Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
If you think about the unthinkable long enough it becomes quite reasonable.Josephine Tey (1896–1952)British author
Have you ever thought about what our life as specialists for rheumatic diseases will look like in 2049? The amount of data gathered from us and our patients is increasing exponentially, and eventually, these data will be used to improve and facilitate patient’s care. We, that means primarily rheumatologists and—as the responsibilities will also change with all data—all other health professionals involved in the care of patients with rheumatic diseases, should know what to expect and actively contribute to this process.
Working routine in 2049
It is 10:00 on a sunny 21 July in 2049 and you are currently at your most favourite place in the world: a small cottage in the mountains, a vineyard in France or your house with a breath-taking view of the sea. You just finished your morning round on your virtual ward and a glimpse at your computer shows that 99% of the 50 000 customers in your virtual practice do not have any complaints and are enjoying their life without health-related limitations. In fact, most of them have never even had any symptoms as they were diagnosed before disease manifestation and preventive measures have successfully been applied. Like every morning, the system reports a few patients that deviate from their normal status. In some patients, the system has already adapted or changed therapy or has given behavioural advice. Most patients do not need any further adjustments. The system has identified two customers who need personal assessment in your virtual clinic, and therefore, an appointment has already been made. Other patients are still on your agenda for the virtual expert meeting this afternoon, as they do not fit into the known disease entities or treatment standards. Your avatar will present these complex cases to the other members of your expert board and together with already established artificial intelligence (AI)-algorithms you will find the best solution. In rare situations, it is still necessary to see the patients in person since the sensitive and trained sensors that track the patients’ condition sometimes miss a rare manifestation that is unknown to them, but relevant for making the diagnosis.
In the age of virtual patient contact, you can work from anywhere you like. However, we believe that there is more to rheumatology care in 2049 than just operating from your favourite place. There are plenty of good reasons to proactively shape our future and we would like to get you on board to discuss in which direction our medicine should evolve in the future and to reflect on your dreams, hopes and fears.
Coping with increasing amount of data nowadays
In only a few years of time, medical problem-solving has evolved quickly and changed drastically: we have a continuously increasing amount of data at our immediate disposal due to the exponential growth in medical knowledge, abundant data acquisition and the easy data availability. In 1950, medical knowledge was doubled within 50 years.1 Today, it takes only 73 days to double. At the time a medical student graduates, he or she acquired about 6% of all medical knowledge.1 Therefore, it is already impossible to keep up to date even in the rather small field of rheumatology. We are already taking the opportunity to use devices to handle new knowledge: a computer like Watson already has all data and facts from PubMed available for medical decisions.2
The vitreous patient
Big data is expected to advance personalised medicine not only in terms of diagnostics but also by improving the care of every individual patient. Terms that are frequently used in this context are the ‘omics’ like genomics, epigenomics or proteomics. Nevertheless, all data is more than the scientific data. All data means data provided by our patient: information regarding his or her medical history provided by health records, body sensors and home camera systems as well as information about all individuals from the entire world that have similar or the same complaints.3–5 In fact, all data means to have access to all data of every individual. Extended health-related information such as sleep quality and activity but also non-health-related data that is, nutritional and environmental information, consumption habits and internet activities. First consultations would be much more effective and efficient as all information would already be available to the physician without important information being lost. Every patient has given written and informed consent to the distribution of their anonymised data and all regulatory questions have been solved. In most cases, our clinical examination is replaced by modern technology as most clinical conditions can be detected by body sensors, automated ultrasound systems, whole body MRIs and skin robots that capture plenty of body parameters. However, a big challenge will be to also integrate interpersonal information such as emotions and expectations into all data, which are and will be a substantial part of rheumatology care.
Reduction of information versus integration of all data
Our perception is characterised by the fact that we constantly try to reduce information diversity by comparing and matching as much information as possible.6 In clinical practice, it is then often few seemingly inappropriate pieces of information that help us progress or we simply recognise patterns, often based on personal experiences and focus. Once a diagnosis has been made, we often stick with it as long as possible until we are forced to change it.
With all data the challenge will be to use the additional information both in the treatment of each individual and, of course, the entire field of rheumatology and beyond, which requires the connection of all data and its interpretation. Imagine that in the future, all our patient’s test results are combined into one outcome and we do not know the individual results. This system of data and their interpretation resembles a so-called complex system7 which is a network of many components that may interact with each other and evoke a complex collective behaviour, discerning information processing and adaption through learning and development.8 We will be unable to fully capture the validation and interpretation of any result and we need to rely on support from AI.
One scenario could be that we as physicians make the diagnosis and report it back to the system, which, in turn, can verify or question the diagnosis and accurately predict the patient’s prognosis. The same may apply to treatment decisions. Situations could occur in which the system disagrees with our diagnosis. In another scenario, the system makes suggestions and we check these for plausibility. The system would then serve as guidance and rather facilitate than dictate decisions. These situations need to be carefully reflected by considering both ethical and legal consequences that would rise from ignoring the system’s recommendation which is based on an infinite pool of data and algorithms. All intermediate results, for example, highly complex information from ‘omics’, will be controlled by experts, and a holistic interpretation of all areas requires all professions together or one specialist who relies on the analyses of the others. AI will take the expert’s part and constantly optimise itself and will support us by recognising and assigning specific patterns. Even today, a standardised, transparent and rigorous report procedure for AI interventions in clinical research is recommended.9 In the scenario outlined at the beginning, therapy will be adjusted by the system. This treatment decision will be based on a correction or shift in data leading to a different outcome of the algorithm. This clinical shift might not be noticed by the patient itself—the mere fact that a better therapeutic option exists, leads to the adjustment. In case of clinical symptoms, the system will react directly, initiate further diagnostic testing if necessary and adjust therapy in accordance with the ideal personalised approach. Each patient would receive the best treatment at the earliest possible time, even before symptoms occur. As it is today, our goal will be to achieve the best results for the patient.
