Artificial Intelligence (AI) in Healthcare

AI is the science, technology, and engineering that makes intelligent machines (Bajwa et al., 2021; Stryker, 2024). AI machines can mimic human cognitive functions through algorithms and rules, such as learning, comprehension, problem solving, decision making, creativity, and autonomy (Bajwa et al., 2021; Stryker, 2024). AI can see and identify objects, understand and respond to human language, learn new information, and act independently without human intelligence or intervention (Stryker, 2024). AI is not one independent technology; rather, it is a conglomerate of several technological subfields that work independently and in combination to add intelligence to applications (Bajwa et al., 2021).

The first human intelligence exhibited by machines was noted in the 1950s (Stryker, 2024). This was followed by the beginning of Machine Learning (ML) in the 1980s (Stryker, 2024). Machine Learning (ML) studies algorithms that automatically allow computer programs to improve (learn) through their experiences. This learning can be further categorized as supervised learning, unsupervised learning, reinforcement learning, and various other sub-fields, including “semi-supervised”, “self-supervised”, and “multi-instance ML” (Bajwa et al., 2021). These subfields are really what are allowing AI integration into applications in human healthcare.

• Supervised learning: This subfield leverages labelled data, such as X-ray images of known tumors, to detect tumors in new images (Bajwa et al., 2021).
• Unsupervised learning: This subfield attempts to extract information from data without annotated information (labels). For example, categorizing patients into groups with similar symptoms to identify a common cause (disease or pathology) (Bajwa et al., 2021).
• Reinforced learning (RL): Learning is acquired through trial and error or expert demonstration. “The algorithm learns by developing a strategy to maximise rewards. Of note, major breakthroughs in AI in recent years have been based on RL (Bajwa et al., 2021).”
• Deep learning (DL): DL started in the 2020s and is the predominant method of AI driving improvements in areas such as speech recognition (Bajwa et al., 2021; Stryker, 2024).  DL uses a class of algorithms that learn by using multilayered neural networks, called deep neural networks, to simulate the complex decision-making power of the human brain. Because it doesn’t require human intervention, the machine can learn at a tremendous scale (Stryker, 2024).
• Generative AI: Generative AI is a type of artificial intelligence that can create new content, such as text, images, audio, and video, based on existing data it has been trained on. This can include pattern recognition, content creation, product design, customer service, educational curriculum development, and healthcare functions.

AI systems have the potential to anticipate problems, deal with issues as they arise, and operate intentionally, intelligently, and adaptively. AI’s strength is learning and recognizing patterns and relationships from large multidimensional data sets much quicker than a human could. For example, AI systems could translate a patient's medical record into a single number representing a likely diagnosis (Bajwa, 2021). Furthermore, AI systems continue learning and adapting as more data becomes available.

Studies have demonstrated AI's ability to meet or exceed human expert performance in image-based diagnosis in many radiology subspecialties (Bajwa et al., 2021). AI has also demonstrated encouraging results with early detection of diseases such as breast and skin cancer, eye disease, and pneumonia using body imaging modalities (Bajwa et al., 2021; Al Kuwaiti et al., 2023). Research continues to examine ways AI can analyze X-rays, CT scans, and MRIs(Alowais et al., 2023).

AI also impacts clinical decision-making and disease diagnosis. AI can improve accuracy, reduce costs, save time compared to traditional diagnostic methods, and reduce the risk of human error. For these purposes, it can process, analyze, and report a large amount of data across different modalities, which can help physicians make better clinical decisions (Alowais et al., 2023; Al Kuwaiti et al., 2023).

One potential drawback is that the increased sensitivity could lead to increased false positives and possibly overdiagnosis by detecting minor abnormalities that might mimic subclinical disease (Al Kuwaiti et al., 2023).

