ReportAId, a start-up specialising in the automated processing of clinical documents which raised €2.2 million in May 2025, conducted a survey of a sample of approximately 1.6 million medical reports produced by private and accredited healthcare facilities across the country, revealing a picture that challenges the sector’s perceptions: 70% of patients do not return to the same facility for the follow-ups prescribed by their doctor when they do not receive active post-report support. At the same time, for every 100 reports analysed, an average of 75 prescriptions for further tests or appointments emerge, which, in almost all cases, go unheeded.
80% of the clinical data produced by healthcare facilities is in unstructured format: free-text content found in medical reports, discharge letters and clinical records. Highly valuable information, diagnoses, prescriptions and follow-up schedules – which traditional management systems are unable to read, interpret or translate into action.
The result is what ReportAId refers to as an ‘operational void’: a systematic gap between the generation of clinical data and its ability to ensure continuity of care. The doctor’s referral does exist; it is written in the report, but it is not communicated to the patient and never results in an appointment.
There are three main reasons for this: logistical barriers, such as appointment slots not being available at the time of the request; information barriers, where the patient is unaware that the service is available at the same facility; and communication barriers, where the recommendation for a follow-up appointment remains buried in the report, without ever being communicated directly to the patient. None of these factors is within the facility’s control; they all stem from a system that was not designed to translate clinical information into action.
The technological solution to this problem does not lie in generative AI, but in a lesser-known and more clinically mature form of it: extractive AI. These are systems based on large language models (LLMs) trained to read and understand clinical documents with a depth comparable to that of an expert doctor, automatically extracting diagnoses, prescriptions, diagnostic tests and follow-up schedules.
The same test may appear in a report under different names – such as cardiac ECG, 24-hour ECG or Holter ECG – depending on the doctor, the medical specialism or the practice of the healthcare facility. An extractive AI system is capable of mapping these variations to a unique code and translating them into a single action for the patient.
Recent scientific research supports this view: a study published in *Scientific Reports* (Nature, 2025) has shown that language models trained on clinical data achieve levels of accuracy comparable to those of human operators when extracting information from unstructured text.
Facilities that have adopted the ReportAId system – according to the findings of the company’s own survey – a hybrid edge-cloud platform that integrates with existing IT systems and operates in real time, are achieving measurable results: an average 21% increase in patients returning for prescribed follow-ups in the months following implementation, and a 70% open rate for automated clinical communications sent to patients. The engagement figure is particularly significant when compared with healthcare sector averages, where open rates for institutional communications typically range between 20% and 30%. According to the company, the difference lies in clinical personalisation: patients do not receive a generic reminder, but a specific message linked to their prescription, sent at the right time, in understandable language.
The analysis raises a strategic issue for the entire Italian private healthcare sector. Facilities are investing increasing amounts of resources in acquiring new patients – through marketing, booking portals and partnership agreements – whilst allowing a significant proportion of those already acquired to slip through the net for purely operational reasons. The cost of this attrition is not merely financial: it is also clinical, because a missed follow-up constitutes an interrupted course of treatment. The platform also addresses clinical business intelligence: by aggregating prescriptions extracted from medical reports, it provides healthcare management with a map of future demand based on reliable data rather than historical estimates. An orthopaedic department that systematically prescribes MRI scans of the lumbar spine generates, within its own reports, an early indicator of demand for the imaging service – information that is currently lost and which could guide more informed investment decisions.
With regard to data protection – a particularly sensitive issue in the healthcare sector – the start-up has implemented an architecture in which clinical data is processed locally, within the facility’s infrastructure, and synchronised securely, ensuring that sensitive information does not leave the institution’s premises, in compliance with GDPR regulations. The platform is up and running in plug-and-play mode within a few weeks, without requiring any system migrations. (photo by National Cancer Institute on Unsplash)
ALL RIGHTS RESERVED ©