Pre-conceived ideas about artificial intelligence
POINT OF VIEW
Artificial intelligence has made considerable progress in handling many tasks or activities that affect a large number of trades and industries. It is thus the subject of many fantasies and fear. The general public imagines artificial intelligence as a machine as intelligent as a human, who is self-aware and can make choices independently.
This representation is very far from reality. Existing devices are not close to being empowered and capable of thinking for themselves. It is true that there are many artificial intelligence tools in the healthcare market, and the extent of their current applications in the field of medical diagnosis seems impressive (oncology, cardiology, ophthalmology, radiology, mental health, detection rare diseases ...). But, overall, the promises are still uncertain at this stage. The scenario of a medicine without doctors, of brutal destruction of the professions of the health, raises today more of the buzz and the runaway media than science. One way to understand this media hype is to ask why artificial intelligence and health "go well together". What is the reality of artificial intelligence today and what are its strong limits? The objective of this article is not to affirm that artificial intelligence is a non-subject, on the contrary, but to make aware that the future of the French health system will be the one we want to promote. tomorrow.
"Preconceived ideas about artificial intelligence - less a threat than a potential opportunity to rethink the health system." Point of view of Salima Benhamiou, Economist at France Stratégie, department of work, employment and skills, in "La revue du praticien" of December 2018.
Why artificial intelligence and health go well together?
A "knowledge intensive" sector
Basically, the health sector is based on knowledge, which itself generates progress, and thus improves medical outcomes. Progress in medical knowledge and treatment, progress in diagnostic techniques, progress in management methods.
There are about two and a half million scientific articles published each year in the field of health. The ability to learn while keeping up with scientific advances is therefore central to this sector, as is the ability to manage learning. The health system
itself, doctors and hospitals, patient records, even automated data collection techniques (eg the number of market steps collected via a connected watch), generate large amounts of data - the so-called big data, whose exploitation is no doubt in its infancy, but it is hoped that smart exploitation can significantly improve the efficiency and effectiveness of health systems. The main obstacle at this stage is the human capacity to process these data, in addition to the challenge of their availability.
A sector that concentrates a lot of resources
The economic weight of the health sector is also considerable. Health expenditure accounts for nearly 10% of the world's gross domestic product (GDP), 17% of US GDP, or 3,200 billion US dollars in 2015 . In France, health expenditure accounts for more than 11% of GDP at least 200 billion euros. Given demographic and epidemiological developments, there is no reason for the rate of growth of these expenditures to slow down. It is therefore an extremely dynamic and buoyant market worldwide, especially as many important countries, for example China and India, have not yet caught up with the developed countries in this respect. . This important economic weight and its predictable dynamism explain why the sector generates so many "appetites".
In the developed world, there are also extremely strong constraints on public funding, and these constraints will continue. Many health actors and systems are therefore seeking solutions to improve the efficiency of the system, especially since, according to recent estimates by the Organization for Economic Co-operation and Development (OECD), at least 20 % of health expenditures would have no impact, or even degrade health status. The health sector is therefore extremely attractive, especially for GAFAM (Google, Apple,
Facebook, Amazon and Microsoft) and Chinese BATX (Baidu, Alibaba, Tencent and Xiaomi) looking for growth opportunities but also for "traditional" companies looking for new business models.
A multiple sector with a wide range of skills
Finally, the health sector, in terms of jobs and skills, is incredibly diverse: from the very high-tech represented for example by very specialized surgeons (as in neurosurgery) to simple but very demanding tasks, such as caregivers for patients suffering from cognitive degeneration, public health intervention specialists or a host of new professions related to coordination, data collection and formatting (case manager especially). Thus, there are a multitude of entry points for applications of artificial intelligence, either for ultrapointus trades, or for "mass" trades. But more fundamentally, the health sector is also a major employer, for which labor costs are an important part of health expenditure. This may lead, in a context of budgetary constraint associated with an increase in needs, to ask whether it would not be possible to replace at least some of these workers with "intelligent" machines, as in the case of was done in the automotive industry.
So there is currently a buzz around artificial intelligence in health, buzz that has even reached the general public, with the high-profile experiences of Watson Healthcare, or even an epidemiology program developed by Google, which would be able to identify outbreaks from words typed in the search engine. The "exploits" of surgical robots are regularly reported in the media.
What is the reality of artificial intelligence today?
