AI may assist: Diagnosis, therapy & prescription remain physician tasks

Artificial intelligence (AI) can support doctors in outpatient care but may neither replace medical decisions nor act independently. This was pointed out by representatives of the National Association of Statutory Health Insurance Physicians (KBV) at an event organized by the Association of Statutory Health Insurance Physicians of Lower Saxony (KVN) on the use of AI in practice.

According to the KBV's assessment, AI is particularly suitable for administrative tasks such as documentation, appointment management, and telephone assistance. However, a prerequisite for actual benefit is deep integration into the practice IT. Dr. Philipp Stachwitz, head of the digitalization department at the KBV, emphasized: “There's no point in having an AI that functions like a better answering machine.” Such a system must be actively integrated into practice workflows. “The system must also have the competence and ultimately the rights to, for example, postpone an appointment,” said Stachwitz. “But for that, the system really needs to have deep access.”

AI pilot projects for issuing repeat prescriptions

The KBV also sees potential for AI in repeat prescriptions – but exclusively for preparatory purposes. “Such a system is quite capable of recognizing whether a patient can reasonably receive a repeat prescription,” for example, based on dosage and the time of the last prescription. However, the decision must always be made by a doctor: “But of course, the machine must not make the final decision alone.”

Stachwitz illustrated practical limitations in the online information event using his experiences from a Medical Care Center (MVZ), a physician-led facility where several doctors work together. “I myself still work one day a week in a large practice at an MVZ where such a machine was once in use,” he reported. However, the system was subsequently abolished “because ultimately the integration of the AI system into our overall systems was not yet good enough.”

Internationally, there are already more advanced approaches. In the US state of Utah, according to media reports, a pilot project has been running since the end of 2025 in which AI systems fully autonomously carry out certain repeat prescription renewals for chronically ill patients.

No ban on AI

Jürgen Schröder, head of the contract physicians' law department at the KBV, explained the legal framework. In principle, the use of AI in doctors' practices is permissible. “There is no legal norm that prohibits the use of AI in doctors' practices,” Schröder clarified. However, the crucial factor is the clear separation between administrative applications and medical AI, for which stricter requirements apply. For medical applications, physician supervision must be guaranteed at all times. “You as a doctor must bear the ultimate responsibility,” said Schröder. Medical decisions may not be transferred to AI; results must be checked and accounted for.

Regarding data protection, Schröder pointed out that for the use of AI in diagnostics and treatment, separate consent from patients is generally not required, as data processing is carried out by medical professionals and is necessary for treatment. This is different for the use of patient data for training AI systems, where particular caution is advised. Furthermore, the location of data processing plays a central role: if it takes place within the EU, the protection of the General Data Protection Regulation applies; for third countries, additional checks are necessary.

Finally, Schröder warned against the use of freely available AI systems in the treatment context. Systems like ChatGPT are not approved as medical devices and therefore may not be used to support treatment. Their use for general research or information purposes without personal reference is to be distinguished from this. In the long term, the KBV sees AI as an important tool for relieving the burden on practices. At the same time, the principle remains: AI is an assistance system; diagnosis, therapy, and prescription are and remain physician tasks.

(mack)