Artificial intelligence can help, but nephrologists need to use it wisely
Click Here to Manage Email Alerts
Artificial intelligence and its role in clinical medicine has taken center stage, with nervous excitement building about its impact on diagnosis and management of illness amid fears of affecting the physician-patient relationship.
Artificial intelligence (AI) has been used for many years in common clinical scenarios, such as computer-generated results for electrocardiograms and white blood cell differential cell counts, as well as for analysis of retinal photographs. Recent applications also include use of natural language processing in evaluating patient experience, billing and coding software, as well as medical scribes, and the use of deep learning algorithms in smartwatches recognizing atrial fibrillation.
Clinical nephrology tends to rely heavily on data points, specifically lab data to measure kidney function, compared with some other specialties of medicine; this combined with the chronicity and burden of kidney disease is a perfect opportunity for leveraging AI in the care of patients with kidney disease.
What AI offers
Prior studies have shown the potential of AI in clinical nephrology, including a) predicting the progression of IgA nephropathy and polycystic kidney disease; b) the role in monitoring and dosing tacrolimus in kidney transplant recipients, as well as prediction of rejection and graft loss; c) dosing drugs and adjusting dialysis prescriptions in patients on hemodialysis; d) image analysis to help diagnosis and prognostic role in kidney biopsy; and e) deep learning techniques for early identification of AKI. There are other AI-assisted techniques in the pipeline focused on quality and safety, patient education and early detection of CKD and AKI among other applications.
Chatbots like ChatGPT use machine-learning algorithms to process large amounts of data to make predictions and generate responses to user queries.
ChatGPT has multiple potential roles in the field of nephrology, including but not limited to diagnosis and treatment; strategic dataset management, including large datasets such as the U.S. Renal Data System and the United Network for Organ Sharing; patient and medical education research; and decreasing physician burden with scribing, billing and coding.
Among the abstracts and posters at ASN Kidney Week, clinicians discussed the impact of AI and ChatGPT on the specialty of nephrology. One such example was using deep learning methods to estimate blood calcium levels using EKGs, also called point-of-care bloodless EKG-K. Using this technique, data from the EKG can be used to assess hyperkalemia in a prompt manner, potentially leading to early interventions.
A group of researchers from the University of Louisville used AI to leverage existing data from dialysis treatment sessions and predict complications with dialysis access. Such efforts can help develop pathways for early interventions to prevent access-related complications.
Multiple groups presented findings on the potential role of ChatGPT in augmenting patient education regarding renal diet, dialysis treatment options and living kidney donation, as well as modifying existing medical information to a below-grade level (the average reading level in the United States).
AI and kidney health
As we think about how to incorporate AI into the practice of nephrology, a few things seem to be clear. AI is ready to take on a big role, and it is up to the nephrology community to find innovative ways to do so with “augmented intelligence.”
Rather than worrying about AI replacing our expertise, we need to find ways to leverage AI to enhance our expertise, with the goal of improving clinical care, research and education. As we develop AI-enabled pathways, we also need to ensure that after the priority of improving patient care, the next priority is to improve the quality of life of health professionals and try and find ways to ease the burden of the electronic medical record and charting along the way.
AI has proven its mettle in data mining and analysis, advanced mathematical calculations to put data analysis to good use and the ability to continuously learn, which is a pillar of any good medical professional. The ability to extract medical knowledge from text and literature and apply that knowledge to clinical scenarios is not an area where AI has performed well consistently; ChatGPT did not perform as well in answering questions from the nephrology self-assessment program (NephSAP) across different areas, such as transplant, hypertension, acid-base and critical care nephrology. While this function will likely improve for ChatGPT with continuous learning, this is also an area where clinicians will take the lead, using the AI to improve their clinical decision-making skills, incorporate the “human” touch and make a plan that best suits the individual patient.
Data privacy
There are undeniable challenges with AI, with the biggest one being the privacy of data and ability to regulate how data are put to best use. Additionally, the lack of uniform and standardized data across different health systems can lead to a common malfunction known as dataset shift. This happens when a machine learning system is used in a dataset, which is different from the dataset where it was developed. There is also a concern of inherent bias toward the product and development private equity partners, who collaborate in building many of the AI platforms. This will need to be navigated with integrity and honesty by the medical community.
The lack of structured training in medical courses about AI likely needs to be revisited, as we continue to grow this specialty. Last, but not the least, there is a significant concern of the effects of AI on the physician-patient relationship. This relationship is like an evolving marriage, which tends to grow with time and adapts to changes. If the medical community embraces AI and empowers our patients to use the tools to improve their health, AI will only strengthen the physician-patient relationship. If we fight it, we likely risk distancing ourselves from our patients and potentially the benefits that AI can bring to health care.
- Reference:
- Krisanapan P, et al. Abstract TH-PO003. Presented at ASN Kidney Week, Nov. 2-5, 2023; Philadelphia.
- For more information:
- Clifford (Ford) Wayne Cleveland III, BA is a second-year medical student at the University of Alabama Heersink School of Medicine.. He can be reached at fcleveland@uab.edu.
- Gaurav Jain, MD, FASN, is the Anderson Family Endowed Professor in Nephrology and associate director of the division of nephrology in the department of medicine at the University of Alabama (UAB) Medicine | UAB Health System. He is also co-director of UAB’s home dialysis program and a member of the Editorial Advisory Board for Healio | Nephrology News & Issues. He can be reached at gjain@uabmc.edu.
Collapse
Read more about
Click Here to Manage Email Alerts