With the advent of artificial intelligence, should the healthcare industry the recent technological advancements for the productivity of patient treatment? And if so, to what extent should medical professionals use artificial intelligence?
Pro
An understaffed industry
It’s no surprise that the healthcare industry is quite understaffed. With physicians and medical staff overloaded, the risk of diagnosis errors increases. The Society to Improve Diagnosis in Medicine estimates that medical errors affect more than 12 million Americans each year, resulting in incorrect treatments and likely incurring costs that exceed $100 billion. By employing artificial intelligence, healthcare organizations can tackle this issue, as AI can analyze reports both more quickly and accurately than human staff.
Furthermore, there has been an increasing popularity of chatbots within many industries, and the medical industry has not been an exception. IBM reports that around 64% of patients are comfortable with interacting with chatbots for basic medical inquiries. AI nurse chatbots can assist patients by answering medication-related questions, forwarding reports to doctors, and scheduling visits. By handling these clerical tasks, these tools can alleviate the workload of clinical staff. This enables them to dedicate more time to direct patient care where human judgment and interaction are crucial.
Adept technologies
Currently, one of the main uses of AI lies in computer vision, which can identify and classify objects within medical scans. When trained on accurate datasets, computer vision models have been extensively used to analyze medical images such as X-rays, CT scans, and MRI scans. These models can detect abnormalities, such as tumors, fractures, or blockages, often with higher accuracy and speed than human radiologists.
Moreover, one study in Nature Medicine found that up to 70% of patients didn’t take insulin as prescribed. An AI-powered tool could be utilized to detect errors in a patient’s use of devices like insulin pens.
Con
Privacy
Although many computer vision models are compelling, they are reliant on bulky datasets. For example, many tumor segmentation models rely on tens of thousands of MRI scans, with each individually acquired from real patients. Furthermore, with the rise in cloud computing servers, hackers may be prone to target cloud-stored medical datasets. Health records, which are both valuable and susceptible to breaches, are often targeted by hackers. Thus, safeguarding the confidentiality of medical records is necessary. The advancement of AI might lead some users to confuse artificial systems with human operators, potentially consenting to more intrusive data collection without realizing it.
Furthermore, healthcare providers might use patient data for AI research without obtaining explicit patient consent. In 2018, Google acquired DeepMind, a leading company in healthcare AI. Google was under scrutiny soon after as it was revealed that the National Health Service had transferred data on 1.6 million patients to DeepMind servers without their consent for developing its Streams app, which includes an algorithm to manage acute kidney injuries.
Lack of legislation
In truth, even the finest machine learning models are imperfect. And that’s an issue, especially in the medical field. While a 95% accuracy for a typical image classification model would be considered “good” by general standards, the missed classifications in a medical setting could lead to serious misdiagnoses. Furthermore, when doctors are not involved in the development or management of the algorithm, it can be challenging to assign blame. Developers seem a bit distanced from the clinic, so it seems inappropriate to hold them accountable. Consequently, in regions like China and Hong Kong, the use of AI for making ethical decisions in healthcare is prohibited. Without legislation ruling on the performance requirements of such medical machine learning models, there lies an ethical dilemma.
Forbes. 28 Apr. 2024, www.forbes.com/sites/shashankagarwal/2024/04/28/
how-ai-driven-diagnostics-can-save-lives-when-humans-cant/?sh=a63981d69046. Accessed 3 May
2024.
IBM. 11 July 2023, www.ibm.com/blog/the-benefits-of-ai-in-healthcare/. Accessed 4 May 2024.
National Library of Medicine. 8 Feb. 2023, www.ncbi.nlm.nih.gov/pmc/articles/PMC9908503/. Accessed 4
May 2024.