Today, artificial intelligence is not a reality but a potentiality. There is no health system or health organization in the world that has been totally transformed, and the real applications of health remain very limited. There is no large-scale, closely related or artificial-related deployment of artificial intelligence except for some very isolated cases like IBM's Watson, Google's Deep Mind or social robots like Paro. On the other hand, what really exists in the most performing hospitals in terms of health (quality of care, patient empowerment, control of health expenditure) is the intelligent and massive use of information technologies and communication and digitalization of tools.
that is much less the case in French hospitals. But most of his investments are aimed at strengthening communication and coordination for better organizational efficiency and to strengthen the human contact between the medical and nursing team and the patient.
At the level of scientific research, there is also little scientific evidence and objectivables today on the effectiveness of artificial intelligence in the various fields of application mentioned above. The most recognized international academic journals in the medical field, such as the Journal of the American Medical Association , The Lancet , the New England Journal of Medicine and the Annals of Internal Medicine , publish few articles on the evaluation of the artificial intelligence. There is also very little medical-economic evaluation of artificial intelligence applications
that would measure their economic and social returns. But the information we have today on the impact of artificial intelligence on diagnosis will cover a sub field of a very specific discipline, for example oncology, which will focus on a certain type of cancer. .
Finally, the quality of patient care (detection of the disease, therapeutic proposal, patient follow-up, etc.) is also a very complex process, which artificial intelligence can only integrate. imperfectly. This complexity is directly related to the discipline itself and to the existence of a strong relationship between the medical team and the patient, and sometimes with his family environment. In this area, the existence of a "typical model" of a patient is generally not sufficient to develop and implement a management strategy that is totally adapted to each patient, the individual patient.
Correlation is not worth causality in the clinical field
It can also be recalled that artificial intelligence feeds on a large mass of data, via algorithms that aim to establish correlations to explain phenomena and explain them (determine their causes) to, for example, derive clinical recommendations. . The robustness of the correlations between several phenomena depends in the artificial intelligence of the mass of data collected. The more important it is, the more robust are the correlations. It is therefore big data that allows artificial intelligence to function and compete with humans through its ability to process data from a mass of continuously updated information. The strength of artificial intelligence is therefore fundamentally based on big data and access to data. Correlation does not necessarily mean causality.
Moreover, there can be a very strong variance between patients. All the complexity of the work of the health professional is precisely to take into account all these specificities, from the detection of the disease or pathologies to the therapeutic proposal. In the clinical field, correlations are not sufficient. Even evidence-based medicine that relies on the most up-to-date evaluation results can only provide clinical knowledge and recommendations based on an "average" modeled patient. The "real" patient who consults a physician does not necessarily correspond to the "average" patient who has been modeled.
Several tasks are not automatable
This is also valid for therapeutic proposals, the doctor must take into account the specific characteristics of the patient to define the right treatment. He must also often "negotiate" with him to adapt his care. Patient adherence is an important lever that influences the healing process as well as the prevention of pathologies. This activity is inherently social and inherently human, and the risk of machine empowerment is much lower than for other "routine" tasks.
Finally, artificial intelligence works not only on simple and mechanical explanatory phenomena such as the detection of fractures but also on complex but highly targeted pathologies, such as the detection of a tumor. But what happens when a patient has several pathologies? For the moment, this issue, which is central in the current debates on the future of the health system, is not answered, whereas the performance of health systems in the future will come largely from their ability to "manage" patients polypathological in the long term and will move more towards a system dedicated to the management of the prevention of polypathologies. Demographic and epidemiological trends go in this direction: tomorrow we will live longer and with several pathologies.
in charge of more and more complex and focused on prevention in which coordination between the professions of health professionals but also the medico-social will be strong.
The detection of pathologies is not measured solely by the volumetry of the data
In short, the quality of a diagnosis is not only measured by the volume of information available but by the quality of interpretation of complex mechanisms that are not based on natural and therefore non-deterministic laws. Big data works well on simple and mechanical explanatory phenomena. But the human being is a being that constantly evolves with its environment, this dynamic process makes that, in the field of medicine and support for patients, artificial intelligence can not replace health professionals, at any time. the least in their overwhelming majority. In addition, the availability of information is not enough to influence behavior: obesity has become a major public health issue, while access to health applications via a smartphone exists.
An opportunity to rethink the health system in a more integrated and systemic way
The discourse that artificial intelligence will lead to the disappearance of the human in the field of health must be rejected. In all likelihood, employment in the health sector will continue to grow, including for the medical professions as a whole. However, this does not mean that artificial intelligence is a "non-subject" in the field of health, nor that certain professions - some highly prestigious and paid (eg radiologists) - are threatened. It is unlikely that artificial intelligence will replace the work of doctors and medical staff before long. Artificial intelligence can not substitute for human work in this area. And they will not be dedicated to being mere executors of algorithms.
