Crisis of urgencies: chatbots to finish with bobology?



Crisis of urgencies: chatbots to finish with bobology?

Emergencies are on the verge of a nervous breakdown. In fact, they have already been for more than ten years, but the lack of resources, and the large number of patients overwhelm them more and more. As a result, several dozens of caregivers across the country are making their voices heard, since they can not really go on strike. If only they had fewer patients to care for, especially those too many who come to treat small sores …

Could these hypochondriacs or allergies to medical appointments be treated differently? Until now, considering it was almost impossible. But since about 5 years, the AI ​​changes the game and lets start thinking about it. What if the solution was robots? No, no humanoid robots capable of moving (at least for now), but chatbots, these conversational robots, which use algorithms and artificial intelligence techniques to speak with their interlocutor most "humanly" possible.

While emergencies have doubled in 20 years, HealthTech startups and doctors are currently developing chatbots that can give you advice on how to prevent illness or treat small colds (without prescription drugs). On Facebook Messenger, "Smart Alfred", the "docbot" of the health platform Betterise Health Tech, offers the user a daily "check-up" to closely monitor his health – a virtual medical coach, in short, "Trained with great doctors, according to the best scientific studies and able to send you a unique and extremely varied content". He asks a lot of questions (especially about potential symptoms), to give his interlocutor sports tips, diets, etc. But other conversational robots go further, and already venture, thanks to AI, in the field of pre-diagnosis.

IA + telemedicine = medical chatbots

If you do not live in a cave, you probably know that the AI ​​is progressing giants in recent years, and that it is now able to "predict", more or less well, diseases and pathologies – especially in analyzing a huge amount of health data using ever more efficient algorithms. Research to fight cancer or to pre-diagnose diseases such as type 2 diabetes or heart failure through artificial intelligence is growing. Watson, IBM's AI, is already able to detect very rare skin cancer or leukemia by analyzing millions of medical records of patients – with a 90% effectiveness rate.

At the same time, "wearable" devices such as wristbands and smartwatches allow you to analyze a lot of constants, like heart rate for example – but their reliability remains to be demonstrated. This is why more and more health professionals, determined to take matters into their own hands and not to let private platforms like Doctolib or Qare "uberise" them, develop telemedicine systems, allowing a patient, through connected medical devices (for example, an otoscope or a blood pressure monitor) and videoconferencing, to perform a more or less complete medical check-up at a distance. In order to save both practitioner and patient time – at least when it comes to "sores" or mild symptoms.

So why not mix the two, telemedicine and AI, to conduct teleconsultations not with humans, but with "intelligent" chatbots? A solution of "pre-diagnosis" for doctors, which would be even more effective to reassure patients (often imaginary) and free medical time, support those who are currently developing "medical robots". In the specialized magazine What's Up Doc, health professionals who are working on the subject in the Pastel Health startup, said that chatbots will soon "collect as many symptoms, feedbacks, impressions of the patient while waiting for the availability doctor ", but also to" sort "between" serious, not serious, urgent or not urgent "cases, before handing over to the caregivers. For this reason, conversational robots ask questions "repetitive and forbidding beginning of interrogation" to patients.

There is no question, therefore, of replacing doctors, but rather of "augmenting" them, thanks to chatbots acting as assistants in their service, saving them time and efficiency, and sometimes even giving them, thanks to the 'IA, diagnostic assistance and treatment proposals. According to Joel Barthelemy, founder and CEO of GlobalMed, a US telemedicine company, "Current AI-based applications, based on the experience of thousands of similar patients with the same symptoms, may already suggest to physicians the type of drug to prescribe or treatment to try, and tomorrow, with robots, patients will no longer need to see a doctor for primary care. Do you take him for an enlightened prophet? You may be wrong, because the chatbots able to make this prediction are already there.

Questions, interrogations and pre-diagnoses

In the UK, in 2017, the National Health Service teamed up with the startup Babylon Health to create an alternative to 111. This issue, which is the target of much criticism, has been put in place by the British government to relieve the number 999 , dedicated to real emergencies, offering to provide advice to those who need "non-urgent" help. But even the 111 is overloaded, and those who are on the line are not necessarily health professionals – hence a significant risk of errors. The Babylon chatbot does not pretend to make a diagnosis, but submits questions to the user – defined in advance, but refined through the AI. For example, "how long does your pain last? Where is this pain? "Or" are there any related symptoms? ". The purpose of this "intelligent" interrogation: to determine whether or not the problem presented requires a "real" examination, carried out (at a distance, by videoconference) by a human doctor, or if it is likely to disappear over time – as all self-respecting bobo. The robot ultimately provides medical advice, or directs the person to a professional.