Given that knowing all data and its connections is impossible for an individual person and even for an expert group, a selection of the most important aspects could be provided by the system on a dashboard as it has already been successfully realised with literature (Blinkist10). Should further information be required at some point, one can look deeper into the specific data by selecting it on the dashboard. The system would provide the current state of data and the specific data used to solve the respective case. This way, complex analytical processes could be broken into smaller segments, which are, in the sense of understandable AI, easier to understand and modify and facilitate the comprehension and accessibility of AI-processes for the user.
What defines us as rheumatologists in the future?
It is on us to define our role and grade of participation in this scenario. As all data is expected to highly improve patient care, ignoring this development cannot be the solution to this challenge. To this day, in addition to the empathy for patients and our communicational skills, it is our knowledge, our insights and our (clinical) experience that define us as rheumatologists. In the setting of all data, information gained from these skills and experiences is likely to get another relevance for decision-making in precise medicine. We have to ensure that relevant information and interactions are not lost when we base our decisions on knowledge not acquired and reflected by us personally but by an impersonal, alien-like ‘expert system’.
Given that it knows all data, we cannot win a knowledge battle against the system. However, rheumatologists should control the interpretation of data, or at least know which ratings they are based on and how they are generated. Being a relevant part of this process of diagnosis and treatment decision, our choices will turn into further information in the system, and algorithms will then be adjusted accordingly. We will remain the gold standard for some time until the system gains more sensitivity and specificity and will take our place by training itself. With time, the system will move away from the dichotomous evaluation of ‘normal’ and ‘abnormal’ and will be able to class everything in continuum from normal to altered by developing new ‘normalities’ of human beings that differ from the mean value of the general population. At this point, the system will live up to the diversity of all human beings and their health-related characteristics and challenges and the aim will be to target outcomes such as well-being and vitality instead of correcting anomalies. Each individual will be considered ‘normal’ in his own cohort of human beings that might be spread all over the world. This will not only change the approach to disease but allows for rare manifestations and characteristics to find a comparative population. This will significantly shape our specialty, as rare diseases will subjectively occur more frequent. Nonetheless, this development will force us to move away from the standard we have believed and confided in. Pattern recognition will become finer and the grading continues to grow as computing power and data volumes increase.
At this point, we could (1) accept less, but digitisable data and hence the omission of data or (2) get involved so that relevant data sources are still available to us. For this, we do not only have to take an active part, but we must be allowed to take decisions. We should be active participants in knowledge management and develop an ethically valuable technology for our customers. Therefore, it is crucial that we understand as much of the system as possible, an aspect that should be included in the curriculum. All data management must be part of the curricular education of every medical student.
We can integrate the system as an extension of our senses, like a blind person learning to see with a retinal implant.
The legal aspect is highly relevant as well: the rights to the system have to be placed in the right hands and every patient should be the personal data owner. The system should be able to carry itself, develop further and the profit should primarily lie with the customer as it is being developed in the European commission funded DECODE project.11
With all data being available, our misjudgements and mistakes inevitably become apparent and part of the system. Hence, they may be corrected by the system, but we should take them as a possibility for our personal and the system’s development. And surely, we will have to learn to handle mistakes differently: neither correcting or justifying mistakes nor reducing our actions and decisions to things we are fully convinced to be capable of, will be the way to success. All data is a challenge with many opportunities to improve healthcare of patients with rare diseases and with many unasked and unanswered questions to reflect on (box 1).
Questions to be addressed regarding all data
How do we want to actively shape the all data development and who do we want to be in the future?
What will be the value of our skills as rheumatologists, such as physical examination, communicational skills and empathy in the era of all data?
How will all data impact the medical training in general?
What will a rheumatologist need to know and understand of all data?
To which extent should we allow that ‘knowing’ and possibly ‘intelligent’ machines take over our genuine power and our tasks?
Will we furthermore be responsible for making diagnoses and treatment decisions?
Will subspecialties as rheumatology still be necessary or will the knowledge and the expertise within specialties be replaced by all data?
Will artificial intelligence take the expert’s part and constantly review and optimise itself or does it only help us by recognising and assigning specific patterns, which the expert must interpret himself?
Which data will be integrated into the system, how will its quality be guaranteed, and measurement errors identified?
Will the peer-review system, as we know from scientific journals, be transferred into system and the journals disappear as the medium for data communication?
Who will guarantee data security?
How will computer viruses and hacker attacks be warded off?
It is on us to shape the new developments and their implementation in our field in order to realise our visions and derive the greatest benefit for our patients and for us from the Big Data Age.
We thank Professor Jutta Richter, Professor Ralph Brinks, Professor Alfons Labisch and Dr Daniel Abrar for interesting and challenging discussions on the topic and the valuable revision of this manuscript.
Handling editor Josef S Smolen
Contributors JM, PS and MS contributed equally to the conception, drafting and writing of the manuscript. All authors read and approved the manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Commissioned; externally peer reviewed.