OPeNet: an AI-based platform implemented to facilitate clinical reasoning by primary care practitioners, as well as the virtuous co-management of chronic patients during and after the COVID-19 pandemic in Italy

To the Editor,

In October 2020, we reported in this journal the description of the Embrace project as an enabler to help General Practitioners (GPs) in their virtuous path of co-management of the chronic patient in the context of the Italian National Healthcare Service (NHS) [1]. From June 2019 to February 2022, the project involved 12,000 GPs, and 21,000 patients were supported through the improvement of their clinical management. Although satisfactory, these results are largely impacted by the evolution of the COVID-19 pandemic, which unfortunately did not stop, but continued its trajectory with the emergence of new variants, such as omicron in late 2021 [2]. The majority of the Italian population, especially the subjects most at risk, have been vaccinated, but at the beginning of the winter of 2021 a significant proportion of them had not yet completed the primary vaccination cycle, or even taken their first dose of COVID-19 vaccine [3]. Therefore, a substantial share of resources was still allocated to the management of the new wave, and once again the management of patients with chronic diseases proved difficult. The digital interaction between GPs and hospital specialists and that between doctors and patients, which had started and spread with the previous waves, was once again promptly activated during the fourth wave. In other words, the COVID-19 pandemic has highlighted the need for a medicine based on proximity (i.e. a closer relationship between the patient and the GP in their common space) and proactivity (i.e. the GP’s self-initiated behavior that aims at facing a clinical issue earlier, or even before it has occurred). This approach finds its maximum expression in the remote management of the chronic patient, so much so that the implementation of new digital systems for the management of chronic patients and healthcare in general had already been clearly identified as priority objectives for the reform of the NHS, as outlined by the Italian Council of Ministers in the National Recovery and Resilience Plan issued in January 2021 [4]. An efficient connection and integration between primary and hospital care is the cornerstone on which the reform is based, and the Embrace project – which was born in the pre-pandemic era with precisely this objective – has also evolved by placing alongside the human element a digital project based on the use of Artificial Intelligence (AI). A collaboration between Novartis Farma, IBM Italia and NetMedica Italia has resulted in “OPeNet”, a multifunctional platform for GPs. The acronym “OPeNet” indeed reflects the conceptual and technical features on which the platform is based: in fact, it’s an OPen tool, because it can integrate and exchange data with the management software already in use by GPs; e-because it exploits an IBM AI tool (Watson) to improve the care path of the chronic patient through information and tools useful to the doctor; and Net because it promotes the network and connection of the main players involved. Watson is an AI tool that mimics the way human beings learn and interact. In particular, it can understand language and data even in unstructured form; it can reason, grasp concepts, formulate hypotheses, deduce and extract ideas; it also never stops learning, because from every piece of data, interaction and result it can develop and refine its skills; finally, it can interact with humans in a natural way. The Watson services used in the OPeNet platform are four: Watson Knowledge Studio, a trainer on the specific knowledge domain related to the diseases in scope with the development of cognitive models that identify entities and relations; Watson Discovery, which allows a classification and understanding of contents within texts and other relevant material; Watson Assistant, able to create a dialogue based on natural language interactions, and therefore to learn from previous conversations; Natural Language Understanding, which allows the analysis of text and unstructured data and the identification of concepts, entities and relations. The platform meets the requirements of the EU General Data Protection Regulation 79/2016 and neither Novartis nor IBM have access directly or indirectly to the clinical records. OPeNet is divided into four use cases (Figure 1):

Figure 1:

OPeNet: the four components of the digital platform.

