The Role Of Artificial Intelligence In Preventing Medical Errors

Introduction

According to The World Health Organisation, 1 in every 10 patients is harmed in health care due to medical errors of which 50% have been deemed preventable. A medical error can be defined as an unintentional act that results in an outcome that deviates from the intended method of care.¹ This deviation often causes harm to the patient and approximately 250000 deaths in the United States alone have been associated with medical errors, making them the third leading cause of death.² Although some medical errors are unpreventable and a result of merely being human, actions need to be taken to enhance patient safety and attempt to minimise the frequency of these errors. The incorporation of Artificial Intelligence (AI) into health care may present a solution to reduce these medical error numbers. There are many ways in which this may be achieved which will be further explored in this article. 

Understanding medical errors

Types of medical errors

Medical errors can be split broadly into active and latent errors where active errors involve a specific event that causes patient harm such as a surgeon making a mistake while operating while latent errors involve a failure within the patient care process such as a malfunctioning ventilator machine.³ While active errors require addressing at an individual level, latent errors need to be addressed at a broader, organisational level. Many different types of both active and latent medical errors are prevalent including diagnostic errors, surgical errors, equipment failures, and medication errors.⁴ 

Causes of medical errors

Human error is natural and unsurprisingly the most common cause of medical errors. However, many of these human errors can be traced back to systemic issues within the healthcare system.⁵ For example fatigue during the work shift due to the long and intensive hours required of healthcare workers can often impede healthcare workers' decision-making and lead to these errors occurring. A study interviewing resident doctors found that 66% reported fatigue and sleep deprivation to have been the cause of their medical errors.⁶ Other causes of error were reported as faulty communication with superiors and inadequate supervising. Lastly, medical errors have also been reported to have occurred due to technological malfunctions such as errors in equipment, devices or implants. 

Artificial intelligence in healthcare 

Artificial intelligence (AI) comes in many forms but broadly refers to complex computer systems which can perform tasks that were historically thought to need humans such as problem-solving. The potential of this intelligent decision-making technology in healthcare can already be seen by its widespread implementation and the three most common forms of AI used are:⁷

• Machine learning
• Natural language processing
• Robotics

Machine learning, which uses analytical algorithms to extract data, can be used to analyse large datasets such as demographics and healthcare records. This not only makes bookkeeping in the medical field increasingly accurate but can also help to personalise treatment by analysing trends in data and producing recommended treatment plans based on patient needs.8 Natural Language Processing (NLP) is another form of AI where computer programs function to understand human language. This can also function to analyse large data sets such as patient records and aid in medical transcription and decision-making.⁹ At present, machine learning and NLP are the most common forms of AI being used in healthcare.¹⁰ Lastly, there are many kinds of robotic technologies being used in various areas of healthcare such as to enhance surgery and aid with rehabilitation.¹¹ All of these variations of AI have shown great potential to lessen medical errors through alleviating workload and thereby preventing physician fatigue as well as generally enhancing the precision of healthcare. 

AI in diagnostic error prevention

Accurate medical diagnosis is essential to ensure the correct course of treatment is chosen for the patient. Errors in medical diagnosis affect approximately 5% of the US population annually and result in excess costs of over 100 billion dollars annually in the US alone. AI, specifically machine learning has been shown to increase the accuracy and speed of interpreting medical images and reduce overall diagnostic errors in this way.¹² For example AI was shown to be able to detect osteoarthritis, a generally undetectable disease using the human eye, from images of patients with 78% accuracy.¹³ This speed of AI diagnosing has also benefited time-sensitive patients such as those being admitted to trauma units and AI methods were shown to appropriately triage patients with 99% accuracy which is 13% more accurate than traditional methods.14 Moreover, AI has shown great potential in the early detection and prediction of diseases such as cancer and sepsis for which patient recovery is dependent on early treatment.¹⁵ In fact, pathologists have demonstrated that using AI decreased their error rate for recognizing cancerous lymph nodes from 3.4 to 0.5%.¹⁶ Overall, AI rapidly considers vast volumes of patient records and accurately scans medical images to produce more accurate diagnoses than possible by human physicians alone. 

AI in preventing medication errors

Medication errors, specifically inaccurate dispensaries, pose a large risk to patients' health outcomes. Physicians have proposed that a hybrid approach where humans act along with AI may best counteract these errors. AI has shown the potential to implement checkpoints that prompt physicians to double-check dosage volume and timing as well as take into consideration patients' personal information such as allergies.¹⁷ Through algorithm analysis, AI can therefore predict adverse reactions to medications such as allergies and alert healthcare workers before the errors occur. 

AI in surgery error prevention

AI-assisted robotics surgery has made significant recent advances and allows surgeons to utilise AI in a hybrid manner during surgical operations. AI has specifically been used to help surgeons perform repetitive techniques during surgery and minimally invasive surgeries. The AI robotics have more precision than human hands prone to tremors and allow for more accurate trajectory, incisions and depth. Moreover, the use of machines allows surgeons to lessen errors caused by fatigue.¹⁸ Some examples of surgeries that commonly utilise AI robotics are Da Vinci Cardiac surgery as well as FAce MOUSe which allows surgeons to control the position of a laparoscope using facial expressions and thereby limiting body movements. Aside from within surgery uses, AI has also been used as a teaching tool to simulate surgery in a virtual reality or 3D modelling manner and allow medical students to better practice techniques and thereby lessen the risk of medical errors within the theatre.¹⁸

Additionally, AI holds the potential to greatly advance post-surgery care which is a critical part of patient recovery where monitoring errors often occur and result in patient harm. Around-the-clock monitoring, though necessary, is taxing on healthcare professionals and AI can reduce fatigue-related errors by assisting with patient monitoring. Moreover, AI can also assist with telemedicine which allows for the treatment of patients remotely. This allows patients in rural areas to access health care more easily and reduces the impact of missed medical appointments. In the case of surgery, this also increases the availability of post-surgery check-ups which are essential to the patient's recovery.¹⁹

Limitations to AI in healthcare 

Despite the many potential benefits that AI usage brings, it also comes with limitations. The first and perhaps most obvious limitation is the financial burden of implementing the technology which could widen the gap in quality of healthcare between places that can and cannot afford it and lessen the standardisation of treatment. AI is also limited in that it aims to mimic human decision-making but lacks human intuition which is critical in healthcare. However, this limitation may actually prove to be an advantage as surgeons do not pose the risk of being replaced by AI technology.¹⁸ An area of the healthcare system which may limit AI’s usage is insufficient patient data collection as well as bias within the data collection process since AI technology can only scan and monitor what information is available in the first place.

Summary

Medical errors cause great harm to patients, often resulting in death, and efforts to lessen the frequency of their occurrence should be made. Some amount of human error is inevitable but a large majority of medical errors arise from healthcare worker fatigue and poor systematic organisation. In these instances, AI can greatly reduce medical errors using machine learning and robotics. AI can monitor vast numbers of patient records and assist healthcare professionals in personalised decision-making that benefits the patient as well as aids in providing the correct diagnoses. AI can also reduce workload, prevent fatigue, and increase accuracy for surgeons during surgical procedures. Although AI in an autonomous manner would lack human intuition which is critical in healthcare, a hybrid approach that combines healthcare workers and AI together can greatly improve medical care and reduce errors occurring.