CHAT-BOT WITH ARTIFICIAL INTELLIGENCE (AI) AND NATURAL LANGUAGE PROCESSING (NLP) FOR CLINICAL FOLLOW-UP AND ACCOMPANIMENT OF COVID-19 PATIENTS AT HOME OR AFTER HOSPITAL DISCHARGE (COBOT).

What? COBOT is a chat-bot with Artificial Intelligence (AI) and Natural Language Processing (NLP) based on clinical protocols of Spanish Health System and evidence based informative resources..

What for? Designed to help the population to understand  COVID-19, cope with mild or moderate illness at home and enable immediate contact and clinical monitoring by the primary care service.

 Whose? Multidisciplinary and experienced team with clinical Family physicians and computer experts in artificial intelligence.

For whom? For general population.

How?  Increasing their accessibility to evidence-based information and the possibility of home management of less severe cases, facilitating an adequate response from the health system and improving patients quality of life.

When? Now

COBOT is a chat-bot with Artificial Intelligence (AI) and Natural Language Processing (NLP) specifically designed by Family physicians of Madrid Health Service and artificial intelligence experts in response to the challenge posed by two circumstances: 1. The large number of hospital discharges of COVID-19 and less severe patients who require home clinical and social follow-up; And 2. The detection program for asymptomatic infected individuals. This high number of patients will test a service already overloaded and decimated, Primary Health Care.

COBOT will help the population to understand  COVID-19, cope with mild or moderate illness at home and enable immediate contact and clinical monitoring by the primary care service.

COBOT incorporates clinical protocols of Spanish Health System, as well as evidence based materials (videos, infographics, pdf, and links) and gamification components.

This computer softwareis hosted on a protected server, and has a conversation interface that can both answer questions posed by the user in a natural language (that commonly used by people with all their variants) and ask them questions (1).

Although it is not a new technology, the recent technical revolution for the interpretation of natural languages (NLP), together with other advantages (speed of adaptation, scalability, privacy)  and certain downsides of apps, make chat-bots ideal tools for the purposes of this intervention. Since chat-bots “understand” requests expressed with the complexity and variability of human language, they provide a component of technological “humanization” that other interfaces based on menus and chat-buttons lack. This way, they are capable of returning a personalized answer and add a component of user loyalty and usability of the tool (2-3).

Therefore, COBOT will process users’ natural language requests (NLP with AI trained to suit needs), analyze the information being received and propose actions to users according to the defined guidelines by the medical team.

A chat-bot is not a software that needs to be installed in a smartphone or tablet, so it respects the privacy of the patient scrupulously, and its learning curve is very short, as it  uses as interface a messaging application for common use (4). The messaging application would provide support for access by millions of users.

In this way, the patient or their relatives will be able to access COBOT through their mobile phones, where they will find reliable and updated information on crucial aspects of the epidemic (preventive measures, isolation, hygiene, clinical course, treatment, and social resources)with immediate accessibility; Continuous clinical monitoring with symptom record and early detection of exacerbation; And alerts for taking medication, registering symptoms and new information.

When a patient uses COBOT, the primary care professional will be able to be more effective in their work by prioritizing the patients who need assistance the most, facilitating the triage and early detection of severe cases or exacerbations, and enabling the immediate referral of urgent cases to hospital.

Currently, the primary care service is reorganizing and adapting its activity to these needs with poorly equipped, depleted and unmotivated human resources.

Team experience: Our team is a multidisciplinary one made up of clinical Family physicians right now on the front line of the fight against the COVID-19 epidemic and with expertise in communication technologies, and also computer experts in artificial intelligence.

The team has been working together last 5 years, during which we designed, launched, and evaluated through a pragmatic clinical trial the effectiveness of a chat-bot for the adult population to quit smoking compared to usual practice (namely the different interventions by primary care nurses and physicians included in the services portfolio of the Madrid Regional Health Service)  (FIS Project  PI17/01942; ClinicalTrials.gov identifier NCT03445507 ; https://dejalobot.es/wordpress/ ).  We have also received two grants to intensify the investigation from the Fundacion para la Investigación e Innovación Biosanitaria de Atención Primaria de Madrid (FIIBAP)

This project is currently in the data analysis phase with promising results.(self-declared abstinence at one month of 10.8% in the control arm and 24.3% in the intervention arm, with odds ratio 2.64 -95% CI, 1.65-4.22-, p = 0 , 00.) These results are provisional, and have not yet been published.