So what impact could artificial intelligence have on the work of health professionals, their organization of work and their working conditions?? Health systems in advanced countries have been designed to deal with acute and well-identified diseases - for example, a fracture, a cancerous tumor. In this model, which is more closely related to mechanical production, artificial intelligence can certainly have a major impact. But is this the future of our health system? It is likely that we are moving much more towards a system dedicated to the management and prevention of long-term polypathologies with a very strong psychological and behavioral component. The goal will be to achieve long-term behavioral changes, and this type of action requires close and "human" interaction with patients.
Artificial intelligence can be an opportunity to improve and evolve the health system towards a more integrated care through the improvement of coordination processes of all actors while minimizing unnecessary acts. Take the example of Kaiser, who invests heavily in artificial intelligence-based tools but also recruits many doctors, caregivers, nurses and even social workers to improve the management of complex pathologies. like obesity or diabetes. In addition, the state of research in this area shows that these diseases are of socio-economic origin, which requires developing proximity within territories to be closer to patients. Because the more the health personnel know his patient and his environment,
useless.
Artificial intelligence can lead to rethink health professions and the organization of work
Many trades as they exist today will evolve, and new trades are likely to emerge. The roles will also change, from the subspecialized physician to the home care aide, the work can be profoundly transformed. More generally, artificial intelligence will encourage the reorganization of the organizational system in which the support involves several actors. To illustrate these transformations at the level of trades, skills and at the organizational level, it is necessary to measure the degree of risk of automation of all the tasks that make up these jobs while taking into account the complexity of the tasks to be performed, the risk of acceptable error associated with each task and the degree of interdependence between all occupations within a health ecosystem.
This is what the France Strategy report published in March 2018 sought to provide as a reading grid by analyzing finely the potential effects of artificial intelligence (the benefits but also the risks) on the work in this sector, in several fields of application where the potentialities of artificial intelligence exist: diagnosis and therapeutic proposal, prevention and public health, follow-up of personalized treatments, surgery, clinical research and the field of "care". There are many benefits but also risks not to be underestimated. The overall impact on trades will largely depend on the use that decision-makers make of artificial intelligence and their vision of the health model they wish to promote
 |
| IDEALOGICAL METHODOLOGY |
POINT OF VIEW
Artificial intelligence has made considerable progress in handling many tasks or activities that affect a large number of trades and industries. It is thus the subject of many fantasies and fear. The general public imagines artificial intelligence as a machine as intelligent as a human, who is self-aware and can make choices independently.
This representation is very far from reality. Existing devices are not close to being empowered and capable of thinking for themselves. It is true that there are many artificial intelligence tools in the healthcare market, and the extent of their current applications in the field of medical diagnosis seems impressive (oncology, cardiology, ophthalmology, radiology, mental health, detection rare diseases ...). But, overall, the promises are still uncertain at this stage. The scenario of a medicine without doctors, of brutal destruction of the professions of the health, raises today more of the buzz and the runaway media than science. One way to understand this media hype is to ask why artificial intelligence and health "go well together". What is the reality of artificial intelligence today and what are its strong limits? The objective of this article is not to affirm that artificial intelligence is a non-subject, on the contrary, but to make aware that the future of the French health system will be the one we want to promote. tomorrow.
 |
| AI in robotic form |
"Preconceived ideas about artificial intelligence - less a threat than a potential opportunity to rethink the health system." Point of view of Salima Benhamiou, Economist at France Stratégie, department of work, employment and skills, in "La revue du praticien" of December 2018.
Why artificial intelligence and health go well together?
A "knowledge intensive" sector
Basically, the health sector is based on knowledge, which itself generates progress, and thus improves medical outcomes. Progress in medical knowledge and treatment, progress in diagnostic techniques, progress in management methods.
There are about two and a half million scientific articles published each year in the field of health. The ability to learn while keeping up with scientific advances is therefore central to this sector, as is the ability to manage learning. The health system
itself, doctors and hospitals, patient records, even automated data collection techniques (eg the number of market steps collected via a connected watch), generate large amounts of data - the so-called big data, whose exploitation is no doubt in its infancy, but it is hoped that smart exploitation can significantly improve the efficiency and effectiveness of health systems. The main obstacle at this stage is the human capacity to process these data, in addition to the challenge of their availability.
