What is AI?
Artificial or artificial intelligence (AI) is defined as intelligent systems that can independently perform tasks in complex environments and improve their own performance by learning from experiences.
This definition of AI is broad and does not immediately give everyone a clear picture of all applications that fall under it. AI is already widely used in the world around us and often without realizing it. For example, the ABS that has been around for years in cars. Another example is indexing websites by Google: the search engine uses AI to help us find what we are looking for. So AI does not have to be thought of only in terms of innovative applications, such as self-driving cars or Google Home, the virtual coach who answers all your questions.
Meanwhile, we see some successful AI applications in healthcare: for example, an algorithm that evaluates1 fundus photography or an algorithm that determines a specific type of lung carcinoma.2 In both cases, the algorithm appears to be faster and more accurate than the doctor. These examples are about pattern recognition in images. The computer is very suitable for interpreting retinal images, radiological scans, histological slices, skin lesions, and ECGs.3
AI support is also promising for diagnosing and is currently used by doctors mainly in difficult cases, for example in a rare condition.4 The computer can also recognize patterns for predicting exacerbations of established conditions, for example within psychiatry.5 For choosing therapy, AI also seems effective, for example the choice of an antibiotic for urinary tract infections.6 Anyone who wants to learn more about AI quickly and easily can do so by taking the free national AI course online.7
Why AI in healthcare?
To optimize our healthcare, we need AI. An important reason why the current healthcare is not optimal without AI is the large amount of available data.3 The human brain is already not capable of processing and interpreting all this data, and this is while the available medical literature doubles every few months. Health data in the EHR will also only increase, especially when, for example, all health measurements from outside the hospital are added for patients (e.g. from wearables or health apps). Crucial valuable information in the large mass of data can be missed by healthcare providers, resulting in incorrect or missed diagnoses and harm to the patient.8
But there is another problem that makes the implementation of AI in our clinical practice necessary: the even faster increasing aging population than previously estimated. The aging population puts more pressure on healthcare: in the Netherlands, we are partly dealing with a general shortage of personnel,9 a higher report of burnout complaints among healthcare workers,10, 11 and an increase in healthcare costs.12
AI can also offer a solution in the healthcare system in reducing the workload of healthcare workers, by reducing administrative work, among others. This creates space for human interaction between the healthcare worker and the patient. AI also reduces avoidable mistakes, waste, and administration.
In all these applications, the adage is now: AI does not replace the healthcare worker, but supports medical policy.
Direction and leadership of the healthcare worker
It is important to provide direction from the heart of healthcare now that the industry is offering technical possibilities. There are already many algorithms available with high accuracy in the preclinical setting, but the proportion of algorithms that are validated and proven clinically useful is disappointing. To demonstrate the clinical usefulness of new applications, healthcare workers and the industry must work together. To prepare healthcare workers as well as possible for this new role in clinical research, an introduction to AI, project management, and medical leadership in education is important. This would also be valuable in postdoctoral training for practicing healthcare workers. This education could possibly take place at technology companies. The Netherlands Association for Clinical Physics recently started an SKMS funded project to familiarize doctors with the applications and consequences of AI for their work.13
Why is the implementation of AI not a given in healthcare?
We assume that healthcare workers are willing to implement innovation such as AI. There may be fear. What if the algorithm makes a mistake? Ethical discussions arise. Does the patient or healthcare worker want a computer to make a decision if it has no grip on how the computer arrived at the answer? Damage to the patient is not acceptable. Not causing harm is part of the Hippocratic oath for doctors and the oath for nurses and caregivers. The same oath also requires responsibility to provide the best possible care. This means that we must deal with new technical possibilities that can make healthcare better, more efficient, and cheaper.
We are convinced that cold technology can support warm care.14 What are you doing to improve healthcare with AI? Let us know!
Authors:
Jasmijn van Balveren, Clinical Chemistry Laboratory Specialist in Training.
Gabrielle Speijer, Radiotherapist-Oncologist.
Joris Arts, Hospital Pharmacist.