Home | About | Events | Contact | Advisory Board | Writer’s Guidelines | Reprints | Organizations | Advertising
July/August 2024
July/August 2024 Issue Artificial Intelligence: AI and Geriatric Medicine Artificial intelligence tools offer opportunities for research that can inform treatment of geriatric patients. In recent years, focus on artificial intelligence (AI) and its potential impact on all aspects of life has increased dramatically. The launching of tools such as ChatGPT has pushed AI into the headlines, fostering discussions on how, when, and to what degree AI should be utilized by various individuals, organizations, and professions. The news cycles and invigorated focus make AI seem very new and underexplored. However, the reality is that AI has been a part of various fields, including geriatric medicine, for many years. The difference is “Nowadays, a layman can use it, so now everybody is aware of it,” says Eugene Lai, MD, PhD, Robert W. Hervey Distinguished Endowed Chair for Parkinson’s Disease and a professor of neurology and neuroscience at the Houston Methodist Neurological Institute. AI is used in a number of ways in geriatric medicine and, in particular, in the research that aims to better understand and inform the field. It allows, for example, research teams to use larger amounts of data—quantities that would be impossible to go through by hand or at least extraordinarily time consuming—in studies. “We have a tremendous amount of health data. Too much for a person [to sort through]. In the old days, a person would look at the records and manually write down [the data]. Now, you can put it into the computer and the computer can sort it out and can learn what is more important,” Lai says. And that’s just one example. There are numerous and wide-ranging ways in which technology such as this can be employed in research. In short, AI is a tool that when integrated properly into geriatric medicine can facilitate important insight. AI vs Machine Learning While generative AI is being used in medical research—though much differently from the way the public is utilizing it—it’s the larger subset of AI—machine learning—that better defines and encompasses past, present, and future geriatric medicine studies that involve AI. Machine learning is a type of AI that uses data and algorithms to make it possible for a computer to learn and make predictions or decisions without being given explicit instructions that would be provided in traditional programming. Facial recognition programs and even the targeted ads most individuals see on social media are examples of machine learning. Within machine learning are more subsets, like generative AI, including deep learning. “Machine learning is a subset of AI that helps people build AI-driven applications, while deep learning is a subset of machine learning that uses vast volumes of data and complex algorithms to train a model. The machine learning methods can address high dimensional data, integrate data from different sources, model the etiological and clinical heterogeneity, and discover new biomarkers,” says Kesheng Wang, PhD, a professor in biostatistics and genetic epidemiology/statistical genetics in the School of Nursing at West Virginia University. “The deep learning [AI] using artificial neural networks (ANNs)—which are inspired by the layered structure of the brain’s neurons—has reached unprecedented prediction performance for complex tasks. Therefore, deep learning has emerged as a versatile approach for predicting complex biological phenomena and displays tremendous potential in biology and medicine. It’s been shown that deep learning techniques, such as deep neural networks (DNNs) and convolutional neural networks, have been reported to be more accurate for Alzheimer’s disease [AD] diagnosis in comparison to conventional machine learning models,” he explains. Predicting Hospitalization Outcomes “We know that geriatric patients with dementia typically have more comorbid health problems and need more medical attention. Our study shows that we can predict the hospitalization outcomes of these patients on the first or second day of admission. The early assessment of their outcome means a timelier intervention, better care coordination, more judicial resource allocation, more focused care management, and timely treatment of those more vulnerable and high-risk patients,” Lai says. In the study, the researchers were able to look at the medical records of more than 8,000 patients within the Houston Methodist health system over a 10-year period. According to Lai, the data collected came from between 15,000 and 16,000 hospital encounters. “We had their complete medical records. The machine takes [the information], categorizes it, and compares patients from one group to another and looks at the risk factors” for poor hospitalization outcomes, he explains. With the data, the researchers were able to identify 20 significant risk factors that led to poor hospitalization outcomes and developed a machine learning model that can analyze data, recognizing the risk factors in an individual patient and ranking their importance. The results of this study could help providers at admitting hospitals make decisions about the care they will provide to geriatric patients with dementia, prioritizing treatment that addresses the most significant risk factors and hopefully leading to more positive hospitalization outcomes. Detecting Dementia “AD, the most common cause of dementia, is a chronic neurodegenerative disease characterized by a progressive cognitive decline resulting in disability and death. The AD etiology may start years or even decades before clinical symptom onset. Therefore, early diagnosis is especially desirable to manage disease progression,” he says. Wang further explains that biomarkers are indicators of the presence or severity of a disease. Identifying biomarkers for a specific disease could enable not only early detection but also inform treatment. “There are four main types of molecular biomarkers, including genomic biomarkers, transcriptomic biomarkers, proteomic biomarkers, and metabolic biomarkers. Metabolic biomarkers have been shown to be sensitive to AD. Early identification of predictors will impact development of future diagnostic strategies for early and definitive forms of AD, while an understanding of AD predictors will contribute to the development of therapeutic interventions to improve cognition in the early stages of AD,” Wang explains. In the study, Wang and his team identified 21 biomarkers from 150 that showed the highest accuracy for predicting AD. They then developed a deep learning model to predict AD using this information. “The present study developed deep learning models based on a multilayer feedforward ANN, also known as DNN or multilayer perceptron. This is the most common type of DNN,” he says. Moving Forward Machine learning and its subsets will continue to be a valuable part of geriatric medicine research, creating opportunities for researchers and subsequently providers to learn more than they could have without the technology. But that’s not to say that the technology will take over. The people utilizing and developing the AI will remain the most vital part of the research team. “We still need personnel and experienced statisticians and computer science personnel to make sure that we are getting correct and meaningful information. We still need the clinician to make sure the data is correct. We need the tech people to make sure that they have the AI computers working correctly,” Lai says. It is the combination of people and AI that will have the most impact on geriatric medicine and its patients. — Sue Coyle, MSW, is a freelance writer in the Philadelphia suburbs.
Reference |