A sector that concentrates a lot of resources
The economic weight of the health sector is also considerable. Health expenditure accounts for nearly 10% of the world's gross domestic product (GDP), 17% of US GDP, or 3,200 billion US dollars in 2015 . In France, health expenditure accounts for more than 11% of GDP at least 200 billion euros. Given demographic and epidemiological developments, there is no reason for the rate of growth of these expenditures to slow down. It is therefore an extremely dynamic and buoyant market worldwide, especially as many important countries, for example China and India, have not yet caught up with the developed countries in this respect. . This important economic weight and its predictable dynamism explain why the sector generates so many "appetites".
In the developed world, there are also extremely strong constraints on public funding, and these constraints will continue. Many health actors and systems are therefore seeking solutions to improve the efficiency of the system, especially since, according to recent estimates by the Organization for Economic Co-operation and Development (OECD), at least 20 % of health expenditures would have no impact, or even degrade health status. The health sector is therefore extremely attractive, especially for GAFAM (Google, Apple,
Facebook, Amazon and Microsoft) and Chinese BATX (Baidu, Alibaba, Tencent and Xiaomi) looking for growth opportunities but also for "traditional" companies looking for new business models.
A multiple sector with a wide range of skills
Finally, the health sector, in terms of jobs and skills, is incredibly diverse: from the very high-tech represented for example by very specialized surgeons (as in neurosurgery) to simple but very demanding tasks, such as caregivers for patients suffering from cognitive degeneration, public health intervention specialists or a host of new professions related to coordination, data collection and formatting (case manager especially). Thus, there are a multitude of entry points for applications of artificial intelligence, either for ultrapointus trades, or for "mass" trades. But more fundamentally, the health sector is also a major employer, for which labor costs are an important part of health expenditure. This may lead, in a context of budgetary constraint associated with an increase in needs, to ask whether it would not be possible to replace at least some of these workers with "intelligent" machines, as in the case of was done in the automotive industry.
So there is currently a buzz around artificial intelligence in health, buzz that has even reached the general public, with the high-profile experiences of Watson Healthcare, or even an epidemiology program developed by Google, which would be able to identify outbreaks from words typed in the search engine. The "exploits" of surgical robots are regularly reported in the media.
What is the reality of artificial intelligence today?
Today, artificial intelligence is not a reality but a potentiality. There is no health system or health organization in the world that has been totally transformed, and the real applications of health remain very limited. There is no large-scale, closely related or artificial-related deployment of artificial intelligence except for some very isolated cases like IBM's Watson, Google's Deep Mind or social robots like Paro. On the other hand, what really exists in the most performing hospitals in terms of health (quality of care, patient empowerment, control of health expenditure) is the intelligent and massive use of information technologies and communication and digitalization of tools.
that is much less the case in French hospitals. But most of his investments are aimed at strengthening communication and coordination for better organizational efficiency and to strengthen the human contact between the medical and nursing team and the patient.
At the level of scientific research, there is also little scientific evidence and objectivables today on the effectiveness of artificial intelligence in the various fields of application mentioned above. The most recognized international academic journals in the medical field, such as the Journal of the American Medical Association , The Lancet , the New England Journal of Medicine and the Annals of Internal Medicine , publish few articles on the evaluation of the artificial intelligence. There is also very little medical-economic evaluation of artificial intelligence applications
that would measure their economic and social returns. But the information we have today on the impact of artificial intelligence on diagnosis will cover a sub field of a very specific discipline, for example oncology, which will focus on a certain type of cancer. .
Finally, the quality of patient care (detection of the disease, therapeutic proposal, patient follow-up, etc.) is also a very complex process, which artificial intelligence can only integrate. imperfectly. This complexity is directly related to the discipline itself and to the existence of a strong relationship between the medical team and the patient, and sometimes with his family environment. In this area, the existence of a "typical model" of a patient is generally not sufficient to develop and implement a management strategy that is totally adapted to each patient, the individual patient.
Correlation is not worth causality in the clinical field
It can also be recalled that artificial intelligence feeds on a large mass of data, via algorithms that aim to establish correlations to explain phenomena and explain them (determine their causes) to, for example, derive clinical recommendations. . The robustness of the correlations between several phenomena depends in the artificial intelligence of the mass of data collected. The more important it is, the more robust are the correlations. It is therefore big data that allows artificial intelligence to function and compete with humans through its ability to process data from a mass of continuously updated information. The strength of artificial intelligence is therefore fundamentally based on big data and access to data. Correlation does not necessarily mean causality.
Moreover, there can be a very strong variance between patients. All the complexity of the work of the health professional is precisely to take into account all these specificities, from the detection of the disease or pathologies to the therapeutic proposal. In the clinical field, correlations are not sufficient. Even evidence-based medicine that relies on the most up-to-date evaluation results can only provide clinical knowledge and recommendations based on an "average" modeled patient. The "real" patient who consults a physician does not necessarily correspond to the "average" patient who has been modeled.