Scientific feasibility:

The boom in information and communication technologies, like the internet or smartphones, open up new therapeutic perspectives. The smart mobile phone has become the majority and most accessible computer in most countries. In 2018, the percentage of households with mobile telephony reached 98% (5). Mobile telephony penetration increased 3.2 percentage points to 116.1 lines / 100 inhabitants (6). In Spain, the smart mobile phone is the main access device for 46.1% of Internet users (+2.7 points compared to 2018 and +22.8 points compared to the 2014 results) (7).

On the other hand, patients aspire to play a greater role in their health management and increasingly search for more information in the internet. The resulting opportunities create a new framework to empower the patient and improve clinical outcomes and health expenditure (8).

At the present time, with the majority of the population incarcerated in their homes, the use of these technologies takes on a special role, as it respects government distance regulations while allowing access to health services.On the other hand, the overload of health services makes it necessary to develop new solutions that allow for greater efficiency of available resources.

In terms of scientific evidence, we have carried out an ad hoc search for this call and we have not found quality studies of similar interventions.We do have found multiple bots launched in response to the pandemic which are merely informative (but no one with a clinical vocation, such as COBOT), as well as articles on the use of AI (9), geographical tracking and mapping (10) peer to peer contact tracing (11) and telehealth (12 ). The only study we have found is one aimed to evaluate the flu-tracking ability of Flu-Report, a new influenza-tracking mobile phone app, done by comparison with the conventional influenza surveillance information and basic information from an existing large-scale influenza-tracking system.(13)

We know that providing science-based evidence on the effectiveness of clinical interventions via information technologies is essential. In addition to being more efficient, the characteristics of these interventions should improve effectiveness, accessibility, and adherence to treatment. From an ethics perspective, this new type of intervention must be backed by scientific evidence to circumvent pressures from the market or particular interests, improve patient safety, and follow the standards of correct practices for clinical interventions. But at the present time, ethical constraints hinder the design of a clinical trial, so we propose a validation project at the level of efficacy, usability and satisfaction by testing with end users, which also will serve to enter into a process of continuous quality improvement. Clinical records database can be useful for studies on the clinical evolution of mild or moderate cases. As far as we know, this should be the first bot to be studied.

Technical feasibility.

COBOT scripts are ready to be used on Telegram app.

As development possibilities, there is integration with common messaging applications (Messenger, WhatsApp…)  and personal assistants (Alexa, Google Assistant, etc), as well as developing automatic speech recognition (ASR).

The modules: diagnosis and geolocation are also partially designed,and could be added to the project if deemed necessary Synergies with social projects could also be developed.

The main beneficiaries will be general population, as COBOT will increase their accessibility  to evidence-based information, monitor CoVid19 disease in infected people in real time, carry out continuous triage of complicated cases with the possibility of directed referral, and increase adherence to treatments by generating alerts on mobile phones.

All in order to contribute to the home management of less severe cases, freeing up resources in primary health care, facilitating an adequate response from the health system and improving the quality of life of patients.

COBOT is a response from Primary Care to Primary Care and the community.

We believe that COBOT will increase the quality of life of patients and their families, allowing them to access quality information and a clinical monitoring program that eliminates barriers between patients and their primary care professionals.

COBOT will also increase the effectiveness and efficiency of Primary Care health care workers, as in the current phase of the SARS-Cov-2 epidemic, a high number of hospital discharges and the detection of a high number of asymptomatic cases of COVID-19 with the so-called «rapid tests» are anticipated. All these people will need close clinical monitoring and active support

COBOT can serve the number of users required, as it is an easily scalable product depending on the needs and available means.

Finally, it is easily adaptable to other future pandemic crisis.