AI has increased efficacy, precision, and efficiency in clinical laboratories for blood cultures, susceptibility testing, and molecular platforms. It has facilitated results within the first 24-48 hours and the selection of suitable antibiotic treatment regimens for patients with positive blood cultures (Alowais et al., 2023).

Remote Monitoring Devices (RMD)

RMDs are devices that allow healthcare providers to monitor, investigate, and track patient health data remotely. Wearable technology, such as smart watches and other wearables, can perform functions like monitoring vital signs, monitoring for and reporting falls, and even performing ECGs. Other devices, like wearable glucose monitors and Bluetooth blood pressure cuffs, allow the healthcare team to receive immediate updates through continuous data collection. Providers can track health data trends, enabling them to prevent patient decline and detect when medical intervention becomes necessary(Alenoghena et al., 2023). RMDs have been utilized to track everything from specific health-related data (mentioned above) to monitoring healthy behaviors, sleep, and general wellness.

Virtual Assistance, chatbots, and other AI interfaces are designed to simulate human conversation. These technologies can help with tasks such as identifying problems based on patient symptoms, providing medical advice, reminding patients when to take their medication, scheduling medical appointments, and monitoring vital signs (Alowais et al., 2023). Patients use AI chatbots to identify symptoms, provide insight, and recommend further action in primary care settings. AI-based apps have even been designed for non-emergency medical consultations (Al Kuwaiti et al., 2023). AI is more likely to complement rather than substitute clinicians.

AI is useful to the research process and can identify gaps in knowledge by analyzing the scholarly literature. AI can generate research ideas from knowledge gaps and suggest topics and hypotheses. AI tools can suggest research designs. AI tools help determine the optimal sample size for a study and calculate the number of participants needed to achieve statistically significant results. Just as AI is useful in research, one must remember that AI cannot determine causality. Only a well-designed study with a large sample size can provide data suggesting cause and effect.

AI is ideally suited to analyzing large amounts of complex data, searching for scientific research works, and integrating data from medical research. AI has been utilized in clinical trials in the pre-trial phase by choosing cohorts, organizing participants, and collecting and analyzing data (Al Kuwaiti et al., 2023). AI can further enhance research by providing personalized support, such as automated messaging or creating chatbots to answer common study FAQs. It can also track progress and predict which participants might be at risk of dropping out of the study. Finally, AI can analyze the information gathered and determine the research's outcome.

AI technologies have been used in drug discovery. They are mainly used to search for candidate molecules, but are likely to be used in drug discovery in the near future. AI can detect lead compounds, rapidly authenticate drug targets, and create better schematics of drug structure design. Additionally, AI can forecast the interaction between drugs and help avoid polypharmacology (Al Kuwaiti et al., 2023).

Researchers are exploring the cellular and molecular basis of disease, collecting massive multimodal datasets that can lead to understanding biological markers for diagnosis, severity, and disease progression. Through AI tools, we may be able to improve understanding of the cellular basis of disease, disease clustering, and patient populations to provide targeted prevention strategies and predict care and treatment options. Bajwa (2021) states, “This will be revolutionary for multiple standards of care, with particular impact in the cancer, neurological and rare disease space, personalising the experience of care for the individual.”

Bajwa (2021) writes that “clinicians could use an ‘AI digital consult’ to examine ‘digital twin’ models of their patients (a truly ‘digital and biomedical’ version of a patient), allowing them to ‘test’ the effectiveness, safety and experience of an intervention (such as a cancer drug) in the digital environment before delivering the intervention to the patient in the real world.”

AI systems can perform predefined tasks with high precision and be continuously active without compromising performance. AI does not suffer from fatigue and burnout as humans do. This feature of AI technology can potentially revolutionize complicated surgeries (Asan et al., 2020). The Da Vinci robotic surgical system can precisely mimic a surgeon’s hand movements (Asan et al., 2020). Significant advancements have been made in robotic AI surgical procedures, such as gynecologic, prostate, spinal, and oral and maxillofacial surgery. Robotic surgery has been shown to enhance surgical precision and predictability (Chustecki, 2024). AI-powered surgical mentorship allows skilled surgeons to offer real-time advice and guidance to other surgeons during procedures, improving surgical outcomes (Chustecki, 2024).