Several tasks are not automatable
This is also valid for therapeutic proposals, the doctor must take into account the specific characteristics of the patient to define the right treatment. He must also often "negotiate" with him to adapt his care. Patient adherence is an important lever that influences the healing process as well as the prevention of pathologies. This activity is inherently social and inherently human, and the risk of machine empowerment is much lower than for other "routine" tasks.
Finally, artificial intelligence works not only on simple and mechanical explanatory phenomena such as the detection of fractures but also on complex but highly targeted pathologies, such as the detection of a tumor. But what happens when a patient has several pathologies? For the moment, this issue, which is central in the current debates on the future of the health system, is not answered, whereas the performance of health systems in the future will come largely from their ability to "manage" patients polypathological in the long term and will move more towards a system dedicated to the management of the prevention of polypathologies. Demographic and epidemiological trends go in this direction: tomorrow we will live longer and with several pathologies.
in charge of more and more complex and focused on prevention in which coordination between the professions of health professionals but also the medico-social will be strong.
The detection of pathologies is not measured solely by the volumetry of the data
In short, the quality of a diagnosis is not only measured by the volume of information available but by the quality of interpretation of complex mechanisms that are not based on natural and therefore non-deterministic laws. Big data works well on simple and mechanical explanatory phenomena. But the human being is a being that constantly evolves with its environment, this dynamic process makes that, in the field of medicine and support for patients, artificial intelligence can not replace health professionals, at any time. the least in their overwhelming majority. In addition, the availability of information is not enough to influence behavior: obesity has become a major public health issue, while access to health applications via a smartphone exists.
 |
| AI prospects |
An opportunity to rethink the health system in a more integrated and systemic way
The discourse that artificial intelligence will lead to the disappearance of the human in the field of health must be rejected. In all likelihood, employment in the health sector will continue to grow, including for the medical professions as a whole. However, this does not mean that artificial intelligence is a "non-subject" in the field of health, nor that certain professions - some highly prestigious and paid (eg radiologists) - are threatened. It is unlikely that artificial intelligence will replace the work of doctors and medical staff before long. Artificial intelligence can not substitute for human work in this area. And they will not be dedicated to being mere executors of algorithms.
So what impact could artificial intelligence have on the work of health professionals, their organization of work and their working conditions?? Health systems in advanced countries have been designed to deal with acute and well-identified diseases - for example, a fracture, a cancerous tumor. In this model, which is more closely related to mechanical production, artificial intelligence can certainly have a major impact. But is this the future of our health system? It is likely that we are moving much more towards a system dedicated to the management and prevention of long-term polypathologies with a very strong psychological and behavioral component. The goal will be to achieve long-term behavioral changes, and this type of action requires close and "human" interaction with patients.
Artificial intelligence can be an opportunity to improve and evolve the health system towards a more integrated care through the improvement of coordination processes of all actors while minimizing unnecessary acts. Take the example of Kaiser, who invests heavily in artificial intelligence-based tools but also recruits many doctors, caregivers, nurses and even social workers to improve the management of complex pathologies. like obesity or diabetes. In addition, the state of research in this area shows that these diseases are of socio-economic origin, which requires developing proximity within territories to be closer to patients. Because the more the health personnel know his patient and his environment,
useless.
Artificial intelligence can lead to rethink health professions and the organization of work
Many trades as they exist today will evolve, and new trades are likely to emerge. The roles will also change, from the subspecialized physician to the home care aide, the work can be profoundly transformed. More generally, artificial intelligence will encourage the reorganization of the organizational system in which the support involves several actors. To illustrate these transformations at the level of trades, skills and at the organizational level, it is necessary to measure the degree of risk of automation of all the tasks that make up these jobs while taking into account the complexity of the tasks to be performed, the risk of acceptable error associated with each task and the degree of interdependence between all occupations within a health ecosystem.
This is what the France Strategy report published in March 2018 sought to provide as a reading grid by analyzing finely the potential effects of artificial intelligence (the benefits but also the risks) on the work in this sector, in several fields of application where the potentialities of artificial intelligence exist: diagnosis and therapeutic proposal, prevention and public health, follow-up of personalized treatments, surgery, clinical research and the field of "care". There are many benefits but also risks not to be underestimated. The overall impact on trades will largely depend on the use that decision-makers make of artificial intelligence and their vision of the health model they wish to promote
Comments
Post a Comment