Use case 1: A virtual assistant based on AI, on which the doctor can rely for up-to-date information on selected chronic diseases. For the deployment of the first version, two chronic diseases with different management characteristics were considered as a model for future integrations: heart failure (HF) and psoriasis (Pso). HF is not a single pathological diagnosis, but a clinical syndrome consisting of cardinal symptoms (e.g. breathlessness, ankle swelling, and fatigue) that may be accompanied by signs (e.g. elevated jugular venous pressure, pulmonary crackles, and peripheral edema). It is due to a structural and/or functional abnormality of the heart resulting in elevated intracardiac pressures and/or inadequate cardiac output at rest and/or during exercise [5]. Currently, the incidence of HF in Europe is about 3/1,000 person-years (all age-groups) or about 5/1,000 person-years in adults [5]; the prevalence of HF among adults appears to be 1–2% [5]. After the initial diagnosis, HF patients are hospitalized once every year on average [5]. Due to population growth, ageing, and the increasing prevalence of comorbidities, the absolute number of hospital admissions for HF is expected to increase considerably in the future, perhaps by as much as 50% over the next 25 years [5]. Hence, an optimal management of outpatients with chronic HF is key to reduce the burden of disease both on patients/caregivers and on the healthcare systems. Pso is a common lifelong disease with a chronic relapsing course, for which most patients require long-term management: therefore, it may have a significant impact on the healthcare system [6]. The management of Pso in daily clinical practice is highly variable, because many psoriasis issues are still being debated and not definitely addressed by evidence-based medicine [6]. Treatment decisions and therapeutic goals are based on the characteristics of the disease (e.g. severity, site, presence of psoriatic arthritis), patient-related features (e.g. age, previous treatment failures) and the characteristics of the treatments (e.g. effectiveness and safety issues) [6]. Therefore, an optimal management of outpatients with Pso requires a close interaction between GPs and dermatologists/rheumatologists. GPs can write any question in their own language – Italian in our case – using their natural parlance in the application module. Watson analyses the question through the analysis models created together with the panel of disease experts and returns an answer consisting of a text identified as the most appropriate on the basis of the question, with the relevant bibliographical references. Using machine learning algorithms, it can improve the quality of the system by grading the answers. Should the virtual assistant not be able to provide a satisfactory answer to the doctor – or should the doctor feel that they wish to explore the topic further by means of an in-person interview in their office, or virtually, via remote connection, through the option of contacting the disease consultant, e.g. the Embrace Advisor – the key element underlying the Embrace project, as outlined in the previous communication, remains valid [1]. To further enhance the value of the disease advisory service, the Embrace project has undergone the UNI EN ISO9001:2015 quality certification.

Use case 2: Thanks to the integration of the AI tool with the GPs’ management software, the medical records of chronic patients suffering from HF or Pso are analyzed. The trained algorithm intercepts variables related to three macro-domains: the diagnosis of the disease, its staging or severity, and the current treatment. For the training in HF the classification elements present in the ESC2021 guidelines [5] were also taken into account, such as the ejection fraction value, the presence or absence of specific symptoms and signs, the performance and result of diagnostic tests (e.g. echocardiography) or laboratory tests (e.g. atrial natriuretic peptides). Similarly, for Pso training, elements from the Italian guidelines [6] were also taken into account, such as the Psoriasis Area Severity Index (PASI) score, the body surface area (BSA) value, the Dermatology Life Quality Index (DLQI) score, the presence or absence of specific symptoms and signs, the location of the lesions, and the presence or not of comorbidities. The AI enables the interpretation, understanding and extraction of meaning and value from the structured and unstructured content available within medical records. When the combination of the information from the structured data and the insights gained from processing and understanding the unstructured data satisfies pre-defined trigger criteria, alerts are generated allowing the GPs to take a proactive role and decide whether to refer their patients for re-evaluation to a hospital center competent for the disease and located close to the patient’s place of residence. This allows the GPs to send the right patient at the right time to the right clinic to perform the necessary investigations, and possibly to get access to innovative treatments.

Use case 3: Through a telemedicine component, the GPs have the possibility to monitor their patients and their main clinical parameters, detected through a mobile application and a set of devices/sensors connected to it.

Use case 4: Through a protected communication channel, the GPs have the opportunity to interact directly with the specialist, to perform a teleconsultation and/or to share the clinical picture of their patient and therapy, using the tele/video consultation tool.

OPeNet is therefore a modular platform that can be integrated with many existing management software solutions. GPs don’t need to change their own software, but only to join the NetMedica cloud. Based on the training received, a specific agent installed on the user’s laptop periodically checks whether there are any reports related to the disease in scope, then notifying the user (Figure 2). OPeNet is simply added to the existing GPs’ tools intended for the care and assistance of their patients, and they have already given their consent for these scopes. No additional specific informed consent is required for Use case 2, since the GP is considered the "Data Controller", whilst for activating the Use case 3 and Use case 4 patients are required to sign specific informed consents.

Figure 2:

Digital architecture of the OPeNet platform.

This makes OPeNet an extremely versatile tool, since it is open both to new diseases and to the use of for research purposes – e.g. real world evidence – with the possibility of carrying out in future studies both at national and regional level. In order for the value of OPeNet to be understood by all the stakeholders involved in the management of chronic patients, a change management process was implemented, capable of providing GPs with the acquisition of the digital knowledge needed to complete their professional profile and with the awareness that it is possible to establish ongoing relations with individual patients without necessarily having to physically meet them, continuing to develop the empathic component necessary to extend the relationship of trust and, through an initiative medicine, to intervene more incisively on the patient’s health. As of February 2022, the OPeNet platform was available to a sample of 1,000 GPs. A mixed technical-scientific commission of experts in IT, project management and disease – including GPs – was formed, in order to assess the impact of the platform on the patient care path. KPIs related to the platform use and the impact on patient management were identified. We will share the results with the scientific community in a future communication.