AI has the potential for innovative applications in the field of rehabilitation.

Similar to wearable devices used in the medical care of patients, wearable technology devices can recognize if patients are performing exercises properly and adhering to a home exercise program. AI devices can use wireless motion trackers to allow a “digital therapist” to give real-time feedback, such as “raise your leg higher”, as well as allowing the clinician to monitor and adjust treatment plans (Mozafaripour, 2022).

AI-driven systems, such as those through XRHealth, analyze data in the medical history, imaging results, and baseline performance tests to tailor exercises specific to the patient's needs (Insidea, 2024). This level of individualization ensures that each step of the patient's recovery aligns with their unique goals and physical condition.

AI is a promising avenue for sports medicine through wearable technologies. Processing data from sensors could monitor patterns in physiological measurements and positional and kinematic data to indicate how athletes can improve their performance, create injury prediction models, and increase the diagnostic precision of risk stratification systems. AI could provide a reliable technique for continuously monitoring patient health data and improving the quality of the patient experience (Al Kuwaiti et al., 2023).

At the Spinal Cord Injury Research at Emory University’s Shepherd Center, AI is used to help patients regain motor function. Through neuromodulation techniques involving transcutaneous spinal stimulation of the spinal cord, nerves are activated to promote motor function. It is envisioned that AI will someday automate the labor-intensive process of refining the TSS intensity, duration, and frequency inputs (Conciatore, 2025).

In addition to medical health benefits, AI in healthcare has other economic and social advantages, such as:

• Reducing administrative burden
• Cost efficiency and economic benefits
• Cost savings through early diagnosis
• Promoting Health Equity
• Support for Public Health Initiatives

Reducing administrative burden

AI can reduce administrative burdens by automatically populating data from therapeutic notes and past medical records, scanning for laboratory result abnormalities, and documenting patient encounters. Additionally, Robotic Process Automation (RPA) can be used for healthcare functions such as clinical records, revenue cycle administration, claims handling, and medical record management (Al Kuwaiti et al., 2023). While chatbots have already been incorporated into patient care for various patient interfaces, they can also be helpful for transactions such as booking appointments, refilling prescriptions, and facility management (Chustecki, 2024; Cutler, 2023; Al Kuwaiti et al., 2023). By reducing the rote processes that can lead to medical provider click fatigue, AI helps save time and resources for medical practitioners, ultimately leading to increased productivity and improved patient care (Chustecki, 2024).

Cost Efficiency and Economic Benefits

Besides reducing administrative burdens, AI can provide significant cost savings and economic benefits to a healthcare system. AI substitution of the rote activities that humans currently perform has the potential to create an estimated savings from $200 billion to $360 billion annually, with 35% of which would be administrative savings (Cutler, 2023). According to Cutler (2023), “This substitution will lead to a reduction in health care employment, but it is likely to be gradual as use of AI expands across areas (billing, management, scheduling, and so on).”

Institutions, such as hospital systems and post-acute facilities, are the most expensive components of medical care. Patients often need care at such an institution because they need continuous monitoring. AI to provide some of this continuous monitoring remotely could help move some of the care to home or a step-down observation unit (Cutler, 2023).

The cost of diagnosing illness may become cheaper as well. Using AI to analyze blood biomarkers, perhaps in combination with less expensive brain imaging, could replace more expensive modalities in clinical research and practice and lead to reductions in high-cost imaging and invasive testing (Cutler, 2023). The speed and accuracy of analyzing medical images can lead to quicker diagnosis, reducing health care costs associated with late-stage diagnosis. Similarly, AI's ability to process and interpret various medical tests rapidly and accurately reduces physicians' errors, further contributing to cost savings (Chustecki, 2024).

