This can help improve the accuracy of diagnoses and optimize patient care.
 

3. Personalized medicine

Another area where NLP and LLMs may play an important role is personalized medicine. By analyzing patient-related data, including medical history and data from wearable devices, these technologies can support the aggregation and structuring of individualized information. This can help medical professionals tailor care pathways and medication strategies to individual patient contexts within established clinical frameworks.

Despite the opportunities offered by NLP and LLMs in healthcare, their use is associated with important challenges and considerations, including the following:
 

1. Data protection and security

The processing of health data is highly sensitive and subject to strict data protection requirements, such as the General Data Protection Regulation (GDPR). Unauthorized access to or misuse of patient data can have serious consequences. Robust technical and organizational measures are therefore essential to ensure data security and integrity.
 

2. Ethics and liability

The use of AI technologies in healthcare raises ethical questions, particularly with regard to responsibility and liability in decision-support scenarios. Clear governance frameworks and regulatory guidance are required to ensure that NLP and LLM applications are used in line with ethical standards and legal requirements.
 

3. Organizational culture

The use of artificial intelligence is already the subject of public debate. This extends beyond concerns about job displacement. Texts generated by Large Language Models can be difficult to distinguish from human-authored content. This requires trust in AI-supported outputs as well as acceptance that interactions between medical professionals and patients may be partially supported by technology. To build trust, it must be transparent why and how AI is used for decision-support purposes. This remains a particular challenge for LLMs, as their complexity can make it difficult—even for developers—to fully explain how specific outputs are generated.
 

4. Human interaction

Digitalization offers many advantages, including more efficient processes and support for addressing workforce shortages. At the same time, human interaction remains essential. Personal care, empathy, and professional judgment are central to healthcare delivery and will continue to play a decisive role alongside technological support.

Averbis sees significant potential for Large Language Models in healthcare, particularly in supporting the reduction of administrative processes and improving interoperability of health data. By structuring and making data more accessible, medical professionals can spend more time on direct patient interaction within their clinical workflows.

Averbis is currently evaluating solution concepts for the data protection–compliant training of LLMs using anonymized and synthetic health data, as well as for operating LLMs in on-premise environments in accordance with regulatory requirements. NLP and LLMs will continue to shape the digital transformation of healthcare, and it is essential that these technologies are applied responsibly and transparently.
 

With the AI platform Health Discovery, NLP can be integrated seamlessly into existing systems and applications. The platform structures health data in real time and processes a wide range of unstructured data formats, including PDFs, Word documents, free-text database entries, and speech-based input. The software extracts more than 50 medical entities—such as diagnoses, medications, laboratory results, and vital parameters—and translates them into established standards including FHIR, SNOMED CT, ICD-10, and LOINC. The platform provides 150 pre-trained AI models and supports more than 15 terminologies.