Rare diseases and digital technologies

The 15th Rare Disease Day was celebrated on 28 February 2022. There are about 300 million rare patients in the world, about 2 of them are in Italy. There are at least 7,000 rare diseases – of which just over 300 have a curative therapy – that individually affect less than 5 people per 10,000 (at the European level), who overall could make up the fourth country in the world for population after China, India and the United States.

Rare diseases are mainly genetic (72%) and generally arise in pediatric age (70% of cases, but they are increasingly showing in adulthood too). They all share a significant diagnostic delay, since 40% of people with rare diseases receive a misdiagnosis, and 25% take 5 to 30 years before receiving a correct diagnosis. This has serious consequences on the course of the pathology and the effectiveness of treatments. With regards to rare disease diagnosis, a key contribution is brought by digital technologies and, in particular, Artificial Intelligence. Such systems allow, to quickly collect, process and "store" a large amount of data (biomedical data and all those derived from connected devices, but also clinical and administrative data of the electronic medical record) speeding up the diagnostic process.

An analysis of the literature on the existing systems to support the diagnosis of rare diseases was carried out by Meridiano Sanità, the Think Tank on healthcare of The European House – Ambrosetti. It shows that the most used are the algorithms of machine learning (used in 2 cases out of 3), a kind of artificial intelligence capable of learning information directly from the data in its possession, without human aid. In the field of machine learning, and more precisely deep learning, the most used models of machine learning for rare diseases are the Support Vector Machines (SVM) and the Artificial Neural Network (ANN), i.e. mathematical/computer calculation models based on the functioning of human biological neural networks that reproduce the typical reasoning of human beings in different situations, which improves its behavioural abilities.

Within the field of machine learning, great potentials are also found in transfer learning. This emerging technique –which consists in adapting and applying the knowledge acquired in a given pathology in another pathology with similar characteristics – represents in fact a partial solution to an existing limit to the diagnosis and treatment of rare patients: the scarcity of available data and the low level of interoperability, connectivity and security of current information systems.

The increasing use of electronic medical records can support the diagnostic and therapeutic decisions of the interconnected devices, such as wearable devices, thanks to the algorithms of text mining and machine learning. It also enables to track and evaluate in real time the Patient Reported Outcomes, augmented reality and virtual reality simulators that play a leading role in the training of rare disease specialists, chatbots and voice assistants. Through Natural Language Processing (NLP), they can understand rare patients and respond to their needs, and are significantly changing this field.

Machine learning is also adopted in the prognostic phase through computerized diagnostic systems (Computed Assisted Diagnosis - CAD) that are able to estimate the probability of having a specific rare genetic disease on the basis of the symptoms of the patients, their facial characteristics or DNA sequencing. In recent years, thanks to second generation sequencing techniques, the costs and timing of genomic analysis have been minimized, while processing has increased exponentially.

Although machine learning seems to be adopted mainly by rare diseases specialists in the phases of diagnostic (40.8%) and prognostic (38.4%), its use in basic research (16.1%) and in treatment (4.7%) should not be underestimated. In the latter case, these systems offer great opportunities for improving the quality of life of the patient, for example by allowing the control of therapeutic adhesion, and the remote measurement of effort or psychological support. Also, telemedicine tools have a relevant role in the treatment and follow-up phase. In addition to being an important data collector, they represent one of the most known technologies and are used in the management of patients suffering from rare diseases.

It therefore is evidently urgent to invest not only on these systems but also on the competences of the operators. The economic resources, thanks to the funds provided by the National Recovery and Resilience Plan (NRRP) and Horizon 2021-2027, do not represent (anymore) an obstacle. An important first step could be dedicating a section of the next National Rare Disease Plan – expected in early 2022, after last year’s adoption of the Rare Disease Single Text – to digital technologies, not only to achieve an improved and more timely diagnosis of diseases but also to ensure a more efficient and "flexible" management, that easily adapts to the changes in the specific needs of patients. 

To learn more about other major healthcare topics, visit the pages of the Meridiano Sanità Think Tank