Support for Public Health Initiatives

AI can help promote awareness about public health concerns, raising awareness and possibly preventing the concern. AI can help promote information on disease prevention online, quickly reaching large numbers of people, and even analyze social media information to predict outbreaks (Mayo Clinic, 2024). Additionally, AI can analyze social media trends to reduce stigma and disinformation related to certain diseases (such as sexually transmitted infections) and use data to improve prevention and care (Chustecki, 2024).

Trust is one mechanism that affects a clinician's or patient's adoption or utilization of AI in healthcare. Trust is the belief in the truthfulness, reliability, or integrity of something (in this case, AI). Providers and patients must have a certain level of trust that AI technology provides accurate, reliable information on which healthcare decisions can be based. The level of trust in the AI significantly impacts how much the user will implement or rely on AI. A lack of trust in AI is a deterrent to adopting this technology in healthcare. This can be influenced by a user's education, past experiences, user biases, and perception toward automation (Asan et al., 2020). “Considering that an AI system might be trained with insufficient and subjective data from multiple sources, AI could generate biased or overfitted outcomes of which the clinical user might not be aware. These concerns hinder the performance of this technology, thus deterring the user’s trust and acceptance of AI systems (Asan et al., 2020).”

Factors that influence trust in AI include (Asan et al., 2020):

• Complex algorithm
• Data Sensitivity
• Cognitive Bias
• Lack of subject knowledge
• Role of AI 

Ways to improve trust in AI (Asan et al., 2020):

• Increasing transparency
• Ensuring robustness
• Encouraging fairness

It is important to note that there must be a balance between overtrust and undertrust in AI. A complete lack of trust in AI systems can significantly hinder the advancement of healthcare and medicine. Alternatively, having maximal trust and always believing and accepting all AI recommendations is not always the best. While in some applications, AI can outperform a human by quickly incorporating large amounts of data from multiple sources, having blind trust in AI may have catastrophic consequences in life-critical applications (Asan et al., 2020). AI algorithms are only as good as the data and evidence used to validate them. AI diagnosis is not always superior to human diagnosis. Asan et al. (2020) propose the concept of optimal trust: "which both the human and the AI have some level of skepticism regarding the other’s decisions since both are capable of making mistakes.” This level of healthy skepticism will “establish and maintain a properly balanced, optimal level of trust from and to the user that matches the capability of the AI system (Asan et al., 2020).”

Other drawbacks include technical intricacy, implementation of AI, and social challenges, including privacy, safety, the right to decide and try, costs, information and consent, access, and efficacy, while integrating AI into healthcare. Healthcare workers may fear a reduction in their occupations, with AI replacing them as healthcare providers.

If not properly trained, AI can lead to bias and discrimination. For example, if AI is trained on electronic health records, it is building only on people with access to healthcare and may perpetuate human bias captured within the records (Mayo Clinic, 2024).

The governance of AI applications is crucial for patient safety and accountability, as well as for raising HCPs’ belief in enhancing acceptance and boosting significant health consequences.

As we have discussed, many benefits of AI can be applied to the primary care setting. HIPAA-compliant AI systems can be used in primary care to lower administrative demands and costs and assist clinicians with gathering and organizing information. AI improves office efficiency by automating appointment scheduling, billing, and claims processing workflows. AI can help providers by identifying and flagging invalid or incomplete claims before they are submitted to payers for reimbursement, thus saving staff time and energy. The best AI systems safeguard patient information with robust privacy controls and industry-leading security features like advanced data encryption.