In China, tech giant Baidu recently designed Melody, a "medical chatbot" whose mission is to help doctors "better diagnose" their patients – integrated with Baidu Doctor, Chinese equivalent of Doctolib, this nurturing conversational robot AI (using deep learning and natural language processing) raises extensive questions about the person's symptoms (as well as his or her background and risk factors), and then provides all this information for human practitioners.

For its part, the Ada Anglo-German Chatbot, developed in 2017 "by more than 100 doctors and scientists", knows "thousands of symptoms and conditions, from colds to rare diseases" and is powered by a "sophisticated AI engine "That uses a" decision tree with probability data ", which allows him to perform a pre-diagnosis – before, again, to propose or not to the user a teleconsultation with a real doctor. This "health companion", used by more than 1.5 million people, provides the rest of the time prevention counseling, and ultimately allows the patient, TechCrunch notes, to "make sure he only consults a doctor when they need it and, more generally, to be proactive in their health care without increasing the caregivers' workload. " A real remedy against the bobology, so … although no study has yet been done on the impact of Ada on the attendance of emergencies in London or Berlin.

The list of AI-fed medical chatbots is already long, but there is also K-Health, which again allows you to "check your symptoms and talk to a doctor if needed". On the Conversational Robot site, it says, "Dr. Google does not know your health, and doctor visits can be expensive and impractical," while "K uses advanced AI to provide you with smarter health information and convenient access to quality health care in minutes. Concretely, he uses algorithms to brew millions of medical data (anonymized), compiled from doctor's notes, laboratory results, hospitalizations, or drug statistics. The AI ​​then allows him to cross all this information with the answers of the user, to detect similar symptoms and perform a "self-diagnosis" – which can, as with Ada, lead to a teleconsultation with a doctor on optionally.

Limits and ethical framework

These conversational robots, capable of "sorting" between patients at the source and thus to one day relieve emergencies by treating "bobology", are obviously not immune to potential errors – not to mention the risks involved. "hacking" that could allow a malicious person to harm you by "diverting" the responses of your chatbot. Thus, there is no guarantee that Ada will not one day make a mistake, interpreting for example your abdominal pain as simple "cramps" when it is in fact an appendicitis. This happened to me personally with a human doctor, so why would a chatbot be different from the mistake? Oh yes, the strength of AI, of course …

Is not the risk, too, to push us to ever more self-medication? By allowing us to perform a "pre-diagnosis", will chatbots induce us indirectly to pass the "human" stage, intended to confirm or not the results of the robot? Ada, K and the rest should be unable to replace the visit with a doctor – the only one able to palpate a stomach, make sure the patient's skin is not discolored, or detect things around the corner. a very human conversation.

This is probably why (besides the fear about the future of health data that gravitate in the clouds) that many patients do not hide their fears vis-à-vis telemedicine and these medical robots. And it is also for this reason that many scientists claim the establishment of a real ethical framework around this virtual medicine. No, chatbots will not replace doctors because they will never have the health professional's medical experience, or their emotional human intelligence, their empathy, their "intuition" and their feelings – which can detect invisible things for algorithms, as noted by MIT researchers. "The health bots will especially come to streamline the organizational problems present today in the medical world, such as making appointments or sharing information.They can be excellent tools for sorting, allowing the doctor to discharge a part of the bobology, "said Dr. Arthur André, neurosurgeon and co-creator of Citizen Doc, a diagnostic aid application, which uses AI and questionnaires to find," from 'a symptom', a 'diagnostic solution'.

To reduce the risk of errors, chatbots will have to remain simple "assistants", tools in the hands of only doctors, able to propose to these "tracks" of diagnosis, and to "prepare the true consultation" between humans. "If it's done right, a chatbot will suggest three possible diagnostic tracks, doctors prefer diagnostic options, and then they refine when they meet the patient," says Dr. Jacques Durand, founder of Pastel Health. Scientists, elected officials and citizens to think about all this, right now, in order to establish a real ethical framework.