AI

AI

Your artificially intelligent doctor will see you now: AI and the impact on healthcare



Categories:

Wojciech Radomski, CEO of StethoMe, weighs up the pros and cons of using AI in primary care

iStock

Broadly speaking, there are two main problems facing primary care. Firstly, the high rate of misdiagnosis that occurs (cases of delayed, missed, and incorrect diagnosis are common – in the range of 10 per cent to 20 per cent and, secondly, the very high number of unnecessary visits (up to 70 per cent in the US). We are beginning to see AI affect the world we live in in a myriad of ways – the healthcare sector is no different. As Dr Bertalan Meskó, MD, PhD, wryly observes, AI will not replace doctors. However, doctors who use AI will replace those who don’t. So how can AI help with the challenges of unnecessary visits and the high rates of misdiagnosis? Let’s take a look.

Any new technologies entering the medical field, AI included, must also integrate with current practices, gain the appropriate regulatory approval and get the ‘buy-in’ from medical staff and patients alike. These challenges have led to a number of emerging trends in AI research and adoption. There has been research that has focused on tasks where AI is able to demonstrate its performance in relation to a human doctor. Such tasks usually have clearly defined inputs and a binary output that is easily validated. For example, when classifying suspicious skin lesions, the input is a digital photograph and the output is a simple binary classification: benign or malignant. Under such conditions, it is easy to demonstrate where AI had superior sensitivity and specificity over dermatologists when classifying previously unseen images of biopsy-validated lesions.

Better than machines

However, machines lack human qualities such as empathy and compassion, so it is important that patients perceive that consultations are being carried out by human doctors. It is human nature to be suspicious of new technology, especially when it could impact on a patient’s diagnosis. It is likely, therefore, that AI will initially handle tasks that are essential, but ultimately leave the primary responsibility of patient management with a human doctor. For example, clinical trials using AI to calculate target zones for head and neck radiotherapy more accurately and more quickly than a human being. The radiologist is still ultimately responsible for delivering the therapy, but AI has a significant background role in protecting the patient from harmful radiation. Another example is a smart stethoscope that records sounds that are then analysed by an AI system and its associated algorithms. The algorithms provide physicians with information regarding detected pathological sounds such as wheezes and rhonchi (characteristic of asthma for example), or fine or coarse crackles (characteristic of bronchitis and pneumonia). The StethoMe stethoscope enables convenient and accurate diagnosis and monitoring of patients without requiring a traditional visit. The doctor is still involved with the analysis of any readings and maintains ongoing contact with the patient but has a much improved diagnosis outlook given the technology used.

The subjectivity of any examination of a patient largely depends on a doctor’s experience. So, let us consider misdiagnosis. If we look at respiratory diseases which are responsible for about a fifth of all deaths worldwide and its prevalence reaches 15 per cent of the world population. What is the diagnostic ability of general practitioners (GPs) working in primary healthcare (PHC) in relation to more prevalent respiratory diseases, such as tuberculosis, asthma or chronic obstructive pulmonary disease (COPD)? In relation to asthma and COPD, studies have shown diagnostic errors leading to overdiagnosis or underdiagnosis depending on the methodology used. The lack of precision for the diagnosis of asthma varied from 54 per cent underdiagnosis to 34 per cent overdiagnosis, whereas for COPD this ranged from 81 per cent for underdiagnosis to 86.1 per cent for overdiagnosis. Such figures demonstrate that PHC, represented by the GP, needs to improve its ability for the diagnosis and management of this group of patients (with respiratory disease). A clear use case for the development of the AI stethoscope outlined earlier – giving doctors the tools to improve diagnosis without removing themselves from primary care. The overall mission of technology like AI in the healthcare system should be to eliminate human error from medical examination by increasing its quality and availability.

Wojciech Radomski

Avoiding the mundane

In the UK the growth of technology in the healthcare sector has witnessed the birth of services such as online pharmacies which serve to alleviate the increasing pressure on the NHS. It is estimated that each GP appointment costs an average of £30, putting the total cost to the NHS of missed appointments at more than £216 million pounds on top of the disruption for staff and fellow patients. Using technology in this way also serves to curb the need for unnecessary visits to the GP which might be better served by phone, email or text – for the mundane and rudimentary stuff clearly.

How medical device CEOs can navigate digital health disruption

In the first of a series of three articles, we get global leaders, McKinsey & Company's insight on the medtech market right now. They give their expert advice to medical device companies, explaining how they can navigate through digital disruption.

READ ARTICLE

Such online pharmacies operate using remote prescribing. For conditions where they can safely diagnose without a physical examination or face-to-face consultation, such pharmacies are able to offer safe and effective treatments. But this type of service is no good if the patient can’t find the treatment they are looking for or have more urgent needs. In such cases, it is always best for the patient to contact their GP or urgent care services. Technology, at the moment anyway, is not a 100 per cent replacement for the traditional GP or physician but new technologies like AI will certainly help towards reducing misdiagnosis and allowed remote treatment for non-urgent conditions.

When it comes to our health – especially in matters of life and death – the promise of artificial intelligence to improve outcomes is very intriguing. While there is still much to overcome to achieve AI-dependent health care, most notably data privacy concerns and fears of mismanaged care due to machine error and lack of human oversight, there is sufficient potential that governments, tech companies, and healthcare providers are willing to invest and test out AI-powered tools and solutions. For healthcare, AI is the future.

About the author

With well over 100 years experience between us, we've been around the editorial and medical blocks a few times. But we're still as keen as any young pup to root out what's new and inspiring.

Expert Insight

The European public-procurement opportunity: delivering value in medtech

The magic of innovation: Interview with Billy Cohn, MD

The MedTech Roadmap: a medical device’s journey from concept to market

An introduction to Beth Susanne – international pitch coach

Be first to know

Key resources

Contribute

You're the expert! Write for The Engine or share your articles, papers and research

Add your content

Add your content

Keep informed

Sign up for Ignition, our regular, ideas-packed newsletter

Sign in with social media

or with a username