Furthermore, AI plays an active role in clinical processes. Esmaeilzadeh (2020) reports that AI can improve care planning, prognosis, and patient diagnostics. AI systems allow providers to focus on the patient, not on a computer screen. When performing a physical examination, an AI tool can transcribe the patient encounter simultaneously as the provider listens, palpates, and observes intently. The AI tool can then summarize the visit’s clinical content using machine learning and natural-language processing. Finally, the AI tool produces a progress note documenting findings and the visit. AI flags crucial information, such as symptoms needing attention. AI programs can predict possible complications and risks if the electronic medical record contains data from other office visits, diagnostic tests, or hospitalizations. This predictive analysis can guide educating the patient, asking additional questions, and possibly referring the patient to other specialists. The AI tool might also find clinical trial matches for patients who are not doing well.

AI can be invaluable when it comes to chronic disease management, such as:

• Seizure detection
• Glucose Monitoring
• Obesity management
• Heart disease and vital signs monitoring
• Sleep health
• Reproductive health

In pediatric healthcare, AI is a powerful tool in bolstering diagnostic accuracy, treatment planning, and long-term disease management for children. Perhaps nowhere is this impact more visible than in the field of medical imaging, where AI-powered systems are enhancing early detection of conditions like congenital heart defects, pediatric cancers, and neurological disorders. For example, recent research suggests that AI algorithms analyzing fetal echocardiograms can achieve detection accuracies exceeding 95%, allowing clinicians to identify structural heart abnormalities that might otherwise go unnoticed in traditional screenings (Day et al., 2021; Ma et al., 2024). Similarly, using AI to interpret pediatric MRI scans has significantly improved the detection of brain tumors. Health care professionals appreciate the accelerated diagnoses, which enabled earlier interventions and improved patient outcomes. Machine learning models show remarkable promise in identifying patterns linked to epilepsy and autism spectrum disorder through the evaluation of EEG data and behavioral indicators. As a result, AI is opening doors to earlier and more individualized treatment approaches (Thabtah & Peebles, 2020).

AI’s growing role extends into neonatal intensive care units (NICUs). Health care professionals appreciate AI’s ability to monitor the fragile health of premature infants. Advanced AI-driven analytics can detect subtle changes in heart rate variability, oxygen saturation, and respiratory patterns while often recognizing warning signs of distress before symptoms become visible. Predictive models can identify neonatal sepsis and respiratory distress syndrome before the onset of symptoms, giving valuable time to intervene (Masino et al., 2019). Some studies report that these systems can predict the onset of neonatal sepsis with up to 85% accuracy, a substantial improvement over conventional monitoring techniques (Helguera-Repetto, 2020; Lyra et al., 2023).

Absent the hospital setting, AI is increasingly integrated into wearable technologies designed to assist in managing chronic pediatric conditions. For children with type 1 diabetes, AI-powered insulin pumps, also known as “artificial pancreas” systems, continuously monitor blood glucose levels and autonomously adjust insulin delivery in real time. These technologies reduce episodes of hypoglycemia while improving overall glycemic control in youngsters (Moon et al., 2021). Similar wearable devices are being developed to detect and monitor seizures in children with epilepsy. These wearables provide real-time alerts to caregivers and medical professionals that may help mitigate severe complications (Bruno et al., 2021).

As with any emerging technology, integrating AI into pediatric healthcare is not without headwinds. Ethical considerations, including patient privacy, the potential for algorithmic bias, and concerns over over-reliance on automated systems, highlight the need for thoughtful and deliberate implementation. AI should serve as a supportive tool that enhances care, rather than as a substitute for passionate human caregivers capable of building relationships and delivering compassionate attention, which is essential to pediatric medicine. Rigorous regulatory oversight and cross-disciplinary collaboration will ensure that these technologies are safe, equitable, and responsive to the needs of children (Moon et al., 2021). These considerations will be equally essential in the geriatric space.

AI is rapidly reshaping how professionals approach care for older adults. While the potential benefits of these technologies are substantial, their implementation raises important questions that deserve careful attention. Increasingly, AI-powered tools are used in elder care to help manage chronic diseases, detect early signs of cognitive decline, and assist with daily caregiving responsibilities (Shiwani et al., 2023).

One of the most promising applications of AI in geriatric medicine lies in predictive analytics for chronic disease management. Research suggests that AI models can assess hospitalization risks in older adults with heart failure with over 90% accuracy, allowing professionals to intervene earlier and allocate resources strategically (Shiwani et al., 2023). AI’s adeptness in anticipating emerging health crises before life-threatening conditions improves quality of life and reduces healthcare costs.

AI is driving innovation in fall detection technology. Smart sensor systems and certain AI-equipped smartphones can analyze subtle movement patterns and identify an increased likelihood of falls. Some AI-powered fall detection systems report accuracy rates higher than 95% in identifying real-time incidents, automatically alerting caregivers or emergency responders when necessary (Kulurkar et al., 2023). For seniors who live alone, this technology is an essential lifeline.

Despite these technological advances, several challenges slow the adoption of AI in elder care. One challenge is that some older adults are not technologically capable and refuse to use AI tools.

Ultimately, while AI has opened exciting new possibilities for enhancing geriatric care, its use depends on acceptance. Once accepted, AI would augment rather than weaken human relationships at the heart of elder care. Ensuring equitable access, protecting privacy, and preserving dignity is essential in shaping AI’s role in supporting healthy aging.

AI offers health professionals new methods for early detection, personalized treatment, and ongoing monitoring of psychiatric patients. Armed with clinical knowledge and empathy algorithms, AI bridges the gap, providing clinicians with innovative solutions for improving care. Research shows that machine learning algorithms can successfully predict major depressive disorder (MDD) with an accuracy rate of approximately 80% (Lee et al., 2021).

AI-Driven Chatbots and Virtual Therapists

Some patients may never see a therapist; AI may be able to help them. AI-driven mental health assistants, such as Woebot, Wysa, and Replika, are designed to offer emotional support and therapeutic advice. These digital tools incorporate cognitive behavioral therapy (CBT) principles to help individuals monitor their mood, manage stress, and develop coping mechanisms. While these tools do not replace the need for mental healthcare professionals, research indicates that individuals using the Woebot chatbot reported a notable decrease in anxiety and depressive symptoms within as little as two weeks of interaction (Karkosz et al., 2024).

Despite advantages, such as round-the-clock availability, cost efficiency, and the ability to serve large populations, AI mental health tools present limitations and challenges. One concern is their inability to identify severe mental health crises that necessitate immediate human intervention. Skeptics argue that while AI can assist with managing symptoms, AI lacks the benefit of human empathy and the therapeutic connection that are essential components of effective mental health care (Lee et al., 2021).

Predictive Analytics and Suicide Prevention

AI has demonstrated significant promise in suicide prevention by recognizing individuals at high risk before a crisis escalates. Machine learning algorithms can evaluate behavioral trends, prescription use, and demographic factors to generate assessments of suicide risk by analyzing past cases of suicide attempts (Le Glaz et al., 2021).

A notable example comes from research conducted at Vanderbilt University Medical Center, where an AI system was developed to predict the likelihood of a suicide attempt within six months, achieving an accuracy rate exceeding 80% through an analysis of patient medical records (Walsh et al., 2017). Furthermore, AI tools designed to examine social media activity and online search patterns have been highly effective in identifying warning signals associated with suicidal ideation, enabling early intervention strategies (Coppersmith et al., 2018).

While case managers excel in primary care, they also shine in hospitals, mental health facilities, home health agencies, workers' compensation companies, schools, and insurance companies. Insurance company case managers might also use AI tools to assist with fraud detection. For example, a nurse working in home health claimed a dog bite injury, but when AI reviewed the intake assessments, none of her clients had dogs.

AI tools can identify irregularities in insurance claims filings, such as double billing, upcoding, unbundling, inconsistent medical narratives, insufficient detail, billing for work not performed, and suspicious data. Examples of suspicious claims would be a circumcision for a woman or antidepressants prescribed for a healthy